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 -fchecking=@var{n} -fdbg-cnt-list @gol
505 -fdbg-cnt=@var{counter-value-list} @gol
506 -fdisable-ipa-@var{pass_name} @gol
507 -fdisable-rtl-@var{pass_name} @gol
508 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
509 -fdisable-tree-@var{pass_name} @gol
510 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
511 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
512 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
513 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
514 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
516 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
517 -fdump-statistics @gol
519 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
520 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
521 -fdump-tree-cfg -fdump-tree-alias @gol
523 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
524 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
525 -fdump-tree-gimple@r{[}-raw@r{]} @gol
526 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
527 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
528 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
529 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
530 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
531 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
532 -fdump-tree-nrv -fdump-tree-vect @gol
533 -fdump-tree-sink @gol
534 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
535 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
536 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
537 -fdump-tree-vtable-verify @gol
538 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
539 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
540 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
541 -fdump-final-insns=@var{file} @gol
542 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
543 -fenable-@var{kind}-@var{pass} @gol
544 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
545 -fira-verbose=@var{n} @gol
546 -flto-report -flto-report-wpa -fmem-report-wpa @gol
547 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
548 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
549 -fprofile-report @gol
550 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
551 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
552 -fstats -fstack-usage -ftime-report @gol
553 -fvar-tracking-assignments-toggle -gtoggle @gol
554 -print-file-name=@var{library} -print-libgcc-file-name @gol
555 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
556 -print-prog-name=@var{program} -print-search-dirs -Q @gol
557 -print-sysroot -print-sysroot-headers-suffix @gol
558 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
560 @item Machine-Dependent Options
561 @xref{Submodel Options,,Machine-Dependent Options}.
562 @c This list is ordered alphanumerically by subsection name.
563 @c Try and put the significant identifier (CPU or system) first,
564 @c so users have a clue at guessing where the ones they want will be.
566 @emph{AArch64 Options}
567 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
568 -mgeneral-regs-only @gol
569 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
571 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
572 -mtls-dialect=desc -mtls-dialect=traditional @gol
573 -mtls-size=@var{size} @gol
574 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
575 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
576 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
577 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
579 @emph{Adapteva Epiphany Options}
580 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
581 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
582 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
583 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
584 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
585 -msplit-vecmove-early -m1reg-@var{reg}}
588 @gccoptlist{-mbarrel-shifter @gol
589 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
590 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
591 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
592 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
593 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
594 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
595 -mlong-calls -mmedium-calls -msdata @gol
596 -mucb-mcount -mvolatile-cache @gol
597 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
598 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
599 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
600 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
601 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
602 -mtune=@var{cpu} -mmultcost=@var{num} @gol
603 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
604 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
607 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
608 -mabi=@var{name} @gol
609 -mapcs-stack-check -mno-apcs-stack-check @gol
610 -mapcs-float -mno-apcs-float @gol
611 -mapcs-reentrant -mno-apcs-reentrant @gol
612 -msched-prolog -mno-sched-prolog @gol
613 -mlittle-endian -mbig-endian @gol
614 -mfloat-abi=@var{name} @gol
615 -mfp16-format=@var{name}
616 -mthumb-interwork -mno-thumb-interwork @gol
617 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
618 -mtune=@var{name} -mprint-tune-info @gol
619 -mstructure-size-boundary=@var{n} @gol
620 -mabort-on-noreturn @gol
621 -mlong-calls -mno-long-calls @gol
622 -msingle-pic-base -mno-single-pic-base @gol
623 -mpic-register=@var{reg} @gol
624 -mnop-fun-dllimport @gol
625 -mpoke-function-name @gol
627 -mtpcs-frame -mtpcs-leaf-frame @gol
628 -mcaller-super-interworking -mcallee-super-interworking @gol
629 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
630 -mword-relocations @gol
631 -mfix-cortex-m3-ldrd @gol
632 -munaligned-access @gol
633 -mneon-for-64bits @gol
634 -mslow-flash-data @gol
635 -masm-syntax-unified @gol
639 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
640 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
641 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
643 @emph{Blackfin Options}
644 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
645 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
646 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
647 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
648 -mno-id-shared-library -mshared-library-id=@var{n} @gol
649 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
650 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
651 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
655 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
656 -msim -msdata=@var{sdata-type}}
659 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
660 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
661 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
662 -mstack-align -mdata-align -mconst-align @gol
663 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
664 -melf -maout -melinux -mlinux -sim -sim2 @gol
665 -mmul-bug-workaround -mno-mul-bug-workaround}
668 @gccoptlist{-mmac @gol
669 -mcr16cplus -mcr16c @gol
670 -msim -mint32 -mbit-ops
671 -mdata-model=@var{model}}
673 @emph{Darwin Options}
674 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
675 -arch_only -bind_at_load -bundle -bundle_loader @gol
676 -client_name -compatibility_version -current_version @gol
678 -dependency-file -dylib_file -dylinker_install_name @gol
679 -dynamic -dynamiclib -exported_symbols_list @gol
680 -filelist -flat_namespace -force_cpusubtype_ALL @gol
681 -force_flat_namespace -headerpad_max_install_names @gol
683 -image_base -init -install_name -keep_private_externs @gol
684 -multi_module -multiply_defined -multiply_defined_unused @gol
685 -noall_load -no_dead_strip_inits_and_terms @gol
686 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
687 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
688 -private_bundle -read_only_relocs -sectalign @gol
689 -sectobjectsymbols -whyload -seg1addr @gol
690 -sectcreate -sectobjectsymbols -sectorder @gol
691 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
692 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
693 -segprot -segs_read_only_addr -segs_read_write_addr @gol
694 -single_module -static -sub_library -sub_umbrella @gol
695 -twolevel_namespace -umbrella -undefined @gol
696 -unexported_symbols_list -weak_reference_mismatches @gol
697 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
698 -mkernel -mone-byte-bool}
700 @emph{DEC Alpha Options}
701 @gccoptlist{-mno-fp-regs -msoft-float @gol
702 -mieee -mieee-with-inexact -mieee-conformant @gol
703 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
704 -mtrap-precision=@var{mode} -mbuild-constants @gol
705 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
706 -mbwx -mmax -mfix -mcix @gol
707 -mfloat-vax -mfloat-ieee @gol
708 -mexplicit-relocs -msmall-data -mlarge-data @gol
709 -msmall-text -mlarge-text @gol
710 -mmemory-latency=@var{time}}
713 @gccoptlist{-msmall-model -mno-lsim}
716 @gccoptlist{-msim -mlra -mnodiv}
719 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
720 -mhard-float -msoft-float @gol
721 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
722 -mdouble -mno-double @gol
723 -mmedia -mno-media -mmuladd -mno-muladd @gol
724 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
725 -mlinked-fp -mlong-calls -malign-labels @gol
726 -mlibrary-pic -macc-4 -macc-8 @gol
727 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
728 -moptimize-membar -mno-optimize-membar @gol
729 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
730 -mvliw-branch -mno-vliw-branch @gol
731 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
732 -mno-nested-cond-exec -mtomcat-stats @gol
736 @emph{GNU/Linux Options}
737 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
738 -tno-android-cc -tno-android-ld}
740 @emph{H8/300 Options}
741 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
744 @gccoptlist{-march=@var{architecture-type} @gol
745 -mdisable-fpregs -mdisable-indexing @gol
746 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
747 -mfixed-range=@var{register-range} @gol
748 -mjump-in-delay -mlinker-opt -mlong-calls @gol
749 -mlong-load-store -mno-disable-fpregs @gol
750 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
751 -mno-jump-in-delay -mno-long-load-store @gol
752 -mno-portable-runtime -mno-soft-float @gol
753 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
754 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
755 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
756 -munix=@var{unix-std} -nolibdld -static -threads}
759 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
760 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
761 -mconstant-gp -mauto-pic -mfused-madd @gol
762 -minline-float-divide-min-latency @gol
763 -minline-float-divide-max-throughput @gol
764 -mno-inline-float-divide @gol
765 -minline-int-divide-min-latency @gol
766 -minline-int-divide-max-throughput @gol
767 -mno-inline-int-divide @gol
768 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
769 -mno-inline-sqrt @gol
770 -mdwarf2-asm -mearly-stop-bits @gol
771 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
772 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
773 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
774 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
775 -msched-spec-ldc -msched-spec-control-ldc @gol
776 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
777 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
778 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
779 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
782 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
783 -msign-extend-enabled -muser-enabled}
785 @emph{M32R/D Options}
786 @gccoptlist{-m32r2 -m32rx -m32r @gol
788 -malign-loops -mno-align-loops @gol
789 -missue-rate=@var{number} @gol
790 -mbranch-cost=@var{number} @gol
791 -mmodel=@var{code-size-model-type} @gol
792 -msdata=@var{sdata-type} @gol
793 -mno-flush-func -mflush-func=@var{name} @gol
794 -mno-flush-trap -mflush-trap=@var{number} @gol
798 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
800 @emph{M680x0 Options}
801 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
802 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
803 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
804 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
805 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
806 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
807 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
808 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
812 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
813 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
814 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
815 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
816 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
819 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
820 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
821 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
822 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
825 @emph{MicroBlaze Options}
826 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
827 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
828 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
829 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
830 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
833 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
834 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
835 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
836 -mips16 -mno-mips16 -mflip-mips16 @gol
837 -minterlink-compressed -mno-interlink-compressed @gol
838 -minterlink-mips16 -mno-interlink-mips16 @gol
839 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
840 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
841 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
842 -mno-float -msingle-float -mdouble-float @gol
843 -modd-spreg -mno-odd-spreg @gol
844 -mabs=@var{mode} -mnan=@var{encoding} @gol
845 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
848 -mvirt -mno-virt @gol
850 -mmicromips -mno-micromips @gol
851 -mfpu=@var{fpu-type} @gol
852 -msmartmips -mno-smartmips @gol
853 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
854 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
855 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
856 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
857 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
858 -membedded-data -mno-embedded-data @gol
859 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
860 -mcode-readable=@var{setting} @gol
861 -msplit-addresses -mno-split-addresses @gol
862 -mexplicit-relocs -mno-explicit-relocs @gol
863 -mcheck-zero-division -mno-check-zero-division @gol
864 -mdivide-traps -mdivide-breaks @gol
865 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
866 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
867 -mfix-24k -mno-fix-24k @gol
868 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
869 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
870 -mfix-vr4120 -mno-fix-vr4120 @gol
871 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
872 -mflush-func=@var{func} -mno-flush-func @gol
873 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
874 -mcompact-branches=@var{policy} @gol
875 -mfp-exceptions -mno-fp-exceptions @gol
876 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
877 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
878 -mframe-header-opt -mno-frame-header-opt}
881 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
882 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
883 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
884 -mno-base-addresses -msingle-exit -mno-single-exit}
886 @emph{MN10300 Options}
887 @gccoptlist{-mmult-bug -mno-mult-bug @gol
888 -mno-am33 -mam33 -mam33-2 -mam34 @gol
889 -mtune=@var{cpu-type} @gol
890 -mreturn-pointer-on-d0 @gol
891 -mno-crt0 -mrelax -mliw -msetlb}
894 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
896 @emph{MSP430 Options}
897 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
899 -mcode-region= -mdata-region= @gol
900 -msilicon-errata= -msilicon-errata-warn= @gol
904 @gccoptlist{-mbig-endian -mlittle-endian @gol
905 -mreduced-regs -mfull-regs @gol
906 -mcmov -mno-cmov @gol
907 -mperf-ext -mno-perf-ext @gol
908 -mv3push -mno-v3push @gol
909 -m16bit -mno-16bit @gol
910 -misr-vector-size=@var{num} @gol
911 -mcache-block-size=@var{num} @gol
912 -march=@var{arch} @gol
913 -mcmodel=@var{code-model} @gol
916 @emph{Nios II Options}
917 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
919 -mno-bypass-cache -mbypass-cache @gol
920 -mno-cache-volatile -mcache-volatile @gol
921 -mno-fast-sw-div -mfast-sw-div @gol
922 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
923 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
924 -mcustom-fpu-cfg=@var{name} @gol
925 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
926 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
928 @emph{Nvidia PTX Options}
929 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
931 @emph{PDP-11 Options}
932 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
933 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
934 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
935 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
936 -mbranch-expensive -mbranch-cheap @gol
937 -munix-asm -mdec-asm}
939 @emph{picoChip Options}
940 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
941 -msymbol-as-address -mno-inefficient-warnings}
943 @emph{PowerPC Options}
944 See RS/6000 and PowerPC Options.
947 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
948 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
949 -m64bit-doubles -m32bit-doubles}
951 @emph{RS/6000 and PowerPC Options}
952 @gccoptlist{-mcpu=@var{cpu-type} @gol
953 -mtune=@var{cpu-type} @gol
954 -mcmodel=@var{code-model} @gol
956 -maltivec -mno-altivec @gol
957 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
958 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
959 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
960 -mfprnd -mno-fprnd @gol
961 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
962 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
963 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
964 -malign-power -malign-natural @gol
965 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
966 -msingle-float -mdouble-float -msimple-fpu @gol
967 -mstring -mno-string -mupdate -mno-update @gol
968 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
969 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
970 -mstrict-align -mno-strict-align -mrelocatable @gol
971 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
972 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
973 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
974 -mprioritize-restricted-insns=@var{priority} @gol
975 -msched-costly-dep=@var{dependence_type} @gol
976 -minsert-sched-nops=@var{scheme} @gol
977 -mcall-sysv -mcall-netbsd @gol
978 -maix-struct-return -msvr4-struct-return @gol
979 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
980 -mblock-move-inline-limit=@var{num} @gol
981 -misel -mno-isel @gol
982 -misel=yes -misel=no @gol
984 -mspe=yes -mspe=no @gol
986 -mgen-cell-microcode -mwarn-cell-microcode @gol
987 -mvrsave -mno-vrsave @gol
988 -mmulhw -mno-mulhw @gol
989 -mdlmzb -mno-dlmzb @gol
990 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
991 -mprototype -mno-prototype @gol
992 -msim -mmvme -mads -myellowknife -memb -msdata @gol
993 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
994 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
995 -mno-recip-precision @gol
996 -mveclibabi=@var{type} -mfriz -mno-friz @gol
997 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
998 -msave-toc-indirect -mno-save-toc-indirect @gol
999 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1000 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
1001 -mquad-memory -mno-quad-memory @gol
1002 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1003 -mcompat-align-parm -mno-compat-align-parm @gol
1004 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1005 -mupper-regs -mno-upper-regs -mmodulo -mno-modulo @gol
1006 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1007 -mpower9-fusion -mno-mpower9-fusion -mpower9-vector -mno-power9-vector}
1010 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1012 -mbig-endian-data -mlittle-endian-data @gol
1015 -mas100-syntax -mno-as100-syntax@gol
1017 -mmax-constant-size=@gol
1020 -mallow-string-insns -mno-allow-string-insns@gol
1022 -mno-warn-multiple-fast-interrupts@gol
1023 -msave-acc-in-interrupts}
1025 @emph{S/390 and zSeries Options}
1026 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1027 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1028 -mlong-double-64 -mlong-double-128 @gol
1029 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1030 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1031 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1032 -mhtm -mvx -mzvector @gol
1033 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1034 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1035 -mhotpatch=@var{halfwords},@var{halfwords}}
1037 @emph{Score Options}
1038 @gccoptlist{-meb -mel @gol
1042 -mscore5 -mscore5u -mscore7 -mscore7d}
1045 @gccoptlist{-m1 -m2 -m2e @gol
1046 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1048 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1049 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1050 -mb -ml -mdalign -mrelax @gol
1051 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1052 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1053 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1054 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1055 -maccumulate-outgoing-args @gol
1056 -matomic-model=@var{atomic-model} @gol
1057 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1058 -mcbranch-force-delay-slot @gol
1059 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1060 -mpretend-cmove -mtas}
1062 @emph{Solaris 2 Options}
1063 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1066 @emph{SPARC Options}
1067 @gccoptlist{-mcpu=@var{cpu-type} @gol
1068 -mtune=@var{cpu-type} @gol
1069 -mcmodel=@var{code-model} @gol
1070 -mmemory-model=@var{mem-model} @gol
1071 -m32 -m64 -mapp-regs -mno-app-regs @gol
1072 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1073 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1074 -mhard-quad-float -msoft-quad-float @gol
1075 -mstack-bias -mno-stack-bias @gol
1076 -mstd-struct-return -mno-std-struct-return @gol
1077 -munaligned-doubles -mno-unaligned-doubles @gol
1078 -muser-mode -mno-user-mode @gol
1079 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1080 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1081 -mcbcond -mno-cbcond @gol
1082 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1083 -mfix-at697f -mfix-ut699}
1086 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1087 -msafe-dma -munsafe-dma @gol
1089 -msmall-mem -mlarge-mem -mstdmain @gol
1090 -mfixed-range=@var{register-range} @gol
1092 -maddress-space-conversion -mno-address-space-conversion @gol
1093 -mcache-size=@var{cache-size} @gol
1094 -matomic-updates -mno-atomic-updates}
1096 @emph{System V Options}
1097 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1099 @emph{TILE-Gx Options}
1100 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1101 -mcmodel=@var{code-model}}
1103 @emph{TILEPro Options}
1104 @gccoptlist{-mcpu=@var{cpu} -m32}
1107 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1108 -mprolog-function -mno-prolog-function -mspace @gol
1109 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1110 -mapp-regs -mno-app-regs @gol
1111 -mdisable-callt -mno-disable-callt @gol
1112 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1113 -mv850e -mv850 -mv850e3v5 @gol
1124 @gccoptlist{-mg -mgnu -munix}
1126 @emph{Visium Options}
1127 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1128 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1131 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1132 -mpointer-size=@var{size}}
1134 @emph{VxWorks Options}
1135 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1136 -Xbind-lazy -Xbind-now}
1139 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1140 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1141 -mfpmath=@var{unit} @gol
1142 -masm=@var{dialect} -mno-fancy-math-387 @gol
1143 -mno-fp-ret-in-387 -msoft-float @gol
1144 -mno-wide-multiply -mrtd -malign-double @gol
1145 -mpreferred-stack-boundary=@var{num} @gol
1146 -mincoming-stack-boundary=@var{num} @gol
1147 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1148 -mrecip -mrecip=@var{opt} @gol
1149 -mvzeroupper -mprefer-avx128 @gol
1150 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1151 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1152 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1153 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1154 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1155 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1156 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1157 -mpku -mthreads @gol
1158 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1159 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1160 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1161 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1162 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1163 -mregparm=@var{num} -msseregparm @gol
1164 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1165 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1166 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1167 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1168 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1169 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1170 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1171 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1174 @emph{x86 Windows Options}
1175 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1176 -mnop-fun-dllimport -mthread @gol
1177 -municode -mwin32 -mwindows -fno-set-stack-executable}
1179 @emph{Xstormy16 Options}
1182 @emph{Xtensa Options}
1183 @gccoptlist{-mconst16 -mno-const16 @gol
1184 -mfused-madd -mno-fused-madd @gol
1186 -mserialize-volatile -mno-serialize-volatile @gol
1187 -mtext-section-literals -mno-text-section-literals @gol
1188 -mauto-litpools -mno-auto-litpools @gol
1189 -mtarget-align -mno-target-align @gol
1190 -mlongcalls -mno-longcalls}
1192 @emph{zSeries Options}
1193 See S/390 and zSeries Options.
1197 @node Overall Options
1198 @section Options Controlling the Kind of Output
1200 Compilation can involve up to four stages: preprocessing, compilation
1201 proper, assembly and linking, always in that order. GCC is capable of
1202 preprocessing and compiling several files either into several
1203 assembler input files, or into one assembler input file; then each
1204 assembler input file produces an object file, and linking combines all
1205 the object files (those newly compiled, and those specified as input)
1206 into an executable file.
1208 @cindex file name suffix
1209 For any given input file, the file name suffix determines what kind of
1210 compilation is done:
1214 C source code that must be preprocessed.
1217 C source code that should not be preprocessed.
1220 C++ source code that should not be preprocessed.
1223 Objective-C source code. Note that you must link with the @file{libobjc}
1224 library to make an Objective-C program work.
1227 Objective-C source code that should not be preprocessed.
1231 Objective-C++ source code. Note that you must link with the @file{libobjc}
1232 library to make an Objective-C++ program work. Note that @samp{.M} refers
1233 to a literal capital M@.
1235 @item @var{file}.mii
1236 Objective-C++ source code that should not be preprocessed.
1239 C, C++, Objective-C or Objective-C++ header file to be turned into a
1240 precompiled header (default), or C, C++ header file to be turned into an
1241 Ada spec (via the @option{-fdump-ada-spec} switch).
1244 @itemx @var{file}.cp
1245 @itemx @var{file}.cxx
1246 @itemx @var{file}.cpp
1247 @itemx @var{file}.CPP
1248 @itemx @var{file}.c++
1250 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1251 the last two letters must both be literally @samp{x}. Likewise,
1252 @samp{.C} refers to a literal capital C@.
1256 Objective-C++ source code that must be preprocessed.
1258 @item @var{file}.mii
1259 Objective-C++ source code that should not be preprocessed.
1263 @itemx @var{file}.hp
1264 @itemx @var{file}.hxx
1265 @itemx @var{file}.hpp
1266 @itemx @var{file}.HPP
1267 @itemx @var{file}.h++
1268 @itemx @var{file}.tcc
1269 C++ header file to be turned into a precompiled header or Ada spec.
1272 @itemx @var{file}.for
1273 @itemx @var{file}.ftn
1274 Fixed form Fortran source code that should not be preprocessed.
1277 @itemx @var{file}.FOR
1278 @itemx @var{file}.fpp
1279 @itemx @var{file}.FPP
1280 @itemx @var{file}.FTN
1281 Fixed form Fortran source code that must be preprocessed (with the traditional
1284 @item @var{file}.f90
1285 @itemx @var{file}.f95
1286 @itemx @var{file}.f03
1287 @itemx @var{file}.f08
1288 Free form Fortran source code that should not be preprocessed.
1290 @item @var{file}.F90
1291 @itemx @var{file}.F95
1292 @itemx @var{file}.F03
1293 @itemx @var{file}.F08
1294 Free form Fortran source code that must be preprocessed (with the
1295 traditional preprocessor).
1300 @c FIXME: Descriptions of Java file types.
1306 @item @var{file}.ads
1307 Ada source code file that contains a library unit declaration (a
1308 declaration of a package, subprogram, or generic, or a generic
1309 instantiation), or a library unit renaming declaration (a package,
1310 generic, or subprogram renaming declaration). Such files are also
1313 @item @var{file}.adb
1314 Ada source code file containing a library unit body (a subprogram or
1315 package body). Such files are also called @dfn{bodies}.
1317 @c GCC also knows about some suffixes for languages not yet included:
1328 @itemx @var{file}.sx
1329 Assembler code that must be preprocessed.
1332 An object file to be fed straight into linking.
1333 Any file name with no recognized suffix is treated this way.
1337 You can specify the input language explicitly with the @option{-x} option:
1340 @item -x @var{language}
1341 Specify explicitly the @var{language} for the following input files
1342 (rather than letting the compiler choose a default based on the file
1343 name suffix). This option applies to all following input files until
1344 the next @option{-x} option. Possible values for @var{language} are:
1346 c c-header cpp-output
1347 c++ c++-header c++-cpp-output
1348 objective-c objective-c-header objective-c-cpp-output
1349 objective-c++ objective-c++-header objective-c++-cpp-output
1350 assembler assembler-with-cpp
1352 f77 f77-cpp-input f95 f95-cpp-input
1358 Turn off any specification of a language, so that subsequent files are
1359 handled according to their file name suffixes (as they are if @option{-x}
1360 has not been used at all).
1363 If you only want some of the stages of compilation, you can use
1364 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1365 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1366 @command{gcc} is to stop. Note that some combinations (for example,
1367 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1372 Compile or assemble the source files, but do not link. The linking
1373 stage simply is not done. The ultimate output is in the form of an
1374 object file for each source file.
1376 By default, the object file name for a source file is made by replacing
1377 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1379 Unrecognized input files, not requiring compilation or assembly, are
1384 Stop after the stage of compilation proper; do not assemble. The output
1385 is in the form of an assembler code file for each non-assembler input
1388 By default, the assembler file name for a source file is made by
1389 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1391 Input files that don't require compilation are ignored.
1395 Stop after the preprocessing stage; do not run the compiler proper. The
1396 output is in the form of preprocessed source code, which is sent to the
1399 Input files that don't require preprocessing are ignored.
1401 @cindex output file option
1404 Place output in file @var{file}. This applies to whatever
1405 sort of output is being produced, whether it be an executable file,
1406 an object file, an assembler file or preprocessed C code.
1408 If @option{-o} is not specified, the default is to put an executable
1409 file in @file{a.out}, the object file for
1410 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1411 assembler file in @file{@var{source}.s}, a precompiled header file in
1412 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1417 Print (on standard error output) the commands executed to run the stages
1418 of compilation. Also print the version number of the compiler driver
1419 program and of the preprocessor and the compiler proper.
1423 Like @option{-v} except the commands are not executed and arguments
1424 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1425 This is useful for shell scripts to capture the driver-generated command lines.
1429 Print (on the standard output) a description of the command-line options
1430 understood by @command{gcc}. If the @option{-v} option is also specified
1431 then @option{--help} is also passed on to the various processes
1432 invoked by @command{gcc}, so that they can display the command-line options
1433 they accept. If the @option{-Wextra} option has also been specified
1434 (prior to the @option{--help} option), then command-line options that
1435 have no documentation associated with them are also displayed.
1438 @opindex target-help
1439 Print (on the standard output) a description of target-specific command-line
1440 options for each tool. For some targets extra target-specific
1441 information may also be printed.
1443 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1444 Print (on the standard output) a description of the command-line
1445 options understood by the compiler that fit into all specified classes
1446 and qualifiers. These are the supported classes:
1449 @item @samp{optimizers}
1450 Display all of the optimization options supported by the
1453 @item @samp{warnings}
1454 Display all of the options controlling warning messages
1455 produced by the compiler.
1458 Display target-specific options. Unlike the
1459 @option{--target-help} option however, target-specific options of the
1460 linker and assembler are not displayed. This is because those
1461 tools do not currently support the extended @option{--help=} syntax.
1464 Display the values recognized by the @option{--param}
1467 @item @var{language}
1468 Display the options supported for @var{language}, where
1469 @var{language} is the name of one of the languages supported in this
1473 Display the options that are common to all languages.
1476 These are the supported qualifiers:
1479 @item @samp{undocumented}
1480 Display only those options that are undocumented.
1483 Display options taking an argument that appears after an equal
1484 sign in the same continuous piece of text, such as:
1485 @samp{--help=target}.
1487 @item @samp{separate}
1488 Display options taking an argument that appears as a separate word
1489 following the original option, such as: @samp{-o output-file}.
1492 Thus for example to display all the undocumented target-specific
1493 switches supported by the compiler, use:
1496 --help=target,undocumented
1499 The sense of a qualifier can be inverted by prefixing it with the
1500 @samp{^} character, so for example to display all binary warning
1501 options (i.e., ones that are either on or off and that do not take an
1502 argument) that have a description, use:
1505 --help=warnings,^joined,^undocumented
1508 The argument to @option{--help=} should not consist solely of inverted
1511 Combining several classes is possible, although this usually
1512 restricts the output so much that there is nothing to display. One
1513 case where it does work, however, is when one of the classes is
1514 @var{target}. For example, to display all the target-specific
1515 optimization options, use:
1518 --help=target,optimizers
1521 The @option{--help=} option can be repeated on the command line. Each
1522 successive use displays its requested class of options, skipping
1523 those that have already been displayed.
1525 If the @option{-Q} option appears on the command line before the
1526 @option{--help=} option, then the descriptive text displayed by
1527 @option{--help=} is changed. Instead of describing the displayed
1528 options, an indication is given as to whether the option is enabled,
1529 disabled or set to a specific value (assuming that the compiler
1530 knows this at the point where the @option{--help=} option is used).
1532 Here is a truncated example from the ARM port of @command{gcc}:
1535 % gcc -Q -mabi=2 --help=target -c
1536 The following options are target specific:
1538 -mabort-on-noreturn [disabled]
1542 The output is sensitive to the effects of previous command-line
1543 options, so for example it is possible to find out which optimizations
1544 are enabled at @option{-O2} by using:
1547 -Q -O2 --help=optimizers
1550 Alternatively you can discover which binary optimizations are enabled
1551 by @option{-O3} by using:
1554 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1555 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1556 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1561 Display the version number and copyrights of the invoked GCC@.
1563 @item -pass-exit-codes
1564 @opindex pass-exit-codes
1565 Normally the @command{gcc} program exits with the code of 1 if any
1566 phase of the compiler returns a non-success return code. If you specify
1567 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1568 the numerically highest error produced by any phase returning an error
1569 indication. The C, C++, and Fortran front ends return 4 if an internal
1570 compiler error is encountered.
1574 Use pipes rather than temporary files for communication between the
1575 various stages of compilation. This fails to work on some systems where
1576 the assembler is unable to read from a pipe; but the GNU assembler has
1579 @item -specs=@var{file}
1581 Process @var{file} after the compiler reads in the standard @file{specs}
1582 file, in order to override the defaults which the @command{gcc} driver
1583 program uses when determining what switches to pass to @command{cc1},
1584 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1585 @option{-specs=@var{file}} can be specified on the command line, and they
1586 are processed in order, from left to right. @xref{Spec Files}, for
1587 information about the format of the @var{file}.
1591 Invoke all subcommands under a wrapper program. The name of the
1592 wrapper program and its parameters are passed as a comma separated
1596 gcc -c t.c -wrapper gdb,--args
1600 This invokes all subprograms of @command{gcc} under
1601 @samp{gdb --args}, thus the invocation of @command{cc1} is
1602 @samp{gdb --args cc1 @dots{}}.
1604 @item -fplugin=@var{name}.so
1606 Load the plugin code in file @var{name}.so, assumed to be a
1607 shared object to be dlopen'd by the compiler. The base name of
1608 the shared object file is used to identify the plugin for the
1609 purposes of argument parsing (See
1610 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1611 Each plugin should define the callback functions specified in the
1614 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1615 @opindex fplugin-arg
1616 Define an argument called @var{key} with a value of @var{value}
1617 for the plugin called @var{name}.
1619 @item -fdump-ada-spec@r{[}-slim@r{]}
1620 @opindex fdump-ada-spec
1621 For C and C++ source and include files, generate corresponding Ada specs.
1622 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1623 GNAT User's Guide}, which provides detailed documentation on this feature.
1625 @item -fada-spec-parent=@var{unit}
1626 @opindex fada-spec-parent
1627 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1628 Ada specs as child units of parent @var{unit}.
1630 @item -fdump-go-spec=@var{file}
1631 @opindex fdump-go-spec
1632 For input files in any language, generate corresponding Go
1633 declarations in @var{file}. This generates Go @code{const},
1634 @code{type}, @code{var}, and @code{func} declarations which may be a
1635 useful way to start writing a Go interface to code written in some
1638 @include @value{srcdir}/../libiberty/at-file.texi
1642 @section Compiling C++ Programs
1644 @cindex suffixes for C++ source
1645 @cindex C++ source file suffixes
1646 C++ source files conventionally use one of the suffixes @samp{.C},
1647 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1648 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1649 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1650 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1651 files with these names and compiles them as C++ programs even if you
1652 call the compiler the same way as for compiling C programs (usually
1653 with the name @command{gcc}).
1657 However, the use of @command{gcc} does not add the C++ library.
1658 @command{g++} is a program that calls GCC and automatically specifies linking
1659 against the C++ library. It treats @samp{.c},
1660 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1661 files unless @option{-x} is used. This program is also useful when
1662 precompiling a C header file with a @samp{.h} extension for use in C++
1663 compilations. On many systems, @command{g++} is also installed with
1664 the name @command{c++}.
1666 @cindex invoking @command{g++}
1667 When you compile C++ programs, you may specify many of the same
1668 command-line options that you use for compiling programs in any
1669 language; or command-line options meaningful for C and related
1670 languages; or options that are meaningful only for C++ programs.
1671 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1672 explanations of options for languages related to C@.
1673 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1674 explanations of options that are meaningful only for C++ programs.
1676 @node C Dialect Options
1677 @section Options Controlling C Dialect
1678 @cindex dialect options
1679 @cindex language dialect options
1680 @cindex options, dialect
1682 The following options control the dialect of C (or languages derived
1683 from C, such as C++, Objective-C and Objective-C++) that the compiler
1687 @cindex ANSI support
1691 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1692 equivalent to @option{-std=c++98}.
1694 This turns off certain features of GCC that are incompatible with ISO
1695 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1696 such as the @code{asm} and @code{typeof} keywords, and
1697 predefined macros such as @code{unix} and @code{vax} that identify the
1698 type of system you are using. It also enables the undesirable and
1699 rarely used ISO trigraph feature. For the C compiler,
1700 it disables recognition of C++ style @samp{//} comments as well as
1701 the @code{inline} keyword.
1703 The alternate keywords @code{__asm__}, @code{__extension__},
1704 @code{__inline__} and @code{__typeof__} continue to work despite
1705 @option{-ansi}. You would not want to use them in an ISO C program, of
1706 course, but it is useful to put them in header files that might be included
1707 in compilations done with @option{-ansi}. Alternate predefined macros
1708 such as @code{__unix__} and @code{__vax__} are also available, with or
1709 without @option{-ansi}.
1711 The @option{-ansi} option does not cause non-ISO programs to be
1712 rejected gratuitously. For that, @option{-Wpedantic} is required in
1713 addition to @option{-ansi}. @xref{Warning Options}.
1715 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1716 option is used. Some header files may notice this macro and refrain
1717 from declaring certain functions or defining certain macros that the
1718 ISO standard doesn't call for; this is to avoid interfering with any
1719 programs that might use these names for other things.
1721 Functions that are normally built in but do not have semantics
1722 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1723 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1724 built-in functions provided by GCC}, for details of the functions
1729 Determine the language standard. @xref{Standards,,Language Standards
1730 Supported by GCC}, for details of these standard versions. This option
1731 is currently only supported when compiling C or C++.
1733 The compiler can accept several base standards, such as @samp{c90} or
1734 @samp{c++98}, and GNU dialects of those standards, such as
1735 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1736 compiler accepts all programs following that standard plus those
1737 using GNU extensions that do not contradict it. For example,
1738 @option{-std=c90} turns off certain features of GCC that are
1739 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1740 keywords, but not other GNU extensions that do not have a meaning in
1741 ISO C90, such as omitting the middle term of a @code{?:}
1742 expression. On the other hand, when a GNU dialect of a standard is
1743 specified, all features supported by the compiler are enabled, even when
1744 those features change the meaning of the base standard. As a result, some
1745 strict-conforming programs may be rejected. The particular standard
1746 is used by @option{-Wpedantic} to identify which features are GNU
1747 extensions given that version of the standard. For example
1748 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1749 comments, while @option{-std=gnu99 -Wpedantic} does not.
1751 A value for this option must be provided; possible values are
1757 Support all ISO C90 programs (certain GNU extensions that conflict
1758 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1760 @item iso9899:199409
1761 ISO C90 as modified in amendment 1.
1767 ISO C99. This standard is substantially completely supported, modulo
1768 bugs and floating-point issues
1769 (mainly but not entirely relating to optional C99 features from
1770 Annexes F and G). See
1771 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1772 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1777 ISO C11, the 2011 revision of the ISO C standard. This standard is
1778 substantially completely supported, modulo bugs, floating-point issues
1779 (mainly but not entirely relating to optional C11 features from
1780 Annexes F and G) and the optional Annexes K (Bounds-checking
1781 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1785 GNU dialect of ISO C90 (including some C99 features).
1789 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1793 GNU dialect of ISO C11. This is the default for C code.
1794 The name @samp{gnu1x} is deprecated.
1798 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1799 additional defect reports. Same as @option{-ansi} for C++ code.
1803 GNU dialect of @option{-std=c++98}.
1807 The 2011 ISO C++ standard plus amendments.
1808 The name @samp{c++0x} is deprecated.
1812 GNU dialect of @option{-std=c++11}.
1813 The name @samp{gnu++0x} is deprecated.
1817 The 2014 ISO C++ standard plus amendments.
1818 The name @samp{c++1y} is deprecated.
1822 GNU dialect of @option{-std=c++14}.
1823 This is the default for C++ code.
1824 The name @samp{gnu++1y} is deprecated.
1827 The next revision of the ISO C++ standard, tentatively planned for
1828 2017. Support is highly experimental, and will almost certainly
1829 change in incompatible ways in future releases.
1832 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1833 and will almost certainly change in incompatible ways in future
1837 @item -fgnu89-inline
1838 @opindex fgnu89-inline
1839 The option @option{-fgnu89-inline} tells GCC to use the traditional
1840 GNU semantics for @code{inline} functions when in C99 mode.
1841 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1842 Using this option is roughly equivalent to adding the
1843 @code{gnu_inline} function attribute to all inline functions
1844 (@pxref{Function Attributes}).
1846 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1847 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1848 specifies the default behavior).
1849 This option is not supported in @option{-std=c90} or
1850 @option{-std=gnu90} mode.
1852 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1853 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1854 in effect for @code{inline} functions. @xref{Common Predefined
1855 Macros,,,cpp,The C Preprocessor}.
1857 @item -aux-info @var{filename}
1859 Output to the given filename prototyped declarations for all functions
1860 declared and/or defined in a translation unit, including those in header
1861 files. This option is silently ignored in any language other than C@.
1863 Besides declarations, the file indicates, in comments, the origin of
1864 each declaration (source file and line), whether the declaration was
1865 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1866 @samp{O} for old, respectively, in the first character after the line
1867 number and the colon), and whether it came from a declaration or a
1868 definition (@samp{C} or @samp{F}, respectively, in the following
1869 character). In the case of function definitions, a K&R-style list of
1870 arguments followed by their declarations is also provided, inside
1871 comments, after the declaration.
1873 @item -fallow-parameterless-variadic-functions
1874 @opindex fallow-parameterless-variadic-functions
1875 Accept variadic functions without named parameters.
1877 Although it is possible to define such a function, this is not very
1878 useful as it is not possible to read the arguments. This is only
1879 supported for C as this construct is allowed by C++.
1883 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1884 keyword, so that code can use these words as identifiers. You can use
1885 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1886 instead. @option{-ansi} implies @option{-fno-asm}.
1888 In C++, this switch only affects the @code{typeof} keyword, since
1889 @code{asm} and @code{inline} are standard keywords. You may want to
1890 use the @option{-fno-gnu-keywords} flag instead, which has the same
1891 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1892 switch only affects the @code{asm} and @code{typeof} keywords, since
1893 @code{inline} is a standard keyword in ISO C99.
1896 @itemx -fno-builtin-@var{function}
1897 @opindex fno-builtin
1898 @cindex built-in functions
1899 Don't recognize built-in functions that do not begin with
1900 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1901 functions provided by GCC}, for details of the functions affected,
1902 including those which are not built-in functions when @option{-ansi} or
1903 @option{-std} options for strict ISO C conformance are used because they
1904 do not have an ISO standard meaning.
1906 GCC normally generates special code to handle certain built-in functions
1907 more efficiently; for instance, calls to @code{alloca} may become single
1908 instructions which adjust the stack directly, and calls to @code{memcpy}
1909 may become inline copy loops. The resulting code is often both smaller
1910 and faster, but since the function calls no longer appear as such, you
1911 cannot set a breakpoint on those calls, nor can you change the behavior
1912 of the functions by linking with a different library. In addition,
1913 when a function is recognized as a built-in function, GCC may use
1914 information about that function to warn about problems with calls to
1915 that function, or to generate more efficient code, even if the
1916 resulting code still contains calls to that function. For example,
1917 warnings are given with @option{-Wformat} for bad calls to
1918 @code{printf} when @code{printf} is built in and @code{strlen} is
1919 known not to modify global memory.
1921 With the @option{-fno-builtin-@var{function}} option
1922 only the built-in function @var{function} is
1923 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1924 function is named that is not built-in in this version of GCC, this
1925 option is ignored. There is no corresponding
1926 @option{-fbuiltin-@var{function}} option; if you wish to enable
1927 built-in functions selectively when using @option{-fno-builtin} or
1928 @option{-ffreestanding}, you may define macros such as:
1931 #define abs(n) __builtin_abs ((n))
1932 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1937 @cindex hosted environment
1939 Assert that compilation targets a hosted environment. This implies
1940 @option{-fbuiltin}. A hosted environment is one in which the
1941 entire standard library is available, and in which @code{main} has a return
1942 type of @code{int}. Examples are nearly everything except a kernel.
1943 This is equivalent to @option{-fno-freestanding}.
1945 @item -ffreestanding
1946 @opindex ffreestanding
1947 @cindex hosted environment
1949 Assert that compilation targets a freestanding environment. This
1950 implies @option{-fno-builtin}. A freestanding environment
1951 is one in which the standard library may not exist, and program startup may
1952 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1953 This is equivalent to @option{-fno-hosted}.
1955 @xref{Standards,,Language Standards Supported by GCC}, for details of
1956 freestanding and hosted environments.
1960 @cindex OpenACC accelerator programming
1961 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1962 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1963 compiler generates accelerated code according to the OpenACC Application
1964 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1965 implies @option{-pthread}, and thus is only supported on targets that
1966 have support for @option{-pthread}.
1968 @item -fopenacc-dim=@var{geom}
1969 @opindex fopenacc-dim
1970 @cindex OpenACC accelerator programming
1971 Specify default compute dimensions for parallel offload regions that do
1972 not explicitly specify. The @var{geom} value is a triple of
1973 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
1974 can be omitted, to use a target-specific default value.
1978 @cindex OpenMP parallel
1979 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1980 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1981 compiler generates parallel code according to the OpenMP Application
1982 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1983 implies @option{-pthread}, and thus is only supported on targets that
1984 have support for @option{-pthread}. @option{-fopenmp} implies
1985 @option{-fopenmp-simd}.
1988 @opindex fopenmp-simd
1991 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1992 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1997 @cindex Enable Cilk Plus
1998 Enable the usage of Cilk Plus language extension features for C/C++.
1999 When the option @option{-fcilkplus} is specified, enable the usage of
2000 the Cilk Plus Language extension features for C/C++. The present
2001 implementation follows ABI version 1.2. This is an experimental
2002 feature that is only partially complete, and whose interface may
2003 change in future versions of GCC as the official specification
2004 changes. Currently, all features but @code{_Cilk_for} have been
2009 When the option @option{-fgnu-tm} is specified, the compiler
2010 generates code for the Linux variant of Intel's current Transactional
2011 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2012 an experimental feature whose interface may change in future versions
2013 of GCC, as the official specification changes. Please note that not
2014 all architectures are supported for this feature.
2016 For more information on GCC's support for transactional memory,
2017 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2018 Transactional Memory Library}.
2020 Note that the transactional memory feature is not supported with
2021 non-call exceptions (@option{-fnon-call-exceptions}).
2023 @item -fms-extensions
2024 @opindex fms-extensions
2025 Accept some non-standard constructs used in Microsoft header files.
2027 In C++ code, this allows member names in structures to be similar
2028 to previous types declarations.
2037 Some cases of unnamed fields in structures and unions are only
2038 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2039 fields within structs/unions}, for details.
2041 Note that this option is off for all targets but x86
2042 targets using ms-abi.
2044 @item -fplan9-extensions
2045 @opindex fplan9-extensions
2046 Accept some non-standard constructs used in Plan 9 code.
2048 This enables @option{-fms-extensions}, permits passing pointers to
2049 structures with anonymous fields to functions that expect pointers to
2050 elements of the type of the field, and permits referring to anonymous
2051 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2052 struct/union fields within structs/unions}, for details. This is only
2053 supported for C, not C++.
2057 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2058 options for strict ISO C conformance) implies @option{-trigraphs}.
2060 @cindex traditional C language
2061 @cindex C language, traditional
2063 @itemx -traditional-cpp
2064 @opindex traditional-cpp
2065 @opindex traditional
2066 Formerly, these options caused GCC to attempt to emulate a pre-standard
2067 C compiler. They are now only supported with the @option{-E} switch.
2068 The preprocessor continues to support a pre-standard mode. See the GNU
2069 CPP manual for details.
2071 @item -fcond-mismatch
2072 @opindex fcond-mismatch
2073 Allow conditional expressions with mismatched types in the second and
2074 third arguments. The value of such an expression is void. This option
2075 is not supported for C++.
2077 @item -flax-vector-conversions
2078 @opindex flax-vector-conversions
2079 Allow implicit conversions between vectors with differing numbers of
2080 elements and/or incompatible element types. This option should not be
2083 @item -funsigned-char
2084 @opindex funsigned-char
2085 Let the type @code{char} be unsigned, like @code{unsigned char}.
2087 Each kind of machine has a default for what @code{char} should
2088 be. It is either like @code{unsigned char} by default or like
2089 @code{signed char} by default.
2091 Ideally, a portable program should always use @code{signed char} or
2092 @code{unsigned char} when it depends on the signedness of an object.
2093 But many programs have been written to use plain @code{char} and
2094 expect it to be signed, or expect it to be unsigned, depending on the
2095 machines they were written for. This option, and its inverse, let you
2096 make such a program work with the opposite default.
2098 The type @code{char} is always a distinct type from each of
2099 @code{signed char} or @code{unsigned char}, even though its behavior
2100 is always just like one of those two.
2103 @opindex fsigned-char
2104 Let the type @code{char} be signed, like @code{signed char}.
2106 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2107 the negative form of @option{-funsigned-char}. Likewise, the option
2108 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2110 @item -fsigned-bitfields
2111 @itemx -funsigned-bitfields
2112 @itemx -fno-signed-bitfields
2113 @itemx -fno-unsigned-bitfields
2114 @opindex fsigned-bitfields
2115 @opindex funsigned-bitfields
2116 @opindex fno-signed-bitfields
2117 @opindex fno-unsigned-bitfields
2118 These options control whether a bit-field is signed or unsigned, when the
2119 declaration does not use either @code{signed} or @code{unsigned}. By
2120 default, such a bit-field is signed, because this is consistent: the
2121 basic integer types such as @code{int} are signed types.
2123 @item -fsso-struct=@var{endianness}
2124 @opindex fsso-struct
2125 Set the default scalar storage order of structures and unions to the
2126 specified endianness. The accepted values are @samp{big-endian} and
2127 @samp{little-endian}. If the option is not passed, the compiler uses
2128 the native endianness of the target. This option is not supported for C++.
2130 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2131 code that is not binary compatible with code generated without it if the
2132 specified endianness is not the native endianness of the target.
2135 @node C++ Dialect Options
2136 @section Options Controlling C++ Dialect
2138 @cindex compiler options, C++
2139 @cindex C++ options, command-line
2140 @cindex options, C++
2141 This section describes the command-line options that are only meaningful
2142 for C++ programs. You can also use most of the GNU compiler options
2143 regardless of what language your program is in. For example, you
2144 might compile a file @file{firstClass.C} like this:
2147 g++ -g -fstrict-enums -O -c firstClass.C
2151 In this example, only @option{-fstrict-enums} is an option meant
2152 only for C++ programs; you can use the other options with any
2153 language supported by GCC@.
2155 Some options for compiling C programs, such as @option{-std}, are also
2156 relevant for C++ programs.
2157 @xref{C Dialect Options,,Options Controlling C Dialect}.
2159 Here is a list of options that are @emph{only} for compiling C++ programs:
2163 @item -fabi-version=@var{n}
2164 @opindex fabi-version
2165 Use version @var{n} of the C++ ABI@. The default is version 0.
2167 Version 0 refers to the version conforming most closely to
2168 the C++ ABI specification. Therefore, the ABI obtained using version 0
2169 will change in different versions of G++ as ABI bugs are fixed.
2171 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2173 Version 2 is the version of the C++ ABI that first appeared in G++
2174 3.4, and was the default through G++ 4.9.
2176 Version 3 corrects an error in mangling a constant address as a
2179 Version 4, which first appeared in G++ 4.5, implements a standard
2180 mangling for vector types.
2182 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2183 attribute const/volatile on function pointer types, decltype of a
2184 plain decl, and use of a function parameter in the declaration of
2187 Version 6, which first appeared in G++ 4.7, corrects the promotion
2188 behavior of C++11 scoped enums and the mangling of template argument
2189 packs, const/static_cast, prefix ++ and --, and a class scope function
2190 used as a template argument.
2192 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2193 builtin type and corrects the mangling of lambdas in default argument
2196 Version 8, which first appeared in G++ 4.9, corrects the substitution
2197 behavior of function types with function-cv-qualifiers.
2199 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2202 Version 10, which first appeared in G++ 6.1, adds mangling of
2203 attributes that affect type identity, such as ia32 calling convention
2204 attributes (e.g. @samp{stdcall}).
2206 See also @option{-Wabi}.
2208 @item -fabi-compat-version=@var{n}
2209 @opindex fabi-compat-version
2210 On targets that support strong aliases, G++
2211 works around mangling changes by creating an alias with the correct
2212 mangled name when defining a symbol with an incorrect mangled name.
2213 This switch specifies which ABI version to use for the alias.
2215 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2216 compatibility). If another ABI version is explicitly selected, this
2217 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2218 use @option{-fabi-compat-version=2}.
2220 If this option is not provided but @option{-Wabi=@var{n}} is, that
2221 version is used for compatibility aliases. If this option is provided
2222 along with @option{-Wabi} (without the version), the version from this
2223 option is used for the warning.
2225 @item -fno-access-control
2226 @opindex fno-access-control
2227 Turn off all access checking. This switch is mainly useful for working
2228 around bugs in the access control code.
2232 Check that the pointer returned by @code{operator new} is non-null
2233 before attempting to modify the storage allocated. This check is
2234 normally unnecessary because the C++ standard specifies that
2235 @code{operator new} only returns @code{0} if it is declared
2236 @code{throw()}, in which case the compiler always checks the
2237 return value even without this option. In all other cases, when
2238 @code{operator new} has a non-empty exception specification, memory
2239 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2240 @samp{new (nothrow)}.
2244 Enable support for the C++ Extensions for Concepts Technical
2245 Specification, ISO 19217 (2015), which allows code like
2248 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2249 template <Addable T> T add (T a, T b) @{ return a + b; @}
2252 @item -fconstexpr-depth=@var{n}
2253 @opindex fconstexpr-depth
2254 Set the maximum nested evaluation depth for C++11 constexpr functions
2255 to @var{n}. A limit is needed to detect endless recursion during
2256 constant expression evaluation. The minimum specified by the standard
2259 @item -fdeduce-init-list
2260 @opindex fdeduce-init-list
2261 Enable deduction of a template type parameter as
2262 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2265 template <class T> auto forward(T t) -> decltype (realfn (t))
2272 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2276 This deduction was implemented as a possible extension to the
2277 originally proposed semantics for the C++11 standard, but was not part
2278 of the final standard, so it is disabled by default. This option is
2279 deprecated, and may be removed in a future version of G++.
2281 @item -ffriend-injection
2282 @opindex ffriend-injection
2283 Inject friend functions into the enclosing namespace, so that they are
2284 visible outside the scope of the class in which they are declared.
2285 Friend functions were documented to work this way in the old Annotated
2286 C++ Reference Manual.
2287 However, in ISO C++ a friend function that is not declared
2288 in an enclosing scope can only be found using argument dependent
2289 lookup. GCC defaults to the standard behavior.
2291 This option is for compatibility, and may be removed in a future
2294 @item -fno-elide-constructors
2295 @opindex fno-elide-constructors
2296 The C++ standard allows an implementation to omit creating a temporary
2297 that is only used to initialize another object of the same type.
2298 Specifying this option disables that optimization, and forces G++ to
2299 call the copy constructor in all cases.
2301 @item -fno-enforce-eh-specs
2302 @opindex fno-enforce-eh-specs
2303 Don't generate code to check for violation of exception specifications
2304 at run time. This option violates the C++ standard, but may be useful
2305 for reducing code size in production builds, much like defining
2306 @code{NDEBUG}. This does not give user code permission to throw
2307 exceptions in violation of the exception specifications; the compiler
2308 still optimizes based on the specifications, so throwing an
2309 unexpected exception results in undefined behavior at run time.
2311 @item -fextern-tls-init
2312 @itemx -fno-extern-tls-init
2313 @opindex fextern-tls-init
2314 @opindex fno-extern-tls-init
2315 The C++11 and OpenMP standards allow @code{thread_local} and
2316 @code{threadprivate} variables to have dynamic (runtime)
2317 initialization. To support this, any use of such a variable goes
2318 through a wrapper function that performs any necessary initialization.
2319 When the use and definition of the variable are in the same
2320 translation unit, this overhead can be optimized away, but when the
2321 use is in a different translation unit there is significant overhead
2322 even if the variable doesn't actually need dynamic initialization. If
2323 the programmer can be sure that no use of the variable in a
2324 non-defining TU needs to trigger dynamic initialization (either
2325 because the variable is statically initialized, or a use of the
2326 variable in the defining TU will be executed before any uses in
2327 another TU), they can avoid this overhead with the
2328 @option{-fno-extern-tls-init} option.
2330 On targets that support symbol aliases, the default is
2331 @option{-fextern-tls-init}. On targets that do not support symbol
2332 aliases, the default is @option{-fno-extern-tls-init}.
2335 @itemx -fno-for-scope
2337 @opindex fno-for-scope
2338 If @option{-ffor-scope} is specified, the scope of variables declared in
2339 a @i{for-init-statement} is limited to the @code{for} loop itself,
2340 as specified by the C++ standard.
2341 If @option{-fno-for-scope} is specified, the scope of variables declared in
2342 a @i{for-init-statement} extends to the end of the enclosing scope,
2343 as was the case in old versions of G++, and other (traditional)
2344 implementations of C++.
2346 If neither flag is given, the default is to follow the standard,
2347 but to allow and give a warning for old-style code that would
2348 otherwise be invalid, or have different behavior.
2350 @item -fno-gnu-keywords
2351 @opindex fno-gnu-keywords
2352 Do not recognize @code{typeof} as a keyword, so that code can use this
2353 word as an identifier. You can use the keyword @code{__typeof__} instead.
2354 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2355 @option{-std=c++98}, @option{-std=c++11}, etc.
2357 @item -fno-implicit-templates
2358 @opindex fno-implicit-templates
2359 Never emit code for non-inline templates that are instantiated
2360 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2361 @xref{Template Instantiation}, for more information.
2363 @item -fno-implicit-inline-templates
2364 @opindex fno-implicit-inline-templates
2365 Don't emit code for implicit instantiations of inline templates, either.
2366 The default is to handle inlines differently so that compiles with and
2367 without optimization need the same set of explicit instantiations.
2369 @item -fno-implement-inlines
2370 @opindex fno-implement-inlines
2371 To save space, do not emit out-of-line copies of inline functions
2372 controlled by @code{#pragma implementation}. This causes linker
2373 errors if these functions are not inlined everywhere they are called.
2375 @item -fms-extensions
2376 @opindex fms-extensions
2377 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2378 int and getting a pointer to member function via non-standard syntax.
2380 @item -fno-nonansi-builtins
2381 @opindex fno-nonansi-builtins
2382 Disable built-in declarations of functions that are not mandated by
2383 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2384 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2387 @opindex fnothrow-opt
2388 Treat a @code{throw()} exception specification as if it were a
2389 @code{noexcept} specification to reduce or eliminate the text size
2390 overhead relative to a function with no exception specification. If
2391 the function has local variables of types with non-trivial
2392 destructors, the exception specification actually makes the
2393 function smaller because the EH cleanups for those variables can be
2394 optimized away. The semantic effect is that an exception thrown out of
2395 a function with such an exception specification results in a call
2396 to @code{terminate} rather than @code{unexpected}.
2398 @item -fno-operator-names
2399 @opindex fno-operator-names
2400 Do not treat the operator name keywords @code{and}, @code{bitand},
2401 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2402 synonyms as keywords.
2404 @item -fno-optional-diags
2405 @opindex fno-optional-diags
2406 Disable diagnostics that the standard says a compiler does not need to
2407 issue. Currently, the only such diagnostic issued by G++ is the one for
2408 a name having multiple meanings within a class.
2411 @opindex fpermissive
2412 Downgrade some diagnostics about nonconformant code from errors to
2413 warnings. Thus, using @option{-fpermissive} allows some
2414 nonconforming code to compile.
2416 @item -fno-pretty-templates
2417 @opindex fno-pretty-templates
2418 When an error message refers to a specialization of a function
2419 template, the compiler normally prints the signature of the
2420 template followed by the template arguments and any typedefs or
2421 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2422 rather than @code{void f(int)}) so that it's clear which template is
2423 involved. When an error message refers to a specialization of a class
2424 template, the compiler omits any template arguments that match
2425 the default template arguments for that template. If either of these
2426 behaviors make it harder to understand the error message rather than
2427 easier, you can use @option{-fno-pretty-templates} to disable them.
2431 Enable automatic template instantiation at link time. This option also
2432 implies @option{-fno-implicit-templates}. @xref{Template
2433 Instantiation}, for more information.
2437 Disable generation of information about every class with virtual
2438 functions for use by the C++ run-time type identification features
2439 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2440 of the language, you can save some space by using this flag. Note that
2441 exception handling uses the same information, but G++ generates it as
2442 needed. The @code{dynamic_cast} operator can still be used for casts that
2443 do not require run-time type information, i.e.@: casts to @code{void *} or to
2444 unambiguous base classes.
2446 @item -fsized-deallocation
2447 @opindex fsized-deallocation
2448 Enable the built-in global declarations
2450 void operator delete (void *, std::size_t) noexcept;
2451 void operator delete[] (void *, std::size_t) noexcept;
2453 as introduced in C++14. This is useful for user-defined replacement
2454 deallocation functions that, for example, use the size of the object
2455 to make deallocation faster. Enabled by default under
2456 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2457 warns about places that might want to add a definition.
2459 @item -fstrict-enums
2460 @opindex fstrict-enums
2461 Allow the compiler to optimize using the assumption that a value of
2462 enumerated type can only be one of the values of the enumeration (as
2463 defined in the C++ standard; basically, a value that can be
2464 represented in the minimum number of bits needed to represent all the
2465 enumerators). This assumption may not be valid if the program uses a
2466 cast to convert an arbitrary integer value to the enumerated type.
2468 @item -ftemplate-backtrace-limit=@var{n}
2469 @opindex ftemplate-backtrace-limit
2470 Set the maximum number of template instantiation notes for a single
2471 warning or error to @var{n}. The default value is 10.
2473 @item -ftemplate-depth=@var{n}
2474 @opindex ftemplate-depth
2475 Set the maximum instantiation depth for template classes to @var{n}.
2476 A limit on the template instantiation depth is needed to detect
2477 endless recursions during template class instantiation. ANSI/ISO C++
2478 conforming programs must not rely on a maximum depth greater than 17
2479 (changed to 1024 in C++11). The default value is 900, as the compiler
2480 can run out of stack space before hitting 1024 in some situations.
2482 @item -fno-threadsafe-statics
2483 @opindex fno-threadsafe-statics
2484 Do not emit the extra code to use the routines specified in the C++
2485 ABI for thread-safe initialization of local statics. You can use this
2486 option to reduce code size slightly in code that doesn't need to be
2489 @item -fuse-cxa-atexit
2490 @opindex fuse-cxa-atexit
2491 Register destructors for objects with static storage duration with the
2492 @code{__cxa_atexit} function rather than the @code{atexit} function.
2493 This option is required for fully standards-compliant handling of static
2494 destructors, but only works if your C library supports
2495 @code{__cxa_atexit}.
2497 @item -fno-use-cxa-get-exception-ptr
2498 @opindex fno-use-cxa-get-exception-ptr
2499 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2500 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2501 if the runtime routine is not available.
2503 @item -fvisibility-inlines-hidden
2504 @opindex fvisibility-inlines-hidden
2505 This switch declares that the user does not attempt to compare
2506 pointers to inline functions or methods where the addresses of the two functions
2507 are taken in different shared objects.
2509 The effect of this is that GCC may, effectively, mark inline methods with
2510 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2511 appear in the export table of a DSO and do not require a PLT indirection
2512 when used within the DSO@. Enabling this option can have a dramatic effect
2513 on load and link times of a DSO as it massively reduces the size of the
2514 dynamic export table when the library makes heavy use of templates.
2516 The behavior of this switch is not quite the same as marking the
2517 methods as hidden directly, because it does not affect static variables
2518 local to the function or cause the compiler to deduce that
2519 the function is defined in only one shared object.
2521 You may mark a method as having a visibility explicitly to negate the
2522 effect of the switch for that method. For example, if you do want to
2523 compare pointers to a particular inline method, you might mark it as
2524 having default visibility. Marking the enclosing class with explicit
2525 visibility has no effect.
2527 Explicitly instantiated inline methods are unaffected by this option
2528 as their linkage might otherwise cross a shared library boundary.
2529 @xref{Template Instantiation}.
2531 @item -fvisibility-ms-compat
2532 @opindex fvisibility-ms-compat
2533 This flag attempts to use visibility settings to make GCC's C++
2534 linkage model compatible with that of Microsoft Visual Studio.
2536 The flag makes these changes to GCC's linkage model:
2540 It sets the default visibility to @code{hidden}, like
2541 @option{-fvisibility=hidden}.
2544 Types, but not their members, are not hidden by default.
2547 The One Definition Rule is relaxed for types without explicit
2548 visibility specifications that are defined in more than one
2549 shared object: those declarations are permitted if they are
2550 permitted when this option is not used.
2553 In new code it is better to use @option{-fvisibility=hidden} and
2554 export those classes that are intended to be externally visible.
2555 Unfortunately it is possible for code to rely, perhaps accidentally,
2556 on the Visual Studio behavior.
2558 Among the consequences of these changes are that static data members
2559 of the same type with the same name but defined in different shared
2560 objects are different, so changing one does not change the other;
2561 and that pointers to function members defined in different shared
2562 objects may not compare equal. When this flag is given, it is a
2563 violation of the ODR to define types with the same name differently.
2567 Do not use weak symbol support, even if it is provided by the linker.
2568 By default, G++ uses weak symbols if they are available. This
2569 option exists only for testing, and should not be used by end-users;
2570 it results in inferior code and has no benefits. This option may
2571 be removed in a future release of G++.
2575 Do not search for header files in the standard directories specific to
2576 C++, but do still search the other standard directories. (This option
2577 is used when building the C++ library.)
2580 In addition, these optimization, warning, and code generation options
2581 have meanings only for C++ programs:
2584 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2587 Warn when G++ it generates code that is probably not compatible with
2588 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2589 ABI with each major release, normally @option{-Wabi} will warn only if
2590 there is a check added later in a release series for an ABI issue
2591 discovered since the initial release. @option{-Wabi} will warn about
2592 more things if an older ABI version is selected (with
2593 @option{-fabi-version=@var{n}}).
2595 @option{-Wabi} can also be used with an explicit version number to
2596 warn about compatibility with a particular @option{-fabi-version}
2597 level, e.g. @option{-Wabi=2} to warn about changes relative to
2598 @option{-fabi-version=2}.
2600 If an explicit version number is provided and
2601 @option{-fabi-compat-version} is not specified, the version number
2602 from this option is used for compatibility aliases. If no explicit
2603 version number is provided with this option, but
2604 @option{-fabi-compat-version} is specified, that version number is
2605 used for ABI warnings.
2607 Although an effort has been made to warn about
2608 all such cases, there are probably some cases that are not warned about,
2609 even though G++ is generating incompatible code. There may also be
2610 cases where warnings are emitted even though the code that is generated
2613 You should rewrite your code to avoid these warnings if you are
2614 concerned about the fact that code generated by G++ may not be binary
2615 compatible with code generated by other compilers.
2617 Known incompatibilities in @option{-fabi-version=2} (which was the
2618 default from GCC 3.4 to 4.9) include:
2623 A template with a non-type template parameter of reference type was
2624 mangled incorrectly:
2627 template <int &> struct S @{@};
2631 This was fixed in @option{-fabi-version=3}.
2634 SIMD vector types declared using @code{__attribute ((vector_size))} were
2635 mangled in a non-standard way that does not allow for overloading of
2636 functions taking vectors of different sizes.
2638 The mangling was changed in @option{-fabi-version=4}.
2641 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2642 qualifiers, and @code{decltype} of a plain declaration was folded away.
2644 These mangling issues were fixed in @option{-fabi-version=5}.
2647 Scoped enumerators passed as arguments to a variadic function are
2648 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2649 On most targets this does not actually affect the parameter passing
2650 ABI, as there is no way to pass an argument smaller than @code{int}.
2652 Also, the ABI changed the mangling of template argument packs,
2653 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2654 a class scope function used as a template argument.
2656 These issues were corrected in @option{-fabi-version=6}.
2659 Lambdas in default argument scope were mangled incorrectly, and the
2660 ABI changed the mangling of @code{nullptr_t}.
2662 These issues were corrected in @option{-fabi-version=7}.
2665 When mangling a function type with function-cv-qualifiers, the
2666 un-qualified function type was incorrectly treated as a substitution
2669 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2672 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2673 unaligned accesses. Note that this did not affect the ABI of a
2674 function with a @code{nullptr_t} parameter, as parameters have a
2677 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2680 Target-specific attributes that affect the identity of a type, such as
2681 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2682 did not affect the mangled name, leading to name collisions when
2683 function pointers were used as template arguments.
2685 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2689 It also warns about psABI-related changes. The known psABI changes at this
2695 For SysV/x86-64, unions with @code{long double} members are
2696 passed in memory as specified in psABI. For example:
2706 @code{union U} is always passed in memory.
2710 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2713 Warn when a type with an ABI tag is used in a context that does not
2714 have that ABI tag. See @ref{C++ Attributes} for more information
2717 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2718 @opindex Wctor-dtor-privacy
2719 @opindex Wno-ctor-dtor-privacy
2720 Warn when a class seems unusable because all the constructors or
2721 destructors in that class are private, and it has neither friends nor
2722 public static member functions. Also warn if there are no non-private
2723 methods, and there's at least one private member function that isn't
2724 a constructor or destructor.
2726 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2727 @opindex Wdelete-non-virtual-dtor
2728 @opindex Wno-delete-non-virtual-dtor
2729 Warn when @code{delete} is used to destroy an instance of a class that
2730 has virtual functions and non-virtual destructor. It is unsafe to delete
2731 an instance of a derived class through a pointer to a base class if the
2732 base class does not have a virtual destructor. This warning is enabled
2735 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2736 @opindex Wliteral-suffix
2737 @opindex Wno-literal-suffix
2738 Warn when a string or character literal is followed by a ud-suffix which does
2739 not begin with an underscore. As a conforming extension, GCC treats such
2740 suffixes as separate preprocessing tokens in order to maintain backwards
2741 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2745 #define __STDC_FORMAT_MACROS
2746 #include <inttypes.h>
2751 printf("My int64: %" PRId64"\n", i64);
2755 In this case, @code{PRId64} is treated as a separate preprocessing token.
2757 This warning is enabled by default.
2759 @item -Wlto-type-mismatch
2760 @opindex Wlto-type-mismatch
2761 @opindex Wno-lto-type-mismatch
2763 During the link-time optimization warn about type mismatches in
2764 global declarations from different compilation units.
2765 Requires @option{-flto} to be enabled. Enabled by default.
2767 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2769 @opindex Wno-narrowing
2770 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2771 conversion prohibited by C++11 occurs within
2775 int i = @{ 2.2 @}; // error: narrowing from double to int
2778 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2780 When a later standard is in effect, e.g. when using @option{-std=c++11},
2781 narrowing conversions are diagnosed by default, as required by the standard.
2782 A narrowing conversion from a constant produces an error,
2783 and a narrowing conversion from a non-constant produces a warning,
2784 but @option{-Wno-narrowing} suppresses the diagnostic.
2785 Note that this does not affect the meaning of well-formed code;
2786 narrowing conversions are still considered ill-formed in SFINAE contexts.
2788 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2790 @opindex Wno-noexcept
2791 Warn when a noexcept-expression evaluates to false because of a call
2792 to a function that does not have a non-throwing exception
2793 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2794 the compiler to never throw an exception.
2796 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2797 @opindex Wnon-virtual-dtor
2798 @opindex Wno-non-virtual-dtor
2799 Warn when a class has virtual functions and an accessible non-virtual
2800 destructor itself or in an accessible polymorphic base class, in which
2801 case it is possible but unsafe to delete an instance of a derived
2802 class through a pointer to the class itself or base class. This
2803 warning is automatically enabled if @option{-Weffc++} is specified.
2805 @item -Wreorder @r{(C++ and Objective-C++ only)}
2807 @opindex Wno-reorder
2808 @cindex reordering, warning
2809 @cindex warning for reordering of member initializers
2810 Warn when the order of member initializers given in the code does not
2811 match the order in which they must be executed. For instance:
2817 A(): j (0), i (1) @{ @}
2822 The compiler rearranges the member initializers for @code{i}
2823 and @code{j} to match the declaration order of the members, emitting
2824 a warning to that effect. This warning is enabled by @option{-Wall}.
2826 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2827 @opindex fext-numeric-literals
2828 @opindex fno-ext-numeric-literals
2829 Accept imaginary, fixed-point, or machine-defined
2830 literal number suffixes as GNU extensions.
2831 When this option is turned off these suffixes are treated
2832 as C++11 user-defined literal numeric suffixes.
2833 This is on by default for all pre-C++11 dialects and all GNU dialects:
2834 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2835 @option{-std=gnu++14}.
2836 This option is off by default
2837 for ISO C++11 onwards (@option{-std=c++11}, ...).
2840 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2843 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2846 Warn about violations of the following style guidelines from Scott Meyers'
2847 @cite{Effective C++} series of books:
2851 Define a copy constructor and an assignment operator for classes
2852 with dynamically-allocated memory.
2855 Prefer initialization to assignment in constructors.
2858 Have @code{operator=} return a reference to @code{*this}.
2861 Don't try to return a reference when you must return an object.
2864 Distinguish between prefix and postfix forms of increment and
2865 decrement operators.
2868 Never overload @code{&&}, @code{||}, or @code{,}.
2872 This option also enables @option{-Wnon-virtual-dtor}, which is also
2873 one of the effective C++ recommendations. However, the check is
2874 extended to warn about the lack of virtual destructor in accessible
2875 non-polymorphic bases classes too.
2877 When selecting this option, be aware that the standard library
2878 headers do not obey all of these guidelines; use @samp{grep -v}
2879 to filter out those warnings.
2881 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2882 @opindex Wstrict-null-sentinel
2883 @opindex Wno-strict-null-sentinel
2884 Warn about the use of an uncasted @code{NULL} as sentinel. When
2885 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2886 to @code{__null}. Although it is a null pointer constant rather than a
2887 null pointer, it is guaranteed to be of the same size as a pointer.
2888 But this use is not portable across different compilers.
2890 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2891 @opindex Wno-non-template-friend
2892 @opindex Wnon-template-friend
2893 Disable warnings when non-templatized friend functions are declared
2894 within a template. Since the advent of explicit template specification
2895 support in G++, if the name of the friend is an unqualified-id (i.e.,
2896 @samp{friend foo(int)}), the C++ language specification demands that the
2897 friend declare or define an ordinary, nontemplate function. (Section
2898 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2899 could be interpreted as a particular specialization of a templatized
2900 function. Because this non-conforming behavior is no longer the default
2901 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2902 check existing code for potential trouble spots and is on by default.
2903 This new compiler behavior can be turned off with
2904 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2905 but disables the helpful warning.
2907 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2908 @opindex Wold-style-cast
2909 @opindex Wno-old-style-cast
2910 Warn if an old-style (C-style) cast to a non-void type is used within
2911 a C++ program. The new-style casts (@code{dynamic_cast},
2912 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2913 less vulnerable to unintended effects and much easier to search for.
2915 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2916 @opindex Woverloaded-virtual
2917 @opindex Wno-overloaded-virtual
2918 @cindex overloaded virtual function, warning
2919 @cindex warning for overloaded virtual function
2920 Warn when a function declaration hides virtual functions from a
2921 base class. For example, in:
2928 struct B: public A @{
2933 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2944 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2945 @opindex Wno-pmf-conversions
2946 @opindex Wpmf-conversions
2947 Disable the diagnostic for converting a bound pointer to member function
2950 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2951 @opindex Wsign-promo
2952 @opindex Wno-sign-promo
2953 Warn when overload resolution chooses a promotion from unsigned or
2954 enumerated type to a signed type, over a conversion to an unsigned type of
2955 the same size. Previous versions of G++ tried to preserve
2956 unsignedness, but the standard mandates the current behavior.
2958 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2960 Warn when a primary template declaration is encountered. Some coding
2961 rules disallow templates, and this may be used to enforce that rule.
2962 The warning is inactive inside a system header file, such as the STL, so
2963 one can still use the STL. One may also instantiate or specialize
2966 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2967 @opindex Wmultiple-inheritance
2968 Warn when a class is defined with multiple direct base classes. Some
2969 coding rules disallow multiple inheritance, and this may be used to
2970 enforce that rule. The warning is inactive inside a system header file,
2971 such as the STL, so one can still use the STL. One may also define
2972 classes that indirectly use multiple inheritance.
2974 @item -Wvirtual-inheritance
2975 @opindex Wvirtual-inheritance
2976 Warn when a class is defined with a virtual direct base classe. Some
2977 coding rules disallow multiple inheritance, and this may be used to
2978 enforce that rule. The warning is inactive inside a system header file,
2979 such as the STL, so one can still use the STL. One may also define
2980 classes that indirectly use virtual inheritance.
2983 @opindex Wnamespaces
2984 Warn when a namespace definition is opened. Some coding rules disallow
2985 namespaces, and this may be used to enforce that rule. The warning is
2986 inactive inside a system header file, such as the STL, so one can still
2987 use the STL. One may also use using directives and qualified names.
2989 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2991 @opindex Wno-terminate
2992 Disable the warning about a throw-expression that will immediately
2993 result in a call to @code{terminate}.
2996 @node Objective-C and Objective-C++ Dialect Options
2997 @section Options Controlling Objective-C and Objective-C++ Dialects
2999 @cindex compiler options, Objective-C and Objective-C++
3000 @cindex Objective-C and Objective-C++ options, command-line
3001 @cindex options, Objective-C and Objective-C++
3002 (NOTE: This manual does not describe the Objective-C and Objective-C++
3003 languages themselves. @xref{Standards,,Language Standards
3004 Supported by GCC}, for references.)
3006 This section describes the command-line options that are only meaningful
3007 for Objective-C and Objective-C++ programs. You can also use most of
3008 the language-independent GNU compiler options.
3009 For example, you might compile a file @file{some_class.m} like this:
3012 gcc -g -fgnu-runtime -O -c some_class.m
3016 In this example, @option{-fgnu-runtime} is an option meant only for
3017 Objective-C and Objective-C++ programs; you can use the other options with
3018 any language supported by GCC@.
3020 Note that since Objective-C is an extension of the C language, Objective-C
3021 compilations may also use options specific to the C front-end (e.g.,
3022 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3023 C++-specific options (e.g., @option{-Wabi}).
3025 Here is a list of options that are @emph{only} for compiling Objective-C
3026 and Objective-C++ programs:
3029 @item -fconstant-string-class=@var{class-name}
3030 @opindex fconstant-string-class
3031 Use @var{class-name} as the name of the class to instantiate for each
3032 literal string specified with the syntax @code{@@"@dots{}"}. The default
3033 class name is @code{NXConstantString} if the GNU runtime is being used, and
3034 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3035 @option{-fconstant-cfstrings} option, if also present, overrides the
3036 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3037 to be laid out as constant CoreFoundation strings.
3040 @opindex fgnu-runtime
3041 Generate object code compatible with the standard GNU Objective-C
3042 runtime. This is the default for most types of systems.
3044 @item -fnext-runtime
3045 @opindex fnext-runtime
3046 Generate output compatible with the NeXT runtime. This is the default
3047 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3048 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3051 @item -fno-nil-receivers
3052 @opindex fno-nil-receivers
3053 Assume that all Objective-C message dispatches (@code{[receiver
3054 message:arg]}) in this translation unit ensure that the receiver is
3055 not @code{nil}. This allows for more efficient entry points in the
3056 runtime to be used. This option is only available in conjunction with
3057 the NeXT runtime and ABI version 0 or 1.
3059 @item -fobjc-abi-version=@var{n}
3060 @opindex fobjc-abi-version
3061 Use version @var{n} of the Objective-C ABI for the selected runtime.
3062 This option is currently supported only for the NeXT runtime. In that
3063 case, Version 0 is the traditional (32-bit) ABI without support for
3064 properties and other Objective-C 2.0 additions. Version 1 is the
3065 traditional (32-bit) ABI with support for properties and other
3066 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3067 nothing is specified, the default is Version 0 on 32-bit target
3068 machines, and Version 2 on 64-bit target machines.
3070 @item -fobjc-call-cxx-cdtors
3071 @opindex fobjc-call-cxx-cdtors
3072 For each Objective-C class, check if any of its instance variables is a
3073 C++ object with a non-trivial default constructor. If so, synthesize a
3074 special @code{- (id) .cxx_construct} instance method which runs
3075 non-trivial default constructors on any such instance variables, in order,
3076 and then return @code{self}. Similarly, check if any instance variable
3077 is a C++ object with a non-trivial destructor, and if so, synthesize a
3078 special @code{- (void) .cxx_destruct} method which runs
3079 all such default destructors, in reverse order.
3081 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3082 methods thusly generated only operate on instance variables
3083 declared in the current Objective-C class, and not those inherited
3084 from superclasses. It is the responsibility of the Objective-C
3085 runtime to invoke all such methods in an object's inheritance
3086 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3087 by the runtime immediately after a new object instance is allocated;
3088 the @code{- (void) .cxx_destruct} methods are invoked immediately
3089 before the runtime deallocates an object instance.
3091 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3092 support for invoking the @code{- (id) .cxx_construct} and
3093 @code{- (void) .cxx_destruct} methods.
3095 @item -fobjc-direct-dispatch
3096 @opindex fobjc-direct-dispatch
3097 Allow fast jumps to the message dispatcher. On Darwin this is
3098 accomplished via the comm page.
3100 @item -fobjc-exceptions
3101 @opindex fobjc-exceptions
3102 Enable syntactic support for structured exception handling in
3103 Objective-C, similar to what is offered by C++ and Java. This option
3104 is required to use the Objective-C keywords @code{@@try},
3105 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3106 @code{@@synchronized}. This option is available with both the GNU
3107 runtime and the NeXT runtime (but not available in conjunction with
3108 the NeXT runtime on Mac OS X 10.2 and earlier).
3112 Enable garbage collection (GC) in Objective-C and Objective-C++
3113 programs. This option is only available with the NeXT runtime; the
3114 GNU runtime has a different garbage collection implementation that
3115 does not require special compiler flags.
3117 @item -fobjc-nilcheck
3118 @opindex fobjc-nilcheck
3119 For the NeXT runtime with version 2 of the ABI, check for a nil
3120 receiver in method invocations before doing the actual method call.
3121 This is the default and can be disabled using
3122 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3123 checked for nil in this way no matter what this flag is set to.
3124 Currently this flag does nothing when the GNU runtime, or an older
3125 version of the NeXT runtime ABI, is used.
3127 @item -fobjc-std=objc1
3129 Conform to the language syntax of Objective-C 1.0, the language
3130 recognized by GCC 4.0. This only affects the Objective-C additions to
3131 the C/C++ language; it does not affect conformance to C/C++ standards,
3132 which is controlled by the separate C/C++ dialect option flags. When
3133 this option is used with the Objective-C or Objective-C++ compiler,
3134 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3135 This is useful if you need to make sure that your Objective-C code can
3136 be compiled with older versions of GCC@.
3138 @item -freplace-objc-classes
3139 @opindex freplace-objc-classes
3140 Emit a special marker instructing @command{ld(1)} not to statically link in
3141 the resulting object file, and allow @command{dyld(1)} to load it in at
3142 run time instead. This is used in conjunction with the Fix-and-Continue
3143 debugging mode, where the object file in question may be recompiled and
3144 dynamically reloaded in the course of program execution, without the need
3145 to restart the program itself. Currently, Fix-and-Continue functionality
3146 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3151 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3152 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3153 compile time) with static class references that get initialized at load time,
3154 which improves run-time performance. Specifying the @option{-fzero-link} flag
3155 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3156 to be retained. This is useful in Zero-Link debugging mode, since it allows
3157 for individual class implementations to be modified during program execution.
3158 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3159 regardless of command-line options.
3161 @item -fno-local-ivars
3162 @opindex fno-local-ivars
3163 @opindex flocal-ivars
3164 By default instance variables in Objective-C can be accessed as if
3165 they were local variables from within the methods of the class they're
3166 declared in. This can lead to shadowing between instance variables
3167 and other variables declared either locally inside a class method or
3168 globally with the same name. Specifying the @option{-fno-local-ivars}
3169 flag disables this behavior thus avoiding variable shadowing issues.
3171 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3172 @opindex fivar-visibility
3173 Set the default instance variable visibility to the specified option
3174 so that instance variables declared outside the scope of any access
3175 modifier directives default to the specified visibility.
3179 Dump interface declarations for all classes seen in the source file to a
3180 file named @file{@var{sourcename}.decl}.
3182 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3183 @opindex Wassign-intercept
3184 @opindex Wno-assign-intercept
3185 Warn whenever an Objective-C assignment is being intercepted by the
3188 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3189 @opindex Wno-protocol
3191 If a class is declared to implement a protocol, a warning is issued for
3192 every method in the protocol that is not implemented by the class. The
3193 default behavior is to issue a warning for every method not explicitly
3194 implemented in the class, even if a method implementation is inherited
3195 from the superclass. If you use the @option{-Wno-protocol} option, then
3196 methods inherited from the superclass are considered to be implemented,
3197 and no warning is issued for them.
3199 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3201 @opindex Wno-selector
3202 Warn if multiple methods of different types for the same selector are
3203 found during compilation. The check is performed on the list of methods
3204 in the final stage of compilation. Additionally, a check is performed
3205 for each selector appearing in a @code{@@selector(@dots{})}
3206 expression, and a corresponding method for that selector has been found
3207 during compilation. Because these checks scan the method table only at
3208 the end of compilation, these warnings are not produced if the final
3209 stage of compilation is not reached, for example because an error is
3210 found during compilation, or because the @option{-fsyntax-only} option is
3213 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3214 @opindex Wstrict-selector-match
3215 @opindex Wno-strict-selector-match
3216 Warn if multiple methods with differing argument and/or return types are
3217 found for a given selector when attempting to send a message using this
3218 selector to a receiver of type @code{id} or @code{Class}. When this flag
3219 is off (which is the default behavior), the compiler omits such warnings
3220 if any differences found are confined to types that share the same size
3223 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3224 @opindex Wundeclared-selector
3225 @opindex Wno-undeclared-selector
3226 Warn if a @code{@@selector(@dots{})} expression referring to an
3227 undeclared selector is found. A selector is considered undeclared if no
3228 method with that name has been declared before the
3229 @code{@@selector(@dots{})} expression, either explicitly in an
3230 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3231 an @code{@@implementation} section. This option always performs its
3232 checks as soon as a @code{@@selector(@dots{})} expression is found,
3233 while @option{-Wselector} only performs its checks in the final stage of
3234 compilation. This also enforces the coding style convention
3235 that methods and selectors must be declared before being used.
3237 @item -print-objc-runtime-info
3238 @opindex print-objc-runtime-info
3239 Generate C header describing the largest structure that is passed by
3244 @node Diagnostic Message Formatting Options
3245 @section Options to Control Diagnostic Messages Formatting
3246 @cindex options to control diagnostics formatting
3247 @cindex diagnostic messages
3248 @cindex message formatting
3250 Traditionally, diagnostic messages have been formatted irrespective of
3251 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3252 options described below
3253 to control the formatting algorithm for diagnostic messages,
3254 e.g.@: how many characters per line, how often source location
3255 information should be reported. Note that some language front ends may not
3256 honor these options.
3259 @item -fmessage-length=@var{n}
3260 @opindex fmessage-length
3261 Try to format error messages so that they fit on lines of about
3262 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3263 done; each error message appears on a single line. This is the
3264 default for all front ends.
3266 @item -fdiagnostics-show-location=once
3267 @opindex fdiagnostics-show-location
3268 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3269 reporter to emit source location information @emph{once}; that is, in
3270 case the message is too long to fit on a single physical line and has to
3271 be wrapped, the source location won't be emitted (as prefix) again,
3272 over and over, in subsequent continuation lines. This is the default
3275 @item -fdiagnostics-show-location=every-line
3276 Only meaningful in line-wrapping mode. Instructs the diagnostic
3277 messages reporter to emit the same source location information (as
3278 prefix) for physical lines that result from the process of breaking
3279 a message which is too long to fit on a single line.
3281 @item -fdiagnostics-color[=@var{WHEN}]
3282 @itemx -fno-diagnostics-color
3283 @opindex fdiagnostics-color
3284 @cindex highlight, color
3285 @vindex GCC_COLORS @r{environment variable}
3286 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3287 or @samp{auto}. The default depends on how the compiler has been configured,
3288 it can be any of the above @var{WHEN} options or also @samp{never}
3289 if @env{GCC_COLORS} environment variable isn't present in the environment,
3290 and @samp{auto} otherwise.
3291 @samp{auto} means to use color only when the standard error is a terminal.
3292 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3293 aliases for @option{-fdiagnostics-color=always} and
3294 @option{-fdiagnostics-color=never}, respectively.
3296 The colors are defined by the environment variable @env{GCC_COLORS}.
3297 Its value is a colon-separated list of capabilities and Select Graphic
3298 Rendition (SGR) substrings. SGR commands are interpreted by the
3299 terminal or terminal emulator. (See the section in the documentation
3300 of your text terminal for permitted values and their meanings as
3301 character attributes.) These substring values are integers in decimal
3302 representation and can be concatenated with semicolons.
3303 Common values to concatenate include
3305 @samp{4} for underline,
3307 @samp{7} for inverse,
3308 @samp{39} for default foreground color,
3309 @samp{30} to @samp{37} for foreground colors,
3310 @samp{90} to @samp{97} for 16-color mode foreground colors,
3311 @samp{38;5;0} to @samp{38;5;255}
3312 for 88-color and 256-color modes foreground colors,
3313 @samp{49} for default background color,
3314 @samp{40} to @samp{47} for background colors,
3315 @samp{100} to @samp{107} for 16-color mode background colors,
3316 and @samp{48;5;0} to @samp{48;5;255}
3317 for 88-color and 256-color modes background colors.
3319 The default @env{GCC_COLORS} is
3321 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3324 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3325 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3326 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3327 string disables colors.
3328 Supported capabilities are as follows.
3332 @vindex error GCC_COLORS @r{capability}
3333 SGR substring for error: markers.
3336 @vindex warning GCC_COLORS @r{capability}
3337 SGR substring for warning: markers.
3340 @vindex note GCC_COLORS @r{capability}
3341 SGR substring for note: markers.
3344 @vindex caret GCC_COLORS @r{capability}
3345 SGR substring for caret line.
3348 @vindex locus GCC_COLORS @r{capability}
3349 SGR substring for location information, @samp{file:line} or
3350 @samp{file:line:column} etc.
3353 @vindex quote GCC_COLORS @r{capability}
3354 SGR substring for information printed within quotes.
3357 @item -fno-diagnostics-show-option
3358 @opindex fno-diagnostics-show-option
3359 @opindex fdiagnostics-show-option
3360 By default, each diagnostic emitted includes text indicating the
3361 command-line option that directly controls the diagnostic (if such an
3362 option is known to the diagnostic machinery). Specifying the
3363 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3365 @item -fno-diagnostics-show-caret
3366 @opindex fno-diagnostics-show-caret
3367 @opindex fdiagnostics-show-caret
3368 By default, each diagnostic emitted includes the original source line
3369 and a caret @samp{^} indicating the column. This option suppresses this
3370 information. The source line is truncated to @var{n} characters, if
3371 the @option{-fmessage-length=n} option is given. When the output is done
3372 to the terminal, the width is limited to the width given by the
3373 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3377 @node Warning Options
3378 @section Options to Request or Suppress Warnings
3379 @cindex options to control warnings
3380 @cindex warning messages
3381 @cindex messages, warning
3382 @cindex suppressing warnings
3384 Warnings are diagnostic messages that report constructions that
3385 are not inherently erroneous but that are risky or suggest there
3386 may have been an error.
3388 The following language-independent options do not enable specific
3389 warnings but control the kinds of diagnostics produced by GCC@.
3392 @cindex syntax checking
3394 @opindex fsyntax-only
3395 Check the code for syntax errors, but don't do anything beyond that.
3397 @item -fmax-errors=@var{n}
3398 @opindex fmax-errors
3399 Limits the maximum number of error messages to @var{n}, at which point
3400 GCC bails out rather than attempting to continue processing the source
3401 code. If @var{n} is 0 (the default), there is no limit on the number
3402 of error messages produced. If @option{-Wfatal-errors} is also
3403 specified, then @option{-Wfatal-errors} takes precedence over this
3408 Inhibit all warning messages.
3413 Make all warnings into errors.
3418 Make the specified warning into an error. The specifier for a warning
3419 is appended; for example @option{-Werror=switch} turns the warnings
3420 controlled by @option{-Wswitch} into errors. This switch takes a
3421 negative form, to be used to negate @option{-Werror} for specific
3422 warnings; for example @option{-Wno-error=switch} makes
3423 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3426 The warning message for each controllable warning includes the
3427 option that controls the warning. That option can then be used with
3428 @option{-Werror=} and @option{-Wno-error=} as described above.
3429 (Printing of the option in the warning message can be disabled using the
3430 @option{-fno-diagnostics-show-option} flag.)
3432 Note that specifying @option{-Werror=}@var{foo} automatically implies
3433 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3436 @item -Wfatal-errors
3437 @opindex Wfatal-errors
3438 @opindex Wno-fatal-errors
3439 This option causes the compiler to abort compilation on the first error
3440 occurred rather than trying to keep going and printing further error
3445 You can request many specific warnings with options beginning with
3446 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3447 implicit declarations. Each of these specific warning options also
3448 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3449 example, @option{-Wno-implicit}. This manual lists only one of the
3450 two forms, whichever is not the default. For further
3451 language-specific options also refer to @ref{C++ Dialect Options} and
3452 @ref{Objective-C and Objective-C++ Dialect Options}.
3454 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3455 options, such as @option{-Wunused}, which may turn on further options,
3456 such as @option{-Wunused-value}. The combined effect of positive and
3457 negative forms is that more specific options have priority over less
3458 specific ones, independently of their position in the command-line. For
3459 options of the same specificity, the last one takes effect. Options
3460 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3461 as if they appeared at the end of the command-line.
3463 When an unrecognized warning option is requested (e.g.,
3464 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3465 that the option is not recognized. However, if the @option{-Wno-} form
3466 is used, the behavior is slightly different: no diagnostic is
3467 produced for @option{-Wno-unknown-warning} unless other diagnostics
3468 are being produced. This allows the use of new @option{-Wno-} options
3469 with old compilers, but if something goes wrong, the compiler
3470 warns that an unrecognized option is present.
3477 Issue all the warnings demanded by strict ISO C and ISO C++;
3478 reject all programs that use forbidden extensions, and some other
3479 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3480 version of the ISO C standard specified by any @option{-std} option used.
3482 Valid ISO C and ISO C++ programs should compile properly with or without
3483 this option (though a rare few require @option{-ansi} or a
3484 @option{-std} option specifying the required version of ISO C)@. However,
3485 without this option, certain GNU extensions and traditional C and C++
3486 features are supported as well. With this option, they are rejected.
3488 @option{-Wpedantic} does not cause warning messages for use of the
3489 alternate keywords whose names begin and end with @samp{__}. Pedantic
3490 warnings are also disabled in the expression that follows
3491 @code{__extension__}. However, only system header files should use
3492 these escape routes; application programs should avoid them.
3493 @xref{Alternate Keywords}.
3495 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3496 C conformance. They soon find that it does not do quite what they want:
3497 it finds some non-ISO practices, but not all---only those for which
3498 ISO C @emph{requires} a diagnostic, and some others for which
3499 diagnostics have been added.
3501 A feature to report any failure to conform to ISO C might be useful in
3502 some instances, but would require considerable additional work and would
3503 be quite different from @option{-Wpedantic}. We don't have plans to
3504 support such a feature in the near future.
3506 Where the standard specified with @option{-std} represents a GNU
3507 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3508 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3509 extended dialect is based. Warnings from @option{-Wpedantic} are given
3510 where they are required by the base standard. (It does not make sense
3511 for such warnings to be given only for features not in the specified GNU
3512 C dialect, since by definition the GNU dialects of C include all
3513 features the compiler supports with the given option, and there would be
3514 nothing to warn about.)
3516 @item -pedantic-errors
3517 @opindex pedantic-errors
3518 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3519 requires a diagnostic, in some cases where there is undefined behavior
3520 at compile-time and in some other cases that do not prevent compilation
3521 of programs that are valid according to the standard. This is not
3522 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3523 by this option and not enabled by the latter and vice versa.
3528 This enables all the warnings about constructions that some users
3529 consider questionable, and that are easy to avoid (or modify to
3530 prevent the warning), even in conjunction with macros. This also
3531 enables some language-specific warnings described in @ref{C++ Dialect
3532 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3534 @option{-Wall} turns on the following warning flags:
3536 @gccoptlist{-Waddress @gol
3537 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3539 -Wc++11-compat -Wc++14-compat@gol
3540 -Wchar-subscripts @gol
3542 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3544 -Wimplicit @r{(C and Objective-C only)} @gol
3545 -Wimplicit-int @r{(C and Objective-C only)} @gol
3546 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3547 -Winit-self @r{(only for C++)} @gol
3548 -Wlogical-not-parentheses
3549 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3550 -Wmaybe-uninitialized @gol
3551 -Wmemset-transposed-args @gol
3552 -Wmisleading-indentation @r{(only for C/C++)} @gol
3553 -Wmissing-braces @r{(only for C/ObjC)} @gol
3554 -Wnarrowing @r{(only for C++)} @gol
3556 -Wnonnull-compare @gol
3562 -Wsequence-point @gol
3563 -Wsign-compare @r{(only in C++)} @gol
3564 -Wsizeof-pointer-memaccess @gol
3565 -Wstrict-aliasing @gol
3566 -Wstrict-overflow=1 @gol
3568 -Wtautological-compare @gol
3570 -Wuninitialized @gol
3571 -Wunknown-pragmas @gol
3572 -Wunused-function @gol
3575 -Wunused-variable @gol
3576 -Wvolatile-register-var @gol
3579 Note that some warning flags are not implied by @option{-Wall}. Some of
3580 them warn about constructions that users generally do not consider
3581 questionable, but which occasionally you might wish to check for;
3582 others warn about constructions that are necessary or hard to avoid in
3583 some cases, and there is no simple way to modify the code to suppress
3584 the warning. Some of them are enabled by @option{-Wextra} but many of
3585 them must be enabled individually.
3591 This enables some extra warning flags that are not enabled by
3592 @option{-Wall}. (This option used to be called @option{-W}. The older
3593 name is still supported, but the newer name is more descriptive.)
3595 @gccoptlist{-Wclobbered @gol
3597 -Wignored-qualifiers @gol
3598 -Wmissing-field-initializers @gol
3599 -Wmissing-parameter-type @r{(C only)} @gol
3600 -Wold-style-declaration @r{(C only)} @gol
3601 -Woverride-init @gol
3602 -Wsign-compare @r{(C only)} @gol
3604 -Wuninitialized @gol
3605 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3606 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3607 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3610 The option @option{-Wextra} also prints warning messages for the
3616 A pointer is compared against integer zero with @code{<}, @code{<=},
3617 @code{>}, or @code{>=}.
3620 (C++ only) An enumerator and a non-enumerator both appear in a
3621 conditional expression.
3624 (C++ only) Ambiguous virtual bases.
3627 (C++ only) Subscripting an array that has been declared @code{register}.
3630 (C++ only) Taking the address of a variable that has been declared
3634 (C++ only) A base class is not initialized in a derived class's copy
3639 @item -Wchar-subscripts
3640 @opindex Wchar-subscripts
3641 @opindex Wno-char-subscripts
3642 Warn if an array subscript has type @code{char}. This is a common cause
3643 of error, as programmers often forget that this type is signed on some
3645 This warning is enabled by @option{-Wall}.
3649 @opindex Wno-comment
3650 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3651 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3652 This warning is enabled by @option{-Wall}.
3654 @item -Wno-coverage-mismatch
3655 @opindex Wno-coverage-mismatch
3656 Warn if feedback profiles do not match when using the
3657 @option{-fprofile-use} option.
3658 If a source file is changed between compiling with @option{-fprofile-gen} and
3659 with @option{-fprofile-use}, the files with the profile feedback can fail
3660 to match the source file and GCC cannot use the profile feedback
3661 information. By default, this warning is enabled and is treated as an
3662 error. @option{-Wno-coverage-mismatch} can be used to disable the
3663 warning or @option{-Wno-error=coverage-mismatch} can be used to
3664 disable the error. Disabling the error for this warning can result in
3665 poorly optimized code and is useful only in the
3666 case of very minor changes such as bug fixes to an existing code-base.
3667 Completely disabling the warning is not recommended.
3670 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3672 Suppress warning messages emitted by @code{#warning} directives.
3674 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3675 @opindex Wdouble-promotion
3676 @opindex Wno-double-promotion
3677 Give a warning when a value of type @code{float} is implicitly
3678 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3679 floating-point unit implement @code{float} in hardware, but emulate
3680 @code{double} in software. On such a machine, doing computations
3681 using @code{double} values is much more expensive because of the
3682 overhead required for software emulation.
3684 It is easy to accidentally do computations with @code{double} because
3685 floating-point literals are implicitly of type @code{double}. For
3689 float area(float radius)
3691 return 3.14159 * radius * radius;
3695 the compiler performs the entire computation with @code{double}
3696 because the floating-point literal is a @code{double}.
3699 @itemx -Wformat=@var{n}
3702 @opindex ffreestanding
3703 @opindex fno-builtin
3705 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3706 the arguments supplied have types appropriate to the format string
3707 specified, and that the conversions specified in the format string make
3708 sense. This includes standard functions, and others specified by format
3709 attributes (@pxref{Function Attributes}), in the @code{printf},
3710 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3711 not in the C standard) families (or other target-specific families).
3712 Which functions are checked without format attributes having been
3713 specified depends on the standard version selected, and such checks of
3714 functions without the attribute specified are disabled by
3715 @option{-ffreestanding} or @option{-fno-builtin}.
3717 The formats are checked against the format features supported by GNU
3718 libc version 2.2. These include all ISO C90 and C99 features, as well
3719 as features from the Single Unix Specification and some BSD and GNU
3720 extensions. Other library implementations may not support all these
3721 features; GCC does not support warning about features that go beyond a
3722 particular library's limitations. However, if @option{-Wpedantic} is used
3723 with @option{-Wformat}, warnings are given about format features not
3724 in the selected standard version (but not for @code{strfmon} formats,
3725 since those are not in any version of the C standard). @xref{C Dialect
3726 Options,,Options Controlling C Dialect}.
3733 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3734 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3735 @option{-Wformat} also checks for null format arguments for several
3736 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3737 aspects of this level of format checking can be disabled by the
3738 options: @option{-Wno-format-contains-nul},
3739 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3740 @option{-Wformat} is enabled by @option{-Wall}.
3742 @item -Wno-format-contains-nul
3743 @opindex Wno-format-contains-nul
3744 @opindex Wformat-contains-nul
3745 If @option{-Wformat} is specified, do not warn about format strings that
3748 @item -Wno-format-extra-args
3749 @opindex Wno-format-extra-args
3750 @opindex Wformat-extra-args
3751 If @option{-Wformat} is specified, do not warn about excess arguments to a
3752 @code{printf} or @code{scanf} format function. The C standard specifies
3753 that such arguments are ignored.
3755 Where the unused arguments lie between used arguments that are
3756 specified with @samp{$} operand number specifications, normally
3757 warnings are still given, since the implementation could not know what
3758 type to pass to @code{va_arg} to skip the unused arguments. However,
3759 in the case of @code{scanf} formats, this option suppresses the
3760 warning if the unused arguments are all pointers, since the Single
3761 Unix Specification says that such unused arguments are allowed.
3763 @item -Wno-format-zero-length
3764 @opindex Wno-format-zero-length
3765 @opindex Wformat-zero-length
3766 If @option{-Wformat} is specified, do not warn about zero-length formats.
3767 The C standard specifies that zero-length formats are allowed.
3772 Enable @option{-Wformat} plus additional format checks. Currently
3773 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3776 @item -Wformat-nonliteral
3777 @opindex Wformat-nonliteral
3778 @opindex Wno-format-nonliteral
3779 If @option{-Wformat} is specified, also warn if the format string is not a
3780 string literal and so cannot be checked, unless the format function
3781 takes its format arguments as a @code{va_list}.
3783 @item -Wformat-security
3784 @opindex Wformat-security
3785 @opindex Wno-format-security
3786 If @option{-Wformat} is specified, also warn about uses of format
3787 functions that represent possible security problems. At present, this
3788 warns about calls to @code{printf} and @code{scanf} functions where the
3789 format string is not a string literal and there are no format arguments,
3790 as in @code{printf (foo);}. This may be a security hole if the format
3791 string came from untrusted input and contains @samp{%n}. (This is
3792 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3793 in future warnings may be added to @option{-Wformat-security} that are not
3794 included in @option{-Wformat-nonliteral}.)
3796 @item -Wformat-signedness
3797 @opindex Wformat-signedness
3798 @opindex Wno-format-signedness
3799 If @option{-Wformat} is specified, also warn if the format string
3800 requires an unsigned argument and the argument is signed and vice versa.
3803 @opindex Wformat-y2k
3804 @opindex Wno-format-y2k
3805 If @option{-Wformat} is specified, also warn about @code{strftime}
3806 formats that may yield only a two-digit year.
3811 @opindex Wno-nonnull
3812 Warn about passing a null pointer for arguments marked as
3813 requiring a non-null value by the @code{nonnull} function attribute.
3815 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3816 can be disabled with the @option{-Wno-nonnull} option.
3818 @item -Wnonnull-compare
3819 @opindex Wnonnull-compare
3820 @opindex Wno-nonnull-compare
3821 Warn when comparing an argument marked with the @code{nonnull}
3822 function attribute against null inside the function.
3824 @option{-Wnonnull-compare} is included in @option{-Wall}. It
3825 can be disabled with the @option{-Wno-nonnull-compare} option.
3827 @item -Wnull-dereference
3828 @opindex Wnull-dereference
3829 @opindex Wno-null-dereference
3830 Warn if the compiler detects paths that trigger erroneous or
3831 undefined behavior due to dereferencing a null pointer. This option
3832 is only active when @option{-fdelete-null-pointer-checks} is active,
3833 which is enabled by optimizations in most targets. The precision of
3834 the warnings depends on the optimization options used.
3836 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3838 @opindex Wno-init-self
3839 Warn about uninitialized variables that are initialized with themselves.
3840 Note this option can only be used with the @option{-Wuninitialized} option.
3842 For example, GCC warns about @code{i} being uninitialized in the
3843 following snippet only when @option{-Winit-self} has been specified:
3854 This warning is enabled by @option{-Wall} in C++.
3856 @item -Wimplicit-int @r{(C and Objective-C only)}
3857 @opindex Wimplicit-int
3858 @opindex Wno-implicit-int
3859 Warn when a declaration does not specify a type.
3860 This warning is enabled by @option{-Wall}.
3862 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3863 @opindex Wimplicit-function-declaration
3864 @opindex Wno-implicit-function-declaration
3865 Give a warning whenever a function is used before being declared. In
3866 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3867 enabled by default and it is made into an error by
3868 @option{-pedantic-errors}. This warning is also enabled by
3871 @item -Wimplicit @r{(C and Objective-C only)}
3873 @opindex Wno-implicit
3874 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3875 This warning is enabled by @option{-Wall}.
3877 @item -Wignored-qualifiers @r{(C and C++ only)}
3878 @opindex Wignored-qualifiers
3879 @opindex Wno-ignored-qualifiers
3880 Warn if the return type of a function has a type qualifier
3881 such as @code{const}. For ISO C such a type qualifier has no effect,
3882 since the value returned by a function is not an lvalue.
3883 For C++, the warning is only emitted for scalar types or @code{void}.
3884 ISO C prohibits qualified @code{void} return types on function
3885 definitions, so such return types always receive a warning
3886 even without this option.
3888 This warning is also enabled by @option{-Wextra}.
3890 @item -Wignored-attributes @r{(C and C++ only)}
3891 @opindex Wignored-attributes
3892 @opindex Wno-ignored-attributes
3893 Warn when an attribute is ignored. This is different from the
3894 @option{-Wattributes} option in that it warns whenever the compiler decides
3895 to drop an attribute, not that the attribute is either unknown, used in a
3896 wrong place, etc. This warning is enabled by default.
3901 Warn if the type of @code{main} is suspicious. @code{main} should be
3902 a function with external linkage, returning int, taking either zero
3903 arguments, two, or three arguments of appropriate types. This warning
3904 is enabled by default in C++ and is enabled by either @option{-Wall}
3905 or @option{-Wpedantic}.
3907 @item -Wmisleading-indentation @r{(C and C++ only)}
3908 @opindex Wmisleading-indentation
3909 @opindex Wno-misleading-indentation
3910 Warn when the indentation of the code does not reflect the block structure.
3911 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3912 @code{for} clauses with a guarded statement that does not use braces,
3913 followed by an unguarded statement with the same indentation.
3915 In the following example, the call to ``bar'' is misleadingly indented as
3916 if it were guarded by the ``if'' conditional.
3919 if (some_condition ())
3921 bar (); /* Gotcha: this is not guarded by the "if". */
3924 In the case of mixed tabs and spaces, the warning uses the
3925 @option{-ftabstop=} option to determine if the statements line up
3928 The warning is not issued for code involving multiline preprocessor logic
3929 such as the following example.
3934 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3940 The warning is not issued after a @code{#line} directive, since this
3941 typically indicates autogenerated code, and no assumptions can be made
3942 about the layout of the file that the directive references.
3944 This warning is enabled by @option{-Wall} in C and C++.
3946 @item -Wmissing-braces
3947 @opindex Wmissing-braces
3948 @opindex Wno-missing-braces
3949 Warn if an aggregate or union initializer is not fully bracketed. In
3950 the following example, the initializer for @code{a} is not fully
3951 bracketed, but that for @code{b} is fully bracketed. This warning is
3952 enabled by @option{-Wall} in C.
3955 int a[2][2] = @{ 0, 1, 2, 3 @};
3956 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3959 This warning is enabled by @option{-Wall}.
3961 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3962 @opindex Wmissing-include-dirs
3963 @opindex Wno-missing-include-dirs
3964 Warn if a user-supplied include directory does not exist.
3967 @opindex Wparentheses
3968 @opindex Wno-parentheses
3969 Warn if parentheses are omitted in certain contexts, such
3970 as when there is an assignment in a context where a truth value
3971 is expected, or when operators are nested whose precedence people
3972 often get confused about.
3974 Also warn if a comparison like @code{x<=y<=z} appears; this is
3975 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3976 interpretation from that of ordinary mathematical notation.
3978 Also warn about constructions where there may be confusion to which
3979 @code{if} statement an @code{else} branch belongs. Here is an example of
3994 In C/C++, every @code{else} branch belongs to the innermost possible
3995 @code{if} statement, which in this example is @code{if (b)}. This is
3996 often not what the programmer expected, as illustrated in the above
3997 example by indentation the programmer chose. When there is the
3998 potential for this confusion, GCC issues a warning when this flag
3999 is specified. To eliminate the warning, add explicit braces around
4000 the innermost @code{if} statement so there is no way the @code{else}
4001 can belong to the enclosing @code{if}. The resulting code
4018 Also warn for dangerous uses of the GNU extension to
4019 @code{?:} with omitted middle operand. When the condition
4020 in the @code{?}: operator is a boolean expression, the omitted value is
4021 always 1. Often programmers expect it to be a value computed
4022 inside the conditional expression instead.
4024 This warning is enabled by @option{-Wall}.
4026 @item -Wsequence-point
4027 @opindex Wsequence-point
4028 @opindex Wno-sequence-point
4029 Warn about code that may have undefined semantics because of violations
4030 of sequence point rules in the C and C++ standards.
4032 The C and C++ standards define the order in which expressions in a C/C++
4033 program are evaluated in terms of @dfn{sequence points}, which represent
4034 a partial ordering between the execution of parts of the program: those
4035 executed before the sequence point, and those executed after it. These
4036 occur after the evaluation of a full expression (one which is not part
4037 of a larger expression), after the evaluation of the first operand of a
4038 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4039 function is called (but after the evaluation of its arguments and the
4040 expression denoting the called function), and in certain other places.
4041 Other than as expressed by the sequence point rules, the order of
4042 evaluation of subexpressions of an expression is not specified. All
4043 these rules describe only a partial order rather than a total order,
4044 since, for example, if two functions are called within one expression
4045 with no sequence point between them, the order in which the functions
4046 are called is not specified. However, the standards committee have
4047 ruled that function calls do not overlap.
4049 It is not specified when between sequence points modifications to the
4050 values of objects take effect. Programs whose behavior depends on this
4051 have undefined behavior; the C and C++ standards specify that ``Between
4052 the previous and next sequence point an object shall have its stored
4053 value modified at most once by the evaluation of an expression.
4054 Furthermore, the prior value shall be read only to determine the value
4055 to be stored.''. If a program breaks these rules, the results on any
4056 particular implementation are entirely unpredictable.
4058 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4059 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4060 diagnosed by this option, and it may give an occasional false positive
4061 result, but in general it has been found fairly effective at detecting
4062 this sort of problem in programs.
4064 The standard is worded confusingly, therefore there is some debate
4065 over the precise meaning of the sequence point rules in subtle cases.
4066 Links to discussions of the problem, including proposed formal
4067 definitions, may be found on the GCC readings page, at
4068 @uref{http://gcc.gnu.org/@/readings.html}.
4070 This warning is enabled by @option{-Wall} for C and C++.
4072 @item -Wno-return-local-addr
4073 @opindex Wno-return-local-addr
4074 @opindex Wreturn-local-addr
4075 Do not warn about returning a pointer (or in C++, a reference) to a
4076 variable that goes out of scope after the function returns.
4079 @opindex Wreturn-type
4080 @opindex Wno-return-type
4081 Warn whenever a function is defined with a return type that defaults
4082 to @code{int}. Also warn about any @code{return} statement with no
4083 return value in a function whose return type is not @code{void}
4084 (falling off the end of the function body is considered returning
4085 without a value), and about a @code{return} statement with an
4086 expression in a function whose return type is @code{void}.
4088 For C++, a function without return type always produces a diagnostic
4089 message, even when @option{-Wno-return-type} is specified. The only
4090 exceptions are @code{main} and functions defined in system headers.
4092 This warning is enabled by @option{-Wall}.
4094 @item -Wshift-count-negative
4095 @opindex Wshift-count-negative
4096 @opindex Wno-shift-count-negative
4097 Warn if shift count is negative. This warning is enabled by default.
4099 @item -Wshift-count-overflow
4100 @opindex Wshift-count-overflow
4101 @opindex Wno-shift-count-overflow
4102 Warn if shift count >= width of type. This warning is enabled by default.
4104 @item -Wshift-negative-value
4105 @opindex Wshift-negative-value
4106 @opindex Wno-shift-negative-value
4107 Warn if left shifting a negative value. This warning is enabled by
4108 @option{-Wextra} in C99 and C++11 modes (and newer).
4110 @item -Wshift-overflow
4111 @itemx -Wshift-overflow=@var{n}
4112 @opindex Wshift-overflow
4113 @opindex Wno-shift-overflow
4114 Warn about left shift overflows. This warning is enabled by
4115 default in C99 and C++11 modes (and newer).
4118 @item -Wshift-overflow=1
4119 This is the warning level of @option{-Wshift-overflow} and is enabled
4120 by default in C99 and C++11 modes (and newer). This warning level does
4121 not warn about left-shifting 1 into the sign bit. (However, in C, such
4122 an overflow is still rejected in contexts where an integer constant expression
4125 @item -Wshift-overflow=2
4126 This warning level also warns about left-shifting 1 into the sign bit,
4127 unless C++14 mode is active.
4133 Warn whenever a @code{switch} statement has an index of enumerated type
4134 and lacks a @code{case} for one or more of the named codes of that
4135 enumeration. (The presence of a @code{default} label prevents this
4136 warning.) @code{case} labels outside the enumeration range also
4137 provoke warnings when this option is used (even if there is a
4138 @code{default} label).
4139 This warning is enabled by @option{-Wall}.
4141 @item -Wswitch-default
4142 @opindex Wswitch-default
4143 @opindex Wno-switch-default
4144 Warn whenever a @code{switch} statement does not have a @code{default}
4148 @opindex Wswitch-enum
4149 @opindex Wno-switch-enum
4150 Warn whenever a @code{switch} statement has an index of enumerated type
4151 and lacks a @code{case} for one or more of the named codes of that
4152 enumeration. @code{case} labels outside the enumeration range also
4153 provoke warnings when this option is used. The only difference
4154 between @option{-Wswitch} and this option is that this option gives a
4155 warning about an omitted enumeration code even if there is a
4156 @code{default} label.
4159 @opindex Wswitch-bool
4160 @opindex Wno-switch-bool
4161 Warn whenever a @code{switch} statement has an index of boolean type
4162 and the case values are outside the range of a boolean type.
4163 It is possible to suppress this warning by casting the controlling
4164 expression to a type other than @code{bool}. For example:
4167 switch ((int) (a == 4))
4173 This warning is enabled by default for C and C++ programs.
4175 @item -Wsync-nand @r{(C and C++ only)}
4177 @opindex Wno-sync-nand
4178 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4179 built-in functions are used. These functions changed semantics in GCC 4.4.
4183 @opindex Wno-trigraphs
4184 Warn if any trigraphs are encountered that might change the meaning of
4185 the program (trigraphs within comments are not warned about).
4186 This warning is enabled by @option{-Wall}.
4188 @item -Wunused-but-set-parameter
4189 @opindex Wunused-but-set-parameter
4190 @opindex Wno-unused-but-set-parameter
4191 Warn whenever a function parameter is assigned to, but otherwise unused
4192 (aside from its declaration).
4194 To suppress this warning use the @code{unused} attribute
4195 (@pxref{Variable Attributes}).
4197 This warning is also enabled by @option{-Wunused} together with
4200 @item -Wunused-but-set-variable
4201 @opindex Wunused-but-set-variable
4202 @opindex Wno-unused-but-set-variable
4203 Warn whenever a local variable is assigned to, but otherwise unused
4204 (aside from its declaration).
4205 This warning is enabled by @option{-Wall}.
4207 To suppress this warning use the @code{unused} attribute
4208 (@pxref{Variable Attributes}).
4210 This warning is also enabled by @option{-Wunused}, which is enabled
4213 @item -Wunused-function
4214 @opindex Wunused-function
4215 @opindex Wno-unused-function
4216 Warn whenever a static function is declared but not defined or a
4217 non-inline static function is unused.
4218 This warning is enabled by @option{-Wall}.
4220 @item -Wunused-label
4221 @opindex Wunused-label
4222 @opindex Wno-unused-label
4223 Warn whenever a label is declared but not used.
4224 This warning is enabled by @option{-Wall}.
4226 To suppress this warning use the @code{unused} attribute
4227 (@pxref{Variable Attributes}).
4229 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4230 @opindex Wunused-local-typedefs
4231 Warn when a typedef locally defined in a function is not used.
4232 This warning is enabled by @option{-Wall}.
4234 @item -Wunused-parameter
4235 @opindex Wunused-parameter
4236 @opindex Wno-unused-parameter
4237 Warn whenever a function parameter is unused aside from its declaration.
4239 To suppress this warning use the @code{unused} attribute
4240 (@pxref{Variable Attributes}).
4242 @item -Wno-unused-result
4243 @opindex Wunused-result
4244 @opindex Wno-unused-result
4245 Do not warn if a caller of a function marked with attribute
4246 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4247 its return value. The default is @option{-Wunused-result}.
4249 @item -Wunused-variable
4250 @opindex Wunused-variable
4251 @opindex Wno-unused-variable
4252 Warn whenever a local or static variable is unused aside from its
4253 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4254 but not for C++. This warning is enabled by @option{-Wall}.
4256 To suppress this warning use the @code{unused} attribute
4257 (@pxref{Variable Attributes}).
4259 @item -Wunused-const-variable
4260 @itemx -Wunused-const-variable=@var{n}
4261 @opindex Wunused-const-variable
4262 @opindex Wno-unused-const-variable
4263 Warn whenever a constant static variable is unused aside from its declaration.
4264 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4265 for C, but not for C++. In C this declares variable storage, but in C++ this
4266 is not an error since const variables take the place of @code{#define}s.
4268 To suppress this warning use the @code{unused} attribute
4269 (@pxref{Variable Attributes}).
4272 @item -Wunused-const-variable=1
4273 This is the warning level that is enabled by @option{-Wunused-variable} for
4274 C. It warns only about unused static const variables defined in the main
4275 compilation unit, but not about static const variables declared in any
4278 @item -Wunused-const-variable=2
4279 This warning level also warns for unused constant static variables in
4280 headers (excluding system headers). This is the warning level of
4281 @option{-Wunused-const-variable} and must be explicitly requested since
4282 in C++ this isn't an error and in C it might be harder to clean up all
4286 @item -Wunused-value
4287 @opindex Wunused-value
4288 @opindex Wno-unused-value
4289 Warn whenever a statement computes a result that is explicitly not
4290 used. To suppress this warning cast the unused expression to
4291 @code{void}. This includes an expression-statement or the left-hand
4292 side of a comma expression that contains no side effects. For example,
4293 an expression such as @code{x[i,j]} causes a warning, while
4294 @code{x[(void)i,j]} does not.
4296 This warning is enabled by @option{-Wall}.
4301 All the above @option{-Wunused} options combined.
4303 In order to get a warning about an unused function parameter, you must
4304 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4305 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4307 @item -Wuninitialized
4308 @opindex Wuninitialized
4309 @opindex Wno-uninitialized
4310 Warn if an automatic variable is used without first being initialized
4311 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4312 warn if a non-static reference or non-static @code{const} member
4313 appears in a class without constructors.
4315 If you want to warn about code that uses the uninitialized value of the
4316 variable in its own initializer, use the @option{-Winit-self} option.
4318 These warnings occur for individual uninitialized or clobbered
4319 elements of structure, union or array variables as well as for
4320 variables that are uninitialized or clobbered as a whole. They do
4321 not occur for variables or elements declared @code{volatile}. Because
4322 these warnings depend on optimization, the exact variables or elements
4323 for which there are warnings depends on the precise optimization
4324 options and version of GCC used.
4326 Note that there may be no warning about a variable that is used only
4327 to compute a value that itself is never used, because such
4328 computations may be deleted by data flow analysis before the warnings
4331 @item -Winvalid-memory-model
4332 @opindex Winvalid-memory-model
4333 @opindex Wno-invalid-memory-model
4334 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4335 and the C11 atomic generic functions with a memory consistency argument
4336 that is either invalid for the operation or outside the range of values
4337 of the @code{memory_order} enumeration. For example, since the
4338 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4339 defined for the relaxed, release, and sequentially consistent memory
4340 orders the following code is diagnosed:
4345 __atomic_store_n (i, 0, memory_order_consume);
4349 @option{-Winvalid-memory-model} is enabled by default.
4351 @item -Wmaybe-uninitialized
4352 @opindex Wmaybe-uninitialized
4353 @opindex Wno-maybe-uninitialized
4354 For an automatic variable, if there exists a path from the function
4355 entry to a use of the variable that is initialized, but there exist
4356 some other paths for which the variable is not initialized, the compiler
4357 emits a warning if it cannot prove the uninitialized paths are not
4358 executed at run time. These warnings are made optional because GCC is
4359 not smart enough to see all the reasons why the code might be correct
4360 in spite of appearing to have an error. Here is one example of how
4381 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4382 always initialized, but GCC doesn't know this. To suppress the
4383 warning, you need to provide a default case with assert(0) or
4386 @cindex @code{longjmp} warnings
4387 This option also warns when a non-volatile automatic variable might be
4388 changed by a call to @code{longjmp}. These warnings as well are possible
4389 only in optimizing compilation.
4391 The compiler sees only the calls to @code{setjmp}. It cannot know
4392 where @code{longjmp} will be called; in fact, a signal handler could
4393 call it at any point in the code. As a result, you may get a warning
4394 even when there is in fact no problem because @code{longjmp} cannot
4395 in fact be called at the place that would cause a problem.
4397 Some spurious warnings can be avoided if you declare all the functions
4398 you use that never return as @code{noreturn}. @xref{Function
4401 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4403 @item -Wunknown-pragmas
4404 @opindex Wunknown-pragmas
4405 @opindex Wno-unknown-pragmas
4406 @cindex warning for unknown pragmas
4407 @cindex unknown pragmas, warning
4408 @cindex pragmas, warning of unknown
4409 Warn when a @code{#pragma} directive is encountered that is not understood by
4410 GCC@. If this command-line option is used, warnings are even issued
4411 for unknown pragmas in system header files. This is not the case if
4412 the warnings are only enabled by the @option{-Wall} command-line option.
4415 @opindex Wno-pragmas
4417 Do not warn about misuses of pragmas, such as incorrect parameters,
4418 invalid syntax, or conflicts between pragmas. See also
4419 @option{-Wunknown-pragmas}.
4421 @item -Wstrict-aliasing
4422 @opindex Wstrict-aliasing
4423 @opindex Wno-strict-aliasing
4424 This option is only active when @option{-fstrict-aliasing} is active.
4425 It warns about code that might break the strict aliasing rules that the
4426 compiler is using for optimization. The warning does not catch all
4427 cases, but does attempt to catch the more common pitfalls. It is
4428 included in @option{-Wall}.
4429 It is equivalent to @option{-Wstrict-aliasing=3}
4431 @item -Wstrict-aliasing=n
4432 @opindex Wstrict-aliasing=n
4433 This option is only active when @option{-fstrict-aliasing} is active.
4434 It warns about code that might break the strict aliasing rules that the
4435 compiler is using for optimization.
4436 Higher levels correspond to higher accuracy (fewer false positives).
4437 Higher levels also correspond to more effort, similar to the way @option{-O}
4439 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4441 Level 1: Most aggressive, quick, least accurate.
4442 Possibly useful when higher levels
4443 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4444 false negatives. However, it has many false positives.
4445 Warns for all pointer conversions between possibly incompatible types,
4446 even if never dereferenced. Runs in the front end only.
4448 Level 2: Aggressive, quick, not too precise.
4449 May still have many false positives (not as many as level 1 though),
4450 and few false negatives (but possibly more than level 1).
4451 Unlike level 1, it only warns when an address is taken. Warns about
4452 incomplete types. Runs in the front end only.
4454 Level 3 (default for @option{-Wstrict-aliasing}):
4455 Should have very few false positives and few false
4456 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4457 Takes care of the common pun+dereference pattern in the front end:
4458 @code{*(int*)&some_float}.
4459 If optimization is enabled, it also runs in the back end, where it deals
4460 with multiple statement cases using flow-sensitive points-to information.
4461 Only warns when the converted pointer is dereferenced.
4462 Does not warn about incomplete types.
4464 @item -Wstrict-overflow
4465 @itemx -Wstrict-overflow=@var{n}
4466 @opindex Wstrict-overflow
4467 @opindex Wno-strict-overflow
4468 This option is only active when @option{-fstrict-overflow} is active.
4469 It warns about cases where the compiler optimizes based on the
4470 assumption that signed overflow does not occur. Note that it does not
4471 warn about all cases where the code might overflow: it only warns
4472 about cases where the compiler implements some optimization. Thus
4473 this warning depends on the optimization level.
4475 An optimization that assumes that signed overflow does not occur is
4476 perfectly safe if the values of the variables involved are such that
4477 overflow never does, in fact, occur. Therefore this warning can
4478 easily give a false positive: a warning about code that is not
4479 actually a problem. To help focus on important issues, several
4480 warning levels are defined. No warnings are issued for the use of
4481 undefined signed overflow when estimating how many iterations a loop
4482 requires, in particular when determining whether a loop will be
4486 @item -Wstrict-overflow=1
4487 Warn about cases that are both questionable and easy to avoid. For
4488 example, with @option{-fstrict-overflow}, the compiler simplifies
4489 @code{x + 1 > x} to @code{1}. This level of
4490 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4491 are not, and must be explicitly requested.
4493 @item -Wstrict-overflow=2
4494 Also warn about other cases where a comparison is simplified to a
4495 constant. For example: @code{abs (x) >= 0}. This can only be
4496 simplified when @option{-fstrict-overflow} is in effect, because
4497 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4498 zero. @option{-Wstrict-overflow} (with no level) is the same as
4499 @option{-Wstrict-overflow=2}.
4501 @item -Wstrict-overflow=3
4502 Also warn about other cases where a comparison is simplified. For
4503 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4505 @item -Wstrict-overflow=4
4506 Also warn about other simplifications not covered by the above cases.
4507 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4509 @item -Wstrict-overflow=5
4510 Also warn about cases where the compiler reduces the magnitude of a
4511 constant involved in a comparison. For example: @code{x + 2 > y} is
4512 simplified to @code{x + 1 >= y}. This is reported only at the
4513 highest warning level because this simplification applies to many
4514 comparisons, so this warning level gives a very large number of
4518 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4519 @opindex Wsuggest-attribute=
4520 @opindex Wno-suggest-attribute=
4521 Warn for cases where adding an attribute may be beneficial. The
4522 attributes currently supported are listed below.
4525 @item -Wsuggest-attribute=pure
4526 @itemx -Wsuggest-attribute=const
4527 @itemx -Wsuggest-attribute=noreturn
4528 @opindex Wsuggest-attribute=pure
4529 @opindex Wno-suggest-attribute=pure
4530 @opindex Wsuggest-attribute=const
4531 @opindex Wno-suggest-attribute=const
4532 @opindex Wsuggest-attribute=noreturn
4533 @opindex Wno-suggest-attribute=noreturn
4535 Warn about functions that might be candidates for attributes
4536 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4537 functions visible in other compilation units or (in the case of @code{pure} and
4538 @code{const}) if it cannot prove that the function returns normally. A function
4539 returns normally if it doesn't contain an infinite loop or return abnormally
4540 by throwing, calling @code{abort} or trapping. This analysis requires option
4541 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4542 higher. Higher optimization levels improve the accuracy of the analysis.
4544 @item -Wsuggest-attribute=format
4545 @itemx -Wmissing-format-attribute
4546 @opindex Wsuggest-attribute=format
4547 @opindex Wmissing-format-attribute
4548 @opindex Wno-suggest-attribute=format
4549 @opindex Wno-missing-format-attribute
4553 Warn about function pointers that might be candidates for @code{format}
4554 attributes. Note these are only possible candidates, not absolute ones.
4555 GCC guesses that function pointers with @code{format} attributes that
4556 are used in assignment, initialization, parameter passing or return
4557 statements should have a corresponding @code{format} attribute in the
4558 resulting type. I.e.@: the left-hand side of the assignment or
4559 initialization, the type of the parameter variable, or the return type
4560 of the containing function respectively should also have a @code{format}
4561 attribute to avoid the warning.
4563 GCC also warns about function definitions that might be
4564 candidates for @code{format} attributes. Again, these are only
4565 possible candidates. GCC guesses that @code{format} attributes
4566 might be appropriate for any function that calls a function like
4567 @code{vprintf} or @code{vscanf}, but this might not always be the
4568 case, and some functions for which @code{format} attributes are
4569 appropriate may not be detected.
4572 @item -Wsuggest-final-types
4573 @opindex Wno-suggest-final-types
4574 @opindex Wsuggest-final-types
4575 Warn about types with virtual methods where code quality would be improved
4576 if the type were declared with the C++11 @code{final} specifier,
4578 declared in an anonymous namespace. This allows GCC to more aggressively
4579 devirtualize the polymorphic calls. This warning is more effective with link
4580 time optimization, where the information about the class hierarchy graph is
4583 @item -Wsuggest-final-methods
4584 @opindex Wno-suggest-final-methods
4585 @opindex Wsuggest-final-methods
4586 Warn about virtual methods where code quality would be improved if the method
4587 were declared with the C++11 @code{final} specifier,
4588 or, if possible, its type were
4589 declared in an anonymous namespace or with the @code{final} specifier.
4591 more effective with link time optimization, where the information about the
4592 class hierarchy graph is more complete. It is recommended to first consider
4593 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4596 @item -Wsuggest-override
4597 Warn about overriding virtual functions that are not marked with the override
4600 @item -Warray-bounds
4601 @itemx -Warray-bounds=@var{n}
4602 @opindex Wno-array-bounds
4603 @opindex Warray-bounds
4604 This option is only active when @option{-ftree-vrp} is active
4605 (default for @option{-O2} and above). It warns about subscripts to arrays
4606 that are always out of bounds. This warning is enabled by @option{-Wall}.
4609 @item -Warray-bounds=1
4610 This is the warning level of @option{-Warray-bounds} and is enabled
4611 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4613 @item -Warray-bounds=2
4614 This warning level also warns about out of bounds access for
4615 arrays at the end of a struct and for arrays accessed through
4616 pointers. This warning level may give a larger number of
4617 false positives and is deactivated by default.
4620 @item -Wbool-compare
4621 @opindex Wno-bool-compare
4622 @opindex Wbool-compare
4623 Warn about boolean expression compared with an integer value different from
4624 @code{true}/@code{false}. For instance, the following comparison is
4629 if ((n > 1) == 2) @{ @dots{} @}
4631 This warning is enabled by @option{-Wall}.
4633 @item -Wduplicated-cond
4634 @opindex Wno-duplicated-cond
4635 @opindex Wduplicated-cond
4636 Warn about duplicated conditions in an if-else-if chain. For instance,
4637 warn for the following code:
4639 if (p->q != NULL) @{ @dots{} @}
4640 else if (p->q != NULL) @{ @dots{} @}
4643 @item -Wframe-address
4644 @opindex Wno-frame-address
4645 @opindex Wframe-address
4646 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4647 is called with an argument greater than 0. Such calls may return indeterminate
4648 values or crash the program. The warning is included in @option{-Wall}.
4650 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4651 @opindex Wno-discarded-qualifiers
4652 @opindex Wdiscarded-qualifiers
4653 Do not warn if type qualifiers on pointers are being discarded.
4654 Typically, the compiler warns if a @code{const char *} variable is
4655 passed to a function that takes a @code{char *} parameter. This option
4656 can be used to suppress such a warning.
4658 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4659 @opindex Wno-discarded-array-qualifiers
4660 @opindex Wdiscarded-array-qualifiers
4661 Do not warn if type qualifiers on arrays which are pointer targets
4662 are being discarded. Typically, the compiler warns if a
4663 @code{const int (*)[]} variable is passed to a function that
4664 takes a @code{int (*)[]} parameter. This option can be used to
4665 suppress such a warning.
4667 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4668 @opindex Wno-incompatible-pointer-types
4669 @opindex Wincompatible-pointer-types
4670 Do not warn when there is a conversion between pointers that have incompatible
4671 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4672 which warns for pointer argument passing or assignment with different
4675 @item -Wno-int-conversion @r{(C and Objective-C only)}
4676 @opindex Wno-int-conversion
4677 @opindex Wint-conversion
4678 Do not warn about incompatible integer to pointer and pointer to integer
4679 conversions. This warning is about implicit conversions; for explicit
4680 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4681 @option{-Wno-pointer-to-int-cast} may be used.
4683 @item -Wno-div-by-zero
4684 @opindex Wno-div-by-zero
4685 @opindex Wdiv-by-zero
4686 Do not warn about compile-time integer division by zero. Floating-point
4687 division by zero is not warned about, as it can be a legitimate way of
4688 obtaining infinities and NaNs.
4690 @item -Wsystem-headers
4691 @opindex Wsystem-headers
4692 @opindex Wno-system-headers
4693 @cindex warnings from system headers
4694 @cindex system headers, warnings from
4695 Print warning messages for constructs found in system header files.
4696 Warnings from system headers are normally suppressed, on the assumption
4697 that they usually do not indicate real problems and would only make the
4698 compiler output harder to read. Using this command-line option tells
4699 GCC to emit warnings from system headers as if they occurred in user
4700 code. However, note that using @option{-Wall} in conjunction with this
4701 option does @emph{not} warn about unknown pragmas in system
4702 headers---for that, @option{-Wunknown-pragmas} must also be used.
4704 @item -Wtautological-compare
4705 @opindex Wtautological-compare
4706 @opindex Wno-tautological-compare
4707 Warn if a self-comparison always evaluates to true or false. This
4708 warning detects various mistakes such as:
4712 if (i > i) @{ @dots{} @}
4714 This warning is enabled by @option{-Wall}.
4717 @opindex Wtrampolines
4718 @opindex Wno-trampolines
4719 Warn about trampolines generated for pointers to nested functions.
4720 A trampoline is a small piece of data or code that is created at run
4721 time on the stack when the address of a nested function is taken, and is
4722 used to call the nested function indirectly. For some targets, it is
4723 made up of data only and thus requires no special treatment. But, for
4724 most targets, it is made up of code and thus requires the stack to be
4725 made executable in order for the program to work properly.
4728 @opindex Wfloat-equal
4729 @opindex Wno-float-equal
4730 Warn if floating-point values are used in equality comparisons.
4732 The idea behind this is that sometimes it is convenient (for the
4733 programmer) to consider floating-point values as approximations to
4734 infinitely precise real numbers. If you are doing this, then you need
4735 to compute (by analyzing the code, or in some other way) the maximum or
4736 likely maximum error that the computation introduces, and allow for it
4737 when performing comparisons (and when producing output, but that's a
4738 different problem). In particular, instead of testing for equality, you
4739 should check to see whether the two values have ranges that overlap; and
4740 this is done with the relational operators, so equality comparisons are
4743 @item -Wtraditional @r{(C and Objective-C only)}
4744 @opindex Wtraditional
4745 @opindex Wno-traditional
4746 Warn about certain constructs that behave differently in traditional and
4747 ISO C@. Also warn about ISO C constructs that have no traditional C
4748 equivalent, and/or problematic constructs that should be avoided.
4752 Macro parameters that appear within string literals in the macro body.
4753 In traditional C macro replacement takes place within string literals,
4754 but in ISO C it does not.
4757 In traditional C, some preprocessor directives did not exist.
4758 Traditional preprocessors only considered a line to be a directive
4759 if the @samp{#} appeared in column 1 on the line. Therefore
4760 @option{-Wtraditional} warns about directives that traditional C
4761 understands but ignores because the @samp{#} does not appear as the
4762 first character on the line. It also suggests you hide directives like
4763 @code{#pragma} not understood by traditional C by indenting them. Some
4764 traditional implementations do not recognize @code{#elif}, so this option
4765 suggests avoiding it altogether.
4768 A function-like macro that appears without arguments.
4771 The unary plus operator.
4774 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4775 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4776 constants.) Note, these suffixes appear in macros defined in the system
4777 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4778 Use of these macros in user code might normally lead to spurious
4779 warnings, however GCC's integrated preprocessor has enough context to
4780 avoid warning in these cases.
4783 A function declared external in one block and then used after the end of
4787 A @code{switch} statement has an operand of type @code{long}.
4790 A non-@code{static} function declaration follows a @code{static} one.
4791 This construct is not accepted by some traditional C compilers.
4794 The ISO type of an integer constant has a different width or
4795 signedness from its traditional type. This warning is only issued if
4796 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4797 typically represent bit patterns, are not warned about.
4800 Usage of ISO string concatenation is detected.
4803 Initialization of automatic aggregates.
4806 Identifier conflicts with labels. Traditional C lacks a separate
4807 namespace for labels.
4810 Initialization of unions. If the initializer is zero, the warning is
4811 omitted. This is done under the assumption that the zero initializer in
4812 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4813 initializer warnings and relies on default initialization to zero in the
4817 Conversions by prototypes between fixed/floating-point values and vice
4818 versa. The absence of these prototypes when compiling with traditional
4819 C causes serious problems. This is a subset of the possible
4820 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4823 Use of ISO C style function definitions. This warning intentionally is
4824 @emph{not} issued for prototype declarations or variadic functions
4825 because these ISO C features appear in your code when using
4826 libiberty's traditional C compatibility macros, @code{PARAMS} and
4827 @code{VPARAMS}. This warning is also bypassed for nested functions
4828 because that feature is already a GCC extension and thus not relevant to
4829 traditional C compatibility.
4832 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4833 @opindex Wtraditional-conversion
4834 @opindex Wno-traditional-conversion
4835 Warn if a prototype causes a type conversion that is different from what
4836 would happen to the same argument in the absence of a prototype. This
4837 includes conversions of fixed point to floating and vice versa, and
4838 conversions changing the width or signedness of a fixed-point argument
4839 except when the same as the default promotion.
4841 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4842 @opindex Wdeclaration-after-statement
4843 @opindex Wno-declaration-after-statement
4844 Warn when a declaration is found after a statement in a block. This
4845 construct, known from C++, was introduced with ISO C99 and is by default
4846 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4851 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4853 @item -Wno-endif-labels
4854 @opindex Wno-endif-labels
4855 @opindex Wendif-labels
4856 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4861 Warn whenever a local variable or type declaration shadows another
4862 variable, parameter, type, class member (in C++), or instance variable
4863 (in Objective-C) or whenever a built-in function is shadowed. Note
4864 that in C++, the compiler warns if a local variable shadows an
4865 explicit typedef, but not if it shadows a struct/class/enum.
4867 @item -Wno-shadow-ivar @r{(Objective-C only)}
4868 @opindex Wno-shadow-ivar
4869 @opindex Wshadow-ivar
4870 Do not warn whenever a local variable shadows an instance variable in an
4873 @item -Wlarger-than=@var{len}
4874 @opindex Wlarger-than=@var{len}
4875 @opindex Wlarger-than-@var{len}
4876 Warn whenever an object of larger than @var{len} bytes is defined.
4878 @item -Wframe-larger-than=@var{len}
4879 @opindex Wframe-larger-than
4880 Warn if the size of a function frame is larger than @var{len} bytes.
4881 The computation done to determine the stack frame size is approximate
4882 and not conservative.
4883 The actual requirements may be somewhat greater than @var{len}
4884 even if you do not get a warning. In addition, any space allocated
4885 via @code{alloca}, variable-length arrays, or related constructs
4886 is not included by the compiler when determining
4887 whether or not to issue a warning.
4889 @item -Wno-free-nonheap-object
4890 @opindex Wno-free-nonheap-object
4891 @opindex Wfree-nonheap-object
4892 Do not warn when attempting to free an object that was not allocated
4895 @item -Wstack-usage=@var{len}
4896 @opindex Wstack-usage
4897 Warn if the stack usage of a function might be larger than @var{len} bytes.
4898 The computation done to determine the stack usage is conservative.
4899 Any space allocated via @code{alloca}, variable-length arrays, or related
4900 constructs is included by the compiler when determining whether or not to
4903 The message is in keeping with the output of @option{-fstack-usage}.
4907 If the stack usage is fully static but exceeds the specified amount, it's:
4910 warning: stack usage is 1120 bytes
4913 If the stack usage is (partly) dynamic but bounded, it's:
4916 warning: stack usage might be 1648 bytes
4919 If the stack usage is (partly) dynamic and not bounded, it's:
4922 warning: stack usage might be unbounded
4926 @item -Wunsafe-loop-optimizations
4927 @opindex Wunsafe-loop-optimizations
4928 @opindex Wno-unsafe-loop-optimizations
4929 Warn if the loop cannot be optimized because the compiler cannot
4930 assume anything on the bounds of the loop indices. With
4931 @option{-funsafe-loop-optimizations} warn if the compiler makes
4934 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4935 @opindex Wno-pedantic-ms-format
4936 @opindex Wpedantic-ms-format
4937 When used in combination with @option{-Wformat}
4938 and @option{-pedantic} without GNU extensions, this option
4939 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4940 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4941 which depend on the MS runtime.
4943 @item -Wplacement-new
4944 @itemx -Wplacement-new=@var{n}
4945 @opindex Wplacement-new
4946 @opindex Wno-placement-new
4947 Warn about placement new expressions with undefined behavior, such as
4948 constructing an object in a buffer that is smaller than the type of
4949 the object. For example, the placement new expression below is diagnosed
4950 because it attempts to construct an array of 64 integers in a buffer only
4956 This warning is enabled by default.
4959 @item -Wplacement-new=1
4960 This is the default warning level of @option{-Wplacement-new}. At this
4961 level the warning is not issued for some strictly undefined constructs that
4962 GCC allows as extensions for compatibility with legacy code. For example,
4963 the following @code{new} expression is not diagnosed at this level even
4964 though it has undefined behavior according to the C++ standard because
4965 it writes past the end of the one-element array.
4967 struct S @{ int n, a[1]; @};
4968 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
4969 new (s->a)int [32]();
4972 @item -Wplacement-new=2
4973 At this level, in addition to diagnosing all the same constructs as at level
4974 1, a diagnostic is also issued for placement new expressions that construct
4975 an object in the last member of structure whose type is an array of a single
4976 element and whose size is less than the size of the object being constructed.
4977 While the previous example would be diagnosed, the following construct makes
4978 use of the flexible member array extension to avoid the warning at level 2.
4980 struct S @{ int n, a[]; @};
4981 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
4982 new (s->a)int [32]();
4987 @item -Wpointer-arith
4988 @opindex Wpointer-arith
4989 @opindex Wno-pointer-arith
4990 Warn about anything that depends on the ``size of'' a function type or
4991 of @code{void}. GNU C assigns these types a size of 1, for
4992 convenience in calculations with @code{void *} pointers and pointers
4993 to functions. In C++, warn also when an arithmetic operation involves
4994 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4997 @opindex Wtype-limits
4998 @opindex Wno-type-limits
4999 Warn if a comparison is always true or always false due to the limited
5000 range of the data type, but do not warn for constant expressions. For
5001 example, warn if an unsigned variable is compared against zero with
5002 @code{<} or @code{>=}. This warning is also enabled by
5005 @item -Wbad-function-cast @r{(C and Objective-C only)}
5006 @opindex Wbad-function-cast
5007 @opindex Wno-bad-function-cast
5008 Warn when a function call is cast to a non-matching type.
5009 For example, warn if a call to a function returning an integer type
5010 is cast to a pointer type.
5012 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5013 @opindex Wc90-c99-compat
5014 @opindex Wno-c90-c99-compat
5015 Warn about features not present in ISO C90, but present in ISO C99.
5016 For instance, warn about use of variable length arrays, @code{long long}
5017 type, @code{bool} type, compound literals, designated initializers, and so
5018 on. This option is independent of the standards mode. Warnings are disabled
5019 in the expression that follows @code{__extension__}.
5021 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5022 @opindex Wc99-c11-compat
5023 @opindex Wno-c99-c11-compat
5024 Warn about features not present in ISO C99, but present in ISO C11.
5025 For instance, warn about use of anonymous structures and unions,
5026 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5027 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5028 and so on. This option is independent of the standards mode. Warnings are
5029 disabled in the expression that follows @code{__extension__}.
5031 @item -Wc++-compat @r{(C and Objective-C only)}
5032 @opindex Wc++-compat
5033 Warn about ISO C constructs that are outside of the common subset of
5034 ISO C and ISO C++, e.g.@: request for implicit conversion from
5035 @code{void *} to a pointer to non-@code{void} type.
5037 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5038 @opindex Wc++11-compat
5039 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5040 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5041 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5042 enabled by @option{-Wall}.
5044 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5045 @opindex Wc++14-compat
5046 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5047 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5051 @opindex Wno-cast-qual
5052 Warn whenever a pointer is cast so as to remove a type qualifier from
5053 the target type. For example, warn if a @code{const char *} is cast
5054 to an ordinary @code{char *}.
5056 Also warn when making a cast that introduces a type qualifier in an
5057 unsafe way. For example, casting @code{char **} to @code{const char **}
5058 is unsafe, as in this example:
5061 /* p is char ** value. */
5062 const char **q = (const char **) p;
5063 /* Assignment of readonly string to const char * is OK. */
5065 /* Now char** pointer points to read-only memory. */
5070 @opindex Wcast-align
5071 @opindex Wno-cast-align
5072 Warn whenever a pointer is cast such that the required alignment of the
5073 target is increased. For example, warn if a @code{char *} is cast to
5074 an @code{int *} on machines where integers can only be accessed at
5075 two- or four-byte boundaries.
5077 @item -Wwrite-strings
5078 @opindex Wwrite-strings
5079 @opindex Wno-write-strings
5080 When compiling C, give string constants the type @code{const
5081 char[@var{length}]} so that copying the address of one into a
5082 non-@code{const} @code{char *} pointer produces a warning. These
5083 warnings help you find at compile time code that can try to write
5084 into a string constant, but only if you have been very careful about
5085 using @code{const} in declarations and prototypes. Otherwise, it is
5086 just a nuisance. This is why we did not make @option{-Wall} request
5089 When compiling C++, warn about the deprecated conversion from string
5090 literals to @code{char *}. This warning is enabled by default for C++
5095 @opindex Wno-clobbered
5096 Warn for variables that might be changed by @code{longjmp} or
5097 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5099 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5100 @opindex Wconditionally-supported
5101 @opindex Wno-conditionally-supported
5102 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5105 @opindex Wconversion
5106 @opindex Wno-conversion
5107 Warn for implicit conversions that may alter a value. This includes
5108 conversions between real and integer, like @code{abs (x)} when
5109 @code{x} is @code{double}; conversions between signed and unsigned,
5110 like @code{unsigned ui = -1}; and conversions to smaller types, like
5111 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5112 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5113 changed by the conversion like in @code{abs (2.0)}. Warnings about
5114 conversions between signed and unsigned integers can be disabled by
5115 using @option{-Wno-sign-conversion}.
5117 For C++, also warn for confusing overload resolution for user-defined
5118 conversions; and conversions that never use a type conversion
5119 operator: conversions to @code{void}, the same type, a base class or a
5120 reference to them. Warnings about conversions between signed and
5121 unsigned integers are disabled by default in C++ unless
5122 @option{-Wsign-conversion} is explicitly enabled.
5124 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5125 @opindex Wconversion-null
5126 @opindex Wno-conversion-null
5127 Do not warn for conversions between @code{NULL} and non-pointer
5128 types. @option{-Wconversion-null} is enabled by default.
5130 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5131 @opindex Wzero-as-null-pointer-constant
5132 @opindex Wno-zero-as-null-pointer-constant
5133 Warn when a literal @samp{0} is used as null pointer constant. This can
5134 be useful to facilitate the conversion to @code{nullptr} in C++11.
5136 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5137 @opindex Wsubobject-linkage
5138 @opindex Wno-subobject-linkage
5139 Warn if a class type has a base or a field whose type uses the anonymous
5140 namespace or depends on a type with no linkage. If a type A depends on
5141 a type B with no or internal linkage, defining it in multiple
5142 translation units would be an ODR violation because the meaning of B
5143 is different in each translation unit. If A only appears in a single
5144 translation unit, the best way to silence the warning is to give it
5145 internal linkage by putting it in an anonymous namespace as well. The
5146 compiler doesn't give this warning for types defined in the main .C
5147 file, as those are unlikely to have multiple definitions.
5148 @option{-Wsubobject-linkage} is enabled by default.
5152 @opindex Wno-date-time
5153 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5154 are encountered as they might prevent bit-wise-identical reproducible
5157 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5158 @opindex Wdelete-incomplete
5159 @opindex Wno-delete-incomplete
5160 Warn when deleting a pointer to incomplete type, which may cause
5161 undefined behavior at runtime. This warning is enabled by default.
5163 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5164 @opindex Wuseless-cast
5165 @opindex Wno-useless-cast
5166 Warn when an expression is casted to its own type.
5169 @opindex Wempty-body
5170 @opindex Wno-empty-body
5171 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5172 while} statement. This warning is also enabled by @option{-Wextra}.
5174 @item -Wenum-compare
5175 @opindex Wenum-compare
5176 @opindex Wno-enum-compare
5177 Warn about a comparison between values of different enumerated types.
5178 In C++ enumeral mismatches in conditional expressions are also
5179 diagnosed and the warning is enabled by default. In C this warning is
5180 enabled by @option{-Wall}.
5182 @item -Wjump-misses-init @r{(C, Objective-C only)}
5183 @opindex Wjump-misses-init
5184 @opindex Wno-jump-misses-init
5185 Warn if a @code{goto} statement or a @code{switch} statement jumps
5186 forward across the initialization of a variable, or jumps backward to a
5187 label after the variable has been initialized. This only warns about
5188 variables that are initialized when they are declared. This warning is
5189 only supported for C and Objective-C; in C++ this sort of branch is an
5192 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5193 can be disabled with the @option{-Wno-jump-misses-init} option.
5195 @item -Wsign-compare
5196 @opindex Wsign-compare
5197 @opindex Wno-sign-compare
5198 @cindex warning for comparison of signed and unsigned values
5199 @cindex comparison of signed and unsigned values, warning
5200 @cindex signed and unsigned values, comparison warning
5201 Warn when a comparison between signed and unsigned values could produce
5202 an incorrect result when the signed value is converted to unsigned.
5203 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5204 also enabled by @option{-Wextra}.
5206 @item -Wsign-conversion
5207 @opindex Wsign-conversion
5208 @opindex Wno-sign-conversion
5209 Warn for implicit conversions that may change the sign of an integer
5210 value, like assigning a signed integer expression to an unsigned
5211 integer variable. An explicit cast silences the warning. In C, this
5212 option is enabled also by @option{-Wconversion}.
5214 @item -Wfloat-conversion
5215 @opindex Wfloat-conversion
5216 @opindex Wno-float-conversion
5217 Warn for implicit conversions that reduce the precision of a real value.
5218 This includes conversions from real to integer, and from higher precision
5219 real to lower precision real values. This option is also enabled by
5220 @option{-Wconversion}.
5222 @item -Wno-scalar-storage-order
5223 @opindex -Wno-scalar-storage-order
5224 @opindex -Wscalar-storage-order
5225 Do not warn on suspicious constructs involving reverse scalar storage order.
5227 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5228 @opindex Wsized-deallocation
5229 @opindex Wno-sized-deallocation
5230 Warn about a definition of an unsized deallocation function
5232 void operator delete (void *) noexcept;
5233 void operator delete[] (void *) noexcept;
5235 without a definition of the corresponding sized deallocation function
5237 void operator delete (void *, std::size_t) noexcept;
5238 void operator delete[] (void *, std::size_t) noexcept;
5240 or vice versa. Enabled by @option{-Wextra} along with
5241 @option{-fsized-deallocation}.
5243 @item -Wsizeof-pointer-memaccess
5244 @opindex Wsizeof-pointer-memaccess
5245 @opindex Wno-sizeof-pointer-memaccess
5246 Warn for suspicious length parameters to certain string and memory built-in
5247 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5248 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5249 but a pointer, and suggests a possible fix, or about
5250 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5253 @item -Wsizeof-array-argument
5254 @opindex Wsizeof-array-argument
5255 @opindex Wno-sizeof-array-argument
5256 Warn when the @code{sizeof} operator is applied to a parameter that is
5257 declared as an array in a function definition. This warning is enabled by
5258 default for C and C++ programs.
5260 @item -Wmemset-transposed-args
5261 @opindex Wmemset-transposed-args
5262 @opindex Wno-memset-transposed-args
5263 Warn for suspicious calls to the @code{memset} built-in function, if the
5264 second argument is not zero and the third argument is zero. This warns e.g.@
5265 about @code{memset (buf, sizeof buf, 0)} where most probably
5266 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5267 is only emitted if the third argument is literal zero. If it is some
5268 expression that is folded to zero, a cast of zero to some type, etc.,
5269 it is far less likely that the user has mistakenly exchanged the arguments
5270 and no warning is emitted. This warning is enabled by @option{-Wall}.
5274 @opindex Wno-address
5275 Warn about suspicious uses of memory addresses. These include using
5276 the address of a function in a conditional expression, such as
5277 @code{void func(void); if (func)}, and comparisons against the memory
5278 address of a string literal, such as @code{if (x == "abc")}. Such
5279 uses typically indicate a programmer error: the address of a function
5280 always evaluates to true, so their use in a conditional usually
5281 indicate that the programmer forgot the parentheses in a function
5282 call; and comparisons against string literals result in unspecified
5283 behavior and are not portable in C, so they usually indicate that the
5284 programmer intended to use @code{strcmp}. This warning is enabled by
5288 @opindex Wlogical-op
5289 @opindex Wno-logical-op
5290 Warn about suspicious uses of logical operators in expressions.
5291 This includes using logical operators in contexts where a
5292 bit-wise operator is likely to be expected. Also warns when
5293 the operands of a logical operator are the same:
5296 if (a < 0 && a < 0) @{ @dots{} @}
5299 @item -Wlogical-not-parentheses
5300 @opindex Wlogical-not-parentheses
5301 @opindex Wno-logical-not-parentheses
5302 Warn about logical not used on the left hand side operand of a comparison.
5303 This option does not warn if the RHS operand is of a boolean type. Its
5304 purpose is to detect suspicious code like the following:
5308 if (!a > 1) @{ @dots{} @}
5311 It is possible to suppress the warning by wrapping the LHS into
5314 if ((!a) > 1) @{ @dots{} @}
5317 This warning is enabled by @option{-Wall}.
5319 @item -Waggregate-return
5320 @opindex Waggregate-return
5321 @opindex Wno-aggregate-return
5322 Warn if any functions that return structures or unions are defined or
5323 called. (In languages where you can return an array, this also elicits
5326 @item -Wno-aggressive-loop-optimizations
5327 @opindex Wno-aggressive-loop-optimizations
5328 @opindex Waggressive-loop-optimizations
5329 Warn if in a loop with constant number of iterations the compiler detects
5330 undefined behavior in some statement during one or more of the iterations.
5332 @item -Wno-attributes
5333 @opindex Wno-attributes
5334 @opindex Wattributes
5335 Do not warn if an unexpected @code{__attribute__} is used, such as
5336 unrecognized attributes, function attributes applied to variables,
5337 etc. This does not stop errors for incorrect use of supported
5340 @item -Wno-builtin-macro-redefined
5341 @opindex Wno-builtin-macro-redefined
5342 @opindex Wbuiltin-macro-redefined
5343 Do not warn if certain built-in macros are redefined. This suppresses
5344 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5345 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5347 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5348 @opindex Wstrict-prototypes
5349 @opindex Wno-strict-prototypes
5350 Warn if a function is declared or defined without specifying the
5351 argument types. (An old-style function definition is permitted without
5352 a warning if preceded by a declaration that specifies the argument
5355 @item -Wold-style-declaration @r{(C and Objective-C only)}
5356 @opindex Wold-style-declaration
5357 @opindex Wno-old-style-declaration
5358 Warn for obsolescent usages, according to the C Standard, in a
5359 declaration. For example, warn if storage-class specifiers like
5360 @code{static} are not the first things in a declaration. This warning
5361 is also enabled by @option{-Wextra}.
5363 @item -Wold-style-definition @r{(C and Objective-C only)}
5364 @opindex Wold-style-definition
5365 @opindex Wno-old-style-definition
5366 Warn if an old-style function definition is used. A warning is given
5367 even if there is a previous prototype.
5369 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5370 @opindex Wmissing-parameter-type
5371 @opindex Wno-missing-parameter-type
5372 A function parameter is declared without a type specifier in K&R-style
5379 This warning is also enabled by @option{-Wextra}.
5381 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5382 @opindex Wmissing-prototypes
5383 @opindex Wno-missing-prototypes
5384 Warn if a global function is defined without a previous prototype
5385 declaration. This warning is issued even if the definition itself
5386 provides a prototype. Use this option to detect global functions
5387 that do not have a matching prototype declaration in a header file.
5388 This option is not valid for C++ because all function declarations
5389 provide prototypes and a non-matching declaration declares an
5390 overload rather than conflict with an earlier declaration.
5391 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5393 @item -Wmissing-declarations
5394 @opindex Wmissing-declarations
5395 @opindex Wno-missing-declarations
5396 Warn if a global function is defined without a previous declaration.
5397 Do so even if the definition itself provides a prototype.
5398 Use this option to detect global functions that are not declared in
5399 header files. In C, no warnings are issued for functions with previous
5400 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5401 missing prototypes. In C++, no warnings are issued for function templates,
5402 or for inline functions, or for functions in anonymous namespaces.
5404 @item -Wmissing-field-initializers
5405 @opindex Wmissing-field-initializers
5406 @opindex Wno-missing-field-initializers
5410 Warn if a structure's initializer has some fields missing. For
5411 example, the following code causes such a warning, because
5412 @code{x.h} is implicitly zero:
5415 struct s @{ int f, g, h; @};
5416 struct s x = @{ 3, 4 @};
5419 This option does not warn about designated initializers, so the following
5420 modification does not trigger a warning:
5423 struct s @{ int f, g, h; @};
5424 struct s x = @{ .f = 3, .g = 4 @};
5427 In C++ this option does not warn either about the empty @{ @}
5428 initializer, for example:
5431 struct s @{ int f, g, h; @};
5435 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5436 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5438 @item -Wno-multichar
5439 @opindex Wno-multichar
5441 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5442 Usually they indicate a typo in the user's code, as they have
5443 implementation-defined values, and should not be used in portable code.
5445 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5446 @opindex Wnormalized=
5447 @opindex Wnormalized
5448 @opindex Wno-normalized
5451 @cindex character set, input normalization
5452 In ISO C and ISO C++, two identifiers are different if they are
5453 different sequences of characters. However, sometimes when characters
5454 outside the basic ASCII character set are used, you can have two
5455 different character sequences that look the same. To avoid confusion,
5456 the ISO 10646 standard sets out some @dfn{normalization rules} which
5457 when applied ensure that two sequences that look the same are turned into
5458 the same sequence. GCC can warn you if you are using identifiers that
5459 have not been normalized; this option controls that warning.
5461 There are four levels of warning supported by GCC@. The default is
5462 @option{-Wnormalized=nfc}, which warns about any identifier that is
5463 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5464 recommended form for most uses. It is equivalent to
5465 @option{-Wnormalized}.
5467 Unfortunately, there are some characters allowed in identifiers by
5468 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5469 identifiers. That is, there's no way to use these symbols in portable
5470 ISO C or C++ and have all your identifiers in NFC@.
5471 @option{-Wnormalized=id} suppresses the warning for these characters.
5472 It is hoped that future versions of the standards involved will correct
5473 this, which is why this option is not the default.
5475 You can switch the warning off for all characters by writing
5476 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5477 only do this if you are using some other normalization scheme (like
5478 ``D''), because otherwise you can easily create bugs that are
5479 literally impossible to see.
5481 Some characters in ISO 10646 have distinct meanings but look identical
5482 in some fonts or display methodologies, especially once formatting has
5483 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5484 LETTER N'', displays just like a regular @code{n} that has been
5485 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5486 normalization scheme to convert all these into a standard form as
5487 well, and GCC warns if your code is not in NFKC if you use
5488 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5489 about every identifier that contains the letter O because it might be
5490 confused with the digit 0, and so is not the default, but may be
5491 useful as a local coding convention if the programming environment
5492 cannot be fixed to display these characters distinctly.
5494 @item -Wno-deprecated
5495 @opindex Wno-deprecated
5496 @opindex Wdeprecated
5497 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5499 @item -Wno-deprecated-declarations
5500 @opindex Wno-deprecated-declarations
5501 @opindex Wdeprecated-declarations
5502 Do not warn about uses of functions (@pxref{Function Attributes}),
5503 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5504 Attributes}) marked as deprecated by using the @code{deprecated}
5508 @opindex Wno-overflow
5510 Do not warn about compile-time overflow in constant expressions.
5515 Warn about One Definition Rule violations during link-time optimization.
5516 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5519 @opindex Wopenm-simd
5520 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5521 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5522 option can be used to relax the cost model.
5524 @item -Woverride-init @r{(C and Objective-C only)}
5525 @opindex Woverride-init
5526 @opindex Wno-override-init
5530 Warn if an initialized field without side effects is overridden when
5531 using designated initializers (@pxref{Designated Inits, , Designated
5534 This warning is included in @option{-Wextra}. To get other
5535 @option{-Wextra} warnings without this one, use @option{-Wextra
5536 -Wno-override-init}.
5538 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5539 @opindex Woverride-init-side-effects
5540 @opindex Wno-override-init-side-effects
5541 Warn if an initialized field with side effects is overridden when
5542 using designated initializers (@pxref{Designated Inits, , Designated
5543 Initializers}). This warning is enabled by default.
5548 Warn if a structure is given the packed attribute, but the packed
5549 attribute has no effect on the layout or size of the structure.
5550 Such structures may be mis-aligned for little benefit. For
5551 instance, in this code, the variable @code{f.x} in @code{struct bar}
5552 is misaligned even though @code{struct bar} does not itself
5553 have the packed attribute:
5560 @} __attribute__((packed));
5568 @item -Wpacked-bitfield-compat
5569 @opindex Wpacked-bitfield-compat
5570 @opindex Wno-packed-bitfield-compat
5571 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5572 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5573 the change can lead to differences in the structure layout. GCC
5574 informs you when the offset of such a field has changed in GCC 4.4.
5575 For example there is no longer a 4-bit padding between field @code{a}
5576 and @code{b} in this structure:
5583 @} __attribute__ ((packed));
5586 This warning is enabled by default. Use
5587 @option{-Wno-packed-bitfield-compat} to disable this warning.
5592 Warn if padding is included in a structure, either to align an element
5593 of the structure or to align the whole structure. Sometimes when this
5594 happens it is possible to rearrange the fields of the structure to
5595 reduce the padding and so make the structure smaller.
5597 @item -Wredundant-decls
5598 @opindex Wredundant-decls
5599 @opindex Wno-redundant-decls
5600 Warn if anything is declared more than once in the same scope, even in
5601 cases where multiple declaration is valid and changes nothing.
5603 @item -Wnested-externs @r{(C and Objective-C only)}
5604 @opindex Wnested-externs
5605 @opindex Wno-nested-externs
5606 Warn if an @code{extern} declaration is encountered within a function.
5608 @item -Wno-inherited-variadic-ctor
5609 @opindex Winherited-variadic-ctor
5610 @opindex Wno-inherited-variadic-ctor
5611 Suppress warnings about use of C++11 inheriting constructors when the
5612 base class inherited from has a C variadic constructor; the warning is
5613 on by default because the ellipsis is not inherited.
5618 Warn if a function that is declared as inline cannot be inlined.
5619 Even with this option, the compiler does not warn about failures to
5620 inline functions declared in system headers.
5622 The compiler uses a variety of heuristics to determine whether or not
5623 to inline a function. For example, the compiler takes into account
5624 the size of the function being inlined and the amount of inlining
5625 that has already been done in the current function. Therefore,
5626 seemingly insignificant changes in the source program can cause the
5627 warnings produced by @option{-Winline} to appear or disappear.
5629 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5630 @opindex Wno-invalid-offsetof
5631 @opindex Winvalid-offsetof
5632 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5633 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5634 to a non-standard-layout type is undefined. In existing C++ implementations,
5635 however, @code{offsetof} typically gives meaningful results.
5636 This flag is for users who are aware that they are
5637 writing nonportable code and who have deliberately chosen to ignore the
5640 The restrictions on @code{offsetof} may be relaxed in a future version
5641 of the C++ standard.
5643 @item -Wno-int-to-pointer-cast
5644 @opindex Wno-int-to-pointer-cast
5645 @opindex Wint-to-pointer-cast
5646 Suppress warnings from casts to pointer type of an integer of a
5647 different size. In C++, casting to a pointer type of smaller size is
5648 an error. @option{Wint-to-pointer-cast} is enabled by default.
5651 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5652 @opindex Wno-pointer-to-int-cast
5653 @opindex Wpointer-to-int-cast
5654 Suppress warnings from casts from a pointer to an integer type of a
5658 @opindex Winvalid-pch
5659 @opindex Wno-invalid-pch
5660 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5661 the search path but can't be used.
5665 @opindex Wno-long-long
5666 Warn if @code{long long} type is used. This is enabled by either
5667 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5668 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5670 @item -Wvariadic-macros
5671 @opindex Wvariadic-macros
5672 @opindex Wno-variadic-macros
5673 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5674 alternate syntax is used in ISO C99 mode. This is enabled by either
5675 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5676 messages, use @option{-Wno-variadic-macros}.
5680 @opindex Wno-varargs
5681 Warn upon questionable usage of the macros used to handle variable
5682 arguments like @code{va_start}. This is default. To inhibit the
5683 warning messages, use @option{-Wno-varargs}.
5685 @item -Wvector-operation-performance
5686 @opindex Wvector-operation-performance
5687 @opindex Wno-vector-operation-performance
5688 Warn if vector operation is not implemented via SIMD capabilities of the
5689 architecture. Mainly useful for the performance tuning.
5690 Vector operation can be implemented @code{piecewise}, which means that the
5691 scalar operation is performed on every vector element;
5692 @code{in parallel}, which means that the vector operation is implemented
5693 using scalars of wider type, which normally is more performance efficient;
5694 and @code{as a single scalar}, which means that vector fits into a
5697 @item -Wno-virtual-move-assign
5698 @opindex Wvirtual-move-assign
5699 @opindex Wno-virtual-move-assign
5700 Suppress warnings about inheriting from a virtual base with a
5701 non-trivial C++11 move assignment operator. This is dangerous because
5702 if the virtual base is reachable along more than one path, it is
5703 moved multiple times, which can mean both objects end up in the
5704 moved-from state. If the move assignment operator is written to avoid
5705 moving from a moved-from object, this warning can be disabled.
5710 Warn if variable length array is used in the code.
5711 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5712 the variable length array.
5714 @item -Wvolatile-register-var
5715 @opindex Wvolatile-register-var
5716 @opindex Wno-volatile-register-var
5717 Warn if a register variable is declared volatile. The volatile
5718 modifier does not inhibit all optimizations that may eliminate reads
5719 and/or writes to register variables. This warning is enabled by
5722 @item -Wdisabled-optimization
5723 @opindex Wdisabled-optimization
5724 @opindex Wno-disabled-optimization
5725 Warn if a requested optimization pass is disabled. This warning does
5726 not generally indicate that there is anything wrong with your code; it
5727 merely indicates that GCC's optimizers are unable to handle the code
5728 effectively. Often, the problem is that your code is too big or too
5729 complex; GCC refuses to optimize programs when the optimization
5730 itself is likely to take inordinate amounts of time.
5732 @item -Wpointer-sign @r{(C and Objective-C only)}
5733 @opindex Wpointer-sign
5734 @opindex Wno-pointer-sign
5735 Warn for pointer argument passing or assignment with different signedness.
5736 This option is only supported for C and Objective-C@. It is implied by
5737 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5738 @option{-Wno-pointer-sign}.
5740 @item -Wstack-protector
5741 @opindex Wstack-protector
5742 @opindex Wno-stack-protector
5743 This option is only active when @option{-fstack-protector} is active. It
5744 warns about functions that are not protected against stack smashing.
5746 @item -Woverlength-strings
5747 @opindex Woverlength-strings
5748 @opindex Wno-overlength-strings
5749 Warn about string constants that are longer than the ``minimum
5750 maximum'' length specified in the C standard. Modern compilers
5751 generally allow string constants that are much longer than the
5752 standard's minimum limit, but very portable programs should avoid
5753 using longer strings.
5755 The limit applies @emph{after} string constant concatenation, and does
5756 not count the trailing NUL@. In C90, the limit was 509 characters; in
5757 C99, it was raised to 4095. C++98 does not specify a normative
5758 minimum maximum, so we do not diagnose overlength strings in C++@.
5760 This option is implied by @option{-Wpedantic}, and can be disabled with
5761 @option{-Wno-overlength-strings}.
5763 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5764 @opindex Wunsuffixed-float-constants
5766 Issue a warning for any floating constant that does not have
5767 a suffix. When used together with @option{-Wsystem-headers} it
5768 warns about such constants in system header files. This can be useful
5769 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5770 from the decimal floating-point extension to C99.
5772 @item -Wno-designated-init @r{(C and Objective-C only)}
5773 Suppress warnings when a positional initializer is used to initialize
5774 a structure that has been marked with the @code{designated_init}
5778 Issue a warning when HSAIL cannot be emitted for the compiled function or
5783 @node Debugging Options
5784 @section Options for Debugging Your Program
5785 @cindex options, debugging
5786 @cindex debugging information options
5788 To tell GCC to emit extra information for use by a debugger, in almost
5789 all cases you need only to add @option{-g} to your other options.
5791 GCC allows you to use @option{-g} with
5792 @option{-O}. The shortcuts taken by optimized code may occasionally
5793 be surprising: some variables you declared may not exist
5794 at all; flow of control may briefly move where you did not expect it;
5795 some statements may not be executed because they compute constant
5796 results or their values are already at hand; some statements may
5797 execute in different places because they have been moved out of loops.
5798 Nevertheless it is possible to debug optimized output. This makes
5799 it reasonable to use the optimizer for programs that might have bugs.
5801 If you are not using some other optimization option, consider
5802 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
5803 With no @option{-O} option at all, some compiler passes that collect
5804 information useful for debugging do not run at all, so that
5805 @option{-Og} may result in a better debugging experience.
5810 Produce debugging information in the operating system's native format
5811 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
5814 On most systems that use stabs format, @option{-g} enables use of extra
5815 debugging information that only GDB can use; this extra information
5816 makes debugging work better in GDB but probably makes other debuggers
5818 refuse to read the program. If you want to control for certain whether
5819 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5820 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5824 Produce debugging information for use by GDB@. This means to use the
5825 most expressive format available (DWARF, stabs, or the native format
5826 if neither of those are supported), including GDB extensions if at all
5830 @itemx -gdwarf-@var{version}
5832 Produce debugging information in DWARF format (if that is supported).
5833 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5834 for most targets is 4. DWARF Version 5 is only experimental.
5836 Note that with DWARF Version 2, some ports require and always
5837 use some non-conflicting DWARF 3 extensions in the unwind tables.
5839 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5840 for maximum benefit.
5842 GCC no longer supports DWARF Version 1, which is substantially
5843 different than Version 2 and later. For historical reasons, some
5844 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
5845 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
5846 in their names, but apply to all currently-supported versions of DWARF.
5850 Produce debugging information in stabs format (if that is supported),
5851 without GDB extensions. This is the format used by DBX on most BSD
5852 systems. On MIPS, Alpha and System V Release 4 systems this option
5853 produces stabs debugging output that is not understood by DBX or SDB@.
5854 On System V Release 4 systems this option requires the GNU assembler.
5858 Produce debugging information in stabs format (if that is supported),
5859 using GNU extensions understood only by the GNU debugger (GDB)@. The
5860 use of these extensions is likely to make other debuggers crash or
5861 refuse to read the program.
5865 Produce debugging information in COFF format (if that is supported).
5866 This is the format used by SDB on most System V systems prior to
5871 Produce debugging information in XCOFF format (if that is supported).
5872 This is the format used by the DBX debugger on IBM RS/6000 systems.
5876 Produce debugging information in XCOFF format (if that is supported),
5877 using GNU extensions understood only by the GNU debugger (GDB)@. The
5878 use of these extensions is likely to make other debuggers crash or
5879 refuse to read the program, and may cause assemblers other than the GNU
5880 assembler (GAS) to fail with an error.
5884 Produce debugging information in Alpha/VMS debug format (if that is
5885 supported). This is the format used by DEBUG on Alpha/VMS systems.
5888 @itemx -ggdb@var{level}
5889 @itemx -gstabs@var{level}
5890 @itemx -gcoff@var{level}
5891 @itemx -gxcoff@var{level}
5892 @itemx -gvms@var{level}
5893 Request debugging information and also use @var{level} to specify how
5894 much information. The default level is 2.
5896 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5899 Level 1 produces minimal information, enough for making backtraces in
5900 parts of the program that you don't plan to debug. This includes
5901 descriptions of functions and external variables, and line number
5902 tables, but no information about local variables.
5904 Level 3 includes extra information, such as all the macro definitions
5905 present in the program. Some debuggers support macro expansion when
5906 you use @option{-g3}.
5908 @option{-gdwarf} does not accept a concatenated debug level, to avoid
5909 confusion with @option{-gdwarf-@var{level}}.
5910 Instead use an additional @option{-g@var{level}} option to change the
5911 debug level for DWARF.
5913 @item -feliminate-unused-debug-symbols
5914 @opindex feliminate-unused-debug-symbols
5915 Produce debugging information in stabs format (if that is supported),
5916 for only symbols that are actually used.
5918 @item -femit-class-debug-always
5919 @opindex femit-class-debug-always
5920 Instead of emitting debugging information for a C++ class in only one
5921 object file, emit it in all object files using the class. This option
5922 should be used only with debuggers that are unable to handle the way GCC
5923 normally emits debugging information for classes because using this
5924 option increases the size of debugging information by as much as a
5927 @item -fno-merge-debug-strings
5928 @opindex fmerge-debug-strings
5929 @opindex fno-merge-debug-strings
5930 Direct the linker to not merge together strings in the debugging
5931 information that are identical in different object files. Merging is
5932 not supported by all assemblers or linkers. Merging decreases the size
5933 of the debug information in the output file at the cost of increasing
5934 link processing time. Merging is enabled by default.
5936 @item -fdebug-prefix-map=@var{old}=@var{new}
5937 @opindex fdebug-prefix-map
5938 When compiling files in directory @file{@var{old}}, record debugging
5939 information describing them as in @file{@var{new}} instead.
5941 @item -fvar-tracking
5942 @opindex fvar-tracking
5943 Run variable tracking pass. It computes where variables are stored at each
5944 position in code. Better debugging information is then generated
5945 (if the debugging information format supports this information).
5947 It is enabled by default when compiling with optimization (@option{-Os},
5948 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5949 the debug info format supports it.
5951 @item -fvar-tracking-assignments
5952 @opindex fvar-tracking-assignments
5953 @opindex fno-var-tracking-assignments
5954 Annotate assignments to user variables early in the compilation and
5955 attempt to carry the annotations over throughout the compilation all the
5956 way to the end, in an attempt to improve debug information while
5957 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5959 It can be enabled even if var-tracking is disabled, in which case
5960 annotations are created and maintained, but discarded at the end.
5961 By default, this flag is enabled together with @option{-fvar-tracking},
5962 except when selective scheduling is enabled.
5965 @opindex gsplit-dwarf
5966 Separate as much DWARF debugging information as possible into a
5967 separate output file with the extension @file{.dwo}. This option allows
5968 the build system to avoid linking files with debug information. To
5969 be useful, this option requires a debugger capable of reading @file{.dwo}
5974 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
5976 @item -ggnu-pubnames
5977 @opindex ggnu-pubnames
5978 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
5979 suitable for conversion into a GDB@ index. This option is only useful
5980 with a linker that can produce GDB@ index version 7.
5982 @item -fdebug-types-section
5983 @opindex fdebug-types-section
5984 @opindex fno-debug-types-section
5985 When using DWARF Version 4 or higher, type DIEs can be put into
5986 their own @code{.debug_types} section instead of making them part of the
5987 @code{.debug_info} section. It is more efficient to put them in a separate
5988 comdat sections since the linker can then remove duplicates.
5989 But not all DWARF consumers support @code{.debug_types} sections yet
5990 and on some objects @code{.debug_types} produces larger instead of smaller
5991 debugging information.
5993 @item -grecord-gcc-switches
5994 @item -gno-record-gcc-switches
5995 @opindex grecord-gcc-switches
5996 @opindex gno-record-gcc-switches
5997 This switch causes the command-line options used to invoke the
5998 compiler that may affect code generation to be appended to the
5999 DW_AT_producer attribute in DWARF debugging information. The options
6000 are concatenated with spaces separating them from each other and from
6001 the compiler version.
6002 It is enabled by default.
6003 See also @option{-frecord-gcc-switches} for another
6004 way of storing compiler options into the object file.
6006 @item -gstrict-dwarf
6007 @opindex gstrict-dwarf
6008 Disallow using extensions of later DWARF standard version than selected
6009 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6010 DWARF extensions from later standard versions is allowed.
6012 @item -gno-strict-dwarf
6013 @opindex gno-strict-dwarf
6014 Allow using extensions of later DWARF standard version than selected with
6015 @option{-gdwarf-@var{version}}.
6017 @item -gz@r{[}=@var{type}@r{]}
6019 Produce compressed debug sections in DWARF format, if that is supported.
6020 If @var{type} is not given, the default type depends on the capabilities
6021 of the assembler and linker used. @var{type} may be one of
6022 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6023 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6024 compression in traditional GNU format). If the linker doesn't support
6025 writing compressed debug sections, the option is rejected. Otherwise,
6026 if the assembler does not support them, @option{-gz} is silently ignored
6027 when producing object files.
6029 @item -feliminate-dwarf2-dups
6030 @opindex feliminate-dwarf2-dups
6031 Compress DWARF debugging information by eliminating duplicated
6032 information about each symbol. This option only makes sense when
6033 generating DWARF debugging information.
6035 @item -femit-struct-debug-baseonly
6036 @opindex femit-struct-debug-baseonly
6037 Emit debug information for struct-like types
6038 only when the base name of the compilation source file
6039 matches the base name of file in which the struct is defined.
6041 This option substantially reduces the size of debugging information,
6042 but at significant potential loss in type information to the debugger.
6043 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6044 See @option{-femit-struct-debug-detailed} for more detailed control.
6046 This option works only with DWARF debug output.
6048 @item -femit-struct-debug-reduced
6049 @opindex femit-struct-debug-reduced
6050 Emit debug information for struct-like types
6051 only when the base name of the compilation source file
6052 matches the base name of file in which the type is defined,
6053 unless the struct is a template or defined in a system header.
6055 This option significantly reduces the size of debugging information,
6056 with some potential loss in type information to the debugger.
6057 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6058 See @option{-femit-struct-debug-detailed} for more detailed control.
6060 This option works only with DWARF debug output.
6062 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6063 @opindex femit-struct-debug-detailed
6064 Specify the struct-like types
6065 for which the compiler generates debug information.
6066 The intent is to reduce duplicate struct debug information
6067 between different object files within the same program.
6069 This option is a detailed version of
6070 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6071 which serves for most needs.
6073 A specification has the syntax@*
6074 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6076 The optional first word limits the specification to
6077 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6078 A struct type is used directly when it is the type of a variable, member.
6079 Indirect uses arise through pointers to structs.
6080 That is, when use of an incomplete struct is valid, the use is indirect.
6082 @samp{struct one direct; struct two * indirect;}.
6084 The optional second word limits the specification to
6085 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6086 Generic structs are a bit complicated to explain.
6087 For C++, these are non-explicit specializations of template classes,
6088 or non-template classes within the above.
6089 Other programming languages have generics,
6090 but @option{-femit-struct-debug-detailed} does not yet implement them.
6092 The third word specifies the source files for those
6093 structs for which the compiler should emit debug information.
6094 The values @samp{none} and @samp{any} have the normal meaning.
6095 The value @samp{base} means that
6096 the base of name of the file in which the type declaration appears
6097 must match the base of the name of the main compilation file.
6098 In practice, this means that when compiling @file{foo.c}, debug information
6099 is generated for types declared in that file and @file{foo.h},
6100 but not other header files.
6101 The value @samp{sys} means those types satisfying @samp{base}
6102 or declared in system or compiler headers.
6104 You may need to experiment to determine the best settings for your application.
6106 The default is @option{-femit-struct-debug-detailed=all}.
6108 This option works only with DWARF debug output.
6110 @item -fno-dwarf2-cfi-asm
6111 @opindex fdwarf2-cfi-asm
6112 @opindex fno-dwarf2-cfi-asm
6113 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6114 instead of using GAS @code{.cfi_*} directives.
6116 @item -fno-eliminate-unused-debug-types
6117 @opindex feliminate-unused-debug-types
6118 @opindex fno-eliminate-unused-debug-types
6119 Normally, when producing DWARF output, GCC avoids producing debug symbol
6120 output for types that are nowhere used in the source file being compiled.
6121 Sometimes it is useful to have GCC emit debugging
6122 information for all types declared in a compilation
6123 unit, regardless of whether or not they are actually used
6124 in that compilation unit, for example
6125 if, in the debugger, you want to cast a value to a type that is
6126 not actually used in your program (but is declared). More often,
6127 however, this results in a significant amount of wasted space.
6130 @node Optimize Options
6131 @section Options That Control Optimization
6132 @cindex optimize options
6133 @cindex options, optimization
6135 These options control various sorts of optimizations.
6137 Without any optimization option, the compiler's goal is to reduce the
6138 cost of compilation and to make debugging produce the expected
6139 results. Statements are independent: if you stop the program with a
6140 breakpoint between statements, you can then assign a new value to any
6141 variable or change the program counter to any other statement in the
6142 function and get exactly the results you expect from the source
6145 Turning on optimization flags makes the compiler attempt to improve
6146 the performance and/or code size at the expense of compilation time
6147 and possibly the ability to debug the program.
6149 The compiler performs optimization based on the knowledge it has of the
6150 program. Compiling multiple files at once to a single output file mode allows
6151 the compiler to use information gained from all of the files when compiling
6154 Not all optimizations are controlled directly by a flag. Only
6155 optimizations that have a flag are listed in this section.
6157 Most optimizations are only enabled if an @option{-O} level is set on
6158 the command line. Otherwise they are disabled, even if individual
6159 optimization flags are specified.
6161 Depending on the target and how GCC was configured, a slightly different
6162 set of optimizations may be enabled at each @option{-O} level than
6163 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6164 to find out the exact set of optimizations that are enabled at each level.
6165 @xref{Overall Options}, for examples.
6172 Optimize. Optimizing compilation takes somewhat more time, and a lot
6173 more memory for a large function.
6175 With @option{-O}, the compiler tries to reduce code size and execution
6176 time, without performing any optimizations that take a great deal of
6179 @option{-O} turns on the following optimization flags:
6182 -fbranch-count-reg @gol
6183 -fcombine-stack-adjustments @gol
6185 -fcprop-registers @gol
6188 -fdelayed-branch @gol
6190 -fforward-propagate @gol
6191 -fguess-branch-probability @gol
6192 -fif-conversion2 @gol
6193 -fif-conversion @gol
6194 -finline-functions-called-once @gol
6195 -fipa-pure-const @gol
6197 -fipa-reference @gol
6198 -fmerge-constants @gol
6199 -fmove-loop-invariants @gol
6200 -freorder-blocks @gol
6202 -fsplit-wide-types @gol
6208 -ftree-coalesce-vars @gol
6209 -ftree-copy-prop @gol
6211 -ftree-dominator-opts @gol
6213 -ftree-forwprop @gol
6223 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6224 where doing so does not interfere with debugging.
6228 Optimize even more. GCC performs nearly all supported optimizations
6229 that do not involve a space-speed tradeoff.
6230 As compared to @option{-O}, this option increases both compilation time
6231 and the performance of the generated code.
6233 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6234 also turns on the following optimization flags:
6235 @gccoptlist{-fthread-jumps @gol
6236 -falign-functions -falign-jumps @gol
6237 -falign-loops -falign-labels @gol
6240 -fcse-follow-jumps -fcse-skip-blocks @gol
6241 -fdelete-null-pointer-checks @gol
6242 -fdevirtualize -fdevirtualize-speculatively @gol
6243 -fexpensive-optimizations @gol
6244 -fgcse -fgcse-lm @gol
6245 -fhoist-adjacent-loads @gol
6246 -finline-small-functions @gol
6247 -findirect-inlining @gol
6249 -fipa-cp-alignment @gol
6252 -fisolate-erroneous-paths-dereference @gol
6254 -foptimize-sibling-calls @gol
6255 -foptimize-strlen @gol
6256 -fpartial-inlining @gol
6258 -freorder-blocks-algorithm=stc @gol
6259 -freorder-blocks-and-partition -freorder-functions @gol
6260 -frerun-cse-after-loop @gol
6261 -fsched-interblock -fsched-spec @gol
6262 -fschedule-insns -fschedule-insns2 @gol
6263 -fstrict-aliasing -fstrict-overflow @gol
6264 -ftree-builtin-call-dce @gol
6265 -ftree-switch-conversion -ftree-tail-merge @gol
6270 Please note the warning under @option{-fgcse} about
6271 invoking @option{-O2} on programs that use computed gotos.
6275 Optimize yet more. @option{-O3} turns on all optimizations specified
6276 by @option{-O2} and also turns on the @option{-finline-functions},
6277 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6278 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6279 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
6280 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6281 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
6285 Reduce compilation time and make debugging produce the expected
6286 results. This is the default.
6290 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6291 do not typically increase code size. It also performs further
6292 optimizations designed to reduce code size.
6294 @option{-Os} disables the following optimization flags:
6295 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6296 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
6297 -freorder-blocks-and-partition -fprefetch-loop-arrays}
6301 Disregard strict standards compliance. @option{-Ofast} enables all
6302 @option{-O3} optimizations. It also enables optimizations that are not
6303 valid for all standard-compliant programs.
6304 It turns on @option{-ffast-math} and the Fortran-specific
6305 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6309 Optimize debugging experience. @option{-Og} enables optimizations
6310 that do not interfere with debugging. It should be the optimization
6311 level of choice for the standard edit-compile-debug cycle, offering
6312 a reasonable level of optimization while maintaining fast compilation
6313 and a good debugging experience.
6316 If you use multiple @option{-O} options, with or without level numbers,
6317 the last such option is the one that is effective.
6319 Options of the form @option{-f@var{flag}} specify machine-independent
6320 flags. Most flags have both positive and negative forms; the negative
6321 form of @option{-ffoo} is @option{-fno-foo}. In the table
6322 below, only one of the forms is listed---the one you typically
6323 use. You can figure out the other form by either removing @samp{no-}
6326 The following options control specific optimizations. They are either
6327 activated by @option{-O} options or are related to ones that are. You
6328 can use the following flags in the rare cases when ``fine-tuning'' of
6329 optimizations to be performed is desired.
6332 @item -fno-defer-pop
6333 @opindex fno-defer-pop
6334 Always pop the arguments to each function call as soon as that function
6335 returns. For machines that must pop arguments after a function call,
6336 the compiler normally lets arguments accumulate on the stack for several
6337 function calls and pops them all at once.
6339 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6341 @item -fforward-propagate
6342 @opindex fforward-propagate
6343 Perform a forward propagation pass on RTL@. The pass tries to combine two
6344 instructions and checks if the result can be simplified. If loop unrolling
6345 is active, two passes are performed and the second is scheduled after
6348 This option is enabled by default at optimization levels @option{-O},
6349 @option{-O2}, @option{-O3}, @option{-Os}.
6351 @item -ffp-contract=@var{style}
6352 @opindex ffp-contract
6353 @option{-ffp-contract=off} disables floating-point expression contraction.
6354 @option{-ffp-contract=fast} enables floating-point expression contraction
6355 such as forming of fused multiply-add operations if the target has
6356 native support for them.
6357 @option{-ffp-contract=on} enables floating-point expression contraction
6358 if allowed by the language standard. This is currently not implemented
6359 and treated equal to @option{-ffp-contract=off}.
6361 The default is @option{-ffp-contract=fast}.
6363 @item -fomit-frame-pointer
6364 @opindex fomit-frame-pointer
6365 Don't keep the frame pointer in a register for functions that
6366 don't need one. This avoids the instructions to save, set up and
6367 restore frame pointers; it also makes an extra register available
6368 in many functions. @strong{It also makes debugging impossible on
6371 On some machines, such as the VAX, this flag has no effect, because
6372 the standard calling sequence automatically handles the frame pointer
6373 and nothing is saved by pretending it doesn't exist. The
6374 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6375 whether a target machine supports this flag. @xref{Registers,,Register
6376 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6378 The default setting (when not optimizing for
6379 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
6380 @option{-fomit-frame-pointer}. You can configure GCC with the
6381 @option{--enable-frame-pointer} configure option to change the default.
6383 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6385 @item -foptimize-sibling-calls
6386 @opindex foptimize-sibling-calls
6387 Optimize sibling and tail recursive calls.
6389 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6391 @item -foptimize-strlen
6392 @opindex foptimize-strlen
6393 Optimize various standard C string functions (e.g. @code{strlen},
6394 @code{strchr} or @code{strcpy}) and
6395 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
6397 Enabled at levels @option{-O2}, @option{-O3}.
6401 Do not expand any functions inline apart from those marked with
6402 the @code{always_inline} attribute. This is the default when not
6405 Single functions can be exempted from inlining by marking them
6406 with the @code{noinline} attribute.
6408 @item -finline-small-functions
6409 @opindex finline-small-functions
6410 Integrate functions into their callers when their body is smaller than expected
6411 function call code (so overall size of program gets smaller). The compiler
6412 heuristically decides which functions are simple enough to be worth integrating
6413 in this way. This inlining applies to all functions, even those not declared
6416 Enabled at level @option{-O2}.
6418 @item -findirect-inlining
6419 @opindex findirect-inlining
6420 Inline also indirect calls that are discovered to be known at compile
6421 time thanks to previous inlining. This option has any effect only
6422 when inlining itself is turned on by the @option{-finline-functions}
6423 or @option{-finline-small-functions} options.
6425 Enabled at level @option{-O2}.
6427 @item -finline-functions
6428 @opindex finline-functions
6429 Consider all functions for inlining, even if they are not declared inline.
6430 The compiler heuristically decides which functions are worth integrating
6433 If all calls to a given function are integrated, and the function is
6434 declared @code{static}, then the function is normally not output as
6435 assembler code in its own right.
6437 Enabled at level @option{-O3}.
6439 @item -finline-functions-called-once
6440 @opindex finline-functions-called-once
6441 Consider all @code{static} functions called once for inlining into their
6442 caller even if they are not marked @code{inline}. If a call to a given
6443 function is integrated, then the function is not output as assembler code
6446 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6448 @item -fearly-inlining
6449 @opindex fearly-inlining
6450 Inline functions marked by @code{always_inline} and functions whose body seems
6451 smaller than the function call overhead early before doing
6452 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6453 makes profiling significantly cheaper and usually inlining faster on programs
6454 having large chains of nested wrapper functions.
6460 Perform interprocedural scalar replacement of aggregates, removal of
6461 unused parameters and replacement of parameters passed by reference
6462 by parameters passed by value.
6464 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6466 @item -finline-limit=@var{n}
6467 @opindex finline-limit
6468 By default, GCC limits the size of functions that can be inlined. This flag
6469 allows coarse control of this limit. @var{n} is the size of functions that
6470 can be inlined in number of pseudo instructions.
6472 Inlining is actually controlled by a number of parameters, which may be
6473 specified individually by using @option{--param @var{name}=@var{value}}.
6474 The @option{-finline-limit=@var{n}} option sets some of these parameters
6478 @item max-inline-insns-single
6479 is set to @var{n}/2.
6480 @item max-inline-insns-auto
6481 is set to @var{n}/2.
6484 See below for a documentation of the individual
6485 parameters controlling inlining and for the defaults of these parameters.
6487 @emph{Note:} there may be no value to @option{-finline-limit} that results
6488 in default behavior.
6490 @emph{Note:} pseudo instruction represents, in this particular context, an
6491 abstract measurement of function's size. In no way does it represent a count
6492 of assembly instructions and as such its exact meaning might change from one
6493 release to an another.
6495 @item -fno-keep-inline-dllexport
6496 @opindex fno-keep-inline-dllexport
6497 This is a more fine-grained version of @option{-fkeep-inline-functions},
6498 which applies only to functions that are declared using the @code{dllexport}
6499 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6502 @item -fkeep-inline-functions
6503 @opindex fkeep-inline-functions
6504 In C, emit @code{static} functions that are declared @code{inline}
6505 into the object file, even if the function has been inlined into all
6506 of its callers. This switch does not affect functions using the
6507 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6508 inline functions into the object file.
6510 @item -fkeep-static-functions
6511 @opindex fkeep-static-functions
6512 Emit @code{static} functions into the object file, even if the function
6515 @item -fkeep-static-consts
6516 @opindex fkeep-static-consts
6517 Emit variables declared @code{static const} when optimization isn't turned
6518 on, even if the variables aren't referenced.
6520 GCC enables this option by default. If you want to force the compiler to
6521 check if a variable is referenced, regardless of whether or not
6522 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6524 @item -fmerge-constants
6525 @opindex fmerge-constants
6526 Attempt to merge identical constants (string constants and floating-point
6527 constants) across compilation units.
6529 This option is the default for optimized compilation if the assembler and
6530 linker support it. Use @option{-fno-merge-constants} to inhibit this
6533 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6535 @item -fmerge-all-constants
6536 @opindex fmerge-all-constants
6537 Attempt to merge identical constants and identical variables.
6539 This option implies @option{-fmerge-constants}. In addition to
6540 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6541 arrays or initialized constant variables with integral or floating-point
6542 types. Languages like C or C++ require each variable, including multiple
6543 instances of the same variable in recursive calls, to have distinct locations,
6544 so using this option results in non-conforming
6547 @item -fmodulo-sched
6548 @opindex fmodulo-sched
6549 Perform swing modulo scheduling immediately before the first scheduling
6550 pass. This pass looks at innermost loops and reorders their
6551 instructions by overlapping different iterations.
6553 @item -fmodulo-sched-allow-regmoves
6554 @opindex fmodulo-sched-allow-regmoves
6555 Perform more aggressive SMS-based modulo scheduling with register moves
6556 allowed. By setting this flag certain anti-dependences edges are
6557 deleted, which triggers the generation of reg-moves based on the
6558 life-range analysis. This option is effective only with
6559 @option{-fmodulo-sched} enabled.
6561 @item -fno-branch-count-reg
6562 @opindex fno-branch-count-reg
6563 Avoid running a pass scanning for opportunities to use ``decrement and
6564 branch'' instructions on a count register instead of generating sequences
6565 of instructions that decrement a register, compare it against zero, and
6566 then branch based upon the result. This option is only meaningful on
6567 architectures that support such instructions, which include x86, PowerPC,
6568 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
6569 doesn't remove the decrement and branch instructions from the generated
6570 instruction stream introduced by other optimization passes.
6572 Enabled by default at @option{-O1} and higher.
6574 The default is @option{-fbranch-count-reg}.
6576 @item -fno-function-cse
6577 @opindex fno-function-cse
6578 Do not put function addresses in registers; make each instruction that
6579 calls a constant function contain the function's address explicitly.
6581 This option results in less efficient code, but some strange hacks
6582 that alter the assembler output may be confused by the optimizations
6583 performed when this option is not used.
6585 The default is @option{-ffunction-cse}
6587 @item -fno-zero-initialized-in-bss
6588 @opindex fno-zero-initialized-in-bss
6589 If the target supports a BSS section, GCC by default puts variables that
6590 are initialized to zero into BSS@. This can save space in the resulting
6593 This option turns off this behavior because some programs explicitly
6594 rely on variables going to the data section---e.g., so that the
6595 resulting executable can find the beginning of that section and/or make
6596 assumptions based on that.
6598 The default is @option{-fzero-initialized-in-bss}.
6600 @item -fthread-jumps
6601 @opindex fthread-jumps
6602 Perform optimizations that check to see if a jump branches to a
6603 location where another comparison subsumed by the first is found. If
6604 so, the first branch is redirected to either the destination of the
6605 second branch or a point immediately following it, depending on whether
6606 the condition is known to be true or false.
6608 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6610 @item -fsplit-wide-types
6611 @opindex fsplit-wide-types
6612 When using a type that occupies multiple registers, such as @code{long
6613 long} on a 32-bit system, split the registers apart and allocate them
6614 independently. This normally generates better code for those types,
6615 but may make debugging more difficult.
6617 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6620 @item -fcse-follow-jumps
6621 @opindex fcse-follow-jumps
6622 In common subexpression elimination (CSE), scan through jump instructions
6623 when the target of the jump is not reached by any other path. For
6624 example, when CSE encounters an @code{if} statement with an
6625 @code{else} clause, CSE follows the jump when the condition
6628 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6630 @item -fcse-skip-blocks
6631 @opindex fcse-skip-blocks
6632 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6633 follow jumps that conditionally skip over blocks. When CSE
6634 encounters a simple @code{if} statement with no else clause,
6635 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6636 body of the @code{if}.
6638 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6640 @item -frerun-cse-after-loop
6641 @opindex frerun-cse-after-loop
6642 Re-run common subexpression elimination after loop optimizations are
6645 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6649 Perform a global common subexpression elimination pass.
6650 This pass also performs global constant and copy propagation.
6652 @emph{Note:} When compiling a program using computed gotos, a GCC
6653 extension, you may get better run-time performance if you disable
6654 the global common subexpression elimination pass by adding
6655 @option{-fno-gcse} to the command line.
6657 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6661 When @option{-fgcse-lm} is enabled, global common subexpression elimination
6662 attempts to move loads that are only killed by stores into themselves. This
6663 allows a loop containing a load/store sequence to be changed to a load outside
6664 the loop, and a copy/store within the loop.
6666 Enabled by default when @option{-fgcse} is enabled.
6670 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6671 global common subexpression elimination. This pass attempts to move
6672 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6673 loops containing a load/store sequence can be changed to a load before
6674 the loop and a store after the loop.
6676 Not enabled at any optimization level.
6680 When @option{-fgcse-las} is enabled, the global common subexpression
6681 elimination pass eliminates redundant loads that come after stores to the
6682 same memory location (both partial and full redundancies).
6684 Not enabled at any optimization level.
6686 @item -fgcse-after-reload
6687 @opindex fgcse-after-reload
6688 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6689 pass is performed after reload. The purpose of this pass is to clean up
6692 @item -faggressive-loop-optimizations
6693 @opindex faggressive-loop-optimizations
6694 This option tells the loop optimizer to use language constraints to
6695 derive bounds for the number of iterations of a loop. This assumes that
6696 loop code does not invoke undefined behavior by for example causing signed
6697 integer overflows or out-of-bound array accesses. The bounds for the
6698 number of iterations of a loop are used to guide loop unrolling and peeling
6699 and loop exit test optimizations.
6700 This option is enabled by default.
6702 @item -funsafe-loop-optimizations
6703 @opindex funsafe-loop-optimizations
6704 This option tells the loop optimizer to assume that loop indices do not
6705 overflow, and that loops with nontrivial exit condition are not
6706 infinite. This enables a wider range of loop optimizations even if
6707 the loop optimizer itself cannot prove that these assumptions are valid.
6708 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
6709 if it finds this kind of loop.
6711 @item -funconstrained-commons
6712 @opindex funconstrained-commons
6713 This option tells the compiler that variables declared in common blocks
6714 (e.g. Fortran) may later be overridden with longer trailing arrays. This
6715 prevents certain optimizations that depend on knowing the array bounds.
6717 @item -fcrossjumping
6718 @opindex fcrossjumping
6719 Perform cross-jumping transformation.
6720 This transformation unifies equivalent code and saves code size. The
6721 resulting code may or may not perform better than without cross-jumping.
6723 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6725 @item -fauto-inc-dec
6726 @opindex fauto-inc-dec
6727 Combine increments or decrements of addresses with memory accesses.
6728 This pass is always skipped on architectures that do not have
6729 instructions to support this. Enabled by default at @option{-O} and
6730 higher on architectures that support this.
6734 Perform dead code elimination (DCE) on RTL@.
6735 Enabled by default at @option{-O} and higher.
6739 Perform dead store elimination (DSE) on RTL@.
6740 Enabled by default at @option{-O} and higher.
6742 @item -fif-conversion
6743 @opindex fif-conversion
6744 Attempt to transform conditional jumps into branch-less equivalents. This
6745 includes use of conditional moves, min, max, set flags and abs instructions, and
6746 some tricks doable by standard arithmetics. The use of conditional execution
6747 on chips where it is available is controlled by @option{-fif-conversion2}.
6749 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6751 @item -fif-conversion2
6752 @opindex fif-conversion2
6753 Use conditional execution (where available) to transform conditional jumps into
6754 branch-less equivalents.
6756 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6758 @item -fdeclone-ctor-dtor
6759 @opindex fdeclone-ctor-dtor
6760 The C++ ABI requires multiple entry points for constructors and
6761 destructors: one for a base subobject, one for a complete object, and
6762 one for a virtual destructor that calls operator delete afterwards.
6763 For a hierarchy with virtual bases, the base and complete variants are
6764 clones, which means two copies of the function. With this option, the
6765 base and complete variants are changed to be thunks that call a common
6768 Enabled by @option{-Os}.
6770 @item -fdelete-null-pointer-checks
6771 @opindex fdelete-null-pointer-checks
6772 Assume that programs cannot safely dereference null pointers, and that
6773 no code or data element resides at address zero.
6774 This option enables simple constant
6775 folding optimizations at all optimization levels. In addition, other
6776 optimization passes in GCC use this flag to control global dataflow
6777 analyses that eliminate useless checks for null pointers; these assume
6778 that a memory access to address zero always results in a trap, so
6779 that if a pointer is checked after it has already been dereferenced,
6782 Note however that in some environments this assumption is not true.
6783 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6784 for programs that depend on that behavior.
6786 This option is enabled by default on most targets. On Nios II ELF, it
6787 defaults to off. On AVR and CR16, this option is completely disabled.
6789 Passes that use the dataflow information
6790 are enabled independently at different optimization levels.
6792 @item -fdevirtualize
6793 @opindex fdevirtualize
6794 Attempt to convert calls to virtual functions to direct calls. This
6795 is done both within a procedure and interprocedurally as part of
6796 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
6797 propagation (@option{-fipa-cp}).
6798 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6800 @item -fdevirtualize-speculatively
6801 @opindex fdevirtualize-speculatively
6802 Attempt to convert calls to virtual functions to speculative direct calls.
6803 Based on the analysis of the type inheritance graph, determine for a given call
6804 the set of likely targets. If the set is small, preferably of size 1, change
6805 the call into a conditional deciding between direct and indirect calls. The
6806 speculative calls enable more optimizations, such as inlining. When they seem
6807 useless after further optimization, they are converted back into original form.
6809 @item -fdevirtualize-at-ltrans
6810 @opindex fdevirtualize-at-ltrans
6811 Stream extra information needed for aggressive devirtualization when running
6812 the link-time optimizer in local transformation mode.
6813 This option enables more devirtualization but
6814 significantly increases the size of streamed data. For this reason it is
6815 disabled by default.
6817 @item -fexpensive-optimizations
6818 @opindex fexpensive-optimizations
6819 Perform a number of minor optimizations that are relatively expensive.
6821 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6825 Attempt to remove redundant extension instructions. This is especially
6826 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
6827 registers after writing to their lower 32-bit half.
6829 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
6830 @option{-O3}, @option{-Os}.
6832 @item -fno-lifetime-dse
6833 @opindex fno-lifetime-dse
6834 In C++ the value of an object is only affected by changes within its
6835 lifetime: when the constructor begins, the object has an indeterminate
6836 value, and any changes during the lifetime of the object are dead when
6837 the object is destroyed. Normally dead store elimination will take
6838 advantage of this; if your code relies on the value of the object
6839 storage persisting beyond the lifetime of the object, you can use this
6840 flag to disable this optimization. To preserve stores before the
6841 constructor starts (e.g. because your operator new clears the object
6842 storage) but still treat the object as dead after the destructor you,
6843 can use @option{-flifetime-dse=1}. The default behavior can be
6844 explicitly selected with @option{-flifetime-dse=2}.
6845 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
6847 @item -flive-range-shrinkage
6848 @opindex flive-range-shrinkage
6849 Attempt to decrease register pressure through register live range
6850 shrinkage. This is helpful for fast processors with small or moderate
6853 @item -fira-algorithm=@var{algorithm}
6854 @opindex fira-algorithm
6855 Use the specified coloring algorithm for the integrated register
6856 allocator. The @var{algorithm} argument can be @samp{priority}, which
6857 specifies Chow's priority coloring, or @samp{CB}, which specifies
6858 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
6859 for all architectures, but for those targets that do support it, it is
6860 the default because it generates better code.
6862 @item -fira-region=@var{region}
6863 @opindex fira-region
6864 Use specified regions for the integrated register allocator. The
6865 @var{region} argument should be one of the following:
6870 Use all loops as register allocation regions.
6871 This can give the best results for machines with a small and/or
6872 irregular register set.
6875 Use all loops except for loops with small register pressure
6876 as the regions. This value usually gives
6877 the best results in most cases and for most architectures,
6878 and is enabled by default when compiling with optimization for speed
6879 (@option{-O}, @option{-O2}, @dots{}).
6882 Use all functions as a single region.
6883 This typically results in the smallest code size, and is enabled by default for
6884 @option{-Os} or @option{-O0}.
6888 @item -fira-hoist-pressure
6889 @opindex fira-hoist-pressure
6890 Use IRA to evaluate register pressure in the code hoisting pass for
6891 decisions to hoist expressions. This option usually results in smaller
6892 code, but it can slow the compiler down.
6894 This option is enabled at level @option{-Os} for all targets.
6896 @item -fira-loop-pressure
6897 @opindex fira-loop-pressure
6898 Use IRA to evaluate register pressure in loops for decisions to move
6899 loop invariants. This option usually results in generation
6900 of faster and smaller code on machines with large register files (>= 32
6901 registers), but it can slow the compiler down.
6903 This option is enabled at level @option{-O3} for some targets.
6905 @item -fno-ira-share-save-slots
6906 @opindex fno-ira-share-save-slots
6907 Disable sharing of stack slots used for saving call-used hard
6908 registers living through a call. Each hard register gets a
6909 separate stack slot, and as a result function stack frames are
6912 @item -fno-ira-share-spill-slots
6913 @opindex fno-ira-share-spill-slots
6914 Disable sharing of stack slots allocated for pseudo-registers. Each
6915 pseudo-register that does not get a hard register gets a separate
6916 stack slot, and as a result function stack frames are larger.
6920 Enable CFG-sensitive rematerialization in LRA. Instead of loading
6921 values of spilled pseudos, LRA tries to rematerialize (recalculate)
6922 values if it is profitable.
6924 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6926 @item -fdelayed-branch
6927 @opindex fdelayed-branch
6928 If supported for the target machine, attempt to reorder instructions
6929 to exploit instruction slots available after delayed branch
6932 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6934 @item -fschedule-insns
6935 @opindex fschedule-insns
6936 If supported for the target machine, attempt to reorder instructions to
6937 eliminate execution stalls due to required data being unavailable. This
6938 helps machines that have slow floating point or memory load instructions
6939 by allowing other instructions to be issued until the result of the load
6940 or floating-point instruction is required.
6942 Enabled at levels @option{-O2}, @option{-O3}.
6944 @item -fschedule-insns2
6945 @opindex fschedule-insns2
6946 Similar to @option{-fschedule-insns}, but requests an additional pass of
6947 instruction scheduling after register allocation has been done. This is
6948 especially useful on machines with a relatively small number of
6949 registers and where memory load instructions take more than one cycle.
6951 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6953 @item -fno-sched-interblock
6954 @opindex fno-sched-interblock
6955 Don't schedule instructions across basic blocks. This is normally
6956 enabled by default when scheduling before register allocation, i.e.@:
6957 with @option{-fschedule-insns} or at @option{-O2} or higher.
6959 @item -fno-sched-spec
6960 @opindex fno-sched-spec
6961 Don't allow speculative motion of non-load instructions. This is normally
6962 enabled by default when scheduling before register allocation, i.e.@:
6963 with @option{-fschedule-insns} or at @option{-O2} or higher.
6965 @item -fsched-pressure
6966 @opindex fsched-pressure
6967 Enable register pressure sensitive insn scheduling before register
6968 allocation. This only makes sense when scheduling before register
6969 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6970 @option{-O2} or higher. Usage of this option can improve the
6971 generated code and decrease its size by preventing register pressure
6972 increase above the number of available hard registers and subsequent
6973 spills in register allocation.
6975 @item -fsched-spec-load
6976 @opindex fsched-spec-load
6977 Allow speculative motion of some load instructions. This only makes
6978 sense when scheduling before register allocation, i.e.@: with
6979 @option{-fschedule-insns} or at @option{-O2} or higher.
6981 @item -fsched-spec-load-dangerous
6982 @opindex fsched-spec-load-dangerous
6983 Allow speculative motion of more load instructions. This only makes
6984 sense when scheduling before register allocation, i.e.@: with
6985 @option{-fschedule-insns} or at @option{-O2} or higher.
6987 @item -fsched-stalled-insns
6988 @itemx -fsched-stalled-insns=@var{n}
6989 @opindex fsched-stalled-insns
6990 Define how many insns (if any) can be moved prematurely from the queue
6991 of stalled insns into the ready list during the second scheduling pass.
6992 @option{-fno-sched-stalled-insns} means that no insns are moved
6993 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6994 on how many queued insns can be moved prematurely.
6995 @option{-fsched-stalled-insns} without a value is equivalent to
6996 @option{-fsched-stalled-insns=1}.
6998 @item -fsched-stalled-insns-dep
6999 @itemx -fsched-stalled-insns-dep=@var{n}
7000 @opindex fsched-stalled-insns-dep
7001 Define how many insn groups (cycles) are examined for a dependency
7002 on a stalled insn that is a candidate for premature removal from the queue
7003 of stalled insns. This has an effect only during the second scheduling pass,
7004 and only if @option{-fsched-stalled-insns} is used.
7005 @option{-fno-sched-stalled-insns-dep} is equivalent to
7006 @option{-fsched-stalled-insns-dep=0}.
7007 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7008 @option{-fsched-stalled-insns-dep=1}.
7010 @item -fsched2-use-superblocks
7011 @opindex fsched2-use-superblocks
7012 When scheduling after register allocation, use superblock scheduling.
7013 This allows motion across basic block boundaries,
7014 resulting in faster schedules. This option is experimental, as not all machine
7015 descriptions used by GCC model the CPU closely enough to avoid unreliable
7016 results from the algorithm.
7018 This only makes sense when scheduling after register allocation, i.e.@: with
7019 @option{-fschedule-insns2} or at @option{-O2} or higher.
7021 @item -fsched-group-heuristic
7022 @opindex fsched-group-heuristic
7023 Enable the group heuristic in the scheduler. This heuristic favors
7024 the instruction that belongs to a schedule group. This is enabled
7025 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7026 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7028 @item -fsched-critical-path-heuristic
7029 @opindex fsched-critical-path-heuristic
7030 Enable the critical-path heuristic in the scheduler. This heuristic favors
7031 instructions on the critical path. This is enabled by default when
7032 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7033 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7035 @item -fsched-spec-insn-heuristic
7036 @opindex fsched-spec-insn-heuristic
7037 Enable the speculative instruction heuristic in the scheduler. This
7038 heuristic favors speculative instructions with greater dependency weakness.
7039 This is enabled by default when scheduling is enabled, i.e.@:
7040 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7041 or at @option{-O2} or higher.
7043 @item -fsched-rank-heuristic
7044 @opindex fsched-rank-heuristic
7045 Enable the rank heuristic in the scheduler. This heuristic favors
7046 the instruction belonging to a basic block with greater size or frequency.
7047 This is enabled by default when scheduling is enabled, i.e.@:
7048 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7049 at @option{-O2} or higher.
7051 @item -fsched-last-insn-heuristic
7052 @opindex fsched-last-insn-heuristic
7053 Enable the last-instruction heuristic in the scheduler. This heuristic
7054 favors the instruction that is less dependent on the last instruction
7055 scheduled. This is enabled by default when scheduling is enabled,
7056 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7057 at @option{-O2} or higher.
7059 @item -fsched-dep-count-heuristic
7060 @opindex fsched-dep-count-heuristic
7061 Enable the dependent-count heuristic in the scheduler. This heuristic
7062 favors the instruction that has more instructions depending on it.
7063 This is enabled by default when scheduling is enabled, i.e.@:
7064 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7065 at @option{-O2} or higher.
7067 @item -freschedule-modulo-scheduled-loops
7068 @opindex freschedule-modulo-scheduled-loops
7069 Modulo scheduling is performed before traditional scheduling. If a loop
7070 is modulo scheduled, later scheduling passes may change its schedule.
7071 Use this option to control that behavior.
7073 @item -fselective-scheduling
7074 @opindex fselective-scheduling
7075 Schedule instructions using selective scheduling algorithm. Selective
7076 scheduling runs instead of the first scheduler pass.
7078 @item -fselective-scheduling2
7079 @opindex fselective-scheduling2
7080 Schedule instructions using selective scheduling algorithm. Selective
7081 scheduling runs instead of the second scheduler pass.
7083 @item -fsel-sched-pipelining
7084 @opindex fsel-sched-pipelining
7085 Enable software pipelining of innermost loops during selective scheduling.
7086 This option has no effect unless one of @option{-fselective-scheduling} or
7087 @option{-fselective-scheduling2} is turned on.
7089 @item -fsel-sched-pipelining-outer-loops
7090 @opindex fsel-sched-pipelining-outer-loops
7091 When pipelining loops during selective scheduling, also pipeline outer loops.
7092 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7094 @item -fsemantic-interposition
7095 @opindex fsemantic-interposition
7096 Some object formats, like ELF, allow interposing of symbols by the
7098 This means that for symbols exported from the DSO, the compiler cannot perform
7099 interprocedural propagation, inlining and other optimizations in anticipation
7100 that the function or variable in question may change. While this feature is
7101 useful, for example, to rewrite memory allocation functions by a debugging
7102 implementation, it is expensive in the terms of code quality.
7103 With @option{-fno-semantic-interposition} the compiler assumes that
7104 if interposition happens for functions the overwriting function will have
7105 precisely the same semantics (and side effects).
7106 Similarly if interposition happens
7107 for variables, the constructor of the variable will be the same. The flag
7108 has no effect for functions explicitly declared inline
7109 (where it is never allowed for interposition to change semantics)
7110 and for symbols explicitly declared weak.
7113 @opindex fshrink-wrap
7114 Emit function prologues only before parts of the function that need it,
7115 rather than at the top of the function. This flag is enabled by default at
7116 @option{-O} and higher.
7118 @item -fcaller-saves
7119 @opindex fcaller-saves
7120 Enable allocation of values to registers that are clobbered by
7121 function calls, by emitting extra instructions to save and restore the
7122 registers around such calls. Such allocation is done only when it
7123 seems to result in better code.
7125 This option is always enabled by default on certain machines, usually
7126 those which have no call-preserved registers to use instead.
7128 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7130 @item -fcombine-stack-adjustments
7131 @opindex fcombine-stack-adjustments
7132 Tracks stack adjustments (pushes and pops) and stack memory references
7133 and then tries to find ways to combine them.
7135 Enabled by default at @option{-O1} and higher.
7139 Use caller save registers for allocation if those registers are not used by
7140 any called function. In that case it is not necessary to save and restore
7141 them around calls. This is only possible if called functions are part of
7142 same compilation unit as current function and they are compiled before it.
7144 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7146 @item -fconserve-stack
7147 @opindex fconserve-stack
7148 Attempt to minimize stack usage. The compiler attempts to use less
7149 stack space, even if that makes the program slower. This option
7150 implies setting the @option{large-stack-frame} parameter to 100
7151 and the @option{large-stack-frame-growth} parameter to 400.
7153 @item -ftree-reassoc
7154 @opindex ftree-reassoc
7155 Perform reassociation on trees. This flag is enabled by default
7156 at @option{-O} and higher.
7160 Perform partial redundancy elimination (PRE) on trees. This flag is
7161 enabled by default at @option{-O2} and @option{-O3}.
7163 @item -ftree-partial-pre
7164 @opindex ftree-partial-pre
7165 Make partial redundancy elimination (PRE) more aggressive. This flag is
7166 enabled by default at @option{-O3}.
7168 @item -ftree-forwprop
7169 @opindex ftree-forwprop
7170 Perform forward propagation on trees. This flag is enabled by default
7171 at @option{-O} and higher.
7175 Perform full redundancy elimination (FRE) on trees. The difference
7176 between FRE and PRE is that FRE only considers expressions
7177 that are computed on all paths leading to the redundant computation.
7178 This analysis is faster than PRE, though it exposes fewer redundancies.
7179 This flag is enabled by default at @option{-O} and higher.
7181 @item -ftree-phiprop
7182 @opindex ftree-phiprop
7183 Perform hoisting of loads from conditional pointers on trees. This
7184 pass is enabled by default at @option{-O} and higher.
7186 @item -fhoist-adjacent-loads
7187 @opindex fhoist-adjacent-loads
7188 Speculatively hoist loads from both branches of an if-then-else if the
7189 loads are from adjacent locations in the same structure and the target
7190 architecture has a conditional move instruction. This flag is enabled
7191 by default at @option{-O2} and higher.
7193 @item -ftree-copy-prop
7194 @opindex ftree-copy-prop
7195 Perform copy propagation on trees. This pass eliminates unnecessary
7196 copy operations. This flag is enabled by default at @option{-O} and
7199 @item -fipa-pure-const
7200 @opindex fipa-pure-const
7201 Discover which functions are pure or constant.
7202 Enabled by default at @option{-O} and higher.
7204 @item -fipa-reference
7205 @opindex fipa-reference
7206 Discover which static variables do not escape the
7208 Enabled by default at @option{-O} and higher.
7212 Perform interprocedural pointer analysis and interprocedural modification
7213 and reference analysis. This option can cause excessive memory and
7214 compile-time usage on large compilation units. It is not enabled by
7215 default at any optimization level.
7218 @opindex fipa-profile
7219 Perform interprocedural profile propagation. The functions called only from
7220 cold functions are marked as cold. Also functions executed once (such as
7221 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7222 functions and loop less parts of functions executed once are then optimized for
7224 Enabled by default at @option{-O} and higher.
7228 Perform interprocedural constant propagation.
7229 This optimization analyzes the program to determine when values passed
7230 to functions are constants and then optimizes accordingly.
7231 This optimization can substantially increase performance
7232 if the application has constants passed to functions.
7233 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7235 @item -fipa-cp-clone
7236 @opindex fipa-cp-clone
7237 Perform function cloning to make interprocedural constant propagation stronger.
7238 When enabled, interprocedural constant propagation performs function cloning
7239 when externally visible function can be called with constant arguments.
7240 Because this optimization can create multiple copies of functions,
7241 it may significantly increase code size
7242 (see @option{--param ipcp-unit-growth=@var{value}}).
7243 This flag is enabled by default at @option{-O3}.
7245 @item -fipa-cp-alignment
7246 @opindex -fipa-cp-alignment
7247 When enabled, this optimization propagates alignment of function
7248 parameters to support better vectorization and string operations.
7250 This flag is enabled by default at @option{-O2} and @option{-Os}. It
7251 requires that @option{-fipa-cp} is enabled.
7255 Perform Identical Code Folding for functions and read-only variables.
7256 The optimization reduces code size and may disturb unwind stacks by replacing
7257 a function by equivalent one with a different name. The optimization works
7258 more effectively with link time optimization enabled.
7260 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
7261 works on different levels and thus the optimizations are not same - there are
7262 equivalences that are found only by GCC and equivalences found only by Gold.
7264 This flag is enabled by default at @option{-O2} and @option{-Os}.
7266 @item -fisolate-erroneous-paths-dereference
7267 @opindex fisolate-erroneous-paths-dereference
7268 Detect paths that trigger erroneous or undefined behavior due to
7269 dereferencing a null pointer. Isolate those paths from the main control
7270 flow and turn the statement with erroneous or undefined behavior into a trap.
7271 This flag is enabled by default at @option{-O2} and higher and depends on
7272 @option{-fdelete-null-pointer-checks} also being enabled.
7274 @item -fisolate-erroneous-paths-attribute
7275 @opindex fisolate-erroneous-paths-attribute
7276 Detect paths that trigger erroneous or undefined behavior due a null value
7277 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
7278 attribute. Isolate those paths from the main control flow and turn the
7279 statement with erroneous or undefined behavior into a trap. This is not
7280 currently enabled, but may be enabled by @option{-O2} in the future.
7284 Perform forward store motion on trees. This flag is
7285 enabled by default at @option{-O} and higher.
7287 @item -ftree-bit-ccp
7288 @opindex ftree-bit-ccp
7289 Perform sparse conditional bit constant propagation on trees and propagate
7290 pointer alignment information.
7291 This pass only operates on local scalar variables and is enabled by default
7292 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7296 Perform sparse conditional constant propagation (CCP) on trees. This
7297 pass only operates on local scalar variables and is enabled by default
7298 at @option{-O} and higher.
7300 @item -fssa-backprop
7301 @opindex fssa-backprop
7302 Propagate information about uses of a value up the definition chain
7303 in order to simplify the definitions. For example, this pass strips
7304 sign operations if the sign of a value never matters. The flag is
7305 enabled by default at @option{-O} and higher.
7308 @opindex fssa-phiopt
7309 Perform pattern matching on SSA PHI nodes to optimize conditional
7310 code. This pass is enabled by default at @option{-O} and higher.
7312 @item -ftree-switch-conversion
7313 @opindex ftree-switch-conversion
7314 Perform conversion of simple initializations in a switch to
7315 initializations from a scalar array. This flag is enabled by default
7316 at @option{-O2} and higher.
7318 @item -ftree-tail-merge
7319 @opindex ftree-tail-merge
7320 Look for identical code sequences. When found, replace one with a jump to the
7321 other. This optimization is known as tail merging or cross jumping. This flag
7322 is enabled by default at @option{-O2} and higher. The compilation time
7324 be limited using @option{max-tail-merge-comparisons} parameter and
7325 @option{max-tail-merge-iterations} parameter.
7329 Perform dead code elimination (DCE) on trees. This flag is enabled by
7330 default at @option{-O} and higher.
7332 @item -ftree-builtin-call-dce
7333 @opindex ftree-builtin-call-dce
7334 Perform conditional dead code elimination (DCE) for calls to built-in functions
7335 that may set @code{errno} but are otherwise side-effect free. This flag is
7336 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7339 @item -ftree-dominator-opts
7340 @opindex ftree-dominator-opts
7341 Perform a variety of simple scalar cleanups (constant/copy
7342 propagation, redundancy elimination, range propagation and expression
7343 simplification) based on a dominator tree traversal. This also
7344 performs jump threading (to reduce jumps to jumps). This flag is
7345 enabled by default at @option{-O} and higher.
7349 Perform dead store elimination (DSE) on trees. A dead store is a store into
7350 a memory location that is later overwritten by another store without
7351 any intervening loads. In this case the earlier store can be deleted. This
7352 flag is enabled by default at @option{-O} and higher.
7356 Perform loop header copying on trees. This is beneficial since it increases
7357 effectiveness of code motion optimizations. It also saves one jump. This flag
7358 is enabled by default at @option{-O} and higher. It is not enabled
7359 for @option{-Os}, since it usually increases code size.
7361 @item -ftree-loop-optimize
7362 @opindex ftree-loop-optimize
7363 Perform loop optimizations on trees. This flag is enabled by default
7364 at @option{-O} and higher.
7366 @item -ftree-loop-linear
7367 @itemx -floop-interchange
7368 @itemx -floop-strip-mine
7370 @itemx -floop-unroll-and-jam
7371 @opindex ftree-loop-linear
7372 @opindex floop-interchange
7373 @opindex floop-strip-mine
7374 @opindex floop-block
7375 @opindex floop-unroll-and-jam
7376 Perform loop nest optimizations. Same as
7377 @option{-floop-nest-optimize}. To use this code transformation, GCC has
7378 to be configured with @option{--with-isl} to enable the Graphite loop
7379 transformation infrastructure.
7381 @item -fgraphite-identity
7382 @opindex fgraphite-identity
7383 Enable the identity transformation for graphite. For every SCoP we generate
7384 the polyhedral representation and transform it back to gimple. Using
7385 @option{-fgraphite-identity} we can check the costs or benefits of the
7386 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7387 are also performed by the code generator isl, like index splitting and
7388 dead code elimination in loops.
7390 @item -floop-nest-optimize
7391 @opindex floop-nest-optimize
7392 Enable the isl based loop nest optimizer. This is a generic loop nest
7393 optimizer based on the Pluto optimization algorithms. It calculates a loop
7394 structure optimized for data-locality and parallelism. This option
7397 @item -floop-parallelize-all
7398 @opindex floop-parallelize-all
7399 Use the Graphite data dependence analysis to identify loops that can
7400 be parallelized. Parallelize all the loops that can be analyzed to
7401 not contain loop carried dependences without checking that it is
7402 profitable to parallelize the loops.
7404 @item -ftree-coalesce-vars
7405 @opindex ftree-coalesce-vars
7406 While transforming the program out of the SSA representation, attempt to
7407 reduce copying by coalescing versions of different user-defined
7408 variables, instead of just compiler temporaries. This may severely
7409 limit the ability to debug an optimized program compiled with
7410 @option{-fno-var-tracking-assignments}. In the negated form, this flag
7411 prevents SSA coalescing of user variables. This option is enabled by
7412 default if optimization is enabled, and it does very little otherwise.
7414 @item -ftree-loop-if-convert
7415 @opindex ftree-loop-if-convert
7416 Attempt to transform conditional jumps in the innermost loops to
7417 branch-less equivalents. The intent is to remove control-flow from
7418 the innermost loops in order to improve the ability of the
7419 vectorization pass to handle these loops. This is enabled by default
7420 if vectorization is enabled.
7422 @item -ftree-loop-if-convert-stores
7423 @opindex ftree-loop-if-convert-stores
7424 Attempt to also if-convert conditional jumps containing memory writes.
7425 This transformation can be unsafe for multi-threaded programs as it
7426 transforms conditional memory writes into unconditional memory writes.
7429 for (i = 0; i < N; i++)
7435 for (i = 0; i < N; i++)
7436 A[i] = cond ? expr : A[i];
7438 potentially producing data races.
7440 @item -ftree-loop-distribution
7441 @opindex ftree-loop-distribution
7442 Perform loop distribution. This flag can improve cache performance on
7443 big loop bodies and allow further loop optimizations, like
7444 parallelization or vectorization, to take place. For example, the loop
7461 @item -ftree-loop-distribute-patterns
7462 @opindex ftree-loop-distribute-patterns
7463 Perform loop distribution of patterns that can be code generated with
7464 calls to a library. This flag is enabled by default at @option{-O3}.
7466 This pass distributes the initialization loops and generates a call to
7467 memset zero. For example, the loop
7483 and the initialization loop is transformed into a call to memset zero.
7485 @item -ftree-loop-im
7486 @opindex ftree-loop-im
7487 Perform loop invariant motion on trees. This pass moves only invariants that
7488 are hard to handle at RTL level (function calls, operations that expand to
7489 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7490 operands of conditions that are invariant out of the loop, so that we can use
7491 just trivial invariantness analysis in loop unswitching. The pass also includes
7494 @item -ftree-loop-ivcanon
7495 @opindex ftree-loop-ivcanon
7496 Create a canonical counter for number of iterations in loops for which
7497 determining number of iterations requires complicated analysis. Later
7498 optimizations then may determine the number easily. Useful especially
7499 in connection with unrolling.
7503 Perform induction variable optimizations (strength reduction, induction
7504 variable merging and induction variable elimination) on trees.
7506 @item -ftree-parallelize-loops=n
7507 @opindex ftree-parallelize-loops
7508 Parallelize loops, i.e., split their iteration space to run in n threads.
7509 This is only possible for loops whose iterations are independent
7510 and can be arbitrarily reordered. The optimization is only
7511 profitable on multiprocessor machines, for loops that are CPU-intensive,
7512 rather than constrained e.g.@: by memory bandwidth. This option
7513 implies @option{-pthread}, and thus is only supported on targets
7514 that have support for @option{-pthread}.
7518 Perform function-local points-to analysis on trees. This flag is
7519 enabled by default at @option{-O} and higher.
7523 Perform scalar replacement of aggregates. This pass replaces structure
7524 references with scalars to prevent committing structures to memory too
7525 early. This flag is enabled by default at @option{-O} and higher.
7529 Perform temporary expression replacement during the SSA->normal phase. Single
7530 use/single def temporaries are replaced at their use location with their
7531 defining expression. This results in non-GIMPLE code, but gives the expanders
7532 much more complex trees to work on resulting in better RTL generation. This is
7533 enabled by default at @option{-O} and higher.
7537 Perform straight-line strength reduction on trees. This recognizes related
7538 expressions involving multiplications and replaces them by less expensive
7539 calculations when possible. This is enabled by default at @option{-O} and
7542 @item -ftree-vectorize
7543 @opindex ftree-vectorize
7544 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
7545 and @option{-ftree-slp-vectorize} if not explicitly specified.
7547 @item -ftree-loop-vectorize
7548 @opindex ftree-loop-vectorize
7549 Perform loop vectorization on trees. This flag is enabled by default at
7550 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7552 @item -ftree-slp-vectorize
7553 @opindex ftree-slp-vectorize
7554 Perform basic block vectorization on trees. This flag is enabled by default at
7555 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7557 @item -fvect-cost-model=@var{model}
7558 @opindex fvect-cost-model
7559 Alter the cost model used for vectorization. The @var{model} argument
7560 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
7561 With the @samp{unlimited} model the vectorized code-path is assumed
7562 to be profitable while with the @samp{dynamic} model a runtime check
7563 guards the vectorized code-path to enable it only for iteration
7564 counts that will likely execute faster than when executing the original
7565 scalar loop. The @samp{cheap} model disables vectorization of
7566 loops where doing so would be cost prohibitive for example due to
7567 required runtime checks for data dependence or alignment but otherwise
7568 is equal to the @samp{dynamic} model.
7569 The default cost model depends on other optimization flags and is
7570 either @samp{dynamic} or @samp{cheap}.
7572 @item -fsimd-cost-model=@var{model}
7573 @opindex fsimd-cost-model
7574 Alter the cost model used for vectorization of loops marked with the OpenMP
7575 or Cilk Plus simd directive. The @var{model} argument should be one of
7576 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
7577 have the same meaning as described in @option{-fvect-cost-model} and by
7578 default a cost model defined with @option{-fvect-cost-model} is used.
7582 Perform Value Range Propagation on trees. This is similar to the
7583 constant propagation pass, but instead of values, ranges of values are
7584 propagated. This allows the optimizers to remove unnecessary range
7585 checks like array bound checks and null pointer checks. This is
7586 enabled by default at @option{-O2} and higher. Null pointer check
7587 elimination is only done if @option{-fdelete-null-pointer-checks} is
7591 @opindex fsplit-paths
7592 Split paths leading to loop backedges. This can improve dead code
7593 elimination and common subexpression elimination. This is enabled by
7594 default at @option{-O2} and above.
7596 @item -fsplit-ivs-in-unroller
7597 @opindex fsplit-ivs-in-unroller
7598 Enables expression of values of induction variables in later iterations
7599 of the unrolled loop using the value in the first iteration. This breaks
7600 long dependency chains, thus improving efficiency of the scheduling passes.
7602 A combination of @option{-fweb} and CSE is often sufficient to obtain the
7603 same effect. However, that is not reliable in cases where the loop body
7604 is more complicated than a single basic block. It also does not work at all
7605 on some architectures due to restrictions in the CSE pass.
7607 This optimization is enabled by default.
7609 @item -fvariable-expansion-in-unroller
7610 @opindex fvariable-expansion-in-unroller
7611 With this option, the compiler creates multiple copies of some
7612 local variables when unrolling a loop, which can result in superior code.
7614 @item -fpartial-inlining
7615 @opindex fpartial-inlining
7616 Inline parts of functions. This option has any effect only
7617 when inlining itself is turned on by the @option{-finline-functions}
7618 or @option{-finline-small-functions} options.
7620 Enabled at level @option{-O2}.
7622 @item -fpredictive-commoning
7623 @opindex fpredictive-commoning
7624 Perform predictive commoning optimization, i.e., reusing computations
7625 (especially memory loads and stores) performed in previous
7626 iterations of loops.
7628 This option is enabled at level @option{-O3}.
7630 @item -fprefetch-loop-arrays
7631 @opindex fprefetch-loop-arrays
7632 If supported by the target machine, generate instructions to prefetch
7633 memory to improve the performance of loops that access large arrays.
7635 This option may generate better or worse code; results are highly
7636 dependent on the structure of loops within the source code.
7638 Disabled at level @option{-Os}.
7641 @itemx -fno-peephole2
7642 @opindex fno-peephole
7643 @opindex fno-peephole2
7644 Disable any machine-specific peephole optimizations. The difference
7645 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7646 are implemented in the compiler; some targets use one, some use the
7647 other, a few use both.
7649 @option{-fpeephole} is enabled by default.
7650 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7652 @item -fno-guess-branch-probability
7653 @opindex fno-guess-branch-probability
7654 Do not guess branch probabilities using heuristics.
7656 GCC uses heuristics to guess branch probabilities if they are
7657 not provided by profiling feedback (@option{-fprofile-arcs}). These
7658 heuristics are based on the control flow graph. If some branch probabilities
7659 are specified by @code{__builtin_expect}, then the heuristics are
7660 used to guess branch probabilities for the rest of the control flow graph,
7661 taking the @code{__builtin_expect} info into account. The interactions
7662 between the heuristics and @code{__builtin_expect} can be complex, and in
7663 some cases, it may be useful to disable the heuristics so that the effects
7664 of @code{__builtin_expect} are easier to understand.
7666 The default is @option{-fguess-branch-probability} at levels
7667 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7669 @item -freorder-blocks
7670 @opindex freorder-blocks
7671 Reorder basic blocks in the compiled function in order to reduce number of
7672 taken branches and improve code locality.
7674 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7676 @item -freorder-blocks-algorithm=@var{algorithm}
7677 @opindex freorder-blocks-algorithm
7678 Use the specified algorithm for basic block reordering. The
7679 @var{algorithm} argument can be @samp{simple}, which does not increase
7680 code size (except sometimes due to secondary effects like alignment),
7681 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
7682 put all often executed code together, minimizing the number of branches
7683 executed by making extra copies of code.
7685 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
7686 @samp{stc} at levels @option{-O2}, @option{-O3}.
7688 @item -freorder-blocks-and-partition
7689 @opindex freorder-blocks-and-partition
7690 In addition to reordering basic blocks in the compiled function, in order
7691 to reduce number of taken branches, partitions hot and cold basic blocks
7692 into separate sections of the assembly and @file{.o} files, to improve
7693 paging and cache locality performance.
7695 This optimization is automatically turned off in the presence of
7696 exception handling, for linkonce sections, for functions with a user-defined
7697 section attribute and on any architecture that does not support named
7700 Enabled for x86 at levels @option{-O2}, @option{-O3}.
7702 @item -freorder-functions
7703 @opindex freorder-functions
7704 Reorder functions in the object file in order to
7705 improve code locality. This is implemented by using special
7706 subsections @code{.text.hot} for most frequently executed functions and
7707 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7708 the linker so object file format must support named sections and linker must
7709 place them in a reasonable way.
7711 Also profile feedback must be available to make this option effective. See
7712 @option{-fprofile-arcs} for details.
7714 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7716 @item -fstrict-aliasing
7717 @opindex fstrict-aliasing
7718 Allow the compiler to assume the strictest aliasing rules applicable to
7719 the language being compiled. For C (and C++), this activates
7720 optimizations based on the type of expressions. In particular, an
7721 object of one type is assumed never to reside at the same address as an
7722 object of a different type, unless the types are almost the same. For
7723 example, an @code{unsigned int} can alias an @code{int}, but not a
7724 @code{void*} or a @code{double}. A character type may alias any other
7727 @anchor{Type-punning}Pay special attention to code like this:
7740 The practice of reading from a different union member than the one most
7741 recently written to (called ``type-punning'') is common. Even with
7742 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7743 is accessed through the union type. So, the code above works as
7744 expected. @xref{Structures unions enumerations and bit-fields
7745 implementation}. However, this code might not:
7756 Similarly, access by taking the address, casting the resulting pointer
7757 and dereferencing the result has undefined behavior, even if the cast
7758 uses a union type, e.g.:
7762 return ((union a_union *) &d)->i;
7766 The @option{-fstrict-aliasing} option is enabled at levels
7767 @option{-O2}, @option{-O3}, @option{-Os}.
7769 @item -fstrict-overflow
7770 @opindex fstrict-overflow
7771 Allow the compiler to assume strict signed overflow rules, depending
7772 on the language being compiled. For C (and C++) this means that
7773 overflow when doing arithmetic with signed numbers is undefined, which
7774 means that the compiler may assume that it does not happen. This
7775 permits various optimizations. For example, the compiler assumes
7776 that an expression like @code{i + 10 > i} is always true for
7777 signed @code{i}. This assumption is only valid if signed overflow is
7778 undefined, as the expression is false if @code{i + 10} overflows when
7779 using twos complement arithmetic. When this option is in effect any
7780 attempt to determine whether an operation on signed numbers
7781 overflows must be written carefully to not actually involve overflow.
7783 This option also allows the compiler to assume strict pointer
7784 semantics: given a pointer to an object, if adding an offset to that
7785 pointer does not produce a pointer to the same object, the addition is
7786 undefined. This permits the compiler to conclude that @code{p + u >
7787 p} is always true for a pointer @code{p} and unsigned integer
7788 @code{u}. This assumption is only valid because pointer wraparound is
7789 undefined, as the expression is false if @code{p + u} overflows using
7790 twos complement arithmetic.
7792 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7793 that integer signed overflow is fully defined: it wraps. When
7794 @option{-fwrapv} is used, there is no difference between
7795 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7796 integers. With @option{-fwrapv} certain types of overflow are
7797 permitted. For example, if the compiler gets an overflow when doing
7798 arithmetic on constants, the overflowed value can still be used with
7799 @option{-fwrapv}, but not otherwise.
7801 The @option{-fstrict-overflow} option is enabled at levels
7802 @option{-O2}, @option{-O3}, @option{-Os}.
7804 @item -falign-functions
7805 @itemx -falign-functions=@var{n}
7806 @opindex falign-functions
7807 Align the start of functions to the next power-of-two greater than
7808 @var{n}, skipping up to @var{n} bytes. For instance,
7809 @option{-falign-functions=32} aligns functions to the next 32-byte
7810 boundary, but @option{-falign-functions=24} aligns to the next
7811 32-byte boundary only if this can be done by skipping 23 bytes or less.
7813 @option{-fno-align-functions} and @option{-falign-functions=1} are
7814 equivalent and mean that functions are not aligned.
7816 Some assemblers only support this flag when @var{n} is a power of two;
7817 in that case, it is rounded up.
7819 If @var{n} is not specified or is zero, use a machine-dependent default.
7821 Enabled at levels @option{-O2}, @option{-O3}.
7823 @item -falign-labels
7824 @itemx -falign-labels=@var{n}
7825 @opindex falign-labels
7826 Align all branch targets to a power-of-two boundary, skipping up to
7827 @var{n} bytes like @option{-falign-functions}. This option can easily
7828 make code slower, because it must insert dummy operations for when the
7829 branch target is reached in the usual flow of the code.
7831 @option{-fno-align-labels} and @option{-falign-labels=1} are
7832 equivalent and mean that labels are not aligned.
7834 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7835 are greater than this value, then their values are used instead.
7837 If @var{n} is not specified or is zero, use a machine-dependent default
7838 which is very likely to be @samp{1}, meaning no alignment.
7840 Enabled at levels @option{-O2}, @option{-O3}.
7843 @itemx -falign-loops=@var{n}
7844 @opindex falign-loops
7845 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7846 like @option{-falign-functions}. If the loops are
7847 executed many times, this makes up for any execution of the dummy
7850 @option{-fno-align-loops} and @option{-falign-loops=1} are
7851 equivalent and mean that loops are not aligned.
7853 If @var{n} is not specified or is zero, use a machine-dependent default.
7855 Enabled at levels @option{-O2}, @option{-O3}.
7858 @itemx -falign-jumps=@var{n}
7859 @opindex falign-jumps
7860 Align branch targets to a power-of-two boundary, for branch targets
7861 where the targets can only be reached by jumping, skipping up to @var{n}
7862 bytes like @option{-falign-functions}. In this case, no dummy operations
7865 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7866 equivalent and mean that loops are not aligned.
7868 If @var{n} is not specified or is zero, use a machine-dependent default.
7870 Enabled at levels @option{-O2}, @option{-O3}.
7872 @item -funit-at-a-time
7873 @opindex funit-at-a-time
7874 This option is left for compatibility reasons. @option{-funit-at-a-time}
7875 has no effect, while @option{-fno-unit-at-a-time} implies
7876 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7880 @item -fno-toplevel-reorder
7881 @opindex fno-toplevel-reorder
7882 Do not reorder top-level functions, variables, and @code{asm}
7883 statements. Output them in the same order that they appear in the
7884 input file. When this option is used, unreferenced static variables
7885 are not removed. This option is intended to support existing code
7886 that relies on a particular ordering. For new code, it is better to
7887 use attributes when possible.
7889 Enabled at level @option{-O0}. When disabled explicitly, it also implies
7890 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
7895 Constructs webs as commonly used for register allocation purposes and assign
7896 each web individual pseudo register. This allows the register allocation pass
7897 to operate on pseudos directly, but also strengthens several other optimization
7898 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7899 however, make debugging impossible, since variables no longer stay in a
7902 Enabled by default with @option{-funroll-loops}.
7904 @item -fwhole-program
7905 @opindex fwhole-program
7906 Assume that the current compilation unit represents the whole program being
7907 compiled. All public functions and variables with the exception of @code{main}
7908 and those merged by attribute @code{externally_visible} become static functions
7909 and in effect are optimized more aggressively by interprocedural optimizers.
7911 This option should not be used in combination with @option{-flto}.
7912 Instead relying on a linker plugin should provide safer and more precise
7915 @item -flto[=@var{n}]
7917 This option runs the standard link-time optimizer. When invoked
7918 with source code, it generates GIMPLE (one of GCC's internal
7919 representations) and writes it to special ELF sections in the object
7920 file. When the object files are linked together, all the function
7921 bodies are read from these ELF sections and instantiated as if they
7922 had been part of the same translation unit.
7924 To use the link-time optimizer, @option{-flto} and optimization
7925 options should be specified at compile time and during the final link.
7926 It is recommended that you compile all the files participating in the
7927 same link with the same options and also specify those options at
7932 gcc -c -O2 -flto foo.c
7933 gcc -c -O2 -flto bar.c
7934 gcc -o myprog -flto -O2 foo.o bar.o
7937 The first two invocations to GCC save a bytecode representation
7938 of GIMPLE into special ELF sections inside @file{foo.o} and
7939 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7940 @file{foo.o} and @file{bar.o}, merges the two files into a single
7941 internal image, and compiles the result as usual. Since both
7942 @file{foo.o} and @file{bar.o} are merged into a single image, this
7943 causes all the interprocedural analyses and optimizations in GCC to
7944 work across the two files as if they were a single one. This means,
7945 for example, that the inliner is able to inline functions in
7946 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7948 Another (simpler) way to enable link-time optimization is:
7951 gcc -o myprog -flto -O2 foo.c bar.c
7954 The above generates bytecode for @file{foo.c} and @file{bar.c},
7955 merges them together into a single GIMPLE representation and optimizes
7956 them as usual to produce @file{myprog}.
7958 The only important thing to keep in mind is that to enable link-time
7959 optimizations you need to use the GCC driver to perform the link step.
7960 GCC then automatically performs link-time optimization if any of the
7961 objects involved were compiled with the @option{-flto} command-line option.
7963 should specify the optimization options to be used for link-time
7964 optimization though GCC tries to be clever at guessing an
7965 optimization level to use from the options used at compile time
7966 if you fail to specify one at link time. You can always override
7967 the automatic decision to do link-time optimization at link time
7968 by passing @option{-fno-lto} to the link command.
7970 To make whole program optimization effective, it is necessary to make
7971 certain whole program assumptions. The compiler needs to know
7972 what functions and variables can be accessed by libraries and runtime
7973 outside of the link-time optimized unit. When supported by the linker,
7974 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7975 to the compiler about used and externally visible symbols. When
7976 the linker plugin is not available, @option{-fwhole-program} should be
7977 used to allow the compiler to make these assumptions, which leads
7978 to more aggressive optimization decisions.
7980 When @option{-fuse-linker-plugin} is not enabled, when a file is
7981 compiled with @option{-flto}, the generated object file is larger than
7982 a regular object file because it contains GIMPLE bytecodes and the usual
7983 final code (see @option{-ffat-lto-objects}. This means that
7984 object files with LTO information can be linked as normal object
7985 files; if @option{-fno-lto} is passed to the linker, no
7986 interprocedural optimizations are applied. Note that when
7987 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
7988 but you cannot perform a regular, non-LTO link on them.
7990 Additionally, the optimization flags used to compile individual files
7991 are not necessarily related to those used at link time. For instance,
7994 gcc -c -O0 -ffat-lto-objects -flto foo.c
7995 gcc -c -O0 -ffat-lto-objects -flto bar.c
7996 gcc -o myprog -O3 foo.o bar.o
7999 This produces individual object files with unoptimized assembler
8000 code, but the resulting binary @file{myprog} is optimized at
8001 @option{-O3}. If, instead, the final binary is generated with
8002 @option{-fno-lto}, then @file{myprog} is not optimized.
8004 When producing the final binary, GCC only
8005 applies link-time optimizations to those files that contain bytecode.
8006 Therefore, you can mix and match object files and libraries with
8007 GIMPLE bytecodes and final object code. GCC automatically selects
8008 which files to optimize in LTO mode and which files to link without
8011 There are some code generation flags preserved by GCC when
8012 generating bytecodes, as they need to be used during the final link
8013 stage. Generally options specified at link time override those
8014 specified at compile time.
8016 If you do not specify an optimization level option @option{-O} at
8017 link time, then GCC uses the highest optimization level
8018 used when compiling the object files.
8020 Currently, the following options and their settings are taken from
8021 the first object file that explicitly specifies them:
8022 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8023 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8024 and all the @option{-m} target flags.
8026 Certain ABI-changing flags are required to match in all compilation units,
8027 and trying to override this at link time with a conflicting value
8028 is ignored. This includes options such as @option{-freg-struct-return}
8029 and @option{-fpcc-struct-return}.
8031 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8032 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8033 are passed through to the link stage and merged conservatively for
8034 conflicting translation units. Specifically
8035 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8036 precedence; and for example @option{-ffp-contract=off} takes precedence
8037 over @option{-ffp-contract=fast}. You can override them at link time.
8039 If LTO encounters objects with C linkage declared with incompatible
8040 types in separate translation units to be linked together (undefined
8041 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8042 issued. The behavior is still undefined at run time. Similar
8043 diagnostics may be raised for other languages.
8045 Another feature of LTO is that it is possible to apply interprocedural
8046 optimizations on files written in different languages:
8051 gfortran -c -flto baz.f90
8052 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8055 Notice that the final link is done with @command{g++} to get the C++
8056 runtime libraries and @option{-lgfortran} is added to get the Fortran
8057 runtime libraries. In general, when mixing languages in LTO mode, you
8058 should use the same link command options as when mixing languages in a
8059 regular (non-LTO) compilation.
8061 If object files containing GIMPLE bytecode are stored in a library archive, say
8062 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8063 are using a linker with plugin support. To create static libraries suitable
8064 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8065 and @command{ranlib};
8066 to show the symbols of object files with GIMPLE bytecode, use
8067 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8068 and @command{nm} have been compiled with plugin support. At link time, use the the
8069 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8070 the LTO optimization process:
8073 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8076 With the linker plugin enabled, the linker extracts the needed
8077 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8078 to make them part of the aggregated GIMPLE image to be optimized.
8080 If you are not using a linker with plugin support and/or do not
8081 enable the linker plugin, then the objects inside @file{libfoo.a}
8082 are extracted and linked as usual, but they do not participate
8083 in the LTO optimization process. In order to make a static library suitable
8084 for both LTO optimization and usual linkage, compile its object files with
8085 @option{-flto} @option{-ffat-lto-objects}.
8087 Link-time optimizations do not require the presence of the whole program to
8088 operate. If the program does not require any symbols to be exported, it is
8089 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8090 the interprocedural optimizers to use more aggressive assumptions which may
8091 lead to improved optimization opportunities.
8092 Use of @option{-fwhole-program} is not needed when linker plugin is
8093 active (see @option{-fuse-linker-plugin}).
8095 The current implementation of LTO makes no
8096 attempt to generate bytecode that is portable between different
8097 types of hosts. The bytecode files are versioned and there is a
8098 strict version check, so bytecode files generated in one version of
8099 GCC do not work with an older or newer version of GCC.
8101 Link-time optimization does not work well with generation of debugging
8102 information. Combining @option{-flto} with
8103 @option{-g} is currently experimental and expected to produce unexpected
8106 If you specify the optional @var{n}, the optimization and code
8107 generation done at link time is executed in parallel using @var{n}
8108 parallel jobs by utilizing an installed @command{make} program. The
8109 environment variable @env{MAKE} may be used to override the program
8110 used. The default value for @var{n} is 1.
8112 You can also specify @option{-flto=jobserver} to use GNU make's
8113 job server mode to determine the number of parallel jobs. This
8114 is useful when the Makefile calling GCC is already executing in parallel.
8115 You must prepend a @samp{+} to the command recipe in the parent Makefile
8116 for this to work. This option likely only works if @env{MAKE} is
8119 @item -flto-partition=@var{alg}
8120 @opindex flto-partition
8121 Specify the partitioning algorithm used by the link-time optimizer.
8122 The value is either @samp{1to1} to specify a partitioning mirroring
8123 the original source files or @samp{balanced} to specify partitioning
8124 into equally sized chunks (whenever possible) or @samp{max} to create
8125 new partition for every symbol where possible. Specifying @samp{none}
8126 as an algorithm disables partitioning and streaming completely.
8127 The default value is @samp{balanced}. While @samp{1to1} can be used
8128 as an workaround for various code ordering issues, the @samp{max}
8129 partitioning is intended for internal testing only.
8130 The value @samp{one} specifies that exactly one partition should be
8131 used while the value @samp{none} bypasses partitioning and executes
8132 the link-time optimization step directly from the WPA phase.
8134 @item -flto-odr-type-merging
8135 @opindex flto-odr-type-merging
8136 Enable streaming of mangled types names of C++ types and their unification
8137 at link time. This increases size of LTO object files, but enables
8138 diagnostics about One Definition Rule violations.
8140 @item -flto-compression-level=@var{n}
8141 @opindex flto-compression-level
8142 This option specifies the level of compression used for intermediate
8143 language written to LTO object files, and is only meaningful in
8144 conjunction with LTO mode (@option{-flto}). Valid
8145 values are 0 (no compression) to 9 (maximum compression). Values
8146 outside this range are clamped to either 0 or 9. If the option is not
8147 given, a default balanced compression setting is used.
8149 @item -fuse-linker-plugin
8150 @opindex fuse-linker-plugin
8151 Enables the use of a linker plugin during link-time optimization. This
8152 option relies on plugin support in the linker, which is available in gold
8153 or in GNU ld 2.21 or newer.
8155 This option enables the extraction of object files with GIMPLE bytecode out
8156 of library archives. This improves the quality of optimization by exposing
8157 more code to the link-time optimizer. This information specifies what
8158 symbols can be accessed externally (by non-LTO object or during dynamic
8159 linking). Resulting code quality improvements on binaries (and shared
8160 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8161 See @option{-flto} for a description of the effect of this flag and how to
8164 This option is enabled by default when LTO support in GCC is enabled
8165 and GCC was configured for use with
8166 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8168 @item -ffat-lto-objects
8169 @opindex ffat-lto-objects
8170 Fat LTO objects are object files that contain both the intermediate language
8171 and the object code. This makes them usable for both LTO linking and normal
8172 linking. This option is effective only when compiling with @option{-flto}
8173 and is ignored at link time.
8175 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8176 requires the complete toolchain to be aware of LTO. It requires a linker with
8177 linker plugin support for basic functionality. Additionally,
8178 @command{nm}, @command{ar} and @command{ranlib}
8179 need to support linker plugins to allow a full-featured build environment
8180 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8181 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8182 to these tools. With non fat LTO makefiles need to be modified to use them.
8184 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8187 @item -fcompare-elim
8188 @opindex fcompare-elim
8189 After register allocation and post-register allocation instruction splitting,
8190 identify arithmetic instructions that compute processor flags similar to a
8191 comparison operation based on that arithmetic. If possible, eliminate the
8192 explicit comparison operation.
8194 This pass only applies to certain targets that cannot explicitly represent
8195 the comparison operation before register allocation is complete.
8197 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8199 @item -fcprop-registers
8200 @opindex fcprop-registers
8201 After register allocation and post-register allocation instruction splitting,
8202 perform a copy-propagation pass to try to reduce scheduling dependencies
8203 and occasionally eliminate the copy.
8205 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8207 @item -fprofile-correction
8208 @opindex fprofile-correction
8209 Profiles collected using an instrumented binary for multi-threaded programs may
8210 be inconsistent due to missed counter updates. When this option is specified,
8211 GCC uses heuristics to correct or smooth out such inconsistencies. By
8212 default, GCC emits an error message when an inconsistent profile is detected.
8215 @itemx -fprofile-use=@var{path}
8216 @opindex fprofile-use
8217 Enable profile feedback-directed optimizations,
8218 and the following optimizations
8219 which are generally profitable only with profile feedback available:
8220 @option{-fbranch-probabilities}, @option{-fvpt},
8221 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8222 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
8224 Before you can use this option, you must first generate profiling information.
8225 @xref{Optimize Options}, for information about the @option{-fprofile-generate}
8228 By default, GCC emits an error message if the feedback profiles do not
8229 match the source code. This error can be turned into a warning by using
8230 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8233 If @var{path} is specified, GCC looks at the @var{path} to find
8234 the profile feedback data files. See @option{-fprofile-dir}.
8236 @item -fauto-profile
8237 @itemx -fauto-profile=@var{path}
8238 @opindex fauto-profile
8239 Enable sampling-based feedback-directed optimizations,
8240 and the following optimizations
8241 which are generally profitable only with profile feedback available:
8242 @option{-fbranch-probabilities}, @option{-fvpt},
8243 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8244 @option{-ftree-vectorize},
8245 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
8246 @option{-fpredictive-commoning}, @option{-funswitch-loops},
8247 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
8249 @var{path} is the name of a file containing AutoFDO profile information.
8250 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
8252 Producing an AutoFDO profile data file requires running your program
8253 with the @command{perf} utility on a supported GNU/Linux target system.
8254 For more information, see @uref{https://perf.wiki.kernel.org/}.
8258 perf record -e br_inst_retired:near_taken -b -o perf.data \
8262 Then use the @command{create_gcov} tool to convert the raw profile data
8263 to a format that can be used by GCC.@ You must also supply the
8264 unstripped binary for your program to this tool.
8265 See @uref{https://github.com/google/autofdo}.
8269 create_gcov --binary=your_program.unstripped --profile=perf.data \
8274 The following options control compiler behavior regarding floating-point
8275 arithmetic. These options trade off between speed and
8276 correctness. All must be specifically enabled.
8280 @opindex ffloat-store
8281 Do not store floating-point variables in registers, and inhibit other
8282 options that might change whether a floating-point value is taken from a
8285 @cindex floating-point precision
8286 This option prevents undesirable excess precision on machines such as
8287 the 68000 where the floating registers (of the 68881) keep more
8288 precision than a @code{double} is supposed to have. Similarly for the
8289 x86 architecture. For most programs, the excess precision does only
8290 good, but a few programs rely on the precise definition of IEEE floating
8291 point. Use @option{-ffloat-store} for such programs, after modifying
8292 them to store all pertinent intermediate computations into variables.
8294 @item -fexcess-precision=@var{style}
8295 @opindex fexcess-precision
8296 This option allows further control over excess precision on machines
8297 where floating-point registers have more precision than the IEEE
8298 @code{float} and @code{double} types and the processor does not
8299 support operations rounding to those types. By default,
8300 @option{-fexcess-precision=fast} is in effect; this means that
8301 operations are carried out in the precision of the registers and that
8302 it is unpredictable when rounding to the types specified in the source
8303 code takes place. When compiling C, if
8304 @option{-fexcess-precision=standard} is specified then excess
8305 precision follows the rules specified in ISO C99; in particular,
8306 both casts and assignments cause values to be rounded to their
8307 semantic types (whereas @option{-ffloat-store} only affects
8308 assignments). This option is enabled by default for C if a strict
8309 conformance option such as @option{-std=c99} is used.
8312 @option{-fexcess-precision=standard} is not implemented for languages
8313 other than C, and has no effect if
8314 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8315 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8316 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8317 semantics apply without excess precision, and in the latter, rounding
8322 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8323 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8324 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8326 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8328 This option is not turned on by any @option{-O} option besides
8329 @option{-Ofast} since it can result in incorrect output for programs
8330 that depend on an exact implementation of IEEE or ISO rules/specifications
8331 for math functions. It may, however, yield faster code for programs
8332 that do not require the guarantees of these specifications.
8334 @item -fno-math-errno
8335 @opindex fno-math-errno
8336 Do not set @code{errno} after calling math functions that are executed
8337 with a single instruction, e.g., @code{sqrt}. A program that relies on
8338 IEEE exceptions for math error handling may want to use this flag
8339 for speed while maintaining IEEE arithmetic compatibility.
8341 This option is not turned on by any @option{-O} option since
8342 it can result in incorrect output for programs that depend on
8343 an exact implementation of IEEE or ISO rules/specifications for
8344 math functions. It may, however, yield faster code for programs
8345 that do not require the guarantees of these specifications.
8347 The default is @option{-fmath-errno}.
8349 On Darwin systems, the math library never sets @code{errno}. There is
8350 therefore no reason for the compiler to consider the possibility that
8351 it might, and @option{-fno-math-errno} is the default.
8353 @item -funsafe-math-optimizations
8354 @opindex funsafe-math-optimizations
8356 Allow optimizations for floating-point arithmetic that (a) assume
8357 that arguments and results are valid and (b) may violate IEEE or
8358 ANSI standards. When used at link time, it may include libraries
8359 or startup files that change the default FPU control word or other
8360 similar optimizations.
8362 This option is not turned on by any @option{-O} option since
8363 it can result in incorrect output for programs that depend on
8364 an exact implementation of IEEE or ISO rules/specifications for
8365 math functions. It may, however, yield faster code for programs
8366 that do not require the guarantees of these specifications.
8367 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8368 @option{-fassociative-math} and @option{-freciprocal-math}.
8370 The default is @option{-fno-unsafe-math-optimizations}.
8372 @item -fassociative-math
8373 @opindex fassociative-math
8375 Allow re-association of operands in series of floating-point operations.
8376 This violates the ISO C and C++ language standard by possibly changing
8377 computation result. NOTE: re-ordering may change the sign of zero as
8378 well as ignore NaNs and inhibit or create underflow or overflow (and
8379 thus cannot be used on code that relies on rounding behavior like
8380 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8381 and thus may not be used when ordered comparisons are required.
8382 This option requires that both @option{-fno-signed-zeros} and
8383 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8384 much sense with @option{-frounding-math}. For Fortran the option
8385 is automatically enabled when both @option{-fno-signed-zeros} and
8386 @option{-fno-trapping-math} are in effect.
8388 The default is @option{-fno-associative-math}.
8390 @item -freciprocal-math
8391 @opindex freciprocal-math
8393 Allow the reciprocal of a value to be used instead of dividing by
8394 the value if this enables optimizations. For example @code{x / y}
8395 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8396 is subject to common subexpression elimination. Note that this loses
8397 precision and increases the number of flops operating on the value.
8399 The default is @option{-fno-reciprocal-math}.
8401 @item -ffinite-math-only
8402 @opindex ffinite-math-only
8403 Allow optimizations for floating-point arithmetic that assume
8404 that arguments and results are not NaNs or +-Infs.
8406 This option is not turned on by any @option{-O} option since
8407 it can result in incorrect output for programs that depend on
8408 an exact implementation of IEEE or ISO rules/specifications for
8409 math functions. It may, however, yield faster code for programs
8410 that do not require the guarantees of these specifications.
8412 The default is @option{-fno-finite-math-only}.
8414 @item -fno-signed-zeros
8415 @opindex fno-signed-zeros
8416 Allow optimizations for floating-point arithmetic that ignore the
8417 signedness of zero. IEEE arithmetic specifies the behavior of
8418 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8419 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8420 This option implies that the sign of a zero result isn't significant.
8422 The default is @option{-fsigned-zeros}.
8424 @item -fno-trapping-math
8425 @opindex fno-trapping-math
8426 Compile code assuming that floating-point operations cannot generate
8427 user-visible traps. These traps include division by zero, overflow,
8428 underflow, inexact result and invalid operation. This option requires
8429 that @option{-fno-signaling-nans} be in effect. Setting this option may
8430 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8432 This option should never be turned on by any @option{-O} option since
8433 it can result in incorrect output for programs that depend on
8434 an exact implementation of IEEE or ISO rules/specifications for
8437 The default is @option{-ftrapping-math}.
8439 @item -frounding-math
8440 @opindex frounding-math
8441 Disable transformations and optimizations that assume default floating-point
8442 rounding behavior. This is round-to-zero for all floating point
8443 to integer conversions, and round-to-nearest for all other arithmetic
8444 truncations. This option should be specified for programs that change
8445 the FP rounding mode dynamically, or that may be executed with a
8446 non-default rounding mode. This option disables constant folding of
8447 floating-point expressions at compile time (which may be affected by
8448 rounding mode) and arithmetic transformations that are unsafe in the
8449 presence of sign-dependent rounding modes.
8451 The default is @option{-fno-rounding-math}.
8453 This option is experimental and does not currently guarantee to
8454 disable all GCC optimizations that are affected by rounding mode.
8455 Future versions of GCC may provide finer control of this setting
8456 using C99's @code{FENV_ACCESS} pragma. This command-line option
8457 will be used to specify the default state for @code{FENV_ACCESS}.
8459 @item -fsignaling-nans
8460 @opindex fsignaling-nans
8461 Compile code assuming that IEEE signaling NaNs may generate user-visible
8462 traps during floating-point operations. Setting this option disables
8463 optimizations that may change the number of exceptions visible with
8464 signaling NaNs. This option implies @option{-ftrapping-math}.
8466 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8469 The default is @option{-fno-signaling-nans}.
8471 This option is experimental and does not currently guarantee to
8472 disable all GCC optimizations that affect signaling NaN behavior.
8474 @item -fsingle-precision-constant
8475 @opindex fsingle-precision-constant
8476 Treat floating-point constants as single precision instead of
8477 implicitly converting them to double-precision constants.
8479 @item -fcx-limited-range
8480 @opindex fcx-limited-range
8481 When enabled, this option states that a range reduction step is not
8482 needed when performing complex division. Also, there is no checking
8483 whether the result of a complex multiplication or division is @code{NaN
8484 + I*NaN}, with an attempt to rescue the situation in that case. The
8485 default is @option{-fno-cx-limited-range}, but is enabled by
8486 @option{-ffast-math}.
8488 This option controls the default setting of the ISO C99
8489 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8492 @item -fcx-fortran-rules
8493 @opindex fcx-fortran-rules
8494 Complex multiplication and division follow Fortran rules. Range
8495 reduction is done as part of complex division, but there is no checking
8496 whether the result of a complex multiplication or division is @code{NaN
8497 + I*NaN}, with an attempt to rescue the situation in that case.
8499 The default is @option{-fno-cx-fortran-rules}.
8503 The following options control optimizations that may improve
8504 performance, but are not enabled by any @option{-O} options. This
8505 section includes experimental options that may produce broken code.
8508 @item -fbranch-probabilities
8509 @opindex fbranch-probabilities
8510 After running a program compiled with @option{-fprofile-arcs}
8511 (@pxref{Instrumentation Options}),
8512 you can compile it a second time using
8513 @option{-fbranch-probabilities}, to improve optimizations based on
8514 the number of times each branch was taken. When a program
8515 compiled with @option{-fprofile-arcs} exits, it saves arc execution
8516 counts to a file called @file{@var{sourcename}.gcda} for each source
8517 file. The information in this data file is very dependent on the
8518 structure of the generated code, so you must use the same source code
8519 and the same optimization options for both compilations.
8521 With @option{-fbranch-probabilities}, GCC puts a
8522 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8523 These can be used to improve optimization. Currently, they are only
8524 used in one place: in @file{reorg.c}, instead of guessing which path a
8525 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8526 exactly determine which path is taken more often.
8528 @item -fprofile-values
8529 @opindex fprofile-values
8530 If combined with @option{-fprofile-arcs}, it adds code so that some
8531 data about values of expressions in the program is gathered.
8533 With @option{-fbranch-probabilities}, it reads back the data gathered
8534 from profiling values of expressions for usage in optimizations.
8536 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8538 @item -fprofile-reorder-functions
8539 @opindex fprofile-reorder-functions
8540 Function reordering based on profile instrumentation collects
8541 first time of execution of a function and orders these functions
8544 Enabled with @option{-fprofile-use}.
8548 If combined with @option{-fprofile-arcs}, this option instructs the compiler
8549 to add code to gather information about values of expressions.
8551 With @option{-fbranch-probabilities}, it reads back the data gathered
8552 and actually performs the optimizations based on them.
8553 Currently the optimizations include specialization of division operations
8554 using the knowledge about the value of the denominator.
8556 @item -frename-registers
8557 @opindex frename-registers
8558 Attempt to avoid false dependencies in scheduled code by making use
8559 of registers left over after register allocation. This optimization
8560 most benefits processors with lots of registers. Depending on the
8561 debug information format adopted by the target, however, it can
8562 make debugging impossible, since variables no longer stay in
8563 a ``home register''.
8565 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8567 @item -fschedule-fusion
8568 @opindex fschedule-fusion
8569 Performs a target dependent pass over the instruction stream to schedule
8570 instructions of same type together because target machine can execute them
8571 more efficiently if they are adjacent to each other in the instruction flow.
8573 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8577 Perform tail duplication to enlarge superblock size. This transformation
8578 simplifies the control flow of the function allowing other optimizations to do
8581 Enabled with @option{-fprofile-use}.
8583 @item -funroll-loops
8584 @opindex funroll-loops
8585 Unroll loops whose number of iterations can be determined at compile time or
8586 upon entry to the loop. @option{-funroll-loops} implies
8587 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8588 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8589 a small constant number of iterations). This option makes code larger, and may
8590 or may not make it run faster.
8592 Enabled with @option{-fprofile-use}.
8594 @item -funroll-all-loops
8595 @opindex funroll-all-loops
8596 Unroll all loops, even if their number of iterations is uncertain when
8597 the loop is entered. This usually makes programs run more slowly.
8598 @option{-funroll-all-loops} implies the same options as
8599 @option{-funroll-loops}.
8602 @opindex fpeel-loops
8603 Peels loops for which there is enough information that they do not
8604 roll much (from profile feedback). It also turns on complete loop peeling
8605 (i.e.@: complete removal of loops with small constant number of iterations).
8607 Enabled with @option{-fprofile-use}.
8609 @item -fmove-loop-invariants
8610 @opindex fmove-loop-invariants
8611 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8612 at level @option{-O1}
8614 @item -funswitch-loops
8615 @opindex funswitch-loops
8616 Move branches with loop invariant conditions out of the loop, with duplicates
8617 of the loop on both branches (modified according to result of the condition).
8619 @item -ffunction-sections
8620 @itemx -fdata-sections
8621 @opindex ffunction-sections
8622 @opindex fdata-sections
8623 Place each function or data item into its own section in the output
8624 file if the target supports arbitrary sections. The name of the
8625 function or the name of the data item determines the section's name
8628 Use these options on systems where the linker can perform optimizations
8629 to improve locality of reference in the instruction space. Most systems
8630 using the ELF object format and SPARC processors running Solaris 2 have
8631 linkers with such optimizations. AIX may have these optimizations in
8634 Only use these options when there are significant benefits from doing
8635 so. When you specify these options, the assembler and linker
8636 create larger object and executable files and are also slower.
8637 You cannot use @command{gprof} on all systems if you
8638 specify this option, and you may have problems with debugging if
8639 you specify both this option and @option{-g}.
8641 @item -fbranch-target-load-optimize
8642 @opindex fbranch-target-load-optimize
8643 Perform branch target register load optimization before prologue / epilogue
8645 The use of target registers can typically be exposed only during reload,
8646 thus hoisting loads out of loops and doing inter-block scheduling needs
8647 a separate optimization pass.
8649 @item -fbranch-target-load-optimize2
8650 @opindex fbranch-target-load-optimize2
8651 Perform branch target register load optimization after prologue / epilogue
8654 @item -fbtr-bb-exclusive
8655 @opindex fbtr-bb-exclusive
8656 When performing branch target register load optimization, don't reuse
8657 branch target registers within any basic block.
8660 @opindex fstdarg-opt
8661 Optimize the prologue of variadic argument functions with respect to usage of
8664 @item -fsection-anchors
8665 @opindex fsection-anchors
8666 Try to reduce the number of symbolic address calculations by using
8667 shared ``anchor'' symbols to address nearby objects. This transformation
8668 can help to reduce the number of GOT entries and GOT accesses on some
8671 For example, the implementation of the following function @code{foo}:
8675 int foo (void) @{ return a + b + c; @}
8679 usually calculates the addresses of all three variables, but if you
8680 compile it with @option{-fsection-anchors}, it accesses the variables
8681 from a common anchor point instead. The effect is similar to the
8682 following pseudocode (which isn't valid C):
8687 register int *xr = &x;
8688 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8692 Not all targets support this option.
8694 @item --param @var{name}=@var{value}
8696 In some places, GCC uses various constants to control the amount of
8697 optimization that is done. For example, GCC does not inline functions
8698 that contain more than a certain number of instructions. You can
8699 control some of these constants on the command line using the
8700 @option{--param} option.
8702 The names of specific parameters, and the meaning of the values, are
8703 tied to the internals of the compiler, and are subject to change
8704 without notice in future releases.
8706 In each case, the @var{value} is an integer. The allowable choices for
8710 @item predictable-branch-outcome
8711 When branch is predicted to be taken with probability lower than this threshold
8712 (in percent), then it is considered well predictable. The default is 10.
8714 @item max-rtl-if-conversion-insns
8715 RTL if-conversion tries to remove conditional branches around a block and
8716 replace them with conditionally executed instructions. This parameter
8717 gives the maximum number of instructions in a block which should be
8718 considered for if-conversion. The default is 10, though the compiler will
8719 also use other heuristics to decide whether if-conversion is likely to be
8722 @item max-crossjump-edges
8723 The maximum number of incoming edges to consider for cross-jumping.
8724 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8725 the number of edges incoming to each block. Increasing values mean
8726 more aggressive optimization, making the compilation time increase with
8727 probably small improvement in executable size.
8729 @item min-crossjump-insns
8730 The minimum number of instructions that must be matched at the end
8731 of two blocks before cross-jumping is performed on them. This
8732 value is ignored in the case where all instructions in the block being
8733 cross-jumped from are matched. The default value is 5.
8735 @item max-grow-copy-bb-insns
8736 The maximum code size expansion factor when copying basic blocks
8737 instead of jumping. The expansion is relative to a jump instruction.
8738 The default value is 8.
8740 @item max-goto-duplication-insns
8741 The maximum number of instructions to duplicate to a block that jumps
8742 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8743 passes, GCC factors computed gotos early in the compilation process,
8744 and unfactors them as late as possible. Only computed jumps at the
8745 end of a basic blocks with no more than max-goto-duplication-insns are
8746 unfactored. The default value is 8.
8748 @item max-delay-slot-insn-search
8749 The maximum number of instructions to consider when looking for an
8750 instruction to fill a delay slot. If more than this arbitrary number of
8751 instructions are searched, the time savings from filling the delay slot
8752 are minimal, so stop searching. Increasing values mean more
8753 aggressive optimization, making the compilation time increase with probably
8754 small improvement in execution time.
8756 @item max-delay-slot-live-search
8757 When trying to fill delay slots, the maximum number of instructions to
8758 consider when searching for a block with valid live register
8759 information. Increasing this arbitrarily chosen value means more
8760 aggressive optimization, increasing the compilation time. This parameter
8761 should be removed when the delay slot code is rewritten to maintain the
8764 @item max-gcse-memory
8765 The approximate maximum amount of memory that can be allocated in
8766 order to perform the global common subexpression elimination
8767 optimization. If more memory than specified is required, the
8768 optimization is not done.
8770 @item max-gcse-insertion-ratio
8771 If the ratio of expression insertions to deletions is larger than this value
8772 for any expression, then RTL PRE inserts or removes the expression and thus
8773 leaves partially redundant computations in the instruction stream. The default value is 20.
8775 @item max-pending-list-length
8776 The maximum number of pending dependencies scheduling allows
8777 before flushing the current state and starting over. Large functions
8778 with few branches or calls can create excessively large lists which
8779 needlessly consume memory and resources.
8781 @item max-modulo-backtrack-attempts
8782 The maximum number of backtrack attempts the scheduler should make
8783 when modulo scheduling a loop. Larger values can exponentially increase
8786 @item max-inline-insns-single
8787 Several parameters control the tree inliner used in GCC@.
8788 This number sets the maximum number of instructions (counted in GCC's
8789 internal representation) in a single function that the tree inliner
8790 considers for inlining. This only affects functions declared
8791 inline and methods implemented in a class declaration (C++).
8792 The default value is 400.
8794 @item max-inline-insns-auto
8795 When you use @option{-finline-functions} (included in @option{-O3}),
8796 a lot of functions that would otherwise not be considered for inlining
8797 by the compiler are investigated. To those functions, a different
8798 (more restrictive) limit compared to functions declared inline can
8800 The default value is 40.
8802 @item inline-min-speedup
8803 When estimated performance improvement of caller + callee runtime exceeds this
8804 threshold (in precent), the function can be inlined regardless the limit on
8805 @option{--param max-inline-insns-single} and @option{--param
8806 max-inline-insns-auto}.
8808 @item large-function-insns
8809 The limit specifying really large functions. For functions larger than this
8810 limit after inlining, inlining is constrained by
8811 @option{--param large-function-growth}. This parameter is useful primarily
8812 to avoid extreme compilation time caused by non-linear algorithms used by the
8814 The default value is 2700.
8816 @item large-function-growth
8817 Specifies maximal growth of large function caused by inlining in percents.
8818 The default value is 100 which limits large function growth to 2.0 times
8821 @item large-unit-insns
8822 The limit specifying large translation unit. Growth caused by inlining of
8823 units larger than this limit is limited by @option{--param inline-unit-growth}.
8824 For small units this might be too tight.
8825 For example, consider a unit consisting of function A
8826 that is inline and B that just calls A three times. If B is small relative to
8827 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8828 large units consisting of small inlineable functions, however, the overall unit
8829 growth limit is needed to avoid exponential explosion of code size. Thus for
8830 smaller units, the size is increased to @option{--param large-unit-insns}
8831 before applying @option{--param inline-unit-growth}. The default is 10000.
8833 @item inline-unit-growth
8834 Specifies maximal overall growth of the compilation unit caused by inlining.
8835 The default value is 20 which limits unit growth to 1.2 times the original
8836 size. Cold functions (either marked cold via an attribute or by profile
8837 feedback) are not accounted into the unit size.
8839 @item ipcp-unit-growth
8840 Specifies maximal overall growth of the compilation unit caused by
8841 interprocedural constant propagation. The default value is 10 which limits
8842 unit growth to 1.1 times the original size.
8844 @item large-stack-frame
8845 The limit specifying large stack frames. While inlining the algorithm is trying
8846 to not grow past this limit too much. The default value is 256 bytes.
8848 @item large-stack-frame-growth
8849 Specifies maximal growth of large stack frames caused by inlining in percents.
8850 The default value is 1000 which limits large stack frame growth to 11 times
8853 @item max-inline-insns-recursive
8854 @itemx max-inline-insns-recursive-auto
8855 Specifies the maximum number of instructions an out-of-line copy of a
8856 self-recursive inline
8857 function can grow into by performing recursive inlining.
8859 @option{--param max-inline-insns-recursive} applies to functions
8861 For functions not declared inline, recursive inlining
8862 happens only when @option{-finline-functions} (included in @option{-O3}) is
8863 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
8864 default value is 450.
8866 @item max-inline-recursive-depth
8867 @itemx max-inline-recursive-depth-auto
8868 Specifies the maximum recursion depth used for recursive inlining.
8870 @option{--param max-inline-recursive-depth} applies to functions
8871 declared inline. For functions not declared inline, recursive inlining
8872 happens only when @option{-finline-functions} (included in @option{-O3}) is
8873 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
8876 @item min-inline-recursive-probability
8877 Recursive inlining is profitable only for function having deep recursion
8878 in average and can hurt for function having little recursion depth by
8879 increasing the prologue size or complexity of function body to other
8882 When profile feedback is available (see @option{-fprofile-generate}) the actual
8883 recursion depth can be guessed from probability that function recurses via a
8884 given call expression. This parameter limits inlining only to call expressions
8885 whose probability exceeds the given threshold (in percents).
8886 The default value is 10.
8888 @item early-inlining-insns
8889 Specify growth that the early inliner can make. In effect it increases
8890 the amount of inlining for code having a large abstraction penalty.
8891 The default value is 14.
8893 @item max-early-inliner-iterations
8894 Limit of iterations of the early inliner. This basically bounds
8895 the number of nested indirect calls the early inliner can resolve.
8896 Deeper chains are still handled by late inlining.
8898 @item comdat-sharing-probability
8899 Probability (in percent) that C++ inline function with comdat visibility
8900 are shared across multiple compilation units. The default value is 20.
8902 @item profile-func-internal-id
8903 A parameter to control whether to use function internal id in profile
8904 database lookup. If the value is 0, the compiler uses an id that
8905 is based on function assembler name and filename, which makes old profile
8906 data more tolerant to source changes such as function reordering etc.
8907 The default value is 0.
8909 @item min-vect-loop-bound
8910 The minimum number of iterations under which loops are not vectorized
8911 when @option{-ftree-vectorize} is used. The number of iterations after
8912 vectorization needs to be greater than the value specified by this option
8913 to allow vectorization. The default value is 0.
8915 @item gcse-cost-distance-ratio
8916 Scaling factor in calculation of maximum distance an expression
8917 can be moved by GCSE optimizations. This is currently supported only in the
8918 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8919 is with simple expressions, i.e., the expressions that have cost
8920 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
8921 hoisting of simple expressions. The default value is 10.
8923 @item gcse-unrestricted-cost
8924 Cost, roughly measured as the cost of a single typical machine
8925 instruction, at which GCSE optimizations do not constrain
8926 the distance an expression can travel. This is currently
8927 supported only in the code hoisting pass. The lesser the cost,
8928 the more aggressive code hoisting is. Specifying 0
8929 allows all expressions to travel unrestricted distances.
8930 The default value is 3.
8932 @item max-hoist-depth
8933 The depth of search in the dominator tree for expressions to hoist.
8934 This is used to avoid quadratic behavior in hoisting algorithm.
8935 The value of 0 does not limit on the search, but may slow down compilation
8936 of huge functions. The default value is 30.
8938 @item max-tail-merge-comparisons
8939 The maximum amount of similar bbs to compare a bb with. This is used to
8940 avoid quadratic behavior in tree tail merging. The default value is 10.
8942 @item max-tail-merge-iterations
8943 The maximum amount of iterations of the pass over the function. This is used to
8944 limit compilation time in tree tail merging. The default value is 2.
8946 @item max-unrolled-insns
8947 The maximum number of instructions that a loop may have to be unrolled.
8948 If a loop is unrolled, this parameter also determines how many times
8949 the loop code is unrolled.
8951 @item max-average-unrolled-insns
8952 The maximum number of instructions biased by probabilities of their execution
8953 that a loop may have to be unrolled. If a loop is unrolled,
8954 this parameter also determines how many times the loop code is unrolled.
8956 @item max-unroll-times
8957 The maximum number of unrollings of a single loop.
8959 @item max-peeled-insns
8960 The maximum number of instructions that a loop may have to be peeled.
8961 If a loop is peeled, this parameter also determines how many times
8962 the loop code is peeled.
8964 @item max-peel-times
8965 The maximum number of peelings of a single loop.
8967 @item max-peel-branches
8968 The maximum number of branches on the hot path through the peeled sequence.
8970 @item max-completely-peeled-insns
8971 The maximum number of insns of a completely peeled loop.
8973 @item max-completely-peel-times
8974 The maximum number of iterations of a loop to be suitable for complete peeling.
8976 @item max-completely-peel-loop-nest-depth
8977 The maximum depth of a loop nest suitable for complete peeling.
8979 @item max-unswitch-insns
8980 The maximum number of insns of an unswitched loop.
8982 @item max-unswitch-level
8983 The maximum number of branches unswitched in a single loop.
8986 The minimum cost of an expensive expression in the loop invariant motion.
8988 @item iv-consider-all-candidates-bound
8989 Bound on number of candidates for induction variables, below which
8990 all candidates are considered for each use in induction variable
8991 optimizations. If there are more candidates than this,
8992 only the most relevant ones are considered to avoid quadratic time complexity.
8994 @item iv-max-considered-uses
8995 The induction variable optimizations give up on loops that contain more
8996 induction variable uses.
8998 @item iv-always-prune-cand-set-bound
8999 If the number of candidates in the set is smaller than this value,
9000 always try to remove unnecessary ivs from the set
9001 when adding a new one.
9003 @item scev-max-expr-size
9004 Bound on size of expressions used in the scalar evolutions analyzer.
9005 Large expressions slow the analyzer.
9007 @item scev-max-expr-complexity
9008 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9009 Complex expressions slow the analyzer.
9011 @item vect-max-version-for-alignment-checks
9012 The maximum number of run-time checks that can be performed when
9013 doing loop versioning for alignment in the vectorizer.
9015 @item vect-max-version-for-alias-checks
9016 The maximum number of run-time checks that can be performed when
9017 doing loop versioning for alias in the vectorizer.
9019 @item vect-max-peeling-for-alignment
9020 The maximum number of loop peels to enhance access alignment
9021 for vectorizer. Value -1 means no limit.
9023 @item max-iterations-to-track
9024 The maximum number of iterations of a loop the brute-force algorithm
9025 for analysis of the number of iterations of the loop tries to evaluate.
9027 @item hot-bb-count-ws-permille
9028 A basic block profile count is considered hot if it contributes to
9029 the given permillage (i.e. 0...1000) of the entire profiled execution.
9031 @item hot-bb-frequency-fraction
9032 Select fraction of the entry block frequency of executions of basic block in
9033 function given basic block needs to have to be considered hot.
9035 @item max-predicted-iterations
9036 The maximum number of loop iterations we predict statically. This is useful
9037 in cases where a function contains a single loop with known bound and
9038 another loop with unknown bound.
9039 The known number of iterations is predicted correctly, while
9040 the unknown number of iterations average to roughly 10. This means that the
9041 loop without bounds appears artificially cold relative to the other one.
9043 @item builtin-expect-probability
9044 Control the probability of the expression having the specified value. This
9045 parameter takes a percentage (i.e. 0 ... 100) as input.
9046 The default probability of 90 is obtained empirically.
9048 @item align-threshold
9050 Select fraction of the maximal frequency of executions of a basic block in
9051 a function to align the basic block.
9053 @item align-loop-iterations
9055 A loop expected to iterate at least the selected number of iterations is
9058 @item tracer-dynamic-coverage
9059 @itemx tracer-dynamic-coverage-feedback
9061 This value is used to limit superblock formation once the given percentage of
9062 executed instructions is covered. This limits unnecessary code size
9065 The @option{tracer-dynamic-coverage-feedback} parameter
9066 is used only when profile
9067 feedback is available. The real profiles (as opposed to statically estimated
9068 ones) are much less balanced allowing the threshold to be larger value.
9070 @item tracer-max-code-growth
9071 Stop tail duplication once code growth has reached given percentage. This is
9072 a rather artificial limit, as most of the duplicates are eliminated later in
9073 cross jumping, so it may be set to much higher values than is the desired code
9076 @item tracer-min-branch-ratio
9078 Stop reverse growth when the reverse probability of best edge is less than this
9079 threshold (in percent).
9081 @item tracer-min-branch-probability
9082 @itemx tracer-min-branch-probability-feedback
9084 Stop forward growth if the best edge has probability lower than this
9087 Similarly to @option{tracer-dynamic-coverage} two parameters are
9088 provided. @option{tracer-min-branch-probability-feedback} is used for
9089 compilation with profile feedback and @option{tracer-min-branch-probability}
9090 compilation without. The value for compilation with profile feedback
9091 needs to be more conservative (higher) in order to make tracer
9094 @item max-cse-path-length
9096 The maximum number of basic blocks on path that CSE considers.
9100 The maximum number of instructions CSE processes before flushing.
9101 The default is 1000.
9103 @item ggc-min-expand
9105 GCC uses a garbage collector to manage its own memory allocation. This
9106 parameter specifies the minimum percentage by which the garbage
9107 collector's heap should be allowed to expand between collections.
9108 Tuning this may improve compilation speed; it has no effect on code
9111 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9112 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9113 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9114 GCC is not able to calculate RAM on a particular platform, the lower
9115 bound of 30% is used. Setting this parameter and
9116 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9117 every opportunity. This is extremely slow, but can be useful for
9120 @item ggc-min-heapsize
9122 Minimum size of the garbage collector's heap before it begins bothering
9123 to collect garbage. The first collection occurs after the heap expands
9124 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9125 tuning this may improve compilation speed, and has no effect on code
9128 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9129 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9130 with a lower bound of 4096 (four megabytes) and an upper bound of
9131 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9132 particular platform, the lower bound is used. Setting this parameter
9133 very large effectively disables garbage collection. Setting this
9134 parameter and @option{ggc-min-expand} to zero causes a full collection
9135 to occur at every opportunity.
9137 @item max-reload-search-insns
9138 The maximum number of instruction reload should look backward for equivalent
9139 register. Increasing values mean more aggressive optimization, making the
9140 compilation time increase with probably slightly better performance.
9141 The default value is 100.
9143 @item max-cselib-memory-locations
9144 The maximum number of memory locations cselib should take into account.
9145 Increasing values mean more aggressive optimization, making the compilation time
9146 increase with probably slightly better performance. The default value is 500.
9148 @item max-sched-ready-insns
9149 The maximum number of instructions ready to be issued the scheduler should
9150 consider at any given time during the first scheduling pass. Increasing
9151 values mean more thorough searches, making the compilation time increase
9152 with probably little benefit. The default value is 100.
9154 @item max-sched-region-blocks
9155 The maximum number of blocks in a region to be considered for
9156 interblock scheduling. The default value is 10.
9158 @item max-pipeline-region-blocks
9159 The maximum number of blocks in a region to be considered for
9160 pipelining in the selective scheduler. The default value is 15.
9162 @item max-sched-region-insns
9163 The maximum number of insns in a region to be considered for
9164 interblock scheduling. The default value is 100.
9166 @item max-pipeline-region-insns
9167 The maximum number of insns in a region to be considered for
9168 pipelining in the selective scheduler. The default value is 200.
9171 The minimum probability (in percents) of reaching a source block
9172 for interblock speculative scheduling. The default value is 40.
9174 @item max-sched-extend-regions-iters
9175 The maximum number of iterations through CFG to extend regions.
9176 A value of 0 (the default) disables region extensions.
9178 @item max-sched-insn-conflict-delay
9179 The maximum conflict delay for an insn to be considered for speculative motion.
9180 The default value is 3.
9182 @item sched-spec-prob-cutoff
9183 The minimal probability of speculation success (in percents), so that
9184 speculative insns are scheduled.
9185 The default value is 40.
9187 @item sched-state-edge-prob-cutoff
9188 The minimum probability an edge must have for the scheduler to save its
9190 The default value is 10.
9192 @item sched-mem-true-dep-cost
9193 Minimal distance (in CPU cycles) between store and load targeting same
9194 memory locations. The default value is 1.
9196 @item selsched-max-lookahead
9197 The maximum size of the lookahead window of selective scheduling. It is a
9198 depth of search for available instructions.
9199 The default value is 50.
9201 @item selsched-max-sched-times
9202 The maximum number of times that an instruction is scheduled during
9203 selective scheduling. This is the limit on the number of iterations
9204 through which the instruction may be pipelined. The default value is 2.
9206 @item selsched-insns-to-rename
9207 The maximum number of best instructions in the ready list that are considered
9208 for renaming in the selective scheduler. The default value is 2.
9211 The minimum value of stage count that swing modulo scheduler
9212 generates. The default value is 2.
9214 @item max-last-value-rtl
9215 The maximum size measured as number of RTLs that can be recorded in an expression
9216 in combiner for a pseudo register as last known value of that register. The default
9219 @item max-combine-insns
9220 The maximum number of instructions the RTL combiner tries to combine.
9221 The default value is 2 at @option{-Og} and 4 otherwise.
9223 @item integer-share-limit
9224 Small integer constants can use a shared data structure, reducing the
9225 compiler's memory usage and increasing its speed. This sets the maximum
9226 value of a shared integer constant. The default value is 256.
9228 @item ssp-buffer-size
9229 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9230 protection when @option{-fstack-protection} is used.
9232 @item min-size-for-stack-sharing
9233 The minimum size of variables taking part in stack slot sharing when not
9234 optimizing. The default value is 32.
9236 @item max-jump-thread-duplication-stmts
9237 Maximum number of statements allowed in a block that needs to be
9238 duplicated when threading jumps.
9240 @item max-fields-for-field-sensitive
9241 Maximum number of fields in a structure treated in
9242 a field sensitive manner during pointer analysis. The default is zero
9243 for @option{-O0} and @option{-O1},
9244 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9246 @item prefetch-latency
9247 Estimate on average number of instructions that are executed before
9248 prefetch finishes. The distance prefetched ahead is proportional
9249 to this constant. Increasing this number may also lead to less
9250 streams being prefetched (see @option{simultaneous-prefetches}).
9252 @item simultaneous-prefetches
9253 Maximum number of prefetches that can run at the same time.
9255 @item l1-cache-line-size
9256 The size of cache line in L1 cache, in bytes.
9259 The size of L1 cache, in kilobytes.
9262 The size of L2 cache, in kilobytes.
9264 @item min-insn-to-prefetch-ratio
9265 The minimum ratio between the number of instructions and the
9266 number of prefetches to enable prefetching in a loop.
9268 @item prefetch-min-insn-to-mem-ratio
9269 The minimum ratio between the number of instructions and the
9270 number of memory references to enable prefetching in a loop.
9272 @item use-canonical-types
9273 Whether the compiler should use the ``canonical'' type system. By
9274 default, this should always be 1, which uses a more efficient internal
9275 mechanism for comparing types in C++ and Objective-C++. However, if
9276 bugs in the canonical type system are causing compilation failures,
9277 set this value to 0 to disable canonical types.
9279 @item switch-conversion-max-branch-ratio
9280 Switch initialization conversion refuses to create arrays that are
9281 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9282 branches in the switch.
9284 @item max-partial-antic-length
9285 Maximum length of the partial antic set computed during the tree
9286 partial redundancy elimination optimization (@option{-ftree-pre}) when
9287 optimizing at @option{-O3} and above. For some sorts of source code
9288 the enhanced partial redundancy elimination optimization can run away,
9289 consuming all of the memory available on the host machine. This
9290 parameter sets a limit on the length of the sets that are computed,
9291 which prevents the runaway behavior. Setting a value of 0 for
9292 this parameter allows an unlimited set length.
9294 @item sccvn-max-scc-size
9295 Maximum size of a strongly connected component (SCC) during SCCVN
9296 processing. If this limit is hit, SCCVN processing for the whole
9297 function is not done and optimizations depending on it are
9298 disabled. The default maximum SCC size is 10000.
9300 @item sccvn-max-alias-queries-per-access
9301 Maximum number of alias-oracle queries we perform when looking for
9302 redundancies for loads and stores. If this limit is hit the search
9303 is aborted and the load or store is not considered redundant. The
9304 number of queries is algorithmically limited to the number of
9305 stores on all paths from the load to the function entry.
9306 The default maximum number of queries is 1000.
9308 @item ira-max-loops-num
9309 IRA uses regional register allocation by default. If a function
9310 contains more loops than the number given by this parameter, only at most
9311 the given number of the most frequently-executed loops form regions
9312 for regional register allocation. The default value of the
9315 @item ira-max-conflict-table-size
9316 Although IRA uses a sophisticated algorithm to compress the conflict
9317 table, the table can still require excessive amounts of memory for
9318 huge functions. If the conflict table for a function could be more
9319 than the size in MB given by this parameter, the register allocator
9320 instead uses a faster, simpler, and lower-quality
9321 algorithm that does not require building a pseudo-register conflict table.
9322 The default value of the parameter is 2000.
9324 @item ira-loop-reserved-regs
9325 IRA can be used to evaluate more accurate register pressure in loops
9326 for decisions to move loop invariants (see @option{-O3}). The number
9327 of available registers reserved for some other purposes is given
9328 by this parameter. The default value of the parameter is 2, which is
9329 the minimal number of registers needed by typical instructions.
9330 This value is the best found from numerous experiments.
9332 @item lra-inheritance-ebb-probability-cutoff
9333 LRA tries to reuse values reloaded in registers in subsequent insns.
9334 This optimization is called inheritance. EBB is used as a region to
9335 do this optimization. The parameter defines a minimal fall-through
9336 edge probability in percentage used to add BB to inheritance EBB in
9337 LRA. The default value of the parameter is 40. The value was chosen
9338 from numerous runs of SPEC2000 on x86-64.
9340 @item loop-invariant-max-bbs-in-loop
9341 Loop invariant motion can be very expensive, both in compilation time and
9342 in amount of needed compile-time memory, with very large loops. Loops
9343 with more basic blocks than this parameter won't have loop invariant
9344 motion optimization performed on them. The default value of the
9345 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
9347 @item loop-max-datarefs-for-datadeps
9348 Building data dependencies is expensive for very large loops. This
9349 parameter limits the number of data references in loops that are
9350 considered for data dependence analysis. These large loops are no
9351 handled by the optimizations using loop data dependencies.
9352 The default value is 1000.
9354 @item max-vartrack-size
9355 Sets a maximum number of hash table slots to use during variable
9356 tracking dataflow analysis of any function. If this limit is exceeded
9357 with variable tracking at assignments enabled, analysis for that
9358 function is retried without it, after removing all debug insns from
9359 the function. If the limit is exceeded even without debug insns, var
9360 tracking analysis is completely disabled for the function. Setting
9361 the parameter to zero makes it unlimited.
9363 @item max-vartrack-expr-depth
9364 Sets a maximum number of recursion levels when attempting to map
9365 variable names or debug temporaries to value expressions. This trades
9366 compilation time for more complete debug information. If this is set too
9367 low, value expressions that are available and could be represented in
9368 debug information may end up not being used; setting this higher may
9369 enable the compiler to find more complex debug expressions, but compile
9370 time and memory use may grow. The default is 12.
9372 @item min-nondebug-insn-uid
9373 Use uids starting at this parameter for nondebug insns. The range below
9374 the parameter is reserved exclusively for debug insns created by
9375 @option{-fvar-tracking-assignments}, but debug insns may get
9376 (non-overlapping) uids above it if the reserved range is exhausted.
9378 @item ipa-sra-ptr-growth-factor
9379 IPA-SRA replaces a pointer to an aggregate with one or more new
9380 parameters only when their cumulative size is less or equal to
9381 @option{ipa-sra-ptr-growth-factor} times the size of the original
9384 @item sra-max-scalarization-size-Ospeed
9385 @item sra-max-scalarization-size-Osize
9386 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
9387 replace scalar parts of aggregates with uses of independent scalar
9388 variables. These parameters control the maximum size, in storage units,
9389 of aggregate which is considered for replacement when compiling for
9391 (@option{sra-max-scalarization-size-Ospeed}) or size
9392 (@option{sra-max-scalarization-size-Osize}) respectively.
9394 @item tm-max-aggregate-size
9395 When making copies of thread-local variables in a transaction, this
9396 parameter specifies the size in bytes after which variables are
9397 saved with the logging functions as opposed to save/restore code
9398 sequence pairs. This option only applies when using
9401 @item graphite-max-nb-scop-params
9402 To avoid exponential effects in the Graphite loop transforms, the
9403 number of parameters in a Static Control Part (SCoP) is bounded. The
9404 default value is 10 parameters. A variable whose value is unknown at
9405 compilation time and defined outside a SCoP is a parameter of the SCoP.
9407 @item graphite-max-bbs-per-function
9408 To avoid exponential effects in the detection of SCoPs, the size of
9409 the functions analyzed by Graphite is bounded. The default value is
9412 @item loop-block-tile-size
9413 Loop blocking or strip mining transforms, enabled with
9414 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9415 loop in the loop nest by a given number of iterations. The strip
9416 length can be changed using the @option{loop-block-tile-size}
9417 parameter. The default value is 51 iterations.
9419 @item loop-unroll-jam-size
9420 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
9423 @item loop-unroll-jam-depth
9424 Specify the dimension to be unrolled (counting from the most inner loop)
9425 for the @option{-floop-unroll-and-jam}. The default value is 2.
9427 @item ipa-cp-value-list-size
9428 IPA-CP attempts to track all possible values and types passed to a function's
9429 parameter in order to propagate them and perform devirtualization.
9430 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9431 stores per one formal parameter of a function.
9433 @item ipa-cp-eval-threshold
9434 IPA-CP calculates its own score of cloning profitability heuristics
9435 and performs those cloning opportunities with scores that exceed
9436 @option{ipa-cp-eval-threshold}.
9438 @item ipa-cp-recursion-penalty
9439 Percentage penalty the recursive functions will receive when they
9440 are evaluated for cloning.
9442 @item ipa-cp-single-call-penalty
9443 Percentage penalty functions containg a single call to another
9444 function will receive when they are evaluated for cloning.
9447 @item ipa-max-agg-items
9448 IPA-CP is also capable to propagate a number of scalar values passed
9449 in an aggregate. @option{ipa-max-agg-items} controls the maximum
9450 number of such values per one parameter.
9452 @item ipa-cp-loop-hint-bonus
9453 When IPA-CP determines that a cloning candidate would make the number
9454 of iterations of a loop known, it adds a bonus of
9455 @option{ipa-cp-loop-hint-bonus} to the profitability score of
9458 @item ipa-cp-array-index-hint-bonus
9459 When IPA-CP determines that a cloning candidate would make the index of
9460 an array access known, it adds a bonus of
9461 @option{ipa-cp-array-index-hint-bonus} to the profitability
9462 score of the candidate.
9464 @item ipa-max-aa-steps
9465 During its analysis of function bodies, IPA-CP employs alias analysis
9466 in order to track values pointed to by function parameters. In order
9467 not spend too much time analyzing huge functions, it gives up and
9468 consider all memory clobbered after examining
9469 @option{ipa-max-aa-steps} statements modifying memory.
9471 @item lto-partitions
9472 Specify desired number of partitions produced during WHOPR compilation.
9473 The number of partitions should exceed the number of CPUs used for compilation.
9474 The default value is 32.
9476 @item lto-min-partition
9477 Size of minimal partition for WHOPR (in estimated instructions).
9478 This prevents expenses of splitting very small programs into too many
9481 @item cxx-max-namespaces-for-diagnostic-help
9482 The maximum number of namespaces to consult for suggestions when C++
9483 name lookup fails for an identifier. The default is 1000.
9485 @item sink-frequency-threshold
9486 The maximum relative execution frequency (in percents) of the target block
9487 relative to a statement's original block to allow statement sinking of a
9488 statement. Larger numbers result in more aggressive statement sinking.
9489 The default value is 75. A small positive adjustment is applied for
9490 statements with memory operands as those are even more profitable so sink.
9492 @item max-stores-to-sink
9493 The maximum number of conditional store pairs that can be sunk. Set to 0
9494 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9495 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9497 @item allow-store-data-races
9498 Allow optimizers to introduce new data races on stores.
9499 Set to 1 to allow, otherwise to 0. This option is enabled by default
9500 at optimization level @option{-Ofast}.
9502 @item case-values-threshold
9503 The smallest number of different values for which it is best to use a
9504 jump-table instead of a tree of conditional branches. If the value is
9505 0, use the default for the machine. The default is 0.
9507 @item tree-reassoc-width
9508 Set the maximum number of instructions executed in parallel in
9509 reassociated tree. This parameter overrides target dependent
9510 heuristics used by default if has non zero value.
9512 @item sched-pressure-algorithm
9513 Choose between the two available implementations of
9514 @option{-fsched-pressure}. Algorithm 1 is the original implementation
9515 and is the more likely to prevent instructions from being reordered.
9516 Algorithm 2 was designed to be a compromise between the relatively
9517 conservative approach taken by algorithm 1 and the rather aggressive
9518 approach taken by the default scheduler. It relies more heavily on
9519 having a regular register file and accurate register pressure classes.
9520 See @file{haifa-sched.c} in the GCC sources for more details.
9522 The default choice depends on the target.
9524 @item max-slsr-cand-scan
9525 Set the maximum number of existing candidates that are considered when
9526 seeking a basis for a new straight-line strength reduction candidate.
9529 Enable buffer overflow detection for global objects. This kind
9530 of protection is enabled by default if you are using
9531 @option{-fsanitize=address} option.
9532 To disable global objects protection use @option{--param asan-globals=0}.
9535 Enable buffer overflow detection for stack objects. This kind of
9536 protection is enabled by default when using @option{-fsanitize=address}.
9537 To disable stack protection use @option{--param asan-stack=0} option.
9539 @item asan-instrument-reads
9540 Enable buffer overflow detection for memory reads. This kind of
9541 protection is enabled by default when using @option{-fsanitize=address}.
9542 To disable memory reads protection use
9543 @option{--param asan-instrument-reads=0}.
9545 @item asan-instrument-writes
9546 Enable buffer overflow detection for memory writes. This kind of
9547 protection is enabled by default when using @option{-fsanitize=address}.
9548 To disable memory writes protection use
9549 @option{--param asan-instrument-writes=0} option.
9551 @item asan-memintrin
9552 Enable detection for built-in functions. This kind of protection
9553 is enabled by default when using @option{-fsanitize=address}.
9554 To disable built-in functions protection use
9555 @option{--param asan-memintrin=0}.
9557 @item asan-use-after-return
9558 Enable detection of use-after-return. This kind of protection
9559 is enabled by default when using @option{-fsanitize=address} option.
9560 To disable use-after-return detection use
9561 @option{--param asan-use-after-return=0}.
9563 @item asan-instrumentation-with-call-threshold
9564 If number of memory accesses in function being instrumented
9565 is greater or equal to this number, use callbacks instead of inline checks.
9566 E.g. to disable inline code use
9567 @option{--param asan-instrumentation-with-call-threshold=0}.
9569 @item chkp-max-ctor-size
9570 Static constructors generated by Pointer Bounds Checker may become very
9571 large and significantly increase compile time at optimization level
9572 @option{-O1} and higher. This parameter is a maximum nubmer of statements
9573 in a single generated constructor. Default value is 5000.
9575 @item max-fsm-thread-path-insns
9576 Maximum number of instructions to copy when duplicating blocks on a
9577 finite state automaton jump thread path. The default is 100.
9579 @item max-fsm-thread-length
9580 Maximum number of basic blocks on a finite state automaton jump thread
9581 path. The default is 10.
9583 @item max-fsm-thread-paths
9584 Maximum number of new jump thread paths to create for a finite state
9585 automaton. The default is 50.
9587 @item parloops-chunk-size
9588 Chunk size of omp schedule for loops parallelized by parloops. The default
9591 @item parloops-schedule
9592 Schedule type of omp schedule for loops parallelized by parloops (static,
9593 dynamic, guided, auto, runtime). The default is static.
9595 @item max-ssa-name-query-depth
9596 Maximum depth of recursion when querying properties of SSA names in things
9597 like fold routines. One level of recursion corresponds to following a
9600 @item hsa-gen-debug-stores
9601 Enable emission of special debug stores within HSA kernels which are
9602 then read and reported by libgomp plugin. Generation of these stores
9603 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
9606 @item max-speculative-devirt-maydefs
9607 The maximum number of may-defs we analyze when looking for a must-def
9608 specifying the dynamic type of an object that invokes a virtual call
9609 we may be able to devirtualize speculatively.
9613 @node Instrumentation Options
9614 @section Program Instrumentation Options
9615 @cindex instrumentation options
9616 @cindex program instrumentation options
9617 @cindex run-time error checking options
9618 @cindex profiling options
9619 @cindex options, program instrumentation
9620 @cindex options, run-time error checking
9621 @cindex options, profiling
9623 GCC supports a number of command-line options that control adding
9624 run-time instrumentation to the code it normally generates.
9625 For example, one purpose of instrumentation is collect profiling
9626 statistics for use in finding program hot spots, code coverage
9627 analysis, or profile-guided optimizations.
9628 Another class of program instrumentation is adding run-time checking
9629 to detect programming errors like invalid pointer
9630 dereferences or out-of-bounds array accesses, as well as deliberately
9631 hostile attacks such as stack smashing or C++ vtable hijacking.
9632 There is also a general hook which can be used to implement other
9633 forms of tracing or function-level instrumentation for debug or
9634 program analysis purposes.
9637 @cindex @command{prof}
9640 Generate extra code to write profile information suitable for the
9641 analysis program @command{prof}. You must use this option when compiling
9642 the source files you want data about, and you must also use it when
9645 @cindex @command{gprof}
9648 Generate extra code to write profile information suitable for the
9649 analysis program @command{gprof}. You must use this option when compiling
9650 the source files you want data about, and you must also use it when
9653 @item -fprofile-arcs
9654 @opindex fprofile-arcs
9655 Add code so that program flow @dfn{arcs} are instrumented. During
9656 execution the program records how many times each branch and call is
9657 executed and how many times it is taken or returns. When the compiled
9658 program exits it saves this data to a file called
9659 @file{@var{auxname}.gcda} for each source file. The data may be used for
9660 profile-directed optimizations (@option{-fbranch-probabilities}), or for
9661 test coverage analysis (@option{-ftest-coverage}). Each object file's
9662 @var{auxname} is generated from the name of the output file, if
9663 explicitly specified and it is not the final executable, otherwise it is
9664 the basename of the source file. In both cases any suffix is removed
9665 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
9666 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
9667 @xref{Cross-profiling}.
9669 @cindex @command{gcov}
9673 This option is used to compile and link code instrumented for coverage
9674 analysis. The option is a synonym for @option{-fprofile-arcs}
9675 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
9676 linking). See the documentation for those options for more details.
9681 Compile the source files with @option{-fprofile-arcs} plus optimization
9682 and code generation options. For test coverage analysis, use the
9683 additional @option{-ftest-coverage} option. You do not need to profile
9684 every source file in a program.
9687 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
9688 (the latter implies the former).
9691 Run the program on a representative workload to generate the arc profile
9692 information. This may be repeated any number of times. You can run
9693 concurrent instances of your program, and provided that the file system
9694 supports locking, the data files will be correctly updated. Also
9695 @code{fork} calls are detected and correctly handled (double counting
9699 For profile-directed optimizations, compile the source files again with
9700 the same optimization and code generation options plus
9701 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
9702 Control Optimization}).
9705 For test coverage analysis, use @command{gcov} to produce human readable
9706 information from the @file{.gcno} and @file{.gcda} files. Refer to the
9707 @command{gcov} documentation for further information.
9711 With @option{-fprofile-arcs}, for each function of your program GCC
9712 creates a program flow graph, then finds a spanning tree for the graph.
9713 Only arcs that are not on the spanning tree have to be instrumented: the
9714 compiler adds code to count the number of times that these arcs are
9715 executed. When an arc is the only exit or only entrance to a block, the
9716 instrumentation code can be added to the block; otherwise, a new basic
9717 block must be created to hold the instrumentation code.
9720 @item -ftest-coverage
9721 @opindex ftest-coverage
9722 Produce a notes file that the @command{gcov} code-coverage utility
9723 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
9724 show program coverage. Each source file's note file is called
9725 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
9726 above for a description of @var{auxname} and instructions on how to
9727 generate test coverage data. Coverage data matches the source files
9728 more closely if you do not optimize.
9730 @item -fprofile-dir=@var{path}
9731 @opindex fprofile-dir
9733 Set the directory to search for the profile data files in to @var{path}.
9734 This option affects only the profile data generated by
9735 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9736 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9737 and its related options. Both absolute and relative paths can be used.
9738 By default, GCC uses the current directory as @var{path}, thus the
9739 profile data file appears in the same directory as the object file.
9741 @item -fprofile-generate
9742 @itemx -fprofile-generate=@var{path}
9743 @opindex fprofile-generate
9745 Enable options usually used for instrumenting application to produce
9746 profile useful for later recompilation with profile feedback based
9747 optimization. You must use @option{-fprofile-generate} both when
9748 compiling and when linking your program.
9750 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9752 If @var{path} is specified, GCC looks at the @var{path} to find
9753 the profile feedback data files. See @option{-fprofile-dir}.
9755 To optimize the program based on the collected profile information, use
9756 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
9758 @item -fsanitize=address
9759 @opindex fsanitize=address
9760 Enable AddressSanitizer, a fast memory error detector.
9761 Memory access instructions are instrumented to detect
9762 out-of-bounds and use-after-free bugs.
9763 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
9764 more details. The run-time behavior can be influenced using the
9765 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
9766 the available options are shown at startup of the instrumented program. See
9767 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
9768 for a list of supported options.
9770 @item -fsanitize=kernel-address
9771 @opindex fsanitize=kernel-address
9772 Enable AddressSanitizer for Linux kernel.
9773 See @uref{https://github.com/google/kasan/wiki} for more details.
9775 @item -fsanitize=thread
9776 @opindex fsanitize=thread
9777 Enable ThreadSanitizer, a fast data race detector.
9778 Memory access instructions are instrumented to detect
9779 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
9780 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
9781 environment variable; see
9782 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
9785 @item -fsanitize=leak
9786 @opindex fsanitize=leak
9787 Enable LeakSanitizer, a memory leak detector.
9788 This option only matters for linking of executables and if neither
9789 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
9790 case the executable is linked against a library that overrides @code{malloc}
9791 and other allocator functions. See
9792 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
9793 details. The run-time behavior can be influenced using the
9794 @env{LSAN_OPTIONS} environment variable.
9796 @item -fsanitize=undefined
9797 @opindex fsanitize=undefined
9798 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
9799 Various computations are instrumented to detect undefined behavior
9800 at runtime. Current suboptions are:
9804 @item -fsanitize=shift
9805 @opindex fsanitize=shift
9806 This option enables checking that the result of a shift operation is
9807 not undefined. Note that what exactly is considered undefined differs
9808 slightly between C and C++, as well as between ISO C90 and C99, etc.
9810 @item -fsanitize=integer-divide-by-zero
9811 @opindex fsanitize=integer-divide-by-zero
9812 Detect integer division by zero as well as @code{INT_MIN / -1} division.
9814 @item -fsanitize=unreachable
9815 @opindex fsanitize=unreachable
9816 With this option, the compiler turns the @code{__builtin_unreachable}
9817 call into a diagnostics message call instead. When reaching the
9818 @code{__builtin_unreachable} call, the behavior is undefined.
9820 @item -fsanitize=vla-bound
9821 @opindex fsanitize=vla-bound
9822 This option instructs the compiler to check that the size of a variable
9823 length array is positive.
9825 @item -fsanitize=null
9826 @opindex fsanitize=null
9827 This option enables pointer checking. Particularly, the application
9828 built with this option turned on will issue an error message when it
9829 tries to dereference a NULL pointer, or if a reference (possibly an
9830 rvalue reference) is bound to a NULL pointer, or if a method is invoked
9831 on an object pointed by a NULL pointer.
9833 @item -fsanitize=return
9834 @opindex fsanitize=return
9835 This option enables return statement checking. Programs
9836 built with this option turned on will issue an error message
9837 when the end of a non-void function is reached without actually
9838 returning a value. This option works in C++ only.
9840 @item -fsanitize=signed-integer-overflow
9841 @opindex fsanitize=signed-integer-overflow
9842 This option enables signed integer overflow checking. We check that
9843 the result of @code{+}, @code{*}, and both unary and binary @code{-}
9844 does not overflow in the signed arithmetics. Note, integer promotion
9845 rules must be taken into account. That is, the following is not an
9848 signed char a = SCHAR_MAX;
9852 @item -fsanitize=bounds
9853 @opindex fsanitize=bounds
9854 This option enables instrumentation of array bounds. Various out of bounds
9855 accesses are detected. Flexible array members, flexible array member-like
9856 arrays, and initializers of variables with static storage are not instrumented.
9858 @item -fsanitize=bounds-strict
9859 @opindex fsanitize=bounds-strict
9860 This option enables strict instrumentation of array bounds. Most out of bounds
9861 accesses are detected, including flexible array members and flexible array
9862 member-like arrays. Initializers of variables with static storage are not
9865 @item -fsanitize=alignment
9866 @opindex fsanitize=alignment
9868 This option enables checking of alignment of pointers when they are
9869 dereferenced, or when a reference is bound to insufficiently aligned target,
9870 or when a method or constructor is invoked on insufficiently aligned object.
9872 @item -fsanitize=object-size
9873 @opindex fsanitize=object-size
9874 This option enables instrumentation of memory references using the
9875 @code{__builtin_object_size} function. Various out of bounds pointer
9876 accesses are detected.
9878 @item -fsanitize=float-divide-by-zero
9879 @opindex fsanitize=float-divide-by-zero
9880 Detect floating-point division by zero. Unlike other similar options,
9881 @option{-fsanitize=float-divide-by-zero} is not enabled by
9882 @option{-fsanitize=undefined}, since floating-point division by zero can
9883 be a legitimate way of obtaining infinities and NaNs.
9885 @item -fsanitize=float-cast-overflow
9886 @opindex fsanitize=float-cast-overflow
9887 This option enables floating-point type to integer conversion checking.
9888 We check that the result of the conversion does not overflow.
9889 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
9890 not enabled by @option{-fsanitize=undefined}.
9891 This option does not work well with @code{FE_INVALID} exceptions enabled.
9893 @item -fsanitize=nonnull-attribute
9894 @opindex fsanitize=nonnull-attribute
9896 This option enables instrumentation of calls, checking whether null values
9897 are not passed to arguments marked as requiring a non-null value by the
9898 @code{nonnull} function attribute.
9900 @item -fsanitize=returns-nonnull-attribute
9901 @opindex fsanitize=returns-nonnull-attribute
9903 This option enables instrumentation of return statements in functions
9904 marked with @code{returns_nonnull} function attribute, to detect returning
9905 of null values from such functions.
9907 @item -fsanitize=bool
9908 @opindex fsanitize=bool
9910 This option enables instrumentation of loads from bool. If a value other
9911 than 0/1 is loaded, a run-time error is issued.
9913 @item -fsanitize=enum
9914 @opindex fsanitize=enum
9916 This option enables instrumentation of loads from an enum type. If
9917 a value outside the range of values for the enum type is loaded,
9918 a run-time error is issued.
9920 @item -fsanitize=vptr
9921 @opindex fsanitize=vptr
9923 This option enables instrumentation of C++ member function calls, member
9924 accesses and some conversions between pointers to base and derived classes,
9925 to verify the referenced object has the correct dynamic type.
9929 While @option{-ftrapv} causes traps for signed overflows to be emitted,
9930 @option{-fsanitize=undefined} gives a diagnostic message.
9931 This currently works only for the C family of languages.
9933 @item -fno-sanitize=all
9934 @opindex fno-sanitize=all
9936 This option disables all previously enabled sanitizers.
9937 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
9940 @item -fasan-shadow-offset=@var{number}
9941 @opindex fasan-shadow-offset
9942 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
9943 It is useful for experimenting with different shadow memory layouts in
9944 Kernel AddressSanitizer.
9946 @item -fsanitize-sections=@var{s1},@var{s2},...
9947 @opindex fsanitize-sections
9948 Sanitize global variables in selected user-defined sections. @var{si} may
9951 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
9952 @opindex fsanitize-recover
9953 @opindex fno-sanitize-recover
9954 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
9955 mentioned in comma-separated list of @var{opts}. Enabling this option
9956 for a sanitizer component causes it to attempt to continue
9957 running the program as if no error happened. This means multiple
9958 runtime errors can be reported in a single program run, and the exit
9959 code of the program may indicate success even when errors
9960 have been reported. The @option{-fno-sanitize-recover=} option
9961 can be used to alter
9962 this behavior: only the first detected error is reported
9963 and program then exits with a non-zero exit code.
9965 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
9966 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
9967 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
9968 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
9969 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
9970 for which this feature is experimental.
9971 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
9972 accepted, the former enables recovery for all sanitizers that support it,
9973 the latter disables recovery for all sanitizers that support it.
9975 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
9977 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
9980 Similarly @option{-fno-sanitize-recover} is equivalent to
9982 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
9985 @item -fsanitize-undefined-trap-on-error
9986 @opindex fsanitize-undefined-trap-on-error
9987 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
9988 report undefined behavior using @code{__builtin_trap} rather than
9989 a @code{libubsan} library routine. The advantage of this is that the
9990 @code{libubsan} library is not needed and is not linked in, so this
9991 is usable even in freestanding environments.
9993 @item -fsanitize-coverage=trace-pc
9994 @opindex fsanitize-coverage=trace-pc
9995 Enable coverage-guided fuzzing code instrumentation.
9996 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
9998 @item -fbounds-check
9999 @opindex fbounds-check
10000 For front ends that support it, generate additional code to check that
10001 indices used to access arrays are within the declared range. This is
10002 currently only supported by the Java and Fortran front ends, where
10003 this option defaults to true and false respectively.
10005 @item -fcheck-pointer-bounds
10006 @opindex fcheck-pointer-bounds
10007 @opindex fno-check-pointer-bounds
10008 @cindex Pointer Bounds Checker options
10009 Enable Pointer Bounds Checker instrumentation. Each memory reference
10010 is instrumented with checks of the pointer used for memory access against
10011 bounds associated with that pointer.
10014 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10015 and @option{-mmpx} are required to enable this feature.
10016 MPX-based instrumentation requires
10017 a runtime library to enable MPX in hardware and handle bounds
10018 violation signals. By default when @option{-fcheck-pointer-bounds}
10019 and @option{-mmpx} options are used to link a program, the GCC driver
10020 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10021 Bounds checking on calls to dynamic libraries requires a linker
10022 with @option{-z bndplt} support; if GCC was configured with a linker
10023 without support for this option (including the Gold linker and older
10024 versions of ld), a warning is given if you link with @option{-mmpx}
10025 without also specifying @option{-static}, since the overall effectiveness
10026 of the bounds checking protection is reduced.
10027 See also @option{-static-libmpxwrappers}.
10029 MPX-based instrumentation
10030 may be used for debugging and also may be included in production code
10031 to increase program security. Depending on usage, you may
10032 have different requirements for the runtime library. The current version
10033 of the MPX runtime library is more oriented for use as a debugging
10034 tool. MPX runtime library usage implies @option{-lpthread}. See
10035 also @option{-static-libmpx}. The runtime library behavior can be
10036 influenced using various @env{CHKP_RT_*} environment variables. See
10037 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10040 Generated instrumentation may be controlled by various
10041 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10042 structure field attribute (@pxref{Type Attributes}) and
10043 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10044 (@pxref{Function Attributes}). GCC also provides a number of built-in
10045 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10046 Bounds Checker builtins}, for more information.
10048 @item -fchkp-check-incomplete-type
10049 @opindex fchkp-check-incomplete-type
10050 @opindex fno-chkp-check-incomplete-type
10051 Generate pointer bounds checks for variables with incomplete type.
10052 Enabled by default.
10054 @item -fchkp-narrow-bounds
10055 @opindex fchkp-narrow-bounds
10056 @opindex fno-chkp-narrow-bounds
10057 Controls bounds used by Pointer Bounds Checker for pointers to object
10058 fields. If narrowing is enabled then field bounds are used. Otherwise
10059 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10060 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10062 @item -fchkp-first-field-has-own-bounds
10063 @opindex fchkp-first-field-has-own-bounds
10064 @opindex fno-chkp-first-field-has-own-bounds
10065 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10066 first field in the structure. By default a pointer to the first field has
10067 the same bounds as a pointer to the whole structure.
10069 @item -fchkp-narrow-to-innermost-array
10070 @opindex fchkp-narrow-to-innermost-array
10071 @opindex fno-chkp-narrow-to-innermost-array
10072 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10073 case of nested static array access. By default this option is disabled and
10074 bounds of the outermost array are used.
10076 @item -fchkp-optimize
10077 @opindex fchkp-optimize
10078 @opindex fno-chkp-optimize
10079 Enables Pointer Bounds Checker optimizations. Enabled by default at
10080 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
10082 @item -fchkp-use-fast-string-functions
10083 @opindex fchkp-use-fast-string-functions
10084 @opindex fno-chkp-use-fast-string-functions
10085 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
10086 by Pointer Bounds Checker. Disabled by default.
10088 @item -fchkp-use-nochk-string-functions
10089 @opindex fchkp-use-nochk-string-functions
10090 @opindex fno-chkp-use-nochk-string-functions
10091 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
10092 by Pointer Bounds Checker. Disabled by default.
10094 @item -fchkp-use-static-bounds
10095 @opindex fchkp-use-static-bounds
10096 @opindex fno-chkp-use-static-bounds
10097 Allow Pointer Bounds Checker to generate static bounds holding
10098 bounds of static variables. Enabled by default.
10100 @item -fchkp-use-static-const-bounds
10101 @opindex fchkp-use-static-const-bounds
10102 @opindex fno-chkp-use-static-const-bounds
10103 Use statically-initialized bounds for constant bounds instead of
10104 generating them each time they are required. By default enabled when
10105 @option{-fchkp-use-static-bounds} is enabled.
10107 @item -fchkp-treat-zero-dynamic-size-as-infinite
10108 @opindex fchkp-treat-zero-dynamic-size-as-infinite
10109 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
10110 With this option, objects with incomplete type whose
10111 dynamically-obtained size is zero are treated as having infinite size
10112 instead by Pointer Bounds
10113 Checker. This option may be helpful if a program is linked with a library
10114 missing size information for some symbols. Disabled by default.
10116 @item -fchkp-check-read
10117 @opindex fchkp-check-read
10118 @opindex fno-chkp-check-read
10119 Instructs Pointer Bounds Checker to generate checks for all read
10120 accesses to memory. Enabled by default.
10122 @item -fchkp-check-write
10123 @opindex fchkp-check-write
10124 @opindex fno-chkp-check-write
10125 Instructs Pointer Bounds Checker to generate checks for all write
10126 accesses to memory. Enabled by default.
10128 @item -fchkp-store-bounds
10129 @opindex fchkp-store-bounds
10130 @opindex fno-chkp-store-bounds
10131 Instructs Pointer Bounds Checker to generate bounds stores for
10132 pointer writes. Enabled by default.
10134 @item -fchkp-instrument-calls
10135 @opindex fchkp-instrument-calls
10136 @opindex fno-chkp-instrument-calls
10137 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
10138 Enabled by default.
10140 @item -fchkp-instrument-marked-only
10141 @opindex fchkp-instrument-marked-only
10142 @opindex fno-chkp-instrument-marked-only
10143 Instructs Pointer Bounds Checker to instrument only functions
10144 marked with the @code{bnd_instrument} attribute
10145 (@pxref{Function Attributes}). Disabled by default.
10147 @item -fchkp-use-wrappers
10148 @opindex fchkp-use-wrappers
10149 @opindex fno-chkp-use-wrappers
10150 Allows Pointer Bounds Checker to replace calls to built-in functions
10151 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
10152 is used to link a program, the GCC driver automatically links
10153 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
10154 Enabled by default.
10156 @item -fstack-protector
10157 @opindex fstack-protector
10158 Emit extra code to check for buffer overflows, such as stack smashing
10159 attacks. This is done by adding a guard variable to functions with
10160 vulnerable objects. This includes functions that call @code{alloca}, and
10161 functions with buffers larger than 8 bytes. The guards are initialized
10162 when a function is entered and then checked when the function exits.
10163 If a guard check fails, an error message is printed and the program exits.
10165 @item -fstack-protector-all
10166 @opindex fstack-protector-all
10167 Like @option{-fstack-protector} except that all functions are protected.
10169 @item -fstack-protector-strong
10170 @opindex fstack-protector-strong
10171 Like @option{-fstack-protector} but includes additional functions to
10172 be protected --- those that have local array definitions, or have
10173 references to local frame addresses.
10175 @item -fstack-protector-explicit
10176 @opindex fstack-protector-explicit
10177 Like @option{-fstack-protector} but only protects those functions which
10178 have the @code{stack_protect} attribute.
10180 @item -fstack-check
10181 @opindex fstack-check
10182 Generate code to verify that you do not go beyond the boundary of the
10183 stack. You should specify this flag if you are running in an
10184 environment with multiple threads, but you only rarely need to specify it in
10185 a single-threaded environment since stack overflow is automatically
10186 detected on nearly all systems if there is only one stack.
10188 Note that this switch does not actually cause checking to be done; the
10189 operating system or the language runtime must do that. The switch causes
10190 generation of code to ensure that they see the stack being extended.
10192 You can additionally specify a string parameter: @samp{no} means no
10193 checking, @samp{generic} means force the use of old-style checking,
10194 @samp{specific} means use the best checking method and is equivalent
10195 to bare @option{-fstack-check}.
10197 Old-style checking is a generic mechanism that requires no specific
10198 target support in the compiler but comes with the following drawbacks:
10202 Modified allocation strategy for large objects: they are always
10203 allocated dynamically if their size exceeds a fixed threshold.
10206 Fixed limit on the size of the static frame of functions: when it is
10207 topped by a particular function, stack checking is not reliable and
10208 a warning is issued by the compiler.
10211 Inefficiency: because of both the modified allocation strategy and the
10212 generic implementation, code performance is hampered.
10215 Note that old-style stack checking is also the fallback method for
10216 @samp{specific} if no target support has been added in the compiler.
10218 @item -fstack-limit-register=@var{reg}
10219 @itemx -fstack-limit-symbol=@var{sym}
10220 @itemx -fno-stack-limit
10221 @opindex fstack-limit-register
10222 @opindex fstack-limit-symbol
10223 @opindex fno-stack-limit
10224 Generate code to ensure that the stack does not grow beyond a certain value,
10225 either the value of a register or the address of a symbol. If a larger
10226 stack is required, a signal is raised at run time. For most targets,
10227 the signal is raised before the stack overruns the boundary, so
10228 it is possible to catch the signal without taking special precautions.
10230 For instance, if the stack starts at absolute address @samp{0x80000000}
10231 and grows downwards, you can use the flags
10232 @option{-fstack-limit-symbol=__stack_limit} and
10233 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
10234 of 128KB@. Note that this may only work with the GNU linker.
10236 You can locally override stack limit checking by using the
10237 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
10239 @item -fsplit-stack
10240 @opindex fsplit-stack
10241 Generate code to automatically split the stack before it overflows.
10242 The resulting program has a discontiguous stack which can only
10243 overflow if the program is unable to allocate any more memory. This
10244 is most useful when running threaded programs, as it is no longer
10245 necessary to calculate a good stack size to use for each thread. This
10246 is currently only implemented for the x86 targets running
10249 When code compiled with @option{-fsplit-stack} calls code compiled
10250 without @option{-fsplit-stack}, there may not be much stack space
10251 available for the latter code to run. If compiling all code,
10252 including library code, with @option{-fsplit-stack} is not an option,
10253 then the linker can fix up these calls so that the code compiled
10254 without @option{-fsplit-stack} always has a large stack. Support for
10255 this is implemented in the gold linker in GNU binutils release 2.21
10258 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
10259 @opindex fvtable-verify
10260 This option is only available when compiling C++ code.
10261 It turns on (or off, if using @option{-fvtable-verify=none}) the security
10262 feature that verifies at run time, for every virtual call, that
10263 the vtable pointer through which the call is made is valid for the type of
10264 the object, and has not been corrupted or overwritten. If an invalid vtable
10265 pointer is detected at run time, an error is reported and execution of the
10266 program is immediately halted.
10268 This option causes run-time data structures to be built at program startup,
10269 which are used for verifying the vtable pointers.
10270 The options @samp{std} and @samp{preinit}
10271 control the timing of when these data structures are built. In both cases the
10272 data structures are built before execution reaches @code{main}. Using
10273 @option{-fvtable-verify=std} causes the data structures to be built after
10274 shared libraries have been loaded and initialized.
10275 @option{-fvtable-verify=preinit} causes them to be built before shared
10276 libraries have been loaded and initialized.
10278 If this option appears multiple times in the command line with different
10279 values specified, @samp{none} takes highest priority over both @samp{std} and
10280 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
10283 @opindex fvtv-debug
10284 When used in conjunction with @option{-fvtable-verify=std} or
10285 @option{-fvtable-verify=preinit}, causes debug versions of the
10286 runtime functions for the vtable verification feature to be called.
10287 This flag also causes the compiler to log information about which
10288 vtable pointers it finds for each class.
10289 This information is written to a file named @file{vtv_set_ptr_data.log}
10290 in the directory named by the environment variable @env{VTV_LOGS_DIR}
10291 if that is defined or the current working directory otherwise.
10293 Note: This feature @emph{appends} data to the log file. If you want a fresh log
10294 file, be sure to delete any existing one.
10297 @opindex fvtv-counts
10298 This is a debugging flag. When used in conjunction with
10299 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
10300 causes the compiler to keep track of the total number of virtual calls
10301 it encounters and the number of verifications it inserts. It also
10302 counts the number of calls to certain run-time library functions
10303 that it inserts and logs this information for each compilation unit.
10304 The compiler writes this information to a file named
10305 @file{vtv_count_data.log} in the directory named by the environment
10306 variable @env{VTV_LOGS_DIR} if that is defined or the current working
10307 directory otherwise. It also counts the size of the vtable pointer sets
10308 for each class, and writes this information to @file{vtv_class_set_sizes.log}
10309 in the same directory.
10311 Note: This feature @emph{appends} data to the log files. To get fresh log
10312 files, be sure to delete any existing ones.
10314 @item -finstrument-functions
10315 @opindex finstrument-functions
10316 Generate instrumentation calls for entry and exit to functions. Just
10317 after function entry and just before function exit, the following
10318 profiling functions are called with the address of the current
10319 function and its call site. (On some platforms,
10320 @code{__builtin_return_address} does not work beyond the current
10321 function, so the call site information may not be available to the
10322 profiling functions otherwise.)
10325 void __cyg_profile_func_enter (void *this_fn,
10327 void __cyg_profile_func_exit (void *this_fn,
10331 The first argument is the address of the start of the current function,
10332 which may be looked up exactly in the symbol table.
10334 This instrumentation is also done for functions expanded inline in other
10335 functions. The profiling calls indicate where, conceptually, the
10336 inline function is entered and exited. This means that addressable
10337 versions of such functions must be available. If all your uses of a
10338 function are expanded inline, this may mean an additional expansion of
10339 code size. If you use @code{extern inline} in your C code, an
10340 addressable version of such functions must be provided. (This is
10341 normally the case anyway, but if you get lucky and the optimizer always
10342 expands the functions inline, you might have gotten away without
10343 providing static copies.)
10345 A function may be given the attribute @code{no_instrument_function}, in
10346 which case this instrumentation is not done. This can be used, for
10347 example, for the profiling functions listed above, high-priority
10348 interrupt routines, and any functions from which the profiling functions
10349 cannot safely be called (perhaps signal handlers, if the profiling
10350 routines generate output or allocate memory).
10352 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
10353 @opindex finstrument-functions-exclude-file-list
10355 Set the list of functions that are excluded from instrumentation (see
10356 the description of @option{-finstrument-functions}). If the file that
10357 contains a function definition matches with one of @var{file}, then
10358 that function is not instrumented. The match is done on substrings:
10359 if the @var{file} parameter is a substring of the file name, it is
10360 considered to be a match.
10365 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
10369 excludes any inline function defined in files whose pathnames
10370 contain @file{/bits/stl} or @file{include/sys}.
10372 If, for some reason, you want to include letter @samp{,} in one of
10373 @var{sym}, write @samp{\,}. For example,
10374 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
10375 (note the single quote surrounding the option).
10377 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
10378 @opindex finstrument-functions-exclude-function-list
10380 This is similar to @option{-finstrument-functions-exclude-file-list},
10381 but this option sets the list of function names to be excluded from
10382 instrumentation. The function name to be matched is its user-visible
10383 name, such as @code{vector<int> blah(const vector<int> &)}, not the
10384 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
10385 match is done on substrings: if the @var{sym} parameter is a substring
10386 of the function name, it is considered to be a match. For C99 and C++
10387 extended identifiers, the function name must be given in UTF-8, not
10388 using universal character names.
10393 @node Preprocessor Options
10394 @section Options Controlling the Preprocessor
10395 @cindex preprocessor options
10396 @cindex options, preprocessor
10398 These options control the C preprocessor, which is run on each C source
10399 file before actual compilation.
10401 If you use the @option{-E} option, nothing is done except preprocessing.
10402 Some of these options make sense only together with @option{-E} because
10403 they cause the preprocessor output to be unsuitable for actual
10407 @item -Wp,@var{option}
10409 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10410 and pass @var{option} directly through to the preprocessor. If
10411 @var{option} contains commas, it is split into multiple options at the
10412 commas. However, many options are modified, translated or interpreted
10413 by the compiler driver before being passed to the preprocessor, and
10414 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10415 interface is undocumented and subject to change, so whenever possible
10416 you should avoid using @option{-Wp} and let the driver handle the
10419 @item -Xpreprocessor @var{option}
10420 @opindex Xpreprocessor
10421 Pass @var{option} as an option to the preprocessor. You can use this to
10422 supply system-specific preprocessor options that GCC does not
10425 If you want to pass an option that takes an argument, you must use
10426 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10428 @item -no-integrated-cpp
10429 @opindex no-integrated-cpp
10430 Perform preprocessing as a separate pass before compilation.
10431 By default, GCC performs preprocessing as an integrated part of
10432 input tokenization and parsing.
10433 If this option is provided, the appropriate language front end
10434 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10435 and Objective-C, respectively) is instead invoked twice,
10436 once for preprocessing only and once for actual compilation
10437 of the preprocessed input.
10438 This option may be useful in conjunction with the @option{-B} or
10439 @option{-wrapper} options to specify an alternate preprocessor or
10440 perform additional processing of the program source between
10441 normal preprocessing and compilation.
10444 @include cppopts.texi
10446 @node Assembler Options
10447 @section Passing Options to the Assembler
10449 @c prevent bad page break with this line
10450 You can pass options to the assembler.
10453 @item -Wa,@var{option}
10455 Pass @var{option} as an option to the assembler. If @var{option}
10456 contains commas, it is split into multiple options at the commas.
10458 @item -Xassembler @var{option}
10459 @opindex Xassembler
10460 Pass @var{option} as an option to the assembler. You can use this to
10461 supply system-specific assembler options that GCC does not
10464 If you want to pass an option that takes an argument, you must use
10465 @option{-Xassembler} twice, once for the option and once for the argument.
10470 @section Options for Linking
10471 @cindex link options
10472 @cindex options, linking
10474 These options come into play when the compiler links object files into
10475 an executable output file. They are meaningless if the compiler is
10476 not doing a link step.
10480 @item @var{object-file-name}
10481 A file name that does not end in a special recognized suffix is
10482 considered to name an object file or library. (Object files are
10483 distinguished from libraries by the linker according to the file
10484 contents.) If linking is done, these object files are used as input
10493 If any of these options is used, then the linker is not run, and
10494 object file names should not be used as arguments. @xref{Overall
10498 @opindex fuse-ld=bfd
10499 Use the @command{bfd} linker instead of the default linker.
10501 @item -fuse-ld=gold
10502 @opindex fuse-ld=gold
10503 Use the @command{gold} linker instead of the default linker.
10506 @item -l@var{library}
10507 @itemx -l @var{library}
10509 Search the library named @var{library} when linking. (The second
10510 alternative with the library as a separate argument is only for
10511 POSIX compliance and is not recommended.)
10513 It makes a difference where in the command you write this option; the
10514 linker searches and processes libraries and object files in the order they
10515 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10516 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10517 to functions in @samp{z}, those functions may not be loaded.
10519 The linker searches a standard list of directories for the library,
10520 which is actually a file named @file{lib@var{library}.a}. The linker
10521 then uses this file as if it had been specified precisely by name.
10523 The directories searched include several standard system directories
10524 plus any that you specify with @option{-L}.
10526 Normally the files found this way are library files---archive files
10527 whose members are object files. The linker handles an archive file by
10528 scanning through it for members which define symbols that have so far
10529 been referenced but not defined. But if the file that is found is an
10530 ordinary object file, it is linked in the usual fashion. The only
10531 difference between using an @option{-l} option and specifying a file name
10532 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10533 and searches several directories.
10537 You need this special case of the @option{-l} option in order to
10538 link an Objective-C or Objective-C++ program.
10540 @item -nostartfiles
10541 @opindex nostartfiles
10542 Do not use the standard system startup files when linking.
10543 The standard system libraries are used normally, unless @option{-nostdlib}
10544 or @option{-nodefaultlibs} is used.
10546 @item -nodefaultlibs
10547 @opindex nodefaultlibs
10548 Do not use the standard system libraries when linking.
10549 Only the libraries you specify are passed to the linker, and options
10550 specifying linkage of the system libraries, such as @option{-static-libgcc}
10551 or @option{-shared-libgcc}, are ignored.
10552 The standard startup files are used normally, unless @option{-nostartfiles}
10555 The compiler may generate calls to @code{memcmp},
10556 @code{memset}, @code{memcpy} and @code{memmove}.
10557 These entries are usually resolved by entries in
10558 libc. These entry points should be supplied through some other
10559 mechanism when this option is specified.
10563 Do not use the standard system startup files or libraries when linking.
10564 No startup files and only the libraries you specify are passed to
10565 the linker, and options specifying linkage of the system libraries, such as
10566 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10568 The compiler may generate calls to @code{memcmp}, @code{memset},
10569 @code{memcpy} and @code{memmove}.
10570 These entries are usually resolved by entries in
10571 libc. These entry points should be supplied through some other
10572 mechanism when this option is specified.
10574 @cindex @option{-lgcc}, use with @option{-nostdlib}
10575 @cindex @option{-nostdlib} and unresolved references
10576 @cindex unresolved references and @option{-nostdlib}
10577 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10578 @cindex @option{-nodefaultlibs} and unresolved references
10579 @cindex unresolved references and @option{-nodefaultlibs}
10580 One of the standard libraries bypassed by @option{-nostdlib} and
10581 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10582 which GCC uses to overcome shortcomings of particular machines, or special
10583 needs for some languages.
10584 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10585 Collection (GCC) Internals},
10586 for more discussion of @file{libgcc.a}.)
10587 In most cases, you need @file{libgcc.a} even when you want to avoid
10588 other standard libraries. In other words, when you specify @option{-nostdlib}
10589 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10590 This ensures that you have no unresolved references to internal GCC
10591 library subroutines.
10592 (An example of such an internal subroutine is @code{__main}, used to ensure C++
10593 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10594 GNU Compiler Collection (GCC) Internals}.)
10598 Produce a position independent executable on targets that support it.
10599 For predictable results, you must also specify the same set of options
10600 used for compilation (@option{-fpie}, @option{-fPIE},
10601 or model suboptions) when you specify this linker option.
10605 Don't produce a position independent executable.
10609 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10610 that support it. This instructs the linker to add all symbols, not
10611 only used ones, to the dynamic symbol table. This option is needed
10612 for some uses of @code{dlopen} or to allow obtaining backtraces
10613 from within a program.
10617 Remove all symbol table and relocation information from the executable.
10621 On systems that support dynamic linking, this prevents linking with the shared
10622 libraries. On other systems, this option has no effect.
10626 Produce a shared object which can then be linked with other objects to
10627 form an executable. Not all systems support this option. For predictable
10628 results, you must also specify the same set of options used for compilation
10629 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10630 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10631 needs to build supplementary stub code for constructors to work. On
10632 multi-libbed systems, @samp{gcc -shared} must select the correct support
10633 libraries to link against. Failing to supply the correct flags may lead
10634 to subtle defects. Supplying them in cases where they are not necessary
10637 @item -shared-libgcc
10638 @itemx -static-libgcc
10639 @opindex shared-libgcc
10640 @opindex static-libgcc
10641 On systems that provide @file{libgcc} as a shared library, these options
10642 force the use of either the shared or static version, respectively.
10643 If no shared version of @file{libgcc} was built when the compiler was
10644 configured, these options have no effect.
10646 There are several situations in which an application should use the
10647 shared @file{libgcc} instead of the static version. The most common
10648 of these is when the application wishes to throw and catch exceptions
10649 across different shared libraries. In that case, each of the libraries
10650 as well as the application itself should use the shared @file{libgcc}.
10652 Therefore, the G++ and GCJ drivers automatically add
10653 @option{-shared-libgcc} whenever you build a shared library or a main
10654 executable, because C++ and Java programs typically use exceptions, so
10655 this is the right thing to do.
10657 If, instead, you use the GCC driver to create shared libraries, you may
10658 find that they are not always linked with the shared @file{libgcc}.
10659 If GCC finds, at its configuration time, that you have a non-GNU linker
10660 or a GNU linker that does not support option @option{--eh-frame-hdr},
10661 it links the shared version of @file{libgcc} into shared libraries
10662 by default. Otherwise, it takes advantage of the linker and optimizes
10663 away the linking with the shared version of @file{libgcc}, linking with
10664 the static version of libgcc by default. This allows exceptions to
10665 propagate through such shared libraries, without incurring relocation
10666 costs at library load time.
10668 However, if a library or main executable is supposed to throw or catch
10669 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10670 for the languages used in the program, or using the option
10671 @option{-shared-libgcc}, such that it is linked with the shared
10674 @item -static-libasan
10675 @opindex static-libasan
10676 When the @option{-fsanitize=address} option is used to link a program,
10677 the GCC driver automatically links against @option{libasan}. If
10678 @file{libasan} is available as a shared library, and the @option{-static}
10679 option is not used, then this links against the shared version of
10680 @file{libasan}. The @option{-static-libasan} option directs the GCC
10681 driver to link @file{libasan} statically, without necessarily linking
10682 other libraries statically.
10684 @item -static-libtsan
10685 @opindex static-libtsan
10686 When the @option{-fsanitize=thread} option is used to link a program,
10687 the GCC driver automatically links against @option{libtsan}. If
10688 @file{libtsan} is available as a shared library, and the @option{-static}
10689 option is not used, then this links against the shared version of
10690 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10691 driver to link @file{libtsan} statically, without necessarily linking
10692 other libraries statically.
10694 @item -static-liblsan
10695 @opindex static-liblsan
10696 When the @option{-fsanitize=leak} option is used to link a program,
10697 the GCC driver automatically links against @option{liblsan}. If
10698 @file{liblsan} is available as a shared library, and the @option{-static}
10699 option is not used, then this links against the shared version of
10700 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10701 driver to link @file{liblsan} statically, without necessarily linking
10702 other libraries statically.
10704 @item -static-libubsan
10705 @opindex static-libubsan
10706 When the @option{-fsanitize=undefined} option is used to link a program,
10707 the GCC driver automatically links against @option{libubsan}. If
10708 @file{libubsan} is available as a shared library, and the @option{-static}
10709 option is not used, then this links against the shared version of
10710 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10711 driver to link @file{libubsan} statically, without necessarily linking
10712 other libraries statically.
10714 @item -static-libmpx
10715 @opindex static-libmpx
10716 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
10717 used to link a program, the GCC driver automatically links against
10718 @file{libmpx}. If @file{libmpx} is available as a shared library,
10719 and the @option{-static} option is not used, then this links against
10720 the shared version of @file{libmpx}. The @option{-static-libmpx}
10721 option directs the GCC driver to link @file{libmpx} statically,
10722 without necessarily linking other libraries statically.
10724 @item -static-libmpxwrappers
10725 @opindex static-libmpxwrappers
10726 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
10727 to link a program without also using @option{-fno-chkp-use-wrappers}, the
10728 GCC driver automatically links against @file{libmpxwrappers}. If
10729 @file{libmpxwrappers} is available as a shared library, and the
10730 @option{-static} option is not used, then this links against the shared
10731 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
10732 option directs the GCC driver to link @file{libmpxwrappers} statically,
10733 without necessarily linking other libraries statically.
10735 @item -static-libstdc++
10736 @opindex static-libstdc++
10737 When the @command{g++} program is used to link a C++ program, it
10738 normally automatically links against @option{libstdc++}. If
10739 @file{libstdc++} is available as a shared library, and the
10740 @option{-static} option is not used, then this links against the
10741 shared version of @file{libstdc++}. That is normally fine. However, it
10742 is sometimes useful to freeze the version of @file{libstdc++} used by
10743 the program without going all the way to a fully static link. The
10744 @option{-static-libstdc++} option directs the @command{g++} driver to
10745 link @file{libstdc++} statically, without necessarily linking other
10746 libraries statically.
10750 Bind references to global symbols when building a shared object. Warn
10751 about any unresolved references (unless overridden by the link editor
10752 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10755 @item -T @var{script}
10757 @cindex linker script
10758 Use @var{script} as the linker script. This option is supported by most
10759 systems using the GNU linker. On some targets, such as bare-board
10760 targets without an operating system, the @option{-T} option may be required
10761 when linking to avoid references to undefined symbols.
10763 @item -Xlinker @var{option}
10765 Pass @var{option} as an option to the linker. You can use this to
10766 supply system-specific linker options that GCC does not recognize.
10768 If you want to pass an option that takes a separate argument, you must use
10769 @option{-Xlinker} twice, once for the option and once for the argument.
10770 For example, to pass @option{-assert definitions}, you must write
10771 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10772 @option{-Xlinker "-assert definitions"}, because this passes the entire
10773 string as a single argument, which is not what the linker expects.
10775 When using the GNU linker, it is usually more convenient to pass
10776 arguments to linker options using the @option{@var{option}=@var{value}}
10777 syntax than as separate arguments. For example, you can specify
10778 @option{-Xlinker -Map=output.map} rather than
10779 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10780 this syntax for command-line options.
10782 @item -Wl,@var{option}
10784 Pass @var{option} as an option to the linker. If @var{option} contains
10785 commas, it is split into multiple options at the commas. You can use this
10786 syntax to pass an argument to the option.
10787 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10788 linker. When using the GNU linker, you can also get the same effect with
10789 @option{-Wl,-Map=output.map}.
10791 @item -u @var{symbol}
10793 Pretend the symbol @var{symbol} is undefined, to force linking of
10794 library modules to define it. You can use @option{-u} multiple times with
10795 different symbols to force loading of additional library modules.
10797 @item -z @var{keyword}
10799 @option{-z} is passed directly on to the linker along with the keyword
10800 @var{keyword}. See the section in the documentation of your linker for
10801 permitted values and their meanings.
10804 @node Directory Options
10805 @section Options for Directory Search
10806 @cindex directory options
10807 @cindex options, directory search
10808 @cindex search path
10810 These options specify directories to search for header files, for
10811 libraries and for parts of the compiler:
10816 Add the directory @var{dir} to the head of the list of directories to be
10817 searched for header files. This can be used to override a system header
10818 file, substituting your own version, since these directories are
10819 searched before the system header file directories. However, you should
10820 not use this option to add directories that contain vendor-supplied
10821 system header files (use @option{-isystem} for that). If you use more than
10822 one @option{-I} option, the directories are scanned in left-to-right
10823 order; the standard system directories come after.
10825 If a standard system include directory, or a directory specified with
10826 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10827 option is ignored. The directory is still searched but as a
10828 system directory at its normal position in the system include chain.
10829 This is to ensure that GCC's procedure to fix buggy system headers and
10830 the ordering for the @code{include_next} directive are not inadvertently changed.
10831 If you really need to change the search order for system directories,
10832 use the @option{-nostdinc} and/or @option{-isystem} options.
10834 @item -iplugindir=@var{dir}
10835 @opindex iplugindir=
10836 Set the directory to search for plugins that are passed
10837 by @option{-fplugin=@var{name}} instead of
10838 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
10839 to be used by the user, but only passed by the driver.
10841 @item -iquote@var{dir}
10843 Add the directory @var{dir} to the head of the list of directories to
10844 be searched for header files only for the case of @code{#include
10845 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
10846 otherwise just like @option{-I}.
10850 Add directory @var{dir} to the list of directories to be searched
10853 @item -B@var{prefix}
10855 This option specifies where to find the executables, libraries,
10856 include files, and data files of the compiler itself.
10858 The compiler driver program runs one or more of the subprograms
10859 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
10860 @var{prefix} as a prefix for each program it tries to run, both with and
10861 without @samp{@var{machine}/@var{version}/} for the corresponding target
10862 machine and compiler version.
10864 For each subprogram to be run, the compiler driver first tries the
10865 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
10866 is not specified, the driver tries two standard prefixes,
10867 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
10868 those results in a file name that is found, the unmodified program
10869 name is searched for using the directories specified in your
10870 @env{PATH} environment variable.
10872 The compiler checks to see if the path provided by @option{-B}
10873 refers to a directory, and if necessary it adds a directory
10874 separator character at the end of the path.
10876 @option{-B} prefixes that effectively specify directory names also apply
10877 to libraries in the linker, because the compiler translates these
10878 options into @option{-L} options for the linker. They also apply to
10879 include files in the preprocessor, because the compiler translates these
10880 options into @option{-isystem} options for the preprocessor. In this case,
10881 the compiler appends @samp{include} to the prefix.
10883 The runtime support file @file{libgcc.a} can also be searched for using
10884 the @option{-B} prefix, if needed. If it is not found there, the two
10885 standard prefixes above are tried, and that is all. The file is left
10886 out of the link if it is not found by those means.
10888 Another way to specify a prefix much like the @option{-B} prefix is to use
10889 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
10892 As a special kludge, if the path provided by @option{-B} is
10893 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
10894 9, then it is replaced by @file{[dir/]include}. This is to help
10895 with boot-strapping the compiler.
10897 @item -no-canonical-prefixes
10898 @opindex no-canonical-prefixes
10899 Do not expand any symbolic links, resolve references to @samp{/../}
10900 or @samp{/./}, or make the path absolute when generating a relative
10903 @item --sysroot=@var{dir}
10905 Use @var{dir} as the logical root directory for headers and libraries.
10906 For example, if the compiler normally searches for headers in
10907 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
10908 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
10910 If you use both this option and the @option{-isysroot} option, then
10911 the @option{--sysroot} option applies to libraries, but the
10912 @option{-isysroot} option applies to header files.
10914 The GNU linker (beginning with version 2.16) has the necessary support
10915 for this option. If your linker does not support this option, the
10916 header file aspect of @option{--sysroot} still works, but the
10917 library aspect does not.
10919 @item --no-sysroot-suffix
10920 @opindex no-sysroot-suffix
10921 For some targets, a suffix is added to the root directory specified
10922 with @option{--sysroot}, depending on the other options used, so that
10923 headers may for example be found in
10924 @file{@var{dir}/@var{suffix}/usr/include} instead of
10925 @file{@var{dir}/usr/include}. This option disables the addition of
10930 This option has been deprecated. Please use @option{-iquote} instead for
10931 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
10933 Any directories you specify with @option{-I} options before the @option{-I-}
10934 option are searched only for the case of @code{#include "@var{file}"};
10935 they are not searched for @code{#include <@var{file}>}.
10937 If additional directories are specified with @option{-I} options after
10938 the @option{-I-} option, these directories are searched for all @code{#include}
10939 directives. (Ordinarily @emph{all} @option{-I} directories are used
10942 In addition, the @option{-I-} option inhibits the use of the current
10943 directory (where the current input file came from) as the first search
10944 directory for @code{#include "@var{file}"}. There is no way to
10945 override this effect of @option{-I-}. With @option{-I.} you can specify
10946 searching the directory that is current when the compiler is
10947 invoked. That is not exactly the same as what the preprocessor does
10948 by default, but it is often satisfactory.
10950 @option{-I-} does not inhibit the use of the standard system directories
10951 for header files. Thus, @option{-I-} and @option{-nostdinc} are
10955 @node Code Gen Options
10956 @section Options for Code Generation Conventions
10957 @cindex code generation conventions
10958 @cindex options, code generation
10959 @cindex run-time options
10961 These machine-independent options control the interface conventions
10962 used in code generation.
10964 Most of them have both positive and negative forms; the negative form
10965 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
10966 one of the forms is listed---the one that is not the default. You
10967 can figure out the other form by either removing @samp{no-} or adding
10971 @item -fstack-reuse=@var{reuse-level}
10972 @opindex fstack_reuse
10973 This option controls stack space reuse for user declared local/auto variables
10974 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
10975 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
10976 local variables and temporaries, @samp{named_vars} enables the reuse only for
10977 user defined local variables with names, and @samp{none} disables stack reuse
10978 completely. The default value is @samp{all}. The option is needed when the
10979 program extends the lifetime of a scoped local variable or a compiler generated
10980 temporary beyond the end point defined by the language. When a lifetime of
10981 a variable ends, and if the variable lives in memory, the optimizing compiler
10982 has the freedom to reuse its stack space with other temporaries or scoped
10983 local variables whose live range does not overlap with it. Legacy code extending
10984 local lifetime is likely to break with the stack reuse optimization.
11003 if (*p == 10) // out of scope use of local1
11014 A(int k) : i(k), j(k) @{ @}
11021 void foo(const A& ar)
11028 foo(A(10)); // temp object's lifetime ends when foo returns
11034 ap->i+= 10; // ap references out of scope temp whose space
11035 // is reused with a. What is the value of ap->i?
11040 The lifetime of a compiler generated temporary is well defined by the C++
11041 standard. When a lifetime of a temporary ends, and if the temporary lives
11042 in memory, the optimizing compiler has the freedom to reuse its stack
11043 space with other temporaries or scoped local variables whose live range
11044 does not overlap with it. However some of the legacy code relies on
11045 the behavior of older compilers in which temporaries' stack space is
11046 not reused, the aggressive stack reuse can lead to runtime errors. This
11047 option is used to control the temporary stack reuse optimization.
11051 This option generates traps for signed overflow on addition, subtraction,
11052 multiplication operations.
11053 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11054 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11055 @option{-fwrapv} being effective. Note that only active options override, so
11056 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11057 results in @option{-ftrapv} being effective.
11061 This option instructs the compiler to assume that signed arithmetic
11062 overflow of addition, subtraction and multiplication wraps around
11063 using twos-complement representation. This flag enables some optimizations
11064 and disables others. This option is enabled by default for the Java
11065 front end, as required by the Java language specification.
11066 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11067 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11068 @option{-fwrapv} being effective. Note that only active options override, so
11069 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11070 results in @option{-ftrapv} being effective.
11073 @opindex fexceptions
11074 Enable exception handling. Generates extra code needed to propagate
11075 exceptions. For some targets, this implies GCC generates frame
11076 unwind information for all functions, which can produce significant data
11077 size overhead, although it does not affect execution. If you do not
11078 specify this option, GCC enables it by default for languages like
11079 C++ that normally require exception handling, and disables it for
11080 languages like C that do not normally require it. However, you may need
11081 to enable this option when compiling C code that needs to interoperate
11082 properly with exception handlers written in C++. You may also wish to
11083 disable this option if you are compiling older C++ programs that don't
11084 use exception handling.
11086 @item -fnon-call-exceptions
11087 @opindex fnon-call-exceptions
11088 Generate code that allows trapping instructions to throw exceptions.
11089 Note that this requires platform-specific runtime support that does
11090 not exist everywhere. Moreover, it only allows @emph{trapping}
11091 instructions to throw exceptions, i.e.@: memory references or floating-point
11092 instructions. It does not allow exceptions to be thrown from
11093 arbitrary signal handlers such as @code{SIGALRM}.
11095 @item -fdelete-dead-exceptions
11096 @opindex fdelete-dead-exceptions
11097 Consider that instructions that may throw exceptions but don't otherwise
11098 contribute to the execution of the program can be optimized away.
11099 This option is enabled by default for the Ada front end, as permitted by
11100 the Ada language specification.
11101 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
11103 @item -funwind-tables
11104 @opindex funwind-tables
11105 Similar to @option{-fexceptions}, except that it just generates any needed
11106 static data, but does not affect the generated code in any other way.
11107 You normally do not need to enable this option; instead, a language processor
11108 that needs this handling enables it on your behalf.
11110 @item -fasynchronous-unwind-tables
11111 @opindex fasynchronous-unwind-tables
11112 Generate unwind table in DWARF format, if supported by target machine. The
11113 table is exact at each instruction boundary, so it can be used for stack
11114 unwinding from asynchronous events (such as debugger or garbage collector).
11116 @item -fno-gnu-unique
11117 @opindex fno-gnu-unique
11118 On systems with recent GNU assembler and C library, the C++ compiler
11119 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
11120 of template static data members and static local variables in inline
11121 functions are unique even in the presence of @code{RTLD_LOCAL}; this
11122 is necessary to avoid problems with a library used by two different
11123 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
11124 therefore disagreeing with the other one about the binding of the
11125 symbol. But this causes @code{dlclose} to be ignored for affected
11126 DSOs; if your program relies on reinitialization of a DSO via
11127 @code{dlclose} and @code{dlopen}, you can use
11128 @option{-fno-gnu-unique}.
11130 @item -fpcc-struct-return
11131 @opindex fpcc-struct-return
11132 Return ``short'' @code{struct} and @code{union} values in memory like
11133 longer ones, rather than in registers. This convention is less
11134 efficient, but it has the advantage of allowing intercallability between
11135 GCC-compiled files and files compiled with other compilers, particularly
11136 the Portable C Compiler (pcc).
11138 The precise convention for returning structures in memory depends
11139 on the target configuration macros.
11141 Short structures and unions are those whose size and alignment match
11142 that of some integer type.
11144 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11145 switch is not binary compatible with code compiled with the
11146 @option{-freg-struct-return} switch.
11147 Use it to conform to a non-default application binary interface.
11149 @item -freg-struct-return
11150 @opindex freg-struct-return
11151 Return @code{struct} and @code{union} values in registers when possible.
11152 This is more efficient for small structures than
11153 @option{-fpcc-struct-return}.
11155 If you specify neither @option{-fpcc-struct-return} nor
11156 @option{-freg-struct-return}, GCC defaults to whichever convention is
11157 standard for the target. If there is no standard convention, GCC
11158 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11159 the principal compiler. In those cases, we can choose the standard, and
11160 we chose the more efficient register return alternative.
11162 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11163 switch is not binary compatible with code compiled with the
11164 @option{-fpcc-struct-return} switch.
11165 Use it to conform to a non-default application binary interface.
11167 @item -fshort-enums
11168 @opindex fshort-enums
11169 Allocate to an @code{enum} type only as many bytes as it needs for the
11170 declared range of possible values. Specifically, the @code{enum} type
11171 is equivalent to the smallest integer type that has enough room.
11173 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11174 code that is not binary compatible with code generated without that switch.
11175 Use it to conform to a non-default application binary interface.
11177 @item -fshort-wchar
11178 @opindex fshort-wchar
11179 Override the underlying type for @code{wchar_t} to be @code{short
11180 unsigned int} instead of the default for the target. This option is
11181 useful for building programs to run under WINE@.
11183 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11184 code that is not binary compatible with code generated without that switch.
11185 Use it to conform to a non-default application binary interface.
11188 @opindex fno-common
11189 In C code, controls the placement of uninitialized global variables.
11190 Unix C compilers have traditionally permitted multiple definitions of
11191 such variables in different compilation units by placing the variables
11193 This is the behavior specified by @option{-fcommon}, and is the default
11194 for GCC on most targets.
11195 On the other hand, this behavior is not required by ISO C, and on some
11196 targets may carry a speed or code size penalty on variable references.
11197 The @option{-fno-common} option specifies that the compiler should place
11198 uninitialized global variables in the data section of the object file,
11199 rather than generating them as common blocks.
11200 This has the effect that if the same variable is declared
11201 (without @code{extern}) in two different compilations,
11202 you get a multiple-definition error when you link them.
11203 In this case, you must compile with @option{-fcommon} instead.
11204 Compiling with @option{-fno-common} is useful on targets for which
11205 it provides better performance, or if you wish to verify that the
11206 program will work on other systems that always treat uninitialized
11207 variable declarations this way.
11211 Ignore the @code{#ident} directive.
11213 @item -finhibit-size-directive
11214 @opindex finhibit-size-directive
11215 Don't output a @code{.size} assembler directive, or anything else that
11216 would cause trouble if the function is split in the middle, and the
11217 two halves are placed at locations far apart in memory. This option is
11218 used when compiling @file{crtstuff.c}; you should not need to use it
11221 @item -fverbose-asm
11222 @opindex fverbose-asm
11223 Put extra commentary information in the generated assembly code to
11224 make it more readable. This option is generally only of use to those
11225 who actually need to read the generated assembly code (perhaps while
11226 debugging the compiler itself).
11228 @option{-fno-verbose-asm}, the default, causes the
11229 extra information to be omitted and is useful when comparing two assembler
11232 @item -frecord-gcc-switches
11233 @opindex frecord-gcc-switches
11234 This switch causes the command line used to invoke the
11235 compiler to be recorded into the object file that is being created.
11236 This switch is only implemented on some targets and the exact format
11237 of the recording is target and binary file format dependent, but it
11238 usually takes the form of a section containing ASCII text. This
11239 switch is related to the @option{-fverbose-asm} switch, but that
11240 switch only records information in the assembler output file as
11241 comments, so it never reaches the object file.
11242 See also @option{-grecord-gcc-switches} for another
11243 way of storing compiler options into the object file.
11247 @cindex global offset table
11249 Generate position-independent code (PIC) suitable for use in a shared
11250 library, if supported for the target machine. Such code accesses all
11251 constant addresses through a global offset table (GOT)@. The dynamic
11252 loader resolves the GOT entries when the program starts (the dynamic
11253 loader is not part of GCC; it is part of the operating system). If
11254 the GOT size for the linked executable exceeds a machine-specific
11255 maximum size, you get an error message from the linker indicating that
11256 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11257 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
11258 on the m68k and RS/6000. The x86 has no such limit.)
11260 Position-independent code requires special support, and therefore works
11261 only on certain machines. For the x86, GCC supports PIC for System V
11262 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11263 position-independent.
11265 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11270 If supported for the target machine, emit position-independent code,
11271 suitable for dynamic linking and avoiding any limit on the size of the
11272 global offset table. This option makes a difference on AArch64, m68k,
11273 PowerPC and SPARC@.
11275 Position-independent code requires special support, and therefore works
11276 only on certain machines.
11278 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11285 These options are similar to @option{-fpic} and @option{-fPIC}, but
11286 generated position independent code can be only linked into executables.
11287 Usually these options are used when @option{-pie} GCC option is
11288 used during linking.
11290 @option{-fpie} and @option{-fPIE} both define the macros
11291 @code{__pie__} and @code{__PIE__}. The macros have the value 1
11292 for @option{-fpie} and 2 for @option{-fPIE}.
11296 Do not use the PLT for external function calls in position-independent code.
11297 Instead, load the callee address at call sites from the GOT and branch to it.
11298 This leads to more efficient code by eliminating PLT stubs and exposing
11299 GOT loads to optimizations. On architectures such as 32-bit x86 where
11300 PLT stubs expect the GOT pointer in a specific register, this gives more
11301 register allocation freedom to the compiler.
11302 Lazy binding requires use of the PLT;
11303 with @option{-fno-plt} all external symbols are resolved at load time.
11305 Alternatively, the function attribute @code{noplt} can be used to avoid calls
11306 through the PLT for specific external functions.
11308 In position-dependent code, a few targets also convert calls to
11309 functions that are marked to not use the PLT to use the GOT instead.
11311 @item -fno-jump-tables
11312 @opindex fno-jump-tables
11313 Do not use jump tables for switch statements even where it would be
11314 more efficient than other code generation strategies. This option is
11315 of use in conjunction with @option{-fpic} or @option{-fPIC} for
11316 building code that forms part of a dynamic linker and cannot
11317 reference the address of a jump table. On some targets, jump tables
11318 do not require a GOT and this option is not needed.
11320 @item -ffixed-@var{reg}
11322 Treat the register named @var{reg} as a fixed register; generated code
11323 should never refer to it (except perhaps as a stack pointer, frame
11324 pointer or in some other fixed role).
11326 @var{reg} must be the name of a register. The register names accepted
11327 are machine-specific and are defined in the @code{REGISTER_NAMES}
11328 macro in the machine description macro file.
11330 This flag does not have a negative form, because it specifies a
11333 @item -fcall-used-@var{reg}
11334 @opindex fcall-used
11335 Treat the register named @var{reg} as an allocable register that is
11336 clobbered by function calls. It may be allocated for temporaries or
11337 variables that do not live across a call. Functions compiled this way
11338 do not save and restore the register @var{reg}.
11340 It is an error to use this flag with the frame pointer or stack pointer.
11341 Use of this flag for other registers that have fixed pervasive roles in
11342 the machine's execution model produces disastrous results.
11344 This flag does not have a negative form, because it specifies a
11347 @item -fcall-saved-@var{reg}
11348 @opindex fcall-saved
11349 Treat the register named @var{reg} as an allocable register saved by
11350 functions. It may be allocated even for temporaries or variables that
11351 live across a call. Functions compiled this way save and restore
11352 the register @var{reg} if they use it.
11354 It is an error to use this flag with the frame pointer or stack pointer.
11355 Use of this flag for other registers that have fixed pervasive roles in
11356 the machine's execution model produces disastrous results.
11358 A different sort of disaster results from the use of this flag for
11359 a register in which function values may be returned.
11361 This flag does not have a negative form, because it specifies a
11364 @item -fpack-struct[=@var{n}]
11365 @opindex fpack-struct
11366 Without a value specified, pack all structure members together without
11367 holes. When a value is specified (which must be a small power of two), pack
11368 structure members according to this value, representing the maximum
11369 alignment (that is, objects with default alignment requirements larger than
11370 this are output potentially unaligned at the next fitting location.
11372 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11373 code that is not binary compatible with code generated without that switch.
11374 Additionally, it makes the code suboptimal.
11375 Use it to conform to a non-default application binary interface.
11377 @item -fleading-underscore
11378 @opindex fleading-underscore
11379 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11380 change the way C symbols are represented in the object file. One use
11381 is to help link with legacy assembly code.
11383 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11384 generate code that is not binary compatible with code generated without that
11385 switch. Use it to conform to a non-default application binary interface.
11386 Not all targets provide complete support for this switch.
11388 @item -ftls-model=@var{model}
11389 @opindex ftls-model
11390 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11391 The @var{model} argument should be one of @samp{global-dynamic},
11392 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
11393 Note that the choice is subject to optimization: the compiler may use
11394 a more efficient model for symbols not visible outside of the translation
11395 unit, or if @option{-fpic} is not given on the command line.
11397 The default without @option{-fpic} is @samp{initial-exec}; with
11398 @option{-fpic} the default is @samp{global-dynamic}.
11400 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
11401 @opindex fvisibility
11402 Set the default ELF image symbol visibility to the specified option---all
11403 symbols are marked with this unless overridden within the code.
11404 Using this feature can very substantially improve linking and
11405 load times of shared object libraries, produce more optimized
11406 code, provide near-perfect API export and prevent symbol clashes.
11407 It is @strong{strongly} recommended that you use this in any shared objects
11410 Despite the nomenclature, @samp{default} always means public; i.e.,
11411 available to be linked against from outside the shared object.
11412 @samp{protected} and @samp{internal} are pretty useless in real-world
11413 usage so the only other commonly used option is @samp{hidden}.
11414 The default if @option{-fvisibility} isn't specified is
11415 @samp{default}, i.e., make every symbol public.
11417 A good explanation of the benefits offered by ensuring ELF
11418 symbols have the correct visibility is given by ``How To Write
11419 Shared Libraries'' by Ulrich Drepper (which can be found at
11420 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
11421 solution made possible by this option to marking things hidden when
11422 the default is public is to make the default hidden and mark things
11423 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
11424 and @code{__attribute__ ((visibility("default")))} instead of
11425 @code{__declspec(dllexport)} you get almost identical semantics with
11426 identical syntax. This is a great boon to those working with
11427 cross-platform projects.
11429 For those adding visibility support to existing code, you may find
11430 @code{#pragma GCC visibility} of use. This works by you enclosing
11431 the declarations you wish to set visibility for with (for example)
11432 @code{#pragma GCC visibility push(hidden)} and
11433 @code{#pragma GCC visibility pop}.
11434 Bear in mind that symbol visibility should be viewed @strong{as
11435 part of the API interface contract} and thus all new code should
11436 always specify visibility when it is not the default; i.e., declarations
11437 only for use within the local DSO should @strong{always} be marked explicitly
11438 as hidden as so to avoid PLT indirection overheads---making this
11439 abundantly clear also aids readability and self-documentation of the code.
11440 Note that due to ISO C++ specification requirements, @code{operator new} and
11441 @code{operator delete} must always be of default visibility.
11443 Be aware that headers from outside your project, in particular system
11444 headers and headers from any other library you use, may not be
11445 expecting to be compiled with visibility other than the default. You
11446 may need to explicitly say @code{#pragma GCC visibility push(default)}
11447 before including any such headers.
11449 @code{extern} declarations are not affected by @option{-fvisibility}, so
11450 a lot of code can be recompiled with @option{-fvisibility=hidden} with
11451 no modifications. However, this means that calls to @code{extern}
11452 functions with no explicit visibility use the PLT, so it is more
11453 effective to use @code{__attribute ((visibility))} and/or
11454 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
11455 declarations should be treated as hidden.
11457 Note that @option{-fvisibility} does affect C++ vague linkage
11458 entities. This means that, for instance, an exception class that is
11459 be thrown between DSOs must be explicitly marked with default
11460 visibility so that the @samp{type_info} nodes are unified between
11463 An overview of these techniques, their benefits and how to use them
11464 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
11466 @item -fstrict-volatile-bitfields
11467 @opindex fstrict-volatile-bitfields
11468 This option should be used if accesses to volatile bit-fields (or other
11469 structure fields, although the compiler usually honors those types
11470 anyway) should use a single access of the width of the
11471 field's type, aligned to a natural alignment if possible. For
11472 example, targets with memory-mapped peripheral registers might require
11473 all such accesses to be 16 bits wide; with this flag you can
11474 declare all peripheral bit-fields as @code{unsigned short} (assuming short
11475 is 16 bits on these targets) to force GCC to use 16-bit accesses
11476 instead of, perhaps, a more efficient 32-bit access.
11478 If this option is disabled, the compiler uses the most efficient
11479 instruction. In the previous example, that might be a 32-bit load
11480 instruction, even though that accesses bytes that do not contain
11481 any portion of the bit-field, or memory-mapped registers unrelated to
11482 the one being updated.
11484 In some cases, such as when the @code{packed} attribute is applied to a
11485 structure field, it may not be possible to access the field with a single
11486 read or write that is correctly aligned for the target machine. In this
11487 case GCC falls back to generating multiple accesses rather than code that
11488 will fault or truncate the result at run time.
11490 Note: Due to restrictions of the C/C++11 memory model, write accesses are
11491 not allowed to touch non bit-field members. It is therefore recommended
11492 to define all bits of the field's type as bit-field members.
11494 The default value of this option is determined by the application binary
11495 interface for the target processor.
11497 @item -fsync-libcalls
11498 @opindex fsync-libcalls
11499 This option controls whether any out-of-line instance of the @code{__sync}
11500 family of functions may be used to implement the C++11 @code{__atomic}
11501 family of functions.
11503 The default value of this option is enabled, thus the only useful form
11504 of the option is @option{-fno-sync-libcalls}. This option is used in
11505 the implementation of the @file{libatomic} runtime library.
11509 @node Developer Options
11510 @section GCC Developer Options
11511 @cindex developer options
11512 @cindex debugging GCC
11513 @cindex debug dump options
11514 @cindex dump options
11515 @cindex compilation statistics
11517 This section describes command-line options that are primarily of
11518 interest to GCC developers, including options to support compiler
11519 testing and investigation of compiler bugs and compile-time
11520 performance problems. This includes options that produce debug dumps
11521 at various points in the compilation; that print statistics such as
11522 memory use and execution time; and that print information about GCC's
11523 configuration, such as where it searches for libraries. You should
11524 rarely need to use any of these options for ordinary compilation and
11529 @item -d@var{letters}
11530 @itemx -fdump-rtl-@var{pass}
11531 @itemx -fdump-rtl-@var{pass}=@var{filename}
11533 @opindex fdump-rtl-@var{pass}
11534 Says to make debugging dumps during compilation at times specified by
11535 @var{letters}. This is used for debugging the RTL-based passes of the
11536 compiler. The file names for most of the dumps are made by appending
11537 a pass number and a word to the @var{dumpname}, and the files are
11538 created in the directory of the output file. In case of
11539 @option{=@var{filename}} option, the dump is output on the given file
11540 instead of the pass numbered dump files. Note that the pass number is
11541 assigned as passes are registered into the pass manager. Most passes
11542 are registered in the order that they will execute and for these passes
11543 the number corresponds to the pass execution order. However, passes
11544 registered by plugins, passes specific to compilation targets, or
11545 passes that are otherwise registered after all the other passes are
11546 numbered higher than a pass named "final", even if they are executed
11547 earlier. @var{dumpname} is generated from the name of the output
11548 file if explicitly specified and not an executable, otherwise it is
11549 the basename of the source file. These switches may have different
11550 effects when @option{-E} is used for preprocessing.
11552 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
11553 @option{-d} option @var{letters}. Here are the possible
11554 letters for use in @var{pass} and @var{letters}, and their meanings:
11558 @item -fdump-rtl-alignments
11559 @opindex fdump-rtl-alignments
11560 Dump after branch alignments have been computed.
11562 @item -fdump-rtl-asmcons
11563 @opindex fdump-rtl-asmcons
11564 Dump after fixing rtl statements that have unsatisfied in/out constraints.
11566 @item -fdump-rtl-auto_inc_dec
11567 @opindex fdump-rtl-auto_inc_dec
11568 Dump after auto-inc-dec discovery. This pass is only run on
11569 architectures that have auto inc or auto dec instructions.
11571 @item -fdump-rtl-barriers
11572 @opindex fdump-rtl-barriers
11573 Dump after cleaning up the barrier instructions.
11575 @item -fdump-rtl-bbpart
11576 @opindex fdump-rtl-bbpart
11577 Dump after partitioning hot and cold basic blocks.
11579 @item -fdump-rtl-bbro
11580 @opindex fdump-rtl-bbro
11581 Dump after block reordering.
11583 @item -fdump-rtl-btl1
11584 @itemx -fdump-rtl-btl2
11585 @opindex fdump-rtl-btl2
11586 @opindex fdump-rtl-btl2
11587 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
11588 after the two branch
11589 target load optimization passes.
11591 @item -fdump-rtl-bypass
11592 @opindex fdump-rtl-bypass
11593 Dump after jump bypassing and control flow optimizations.
11595 @item -fdump-rtl-combine
11596 @opindex fdump-rtl-combine
11597 Dump after the RTL instruction combination pass.
11599 @item -fdump-rtl-compgotos
11600 @opindex fdump-rtl-compgotos
11601 Dump after duplicating the computed gotos.
11603 @item -fdump-rtl-ce1
11604 @itemx -fdump-rtl-ce2
11605 @itemx -fdump-rtl-ce3
11606 @opindex fdump-rtl-ce1
11607 @opindex fdump-rtl-ce2
11608 @opindex fdump-rtl-ce3
11609 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
11610 @option{-fdump-rtl-ce3} enable dumping after the three
11611 if conversion passes.
11613 @item -fdump-rtl-cprop_hardreg
11614 @opindex fdump-rtl-cprop_hardreg
11615 Dump after hard register copy propagation.
11617 @item -fdump-rtl-csa
11618 @opindex fdump-rtl-csa
11619 Dump after combining stack adjustments.
11621 @item -fdump-rtl-cse1
11622 @itemx -fdump-rtl-cse2
11623 @opindex fdump-rtl-cse1
11624 @opindex fdump-rtl-cse2
11625 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
11626 the two common subexpression elimination passes.
11628 @item -fdump-rtl-dce
11629 @opindex fdump-rtl-dce
11630 Dump after the standalone dead code elimination passes.
11632 @item -fdump-rtl-dbr
11633 @opindex fdump-rtl-dbr
11634 Dump after delayed branch scheduling.
11636 @item -fdump-rtl-dce1
11637 @itemx -fdump-rtl-dce2
11638 @opindex fdump-rtl-dce1
11639 @opindex fdump-rtl-dce2
11640 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
11641 the two dead store elimination passes.
11643 @item -fdump-rtl-eh
11644 @opindex fdump-rtl-eh
11645 Dump after finalization of EH handling code.
11647 @item -fdump-rtl-eh_ranges
11648 @opindex fdump-rtl-eh_ranges
11649 Dump after conversion of EH handling range regions.
11651 @item -fdump-rtl-expand
11652 @opindex fdump-rtl-expand
11653 Dump after RTL generation.
11655 @item -fdump-rtl-fwprop1
11656 @itemx -fdump-rtl-fwprop2
11657 @opindex fdump-rtl-fwprop1
11658 @opindex fdump-rtl-fwprop2
11659 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
11660 dumping after the two forward propagation passes.
11662 @item -fdump-rtl-gcse1
11663 @itemx -fdump-rtl-gcse2
11664 @opindex fdump-rtl-gcse1
11665 @opindex fdump-rtl-gcse2
11666 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
11667 after global common subexpression elimination.
11669 @item -fdump-rtl-init-regs
11670 @opindex fdump-rtl-init-regs
11671 Dump after the initialization of the registers.
11673 @item -fdump-rtl-initvals
11674 @opindex fdump-rtl-initvals
11675 Dump after the computation of the initial value sets.
11677 @item -fdump-rtl-into_cfglayout
11678 @opindex fdump-rtl-into_cfglayout
11679 Dump after converting to cfglayout mode.
11681 @item -fdump-rtl-ira
11682 @opindex fdump-rtl-ira
11683 Dump after iterated register allocation.
11685 @item -fdump-rtl-jump
11686 @opindex fdump-rtl-jump
11687 Dump after the second jump optimization.
11689 @item -fdump-rtl-loop2
11690 @opindex fdump-rtl-loop2
11691 @option{-fdump-rtl-loop2} enables dumping after the rtl
11692 loop optimization passes.
11694 @item -fdump-rtl-mach
11695 @opindex fdump-rtl-mach
11696 Dump after performing the machine dependent reorganization pass, if that
11699 @item -fdump-rtl-mode_sw
11700 @opindex fdump-rtl-mode_sw
11701 Dump after removing redundant mode switches.
11703 @item -fdump-rtl-rnreg
11704 @opindex fdump-rtl-rnreg
11705 Dump after register renumbering.
11707 @item -fdump-rtl-outof_cfglayout
11708 @opindex fdump-rtl-outof_cfglayout
11709 Dump after converting from cfglayout mode.
11711 @item -fdump-rtl-peephole2
11712 @opindex fdump-rtl-peephole2
11713 Dump after the peephole pass.
11715 @item -fdump-rtl-postreload
11716 @opindex fdump-rtl-postreload
11717 Dump after post-reload optimizations.
11719 @item -fdump-rtl-pro_and_epilogue
11720 @opindex fdump-rtl-pro_and_epilogue
11721 Dump after generating the function prologues and epilogues.
11723 @item -fdump-rtl-sched1
11724 @itemx -fdump-rtl-sched2
11725 @opindex fdump-rtl-sched1
11726 @opindex fdump-rtl-sched2
11727 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
11728 after the basic block scheduling passes.
11730 @item -fdump-rtl-ree
11731 @opindex fdump-rtl-ree
11732 Dump after sign/zero extension elimination.
11734 @item -fdump-rtl-seqabstr
11735 @opindex fdump-rtl-seqabstr
11736 Dump after common sequence discovery.
11738 @item -fdump-rtl-shorten
11739 @opindex fdump-rtl-shorten
11740 Dump after shortening branches.
11742 @item -fdump-rtl-sibling
11743 @opindex fdump-rtl-sibling
11744 Dump after sibling call optimizations.
11746 @item -fdump-rtl-split1
11747 @itemx -fdump-rtl-split2
11748 @itemx -fdump-rtl-split3
11749 @itemx -fdump-rtl-split4
11750 @itemx -fdump-rtl-split5
11751 @opindex fdump-rtl-split1
11752 @opindex fdump-rtl-split2
11753 @opindex fdump-rtl-split3
11754 @opindex fdump-rtl-split4
11755 @opindex fdump-rtl-split5
11756 These options enable dumping after five rounds of
11757 instruction splitting.
11759 @item -fdump-rtl-sms
11760 @opindex fdump-rtl-sms
11761 Dump after modulo scheduling. This pass is only run on some
11764 @item -fdump-rtl-stack
11765 @opindex fdump-rtl-stack
11766 Dump after conversion from GCC's ``flat register file'' registers to the
11767 x87's stack-like registers. This pass is only run on x86 variants.
11769 @item -fdump-rtl-subreg1
11770 @itemx -fdump-rtl-subreg2
11771 @opindex fdump-rtl-subreg1
11772 @opindex fdump-rtl-subreg2
11773 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
11774 the two subreg expansion passes.
11776 @item -fdump-rtl-unshare
11777 @opindex fdump-rtl-unshare
11778 Dump after all rtl has been unshared.
11780 @item -fdump-rtl-vartrack
11781 @opindex fdump-rtl-vartrack
11782 Dump after variable tracking.
11784 @item -fdump-rtl-vregs
11785 @opindex fdump-rtl-vregs
11786 Dump after converting virtual registers to hard registers.
11788 @item -fdump-rtl-web
11789 @opindex fdump-rtl-web
11790 Dump after live range splitting.
11792 @item -fdump-rtl-regclass
11793 @itemx -fdump-rtl-subregs_of_mode_init
11794 @itemx -fdump-rtl-subregs_of_mode_finish
11795 @itemx -fdump-rtl-dfinit
11796 @itemx -fdump-rtl-dfinish
11797 @opindex fdump-rtl-regclass
11798 @opindex fdump-rtl-subregs_of_mode_init
11799 @opindex fdump-rtl-subregs_of_mode_finish
11800 @opindex fdump-rtl-dfinit
11801 @opindex fdump-rtl-dfinish
11802 These dumps are defined but always produce empty files.
11805 @itemx -fdump-rtl-all
11807 @opindex fdump-rtl-all
11808 Produce all the dumps listed above.
11812 Annotate the assembler output with miscellaneous debugging information.
11816 Dump all macro definitions, at the end of preprocessing, in addition to
11821 Produce a core dump whenever an error occurs.
11825 Annotate the assembler output with a comment indicating which
11826 pattern and alternative is used. The length of each instruction is
11831 Dump the RTL in the assembler output as a comment before each instruction.
11832 Also turns on @option{-dp} annotation.
11836 Just generate RTL for a function instead of compiling it. Usually used
11837 with @option{-fdump-rtl-expand}.
11840 @item -fdump-noaddr
11841 @opindex fdump-noaddr
11842 When doing debugging dumps, suppress address output. This makes it more
11843 feasible to use diff on debugging dumps for compiler invocations with
11844 different compiler binaries and/or different
11845 text / bss / data / heap / stack / dso start locations.
11848 @opindex freport-bug
11849 Collect and dump debug information into a temporary file if an
11850 internal compiler error (ICE) occurs.
11852 @item -fdump-unnumbered
11853 @opindex fdump-unnumbered
11854 When doing debugging dumps, suppress instruction numbers and address output.
11855 This makes it more feasible to use diff on debugging dumps for compiler
11856 invocations with different options, in particular with and without
11859 @item -fdump-unnumbered-links
11860 @opindex fdump-unnumbered-links
11861 When doing debugging dumps (see @option{-d} option above), suppress
11862 instruction numbers for the links to the previous and next instructions
11865 @item -fdump-translation-unit @r{(C++ only)}
11866 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
11867 @opindex fdump-translation-unit
11868 Dump a representation of the tree structure for the entire translation
11869 unit to a file. The file name is made by appending @file{.tu} to the
11870 source file name, and the file is created in the same directory as the
11871 output file. If the @samp{-@var{options}} form is used, @var{options}
11872 controls the details of the dump as described for the
11873 @option{-fdump-tree} options.
11875 @item -fdump-class-hierarchy @r{(C++ only)}
11876 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
11877 @opindex fdump-class-hierarchy
11878 Dump a representation of each class's hierarchy and virtual function
11879 table layout to a file. The file name is made by appending
11880 @file{.class} to the source file name, and the file is created in the
11881 same directory as the output file. If the @samp{-@var{options}} form
11882 is used, @var{options} controls the details of the dump as described
11883 for the @option{-fdump-tree} options.
11885 @item -fdump-ipa-@var{switch}
11887 Control the dumping at various stages of inter-procedural analysis
11888 language tree to a file. The file name is generated by appending a
11889 switch specific suffix to the source file name, and the file is created
11890 in the same directory as the output file. The following dumps are
11895 Enables all inter-procedural analysis dumps.
11898 Dumps information about call-graph optimization, unused function removal,
11899 and inlining decisions.
11902 Dump after function inlining.
11906 @item -fdump-passes
11907 @opindex fdump-passes
11908 Dump the list of optimization passes that are turned on and off by
11909 the current command-line options.
11911 @item -fdump-statistics-@var{option}
11912 @opindex fdump-statistics
11913 Enable and control dumping of pass statistics in a separate file. The
11914 file name is generated by appending a suffix ending in
11915 @samp{.statistics} to the source file name, and the file is created in
11916 the same directory as the output file. If the @samp{-@var{option}}
11917 form is used, @samp{-stats} causes counters to be summed over the
11918 whole compilation unit while @samp{-details} dumps every event as
11919 the passes generate them. The default with no option is to sum
11920 counters for each function compiled.
11922 @item -fdump-tree-@var{switch}
11923 @itemx -fdump-tree-@var{switch}-@var{options}
11924 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
11925 @opindex fdump-tree
11926 Control the dumping at various stages of processing the intermediate
11927 language tree to a file. The file name is generated by appending a
11928 switch-specific suffix to the source file name, and the file is
11929 created in the same directory as the output file. In case of
11930 @option{=@var{filename}} option, the dump is output on the given file
11931 instead of the auto named dump files. If the @samp{-@var{options}}
11932 form is used, @var{options} is a list of @samp{-} separated options
11933 which control the details of the dump. Not all options are applicable
11934 to all dumps; those that are not meaningful are ignored. The
11935 following options are available
11939 Print the address of each node. Usually this is not meaningful as it
11940 changes according to the environment and source file. Its primary use
11941 is for tying up a dump file with a debug environment.
11943 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
11944 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
11945 use working backward from mangled names in the assembly file.
11947 When dumping front-end intermediate representations, inhibit dumping
11948 of members of a scope or body of a function merely because that scope
11949 has been reached. Only dump such items when they are directly reachable
11950 by some other path.
11952 When dumping pretty-printed trees, this option inhibits dumping the
11953 bodies of control structures.
11955 When dumping RTL, print the RTL in slim (condensed) form instead of
11956 the default LISP-like representation.
11958 Print a raw representation of the tree. By default, trees are
11959 pretty-printed into a C-like representation.
11961 Enable more detailed dumps (not honored by every dump option). Also
11962 include information from the optimization passes.
11964 Enable dumping various statistics about the pass (not honored by every dump
11967 Enable showing basic block boundaries (disabled in raw dumps).
11969 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
11970 dump a representation of the control flow graph suitable for viewing with
11971 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
11972 the file is pretty-printed as a subgraph, so that GraphViz can render them
11973 all in a single plot.
11975 This option currently only works for RTL dumps, and the RTL is always
11976 dumped in slim form.
11978 Enable showing virtual operands for every statement.
11980 Enable showing line numbers for statements.
11982 Enable showing the unique ID (@code{DECL_UID}) for each variable.
11984 Enable showing the tree dump for each statement.
11986 Enable showing the EH region number holding each statement.
11988 Enable showing scalar evolution analysis details.
11990 Enable showing optimization information (only available in certain
11993 Enable showing missed optimization information (only available in certain
11996 Enable other detailed optimization information (only available in
11998 @item =@var{filename}
11999 Instead of an auto named dump file, output into the given file
12000 name. The file names @file{stdout} and @file{stderr} are treated
12001 specially and are considered already open standard streams. For
12005 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
12006 -fdump-tree-pre=stderr file.c
12009 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
12010 output on to @file{stderr}. If two conflicting dump filenames are
12011 given for the same pass, then the latter option overrides the earlier
12015 @opindex fdump-tree-split-paths
12016 Dump each function after splitting paths to loop backedges. The file
12017 name is made by appending @file{.split-paths} to the source file name.
12020 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
12021 and @option{lineno}.
12024 Turn on all optimization options, i.e., @option{optimized},
12025 @option{missed}, and @option{note}.
12028 The following tree dumps are possible:
12032 @opindex fdump-tree-original
12033 Dump before any tree based optimization, to @file{@var{file}.original}.
12036 @opindex fdump-tree-optimized
12037 Dump after all tree based optimization, to @file{@var{file}.optimized}.
12040 @opindex fdump-tree-gimple
12041 Dump each function before and after the gimplification pass to a file. The
12042 file name is made by appending @file{.gimple} to the source file name.
12045 @opindex fdump-tree-cfg
12046 Dump the control flow graph of each function to a file. The file name is
12047 made by appending @file{.cfg} to the source file name.
12050 @opindex fdump-tree-ch
12051 Dump each function after copying loop headers. The file name is made by
12052 appending @file{.ch} to the source file name.
12055 @opindex fdump-tree-ssa
12056 Dump SSA related information to a file. The file name is made by appending
12057 @file{.ssa} to the source file name.
12060 @opindex fdump-tree-alias
12061 Dump aliasing information for each function. The file name is made by
12062 appending @file{.alias} to the source file name.
12065 @opindex fdump-tree-ccp
12066 Dump each function after CCP@. The file name is made by appending
12067 @file{.ccp} to the source file name.
12070 @opindex fdump-tree-storeccp
12071 Dump each function after STORE-CCP@. The file name is made by appending
12072 @file{.storeccp} to the source file name.
12075 @opindex fdump-tree-pre
12076 Dump trees after partial redundancy elimination. The file name is made
12077 by appending @file{.pre} to the source file name.
12080 @opindex fdump-tree-fre
12081 Dump trees after full redundancy elimination. The file name is made
12082 by appending @file{.fre} to the source file name.
12085 @opindex fdump-tree-copyprop
12086 Dump trees after copy propagation. The file name is made
12087 by appending @file{.copyprop} to the source file name.
12089 @item store_copyprop
12090 @opindex fdump-tree-store_copyprop
12091 Dump trees after store copy-propagation. The file name is made
12092 by appending @file{.store_copyprop} to the source file name.
12095 @opindex fdump-tree-dce
12096 Dump each function after dead code elimination. The file name is made by
12097 appending @file{.dce} to the source file name.
12100 @opindex fdump-tree-sra
12101 Dump each function after performing scalar replacement of aggregates. The
12102 file name is made by appending @file{.sra} to the source file name.
12105 @opindex fdump-tree-sink
12106 Dump each function after performing code sinking. The file name is made
12107 by appending @file{.sink} to the source file name.
12110 @opindex fdump-tree-dom
12111 Dump each function after applying dominator tree optimizations. The file
12112 name is made by appending @file{.dom} to the source file name.
12115 @opindex fdump-tree-dse
12116 Dump each function after applying dead store elimination. The file
12117 name is made by appending @file{.dse} to the source file name.
12120 @opindex fdump-tree-phiopt
12121 Dump each function after optimizing PHI nodes into straightline code. The file
12122 name is made by appending @file{.phiopt} to the source file name.
12125 @opindex fdump-tree-backprop
12126 Dump each function after back-propagating use information up the definition
12127 chain. The file name is made by appending @file{.backprop} to the
12131 @opindex fdump-tree-forwprop
12132 Dump each function after forward propagating single use variables. The file
12133 name is made by appending @file{.forwprop} to the source file name.
12136 @opindex fdump-tree-nrv
12137 Dump each function after applying the named return value optimization on
12138 generic trees. The file name is made by appending @file{.nrv} to the source
12142 @opindex fdump-tree-vect
12143 Dump each function after applying vectorization of loops. The file name is
12144 made by appending @file{.vect} to the source file name.
12147 @opindex fdump-tree-slp
12148 Dump each function after applying vectorization of basic blocks. The file name
12149 is made by appending @file{.slp} to the source file name.
12152 @opindex fdump-tree-vrp
12153 Dump each function after Value Range Propagation (VRP). The file name
12154 is made by appending @file{.vrp} to the source file name.
12157 @opindex fdump-tree-oaccdevlow
12158 Dump each function after applying device-specific OpenACC transformations.
12159 The file name is made by appending @file{.oaccdevlow} to the source file name.
12162 @opindex fdump-tree-all
12163 Enable all the available tree dumps with the flags provided in this option.
12167 @itemx -fopt-info-@var{options}
12168 @itemx -fopt-info-@var{options}=@var{filename}
12170 Controls optimization dumps from various optimization passes. If the
12171 @samp{-@var{options}} form is used, @var{options} is a list of
12172 @samp{-} separated option keywords to select the dump details and
12175 The @var{options} can be divided into two groups: options describing the
12176 verbosity of the dump, and options describing which optimizations
12177 should be included. The options from both the groups can be freely
12178 mixed as they are non-overlapping. However, in case of any conflicts,
12179 the later options override the earlier options on the command
12182 The following options control the dump verbosity:
12186 Print information when an optimization is successfully applied. It is
12187 up to a pass to decide which information is relevant. For example, the
12188 vectorizer passes print the source location of loops which are
12189 successfully vectorized.
12191 Print information about missed optimizations. Individual passes
12192 control which information to include in the output.
12194 Print verbose information about optimizations, such as certain
12195 transformations, more detailed messages about decisions etc.
12197 Print detailed optimization information. This includes
12198 @samp{optimized}, @samp{missed}, and @samp{note}.
12201 One or more of the following option keywords can be used to describe a
12202 group of optimizations:
12206 Enable dumps from all interprocedural optimizations.
12208 Enable dumps from all loop optimizations.
12210 Enable dumps from all inlining optimizations.
12212 Enable dumps from all vectorization optimizations.
12214 Enable dumps from all optimizations. This is a superset of
12215 the optimization groups listed above.
12218 If @var{options} is
12219 omitted, it defaults to @samp{optimized-optall}, which means to dump all
12220 info about successful optimizations from all the passes.
12222 If the @var{filename} is provided, then the dumps from all the
12223 applicable optimizations are concatenated into the @var{filename}.
12224 Otherwise the dump is output onto @file{stderr}. Though multiple
12225 @option{-fopt-info} options are accepted, only one of them can include
12226 a @var{filename}. If other filenames are provided then all but the
12227 first such option are ignored.
12229 Note that the output @var{filename} is overwritten
12230 in case of multiple translation units. If a combined output from
12231 multiple translation units is desired, @file{stderr} should be used
12234 In the following example, the optimization info is output to
12243 gcc -O3 -fopt-info-missed=missed.all
12247 outputs missed optimization report from all the passes into
12248 @file{missed.all}, and this one:
12251 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
12255 prints information about missed optimization opportunities from
12256 vectorization passes on @file{stderr}.
12257 Note that @option{-fopt-info-vec-missed} is equivalent to
12258 @option{-fopt-info-missed-vec}.
12260 As another example,
12262 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
12266 outputs information about missed optimizations as well as
12267 optimized locations from all the inlining passes into
12273 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
12277 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
12278 in conflict since only one output file is allowed. In this case, only
12279 the first option takes effect and the subsequent options are
12280 ignored. Thus only @file{vec.miss} is produced which contains
12281 dumps from the vectorizer about missed opportunities.
12283 @item -fsched-verbose=@var{n}
12284 @opindex fsched-verbose
12285 On targets that use instruction scheduling, this option controls the
12286 amount of debugging output the scheduler prints to the dump files.
12288 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
12289 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
12290 For @var{n} greater than one, it also output basic block probabilities,
12291 detailed ready list information and unit/insn info. For @var{n} greater
12292 than two, it includes RTL at abort point, control-flow and regions info.
12293 And for @var{n} over four, @option{-fsched-verbose} also includes
12298 @item -fenable-@var{kind}-@var{pass}
12299 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
12303 This is a set of options that are used to explicitly disable/enable
12304 optimization passes. These options are intended for use for debugging GCC.
12305 Compiler users should use regular options for enabling/disabling
12310 @item -fdisable-ipa-@var{pass}
12311 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12312 statically invoked in the compiler multiple times, the pass name should be
12313 appended with a sequential number starting from 1.
12315 @item -fdisable-rtl-@var{pass}
12316 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
12317 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
12318 statically invoked in the compiler multiple times, the pass name should be
12319 appended with a sequential number starting from 1. @var{range-list} is a
12320 comma-separated list of function ranges or assembler names. Each range is a number
12321 pair separated by a colon. The range is inclusive in both ends. If the range
12322 is trivial, the number pair can be simplified as a single number. If the
12323 function's call graph node's @var{uid} falls within one of the specified ranges,
12324 the @var{pass} is disabled for that function. The @var{uid} is shown in the
12325 function header of a dump file, and the pass names can be dumped by using
12326 option @option{-fdump-passes}.
12328 @item -fdisable-tree-@var{pass}
12329 @itemx -fdisable-tree-@var{pass}=@var{range-list}
12330 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
12333 @item -fenable-ipa-@var{pass}
12334 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12335 statically invoked in the compiler multiple times, the pass name should be
12336 appended with a sequential number starting from 1.
12338 @item -fenable-rtl-@var{pass}
12339 @itemx -fenable-rtl-@var{pass}=@var{range-list}
12340 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
12341 description and examples.
12343 @item -fenable-tree-@var{pass}
12344 @itemx -fenable-tree-@var{pass}=@var{range-list}
12345 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
12346 of option arguments.
12350 Here are some examples showing uses of these options.
12354 # disable ccp1 for all functions
12355 -fdisable-tree-ccp1
12356 # disable complete unroll for function whose cgraph node uid is 1
12357 -fenable-tree-cunroll=1
12358 # disable gcse2 for functions at the following ranges [1,1],
12359 # [300,400], and [400,1000]
12360 # disable gcse2 for functions foo and foo2
12361 -fdisable-rtl-gcse2=foo,foo2
12362 # disable early inlining
12363 -fdisable-tree-einline
12364 # disable ipa inlining
12365 -fdisable-ipa-inline
12366 # enable tree full unroll
12367 -fenable-tree-unroll
12372 @itemx -fchecking=@var{n}
12374 @opindex fno-checking
12375 Enable internal consistency checking. The default depends on
12376 the compiler configuration. @option{-fchecking=2} enables further
12377 internal consistency checking that might affect code generation.
12379 @item -frandom-seed=@var{string}
12380 @opindex frandom-seed
12381 This option provides a seed that GCC uses in place of
12382 random numbers in generating certain symbol names
12383 that have to be different in every compiled file. It is also used to
12384 place unique stamps in coverage data files and the object files that
12385 produce them. You can use the @option{-frandom-seed} option to produce
12386 reproducibly identical object files.
12388 The @var{string} can either be a number (decimal, octal or hex) or an
12389 arbitrary string (in which case it's converted to a number by
12392 The @var{string} should be different for every file you compile.
12395 @itemx -save-temps=cwd
12396 @opindex save-temps
12397 Store the usual ``temporary'' intermediate files permanently; place them
12398 in the current directory and name them based on the source file. Thus,
12399 compiling @file{foo.c} with @option{-c -save-temps} produces files
12400 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
12401 preprocessed @file{foo.i} output file even though the compiler now
12402 normally uses an integrated preprocessor.
12404 When used in combination with the @option{-x} command-line option,
12405 @option{-save-temps} is sensible enough to avoid over writing an
12406 input source file with the same extension as an intermediate file.
12407 The corresponding intermediate file may be obtained by renaming the
12408 source file before using @option{-save-temps}.
12410 If you invoke GCC in parallel, compiling several different source
12411 files that share a common base name in different subdirectories or the
12412 same source file compiled for multiple output destinations, it is
12413 likely that the different parallel compilers will interfere with each
12414 other, and overwrite the temporary files. For instance:
12417 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
12418 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
12421 may result in @file{foo.i} and @file{foo.o} being written to
12422 simultaneously by both compilers.
12424 @item -save-temps=obj
12425 @opindex save-temps=obj
12426 Store the usual ``temporary'' intermediate files permanently. If the
12427 @option{-o} option is used, the temporary files are based on the
12428 object file. If the @option{-o} option is not used, the
12429 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
12434 gcc -save-temps=obj -c foo.c
12435 gcc -save-temps=obj -c bar.c -o dir/xbar.o
12436 gcc -save-temps=obj foobar.c -o dir2/yfoobar
12440 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
12441 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
12442 @file{dir2/yfoobar.o}.
12444 @item -time@r{[}=@var{file}@r{]}
12446 Report the CPU time taken by each subprocess in the compilation
12447 sequence. For C source files, this is the compiler proper and assembler
12448 (plus the linker if linking is done).
12450 Without the specification of an output file, the output looks like this:
12457 The first number on each line is the ``user time'', that is time spent
12458 executing the program itself. The second number is ``system time'',
12459 time spent executing operating system routines on behalf of the program.
12460 Both numbers are in seconds.
12462 With the specification of an output file, the output is appended to the
12463 named file, and it looks like this:
12466 0.12 0.01 cc1 @var{options}
12467 0.00 0.01 as @var{options}
12470 The ``user time'' and the ``system time'' are moved before the program
12471 name, and the options passed to the program are displayed, so that one
12472 can later tell what file was being compiled, and with which options.
12474 @item -fdump-final-insns@r{[}=@var{file}@r{]}
12475 @opindex fdump-final-insns
12476 Dump the final internal representation (RTL) to @var{file}. If the
12477 optional argument is omitted (or if @var{file} is @code{.}), the name
12478 of the dump file is determined by appending @code{.gkd} to the
12479 compilation output file name.
12481 @item -fcompare-debug@r{[}=@var{opts}@r{]}
12482 @opindex fcompare-debug
12483 @opindex fno-compare-debug
12484 If no error occurs during compilation, run the compiler a second time,
12485 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
12486 passed to the second compilation. Dump the final internal
12487 representation in both compilations, and print an error if they differ.
12489 If the equal sign is omitted, the default @option{-gtoggle} is used.
12491 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
12492 and nonzero, implicitly enables @option{-fcompare-debug}. If
12493 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
12494 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
12497 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
12498 is equivalent to @option{-fno-compare-debug}, which disables the dumping
12499 of the final representation and the second compilation, preventing even
12500 @env{GCC_COMPARE_DEBUG} from taking effect.
12502 To verify full coverage during @option{-fcompare-debug} testing, set
12503 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
12504 which GCC rejects as an invalid option in any actual compilation
12505 (rather than preprocessing, assembly or linking). To get just a
12506 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
12507 not overridden} will do.
12509 @item -fcompare-debug-second
12510 @opindex fcompare-debug-second
12511 This option is implicitly passed to the compiler for the second
12512 compilation requested by @option{-fcompare-debug}, along with options to
12513 silence warnings, and omitting other options that would cause
12514 side-effect compiler outputs to files or to the standard output. Dump
12515 files and preserved temporary files are renamed so as to contain the
12516 @code{.gk} additional extension during the second compilation, to avoid
12517 overwriting those generated by the first.
12519 When this option is passed to the compiler driver, it causes the
12520 @emph{first} compilation to be skipped, which makes it useful for little
12521 other than debugging the compiler proper.
12525 Turn off generation of debug info, if leaving out this option
12526 generates it, or turn it on at level 2 otherwise. The position of this
12527 argument in the command line does not matter; it takes effect after all
12528 other options are processed, and it does so only once, no matter how
12529 many times it is given. This is mainly intended to be used with
12530 @option{-fcompare-debug}.
12532 @item -fvar-tracking-assignments-toggle
12533 @opindex fvar-tracking-assignments-toggle
12534 @opindex fno-var-tracking-assignments-toggle
12535 Toggle @option{-fvar-tracking-assignments}, in the same way that
12536 @option{-gtoggle} toggles @option{-g}.
12540 Makes the compiler print out each function name as it is compiled, and
12541 print some statistics about each pass when it finishes.
12543 @item -ftime-report
12544 @opindex ftime-report
12545 Makes the compiler print some statistics about the time consumed by each
12546 pass when it finishes.
12548 @item -fira-verbose=@var{n}
12549 @opindex fira-verbose
12550 Control the verbosity of the dump file for the integrated register allocator.
12551 The default value is 5. If the value @var{n} is greater or equal to 10,
12552 the dump output is sent to stderr using the same format as @var{n} minus 10.
12555 @opindex flto-report
12556 Prints a report with internal details on the workings of the link-time
12557 optimizer. The contents of this report vary from version to version.
12558 It is meant to be useful to GCC developers when processing object
12559 files in LTO mode (via @option{-flto}).
12561 Disabled by default.
12563 @item -flto-report-wpa
12564 @opindex flto-report-wpa
12565 Like @option{-flto-report}, but only print for the WPA phase of Link
12569 @opindex fmem-report
12570 Makes the compiler print some statistics about permanent memory
12571 allocation when it finishes.
12573 @item -fmem-report-wpa
12574 @opindex fmem-report-wpa
12575 Makes the compiler print some statistics about permanent memory
12576 allocation for the WPA phase only.
12578 @item -fpre-ipa-mem-report
12579 @opindex fpre-ipa-mem-report
12580 @item -fpost-ipa-mem-report
12581 @opindex fpost-ipa-mem-report
12582 Makes the compiler print some statistics about permanent memory
12583 allocation before or after interprocedural optimization.
12585 @item -fprofile-report
12586 @opindex fprofile-report
12587 Makes the compiler print some statistics about consistency of the
12588 (estimated) profile and effect of individual passes.
12590 @item -fstack-usage
12591 @opindex fstack-usage
12592 Makes the compiler output stack usage information for the program, on a
12593 per-function basis. The filename for the dump is made by appending
12594 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
12595 the output file, if explicitly specified and it is not an executable,
12596 otherwise it is the basename of the source file. An entry is made up
12601 The name of the function.
12605 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
12608 The qualifier @code{static} means that the function manipulates the stack
12609 statically: a fixed number of bytes are allocated for the frame on function
12610 entry and released on function exit; no stack adjustments are otherwise made
12611 in the function. The second field is this fixed number of bytes.
12613 The qualifier @code{dynamic} means that the function manipulates the stack
12614 dynamically: in addition to the static allocation described above, stack
12615 adjustments are made in the body of the function, for example to push/pop
12616 arguments around function calls. If the qualifier @code{bounded} is also
12617 present, the amount of these adjustments is bounded at compile time and
12618 the second field is an upper bound of the total amount of stack used by
12619 the function. If it is not present, the amount of these adjustments is
12620 not bounded at compile time and the second field only represents the
12625 Emit statistics about front-end processing at the end of the compilation.
12626 This option is supported only by the C++ front end, and
12627 the information is generally only useful to the G++ development team.
12629 @item -fdbg-cnt-list
12630 @opindex fdbg-cnt-list
12631 Print the name and the counter upper bound for all debug counters.
12634 @item -fdbg-cnt=@var{counter-value-list}
12636 Set the internal debug counter upper bound. @var{counter-value-list}
12637 is a comma-separated list of @var{name}:@var{value} pairs
12638 which sets the upper bound of each debug counter @var{name} to @var{value}.
12639 All debug counters have the initial upper bound of @code{UINT_MAX};
12640 thus @code{dbg_cnt} returns true always unless the upper bound
12641 is set by this option.
12642 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
12643 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
12645 @item -print-file-name=@var{library}
12646 @opindex print-file-name
12647 Print the full absolute name of the library file @var{library} that
12648 would be used when linking---and don't do anything else. With this
12649 option, GCC does not compile or link anything; it just prints the
12652 @item -print-multi-directory
12653 @opindex print-multi-directory
12654 Print the directory name corresponding to the multilib selected by any
12655 other switches present in the command line. This directory is supposed
12656 to exist in @env{GCC_EXEC_PREFIX}.
12658 @item -print-multi-lib
12659 @opindex print-multi-lib
12660 Print the mapping from multilib directory names to compiler switches
12661 that enable them. The directory name is separated from the switches by
12662 @samp{;}, and each switch starts with an @samp{@@} instead of the
12663 @samp{-}, without spaces between multiple switches. This is supposed to
12664 ease shell processing.
12666 @item -print-multi-os-directory
12667 @opindex print-multi-os-directory
12668 Print the path to OS libraries for the selected
12669 multilib, relative to some @file{lib} subdirectory. If OS libraries are
12670 present in the @file{lib} subdirectory and no multilibs are used, this is
12671 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
12672 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
12673 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
12674 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
12676 @item -print-multiarch
12677 @opindex print-multiarch
12678 Print the path to OS libraries for the selected multiarch,
12679 relative to some @file{lib} subdirectory.
12681 @item -print-prog-name=@var{program}
12682 @opindex print-prog-name
12683 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
12685 @item -print-libgcc-file-name
12686 @opindex print-libgcc-file-name
12687 Same as @option{-print-file-name=libgcc.a}.
12689 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
12690 but you do want to link with @file{libgcc.a}. You can do:
12693 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
12696 @item -print-search-dirs
12697 @opindex print-search-dirs
12698 Print the name of the configured installation directory and a list of
12699 program and library directories @command{gcc} searches---and don't do anything else.
12701 This is useful when @command{gcc} prints the error message
12702 @samp{installation problem, cannot exec cpp0: No such file or directory}.
12703 To resolve this you either need to put @file{cpp0} and the other compiler
12704 components where @command{gcc} expects to find them, or you can set the environment
12705 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
12706 Don't forget the trailing @samp{/}.
12707 @xref{Environment Variables}.
12709 @item -print-sysroot
12710 @opindex print-sysroot
12711 Print the target sysroot directory that is used during
12712 compilation. This is the target sysroot specified either at configure
12713 time or using the @option{--sysroot} option, possibly with an extra
12714 suffix that depends on compilation options. If no target sysroot is
12715 specified, the option prints nothing.
12717 @item -print-sysroot-headers-suffix
12718 @opindex print-sysroot-headers-suffix
12719 Print the suffix added to the target sysroot when searching for
12720 headers, or give an error if the compiler is not configured with such
12721 a suffix---and don't do anything else.
12724 @opindex dumpmachine
12725 Print the compiler's target machine (for example,
12726 @samp{i686-pc-linux-gnu})---and don't do anything else.
12729 @opindex dumpversion
12730 Print the compiler version (for example, @code{3.0})---and don't do
12735 Print the compiler's built-in specs---and don't do anything else. (This
12736 is used when GCC itself is being built.) @xref{Spec Files}.
12739 @node Submodel Options
12740 @section Machine-Dependent Options
12741 @cindex submodel options
12742 @cindex specifying hardware config
12743 @cindex hardware models and configurations, specifying
12744 @cindex target-dependent options
12745 @cindex machine-dependent options
12747 Each target machine supported by GCC can have its own options---for
12748 example, to allow you to compile for a particular processor variant or
12749 ABI, or to control optimizations specific to that machine. By
12750 convention, the names of machine-specific options start with
12753 Some configurations of the compiler also support additional target-specific
12754 options, usually for compatibility with other compilers on the same
12757 @c This list is ordered alphanumerically by subsection name.
12758 @c It should be the same order and spelling as these options are listed
12759 @c in Machine Dependent Options
12762 * AArch64 Options::
12763 * Adapteva Epiphany Options::
12767 * Blackfin Options::
12772 * DEC Alpha Options::
12776 * GNU/Linux Options::
12786 * MicroBlaze Options::
12789 * MN10300 Options::
12793 * Nios II Options::
12794 * Nvidia PTX Options::
12796 * picoChip Options::
12797 * PowerPC Options::
12799 * RS/6000 and PowerPC Options::
12801 * S/390 and zSeries Options::
12804 * Solaris 2 Options::
12807 * System V Options::
12808 * TILE-Gx Options::
12809 * TILEPro Options::
12814 * VxWorks Options::
12816 * x86 Windows Options::
12817 * Xstormy16 Options::
12819 * zSeries Options::
12822 @node AArch64 Options
12823 @subsection AArch64 Options
12824 @cindex AArch64 Options
12826 These options are defined for AArch64 implementations:
12830 @item -mabi=@var{name}
12832 Generate code for the specified data model. Permissible values
12833 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12834 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12835 but long int and pointer are 64-bit.
12837 The default depends on the specific target configuration. Note that
12838 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12839 entire program with the same ABI, and link with a compatible set of libraries.
12842 @opindex mbig-endian
12843 Generate big-endian code. This is the default when GCC is configured for an
12844 @samp{aarch64_be-*-*} target.
12846 @item -mgeneral-regs-only
12847 @opindex mgeneral-regs-only
12848 Generate code which uses only the general-purpose registers. This will prevent
12849 the compiler from using floating-point and Advanced SIMD registers but will not
12850 impose any restrictions on the assembler.
12852 @item -mlittle-endian
12853 @opindex mlittle-endian
12854 Generate little-endian code. This is the default when GCC is configured for an
12855 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12857 @item -mcmodel=tiny
12858 @opindex mcmodel=tiny
12859 Generate code for the tiny code model. The program and its statically defined
12860 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12861 be statically or dynamically linked. This model is not fully implemented and
12862 mostly treated as @samp{small}.
12864 @item -mcmodel=small
12865 @opindex mcmodel=small
12866 Generate code for the small code model. The program and its statically defined
12867 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12868 be statically or dynamically linked. This is the default code model.
12870 @item -mcmodel=large
12871 @opindex mcmodel=large
12872 Generate code for the large code model. This makes no assumptions about
12873 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12874 statically linked only.
12876 @item -mstrict-align
12877 @opindex mstrict-align
12878 Do not assume that unaligned memory references are handled by the system.
12880 @item -momit-leaf-frame-pointer
12881 @itemx -mno-omit-leaf-frame-pointer
12882 @opindex momit-leaf-frame-pointer
12883 @opindex mno-omit-leaf-frame-pointer
12884 Omit or keep the frame pointer in leaf functions. The former behavior is the
12887 @item -mtls-dialect=desc
12888 @opindex mtls-dialect=desc
12889 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12890 of TLS variables. This is the default.
12892 @item -mtls-dialect=traditional
12893 @opindex mtls-dialect=traditional
12894 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12897 @item -mtls-size=@var{size}
12899 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
12900 This option depends on binutils higher than 2.25.
12902 @item -mfix-cortex-a53-835769
12903 @itemx -mno-fix-cortex-a53-835769
12904 @opindex mfix-cortex-a53-835769
12905 @opindex mno-fix-cortex-a53-835769
12906 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12907 This involves inserting a NOP instruction between memory instructions and
12908 64-bit integer multiply-accumulate instructions.
12910 @item -mfix-cortex-a53-843419
12911 @itemx -mno-fix-cortex-a53-843419
12912 @opindex mfix-cortex-a53-843419
12913 @opindex mno-fix-cortex-a53-843419
12914 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12915 This erratum workaround is made at link time and this will only pass the
12916 corresponding flag to the linker.
12918 @item -mlow-precision-recip-sqrt
12919 @item -mno-low-precision-recip-sqrt
12920 @opindex -mlow-precision-recip-sqrt
12921 @opindex -mno-low-precision-recip-sqrt
12922 When calculating the reciprocal square root approximation,
12923 uses one less step than otherwise, thus reducing latency and precision.
12924 This is only relevant if @option{-ffast-math} enables the reciprocal square root
12925 approximation, which in turn depends on the target processor.
12927 @item -march=@var{name}
12929 Specify the name of the target architecture and, optionally, one or
12930 more feature modifiers. This option has the form
12931 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12933 The permissible values for @var{arch} are @samp{armv8-a},
12934 @samp{armv8.1-a} or @var{native}.
12936 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
12937 support for the ARMv8.1 architecture extension. In particular, it
12938 enables the @samp{+crc} and @samp{+lse} features.
12940 The value @samp{native} is available on native AArch64 GNU/Linux and
12941 causes the compiler to pick the architecture of the host system. This
12942 option has no effect if the compiler is unable to recognize the
12943 architecture of the host system,
12945 The permissible values for @var{feature} are listed in the sub-section
12946 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12947 Feature Modifiers}. Where conflicting feature modifiers are
12948 specified, the right-most feature is used.
12950 GCC uses @var{name} to determine what kind of instructions it can emit
12951 when generating assembly code. If @option{-march} is specified
12952 without either of @option{-mtune} or @option{-mcpu} also being
12953 specified, the code is tuned to perform well across a range of target
12954 processors implementing the target architecture.
12956 @item -mtune=@var{name}
12958 Specify the name of the target processor for which GCC should tune the
12959 performance of the code. Permissible values for this option are:
12960 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
12961 @samp{cortex-a72}, @samp{exynos-m1}, @samp{qdf24xx}, @samp{thunderx},
12964 Additionally, this option can specify that GCC should tune the performance
12965 of the code for a big.LITTLE system. Permissible values for this
12966 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12968 Additionally on native AArch64 GNU/Linux systems the value
12969 @samp{native} is available. This option causes the compiler to pick
12970 the architecture of and tune the performance of the code for the
12971 processor of the host system. This option has no effect if the
12972 compiler is unable to recognize the architecture of the host system.
12974 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12975 are specified, the code is tuned to perform well across a range
12976 of target processors.
12978 This option cannot be suffixed by feature modifiers.
12980 @item -mcpu=@var{name}
12982 Specify the name of the target processor, optionally suffixed by one
12983 or more feature modifiers. This option has the form
12984 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
12985 the permissible values for @var{cpu} are the same as those available
12986 for @option{-mtune}. The permissible values for @var{feature} are
12987 documented in the sub-section on
12988 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12989 Feature Modifiers}. Where conflicting feature modifiers are
12990 specified, the right-most feature is used.
12992 Additionally on native AArch64 GNU/Linux systems the value
12993 @samp{native} is available. This option causes the compiler to tune
12994 the performance of the code for the processor of the host system.
12995 This option has no effect if the compiler is unable to recognize the
12996 architecture of the host system.
12998 GCC uses @var{name} to determine what kind of instructions it can emit when
12999 generating assembly code (as if by @option{-march}) and to determine
13000 the target processor for which to tune for performance (as if
13001 by @option{-mtune}). Where this option is used in conjunction
13002 with @option{-march} or @option{-mtune}, those options take precedence
13003 over the appropriate part of this option.
13005 @item -moverride=@var{string}
13007 Override tuning decisions made by the back-end in response to a
13008 @option{-mtune=} switch. The syntax, semantics, and accepted values
13009 for @var{string} in this option are not guaranteed to be consistent
13012 This option is only intended to be useful when developing GCC.
13014 @item -mpc-relative-literal-loads
13015 @opindex mpcrelativeliteralloads
13016 Enable PC relative literal loads. If this option is used, literal
13017 pools are assumed to have a range of up to 1MiB and an appropriate
13018 instruction sequence is used. This option has no impact when used
13019 with @option{-mcmodel=tiny}.
13023 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
13024 @anchor{aarch64-feature-modifiers}
13025 @cindex @option{-march} feature modifiers
13026 @cindex @option{-mcpu} feature modifiers
13027 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
13028 the following and their inverses @option{no@var{feature}}:
13032 Enable CRC extension. This is on by default for
13033 @option{-march=armv8.1-a}.
13035 Enable Crypto extension. This also enables Advanced SIMD and floating-point
13038 Enable floating-point instructions. This is on by default for all possible
13039 values for options @option{-march} and @option{-mcpu}.
13041 Enable Advanced SIMD instructions. This also enables floating-point
13042 instructions. This is on by default for all possible values for options
13043 @option{-march} and @option{-mcpu}.
13045 Enable Large System Extension instructions. This is on by default for
13046 @option{-march=armv8.1-a}.
13050 That is, @option{crypto} implies @option{simd} implies @option{fp}.
13051 Conversely, @option{nofp} (or equivalently, @option{-mgeneral-regs-only})
13052 implies @option{nosimd} implies @option{nocrypto}.
13054 @node Adapteva Epiphany Options
13055 @subsection Adapteva Epiphany Options
13057 These @samp{-m} options are defined for Adapteva Epiphany:
13060 @item -mhalf-reg-file
13061 @opindex mhalf-reg-file
13062 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
13063 That allows code to run on hardware variants that lack these registers.
13065 @item -mprefer-short-insn-regs
13066 @opindex mprefer-short-insn-regs
13067 Preferentially allocate registers that allow short instruction generation.
13068 This can result in increased instruction count, so this may either reduce or
13069 increase overall code size.
13071 @item -mbranch-cost=@var{num}
13072 @opindex mbranch-cost
13073 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13074 This cost is only a heuristic and is not guaranteed to produce
13075 consistent results across releases.
13079 Enable the generation of conditional moves.
13081 @item -mnops=@var{num}
13083 Emit @var{num} NOPs before every other generated instruction.
13085 @item -mno-soft-cmpsf
13086 @opindex mno-soft-cmpsf
13087 For single-precision floating-point comparisons, emit an @code{fsub} instruction
13088 and test the flags. This is faster than a software comparison, but can
13089 get incorrect results in the presence of NaNs, or when two different small
13090 numbers are compared such that their difference is calculated as zero.
13091 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
13092 software comparisons.
13094 @item -mstack-offset=@var{num}
13095 @opindex mstack-offset
13096 Set the offset between the top of the stack and the stack pointer.
13097 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
13098 can be used by leaf functions without stack allocation.
13099 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
13100 Note also that this option changes the ABI; compiling a program with a
13101 different stack offset than the libraries have been compiled with
13102 generally does not work.
13103 This option can be useful if you want to evaluate if a different stack
13104 offset would give you better code, but to actually use a different stack
13105 offset to build working programs, it is recommended to configure the
13106 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
13108 @item -mno-round-nearest
13109 @opindex mno-round-nearest
13110 Make the scheduler assume that the rounding mode has been set to
13111 truncating. The default is @option{-mround-nearest}.
13114 @opindex mlong-calls
13115 If not otherwise specified by an attribute, assume all calls might be beyond
13116 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
13117 function address into a register before performing a (otherwise direct) call.
13118 This is the default.
13120 @item -mshort-calls
13121 @opindex short-calls
13122 If not otherwise specified by an attribute, assume all direct calls are
13123 in the range of the @code{b} / @code{bl} instructions, so use these instructions
13124 for direct calls. The default is @option{-mlong-calls}.
13128 Assume addresses can be loaded as 16-bit unsigned values. This does not
13129 apply to function addresses for which @option{-mlong-calls} semantics
13132 @item -mfp-mode=@var{mode}
13134 Set the prevailing mode of the floating-point unit.
13135 This determines the floating-point mode that is provided and expected
13136 at function call and return time. Making this mode match the mode you
13137 predominantly need at function start can make your programs smaller and
13138 faster by avoiding unnecessary mode switches.
13140 @var{mode} can be set to one the following values:
13144 Any mode at function entry is valid, and retained or restored when
13145 the function returns, and when it calls other functions.
13146 This mode is useful for compiling libraries or other compilation units
13147 you might want to incorporate into different programs with different
13148 prevailing FPU modes, and the convenience of being able to use a single
13149 object file outweighs the size and speed overhead for any extra
13150 mode switching that might be needed, compared with what would be needed
13151 with a more specific choice of prevailing FPU mode.
13154 This is the mode used for floating-point calculations with
13155 truncating (i.e.@: round towards zero) rounding mode. That includes
13156 conversion from floating point to integer.
13158 @item round-nearest
13159 This is the mode used for floating-point calculations with
13160 round-to-nearest-or-even rounding mode.
13163 This is the mode used to perform integer calculations in the FPU, e.g.@:
13164 integer multiply, or integer multiply-and-accumulate.
13167 The default is @option{-mfp-mode=caller}
13169 @item -mnosplit-lohi
13170 @itemx -mno-postinc
13171 @itemx -mno-postmodify
13172 @opindex mnosplit-lohi
13173 @opindex mno-postinc
13174 @opindex mno-postmodify
13175 Code generation tweaks that disable, respectively, splitting of 32-bit
13176 loads, generation of post-increment addresses, and generation of
13177 post-modify addresses. The defaults are @option{msplit-lohi},
13178 @option{-mpost-inc}, and @option{-mpost-modify}.
13180 @item -mnovect-double
13181 @opindex mno-vect-double
13182 Change the preferred SIMD mode to SImode. The default is
13183 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
13185 @item -max-vect-align=@var{num}
13186 @opindex max-vect-align
13187 The maximum alignment for SIMD vector mode types.
13188 @var{num} may be 4 or 8. The default is 8.
13189 Note that this is an ABI change, even though many library function
13190 interfaces are unaffected if they don't use SIMD vector modes
13191 in places that affect size and/or alignment of relevant types.
13193 @item -msplit-vecmove-early
13194 @opindex msplit-vecmove-early
13195 Split vector moves into single word moves before reload. In theory this
13196 can give better register allocation, but so far the reverse seems to be
13197 generally the case.
13199 @item -m1reg-@var{reg}
13201 Specify a register to hold the constant @minus{}1, which makes loading small negative
13202 constants and certain bitmasks faster.
13203 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
13204 which specify use of that register as a fixed register,
13205 and @samp{none}, which means that no register is used for this
13206 purpose. The default is @option{-m1reg-none}.
13211 @subsection ARC Options
13212 @cindex ARC options
13214 The following options control the architecture variant for which code
13217 @c architecture variants
13220 @item -mbarrel-shifter
13221 @opindex mbarrel-shifter
13222 Generate instructions supported by barrel shifter. This is the default
13223 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
13225 @item -mcpu=@var{cpu}
13227 Set architecture type, register usage, and instruction scheduling
13228 parameters for @var{cpu}. There are also shortcut alias options
13229 available for backward compatibility and convenience. Supported
13230 values for @var{cpu} are
13237 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
13242 Compile for ARC601. Alias: @option{-mARC601}.
13248 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
13249 This is the default when configured with @option{--with-cpu=arc700}@.
13253 Compile for ARC EM.
13257 Compile for ARC HS.
13262 @itemx -mdpfp-compact
13263 @opindex mdpfp-compact
13264 FPX: Generate Double Precision FPX instructions, tuned for the compact
13268 @opindex mdpfp-fast
13269 FPX: Generate Double Precision FPX instructions, tuned for the fast
13272 @item -mno-dpfp-lrsr
13273 @opindex mno-dpfp-lrsr
13274 Disable LR and SR instructions from using FPX extension aux registers.
13278 Generate Extended arithmetic instructions. Currently only
13279 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
13280 supported. This is always enabled for @option{-mcpu=ARC700}.
13284 Do not generate mpy instructions for ARC700.
13288 Generate 32x16 bit multiply and mac instructions.
13292 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
13296 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
13301 @itemx -mspfp-compact
13302 @opindex mspfp-compact
13303 FPX: Generate Single Precision FPX instructions, tuned for the compact
13307 @opindex mspfp-fast
13308 FPX: Generate Single Precision FPX instructions, tuned for the fast
13313 Enable generation of ARC SIMD instructions via target-specific
13314 builtins. Only valid for @option{-mcpu=ARC700}.
13317 @opindex msoft-float
13318 This option ignored; it is provided for compatibility purposes only.
13319 Software floating point code is emitted by default, and this default
13320 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
13321 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
13322 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
13326 Generate swap instructions.
13330 This enables Locked Load/Store Conditional extension to implement
13331 atomic memopry built-in functions. Not available for ARC 6xx or ARC
13336 Enable DIV/REM instructions for ARCv2 cores.
13338 @item -mcode-density
13339 @opindex mcode-density
13340 Enable code density instructions for ARC EM, default on for ARC HS.
13344 Enable double load/store operations for ARC HS cores.
13346 @item -mmpy-option=@var{multo}
13347 @opindex mmpy-option
13348 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
13349 the default value. The recognized values for @var{multo} are:
13353 No multiplier available.
13357 The multiply option is set to w: 16x16 multiplier, fully pipelined.
13358 The following instructions are enabled: MPYW, and MPYUW.
13362 The multiply option is set to wlh1: 32x32 multiplier, fully
13363 pipelined (1 stage). The following instructions are additionally
13364 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
13368 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
13369 (2 stages). The following instructions are additionally enabled: MPY,
13370 MPYU, MPYM, MPYMU, and MPY_S.
13374 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
13375 sequential. The following instructions are additionally enabled: MPY,
13376 MPYU, MPYM, MPYMU, and MPY_S.
13380 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
13381 sequential. The following instructions are additionally enabled: MPY,
13382 MPYU, MPYM, MPYMU, and MPY_S.
13386 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
13387 sequential. The following instructions are additionally enabled: MPY,
13388 MPYU, MPYM, MPYMU, and MPY_S.
13392 This option is only available for ARCv2 cores@.
13394 @item -mfpu=@var{fpu}
13396 Enables specific floating-point hardware extension for ARCv2
13397 core. Supported values for @var{fpu} are:
13403 Enables support for single precision floating point hardware
13408 Enables support for double precision floating point hardware
13409 extensions. The single precision floating point extension is also
13410 enabled. Not available for ARC EM@.
13414 Enables support for double precision floating point hardware
13415 extensions using double precision assist instructions. The single
13416 precision floating point extension is also enabled. This option is
13417 only available for ARC EM@.
13421 Enables support for double precision floating point hardware
13422 extensions using double precision assist instructions, and simple
13423 precision square-root and divide hardware extensions. The single
13424 precision floating point extension is also enabled. This option is
13425 only available for ARC EM@.
13429 Enables support for double precision floating point hardware
13430 extensions using double precision assist instructions, and simple
13431 precision fused multiple and add hardware extension. The single
13432 precision floating point extension is also enabled. This option is
13433 only available for ARC EM@.
13437 Enables support for double precision floating point hardware
13438 extensions using double precision assist instructions, and all simple
13439 precision hardware extensions. The single precision floating point
13440 extension is also enabled. This option is only available for ARC EM@.
13444 Enables support for single precision floating point, and single
13445 precision square-root and divide hardware extensions@.
13449 Enables support for double precision floating point, and double
13450 precision square-root and divide hardware extensions. This option
13451 includes option @samp{fpus_div}. Not available for ARC EM@.
13455 Enables support for single precision floating point, and single
13456 precision fused multiple and add hardware extensions@.
13460 Enables support for double precision floating point, and double
13461 precision fused multiple and add hardware extensions. This option
13462 includes option @samp{fpus_fma}. Not available for ARC EM@.
13466 Enables support for all single precision floating point hardware
13471 Enables support for all single and double precision floating point
13472 hardware extensions. Not available for ARC EM@.
13478 The following options are passed through to the assembler, and also
13479 define preprocessor macro symbols.
13481 @c Flags used by the assembler, but for which we define preprocessor
13482 @c macro symbols as well.
13485 @opindex mdsp-packa
13486 Passed down to the assembler to enable the DSP Pack A extensions.
13487 Also sets the preprocessor symbol @code{__Xdsp_packa}.
13491 Passed down to the assembler to enable the dual viterbi butterfly
13492 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
13494 @c ARC700 4.10 extension instruction
13497 Passed down to the assembler to enable the Locked Load/Store
13498 Conditional extension. Also sets the preprocessor symbol
13503 Passed down to the assembler. Also sets the preprocessor symbol
13504 @code{__Xxmac_d16}.
13508 Passed down to the assembler. Also sets the preprocessor symbol
13511 @c ARC700 4.10 extension instruction
13514 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
13515 extension instruction. Also sets the preprocessor symbol
13518 @c ARC700 4.10 extension instruction
13521 Passed down to the assembler to enable the swap byte ordering
13522 extension instruction. Also sets the preprocessor symbol
13526 @opindex mtelephony
13527 Passed down to the assembler to enable dual and single operand
13528 instructions for telephony. Also sets the preprocessor symbol
13529 @code{__Xtelephony}.
13533 Passed down to the assembler to enable the XY Memory extension. Also
13534 sets the preprocessor symbol @code{__Xxy}.
13538 The following options control how the assembly code is annotated:
13540 @c Assembly annotation options
13544 Annotate assembler instructions with estimated addresses.
13546 @item -mannotate-align
13547 @opindex mannotate-align
13548 Explain what alignment considerations lead to the decision to make an
13549 instruction short or long.
13553 The following options are passed through to the linker:
13555 @c options passed through to the linker
13559 Passed through to the linker, to specify use of the @code{arclinux} emulation.
13560 This option is enabled by default in tool chains built for
13561 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
13562 when profiling is not requested.
13564 @item -marclinux_prof
13565 @opindex marclinux_prof
13566 Passed through to the linker, to specify use of the
13567 @code{arclinux_prof} emulation. This option is enabled by default in
13568 tool chains built for @w{@code{arc-linux-uclibc}} and
13569 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
13573 The following options control the semantics of generated code:
13575 @c semantically relevant code generation options
13578 @opindex mlong-calls
13579 Generate call insns as register indirect calls, thus providing access
13580 to the full 32-bit address range.
13582 @item -mmedium-calls
13583 @opindex mmedium-calls
13584 Don't use less than 25 bit addressing range for calls, which is the
13585 offset available for an unconditional branch-and-link
13586 instruction. Conditional execution of function calls is suppressed, to
13587 allow use of the 25-bit range, rather than the 21-bit range with
13588 conditional branch-and-link. This is the default for tool chains built
13589 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13593 Do not generate sdata references. This is the default for tool chains
13594 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13598 @opindex mucb-mcount
13599 Instrument with mcount calls as used in UCB code. I.e. do the
13600 counting in the callee, not the caller. By default ARC instrumentation
13601 counts in the caller.
13603 @item -mvolatile-cache
13604 @opindex mvolatile-cache
13605 Use ordinarily cached memory accesses for volatile references. This is the
13608 @item -mno-volatile-cache
13609 @opindex mno-volatile-cache
13610 Enable cache bypass for volatile references.
13614 The following options fine tune code generation:
13615 @c code generation tuning options
13618 @opindex malign-call
13619 Do alignment optimizations for call instructions.
13621 @item -mauto-modify-reg
13622 @opindex mauto-modify-reg
13623 Enable the use of pre/post modify with register displacement.
13625 @item -mbbit-peephole
13626 @opindex mbbit-peephole
13627 Enable bbit peephole2.
13631 This option disables a target-specific pass in @file{arc_reorg} to
13632 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13633 generation driven by the combiner pass.
13635 @item -mcase-vector-pcrel
13636 @opindex mcase-vector-pcrel
13637 Use pc-relative switch case tables - this enables case table shortening.
13638 This is the default for @option{-Os}.
13640 @item -mcompact-casesi
13641 @opindex mcompact-casesi
13642 Enable compact casesi pattern.
13643 This is the default for @option{-Os}.
13645 @item -mno-cond-exec
13646 @opindex mno-cond-exec
13647 Disable ARCompact specific pass to generate conditional execution instructions.
13648 Due to delay slot scheduling and interactions between operand numbers,
13649 literal sizes, instruction lengths, and the support for conditional execution,
13650 the target-independent pass to generate conditional execution is often lacking,
13651 so the ARC port has kept a special pass around that tries to find more
13652 conditional execution generating opportunities after register allocation,
13653 branch shortening, and delay slot scheduling have been done. This pass
13654 generally, but not always, improves performance and code size, at the cost of
13655 extra compilation time, which is why there is an option to switch it off.
13656 If you have a problem with call instructions exceeding their allowable
13657 offset range because they are conditionalized, you should consider using
13658 @option{-mmedium-calls} instead.
13660 @item -mearly-cbranchsi
13661 @opindex mearly-cbranchsi
13662 Enable pre-reload use of the cbranchsi pattern.
13664 @item -mexpand-adddi
13665 @opindex mexpand-adddi
13666 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13667 @code{add.f}, @code{adc} etc.
13669 @item -mindexed-loads
13670 @opindex mindexed-loads
13671 Enable the use of indexed loads. This can be problematic because some
13672 optimizers then assume that indexed stores exist, which is not
13677 Enable Local Register Allocation. This is still experimental for ARC,
13678 so by default the compiler uses standard reload
13679 (i.e. @option{-mno-lra}).
13681 @item -mlra-priority-none
13682 @opindex mlra-priority-none
13683 Don't indicate any priority for target registers.
13685 @item -mlra-priority-compact
13686 @opindex mlra-priority-compact
13687 Indicate target register priority for r0..r3 / r12..r15.
13689 @item -mlra-priority-noncompact
13690 @opindex mlra-priority-noncompact
13691 Reduce target register priority for r0..r3 / r12..r15.
13693 @item -mno-millicode
13694 @opindex mno-millicode
13695 When optimizing for size (using @option{-Os}), prologues and epilogues
13696 that have to save or restore a large number of registers are often
13697 shortened by using call to a special function in libgcc; this is
13698 referred to as a @emph{millicode} call. As these calls can pose
13699 performance issues, and/or cause linking issues when linking in a
13700 nonstandard way, this option is provided to turn off millicode call
13704 @opindex mmixed-code
13705 Tweak register allocation to help 16-bit instruction generation.
13706 This generally has the effect of decreasing the average instruction size
13707 while increasing the instruction count.
13711 Enable 'q' instruction alternatives.
13712 This is the default for @option{-Os}.
13716 Enable Rcq constraint handling - most short code generation depends on this.
13717 This is the default.
13721 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13722 This is the default.
13724 @item -msize-level=@var{level}
13725 @opindex msize-level
13726 Fine-tune size optimization with regards to instruction lengths and alignment.
13727 The recognized values for @var{level} are:
13730 No size optimization. This level is deprecated and treated like @samp{1}.
13733 Short instructions are used opportunistically.
13736 In addition, alignment of loops and of code after barriers are dropped.
13739 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13743 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13744 the behavior when this is not set is equivalent to level @samp{1}.
13746 @item -mtune=@var{cpu}
13748 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13749 by @option{-mcpu=}.
13751 Supported values for @var{cpu} are
13755 Tune for ARC600 cpu.
13758 Tune for ARC601 cpu.
13761 Tune for ARC700 cpu with standard multiplier block.
13764 Tune for ARC700 cpu with XMAC block.
13767 Tune for ARC725D cpu.
13770 Tune for ARC750D cpu.
13774 @item -mmultcost=@var{num}
13776 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13777 normal instruction.
13779 @item -munalign-prob-threshold=@var{probability}
13780 @opindex munalign-prob-threshold
13781 Set probability threshold for unaligning branches.
13782 When tuning for @samp{ARC700} and optimizing for speed, branches without
13783 filled delay slot are preferably emitted unaligned and long, unless
13784 profiling indicates that the probability for the branch to be taken
13785 is below @var{probability}. @xref{Cross-profiling}.
13786 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13790 The following options are maintained for backward compatibility, but
13791 are now deprecated and will be removed in a future release:
13793 @c Deprecated options
13801 @opindex mbig-endian
13804 Compile code for big endian targets. Use of these options is now
13805 deprecated. Users wanting big-endian code, should use the
13806 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13807 building the tool chain, for which big-endian is the default.
13809 @item -mlittle-endian
13810 @opindex mlittle-endian
13813 Compile code for little endian targets. Use of these options is now
13814 deprecated. Users wanting little-endian code should use the
13815 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13816 building the tool chain, for which little-endian is the default.
13818 @item -mbarrel_shifter
13819 @opindex mbarrel_shifter
13820 Replaced by @option{-mbarrel-shifter}.
13822 @item -mdpfp_compact
13823 @opindex mdpfp_compact
13824 Replaced by @option{-mdpfp-compact}.
13827 @opindex mdpfp_fast
13828 Replaced by @option{-mdpfp-fast}.
13831 @opindex mdsp_packa
13832 Replaced by @option{-mdsp-packa}.
13836 Replaced by @option{-mea}.
13840 Replaced by @option{-mmac-24}.
13844 Replaced by @option{-mmac-d16}.
13846 @item -mspfp_compact
13847 @opindex mspfp_compact
13848 Replaced by @option{-mspfp-compact}.
13851 @opindex mspfp_fast
13852 Replaced by @option{-mspfp-fast}.
13854 @item -mtune=@var{cpu}
13856 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13857 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13858 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13860 @item -multcost=@var{num}
13862 Replaced by @option{-mmultcost}.
13867 @subsection ARM Options
13868 @cindex ARM options
13870 These @samp{-m} options are defined for the ARM port:
13873 @item -mabi=@var{name}
13875 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13876 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13879 @opindex mapcs-frame
13880 Generate a stack frame that is compliant with the ARM Procedure Call
13881 Standard for all functions, even if this is not strictly necessary for
13882 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13883 with this option causes the stack frames not to be generated for
13884 leaf functions. The default is @option{-mno-apcs-frame}.
13885 This option is deprecated.
13889 This is a synonym for @option{-mapcs-frame} and is deprecated.
13892 @c not currently implemented
13893 @item -mapcs-stack-check
13894 @opindex mapcs-stack-check
13895 Generate code to check the amount of stack space available upon entry to
13896 every function (that actually uses some stack space). If there is
13897 insufficient space available then either the function
13898 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13899 called, depending upon the amount of stack space required. The runtime
13900 system is required to provide these functions. The default is
13901 @option{-mno-apcs-stack-check}, since this produces smaller code.
13903 @c not currently implemented
13905 @opindex mapcs-float
13906 Pass floating-point arguments using the floating-point registers. This is
13907 one of the variants of the APCS@. This option is recommended if the
13908 target hardware has a floating-point unit or if a lot of floating-point
13909 arithmetic is going to be performed by the code. The default is
13910 @option{-mno-apcs-float}, since the size of integer-only code is
13911 slightly increased if @option{-mapcs-float} is used.
13913 @c not currently implemented
13914 @item -mapcs-reentrant
13915 @opindex mapcs-reentrant
13916 Generate reentrant, position-independent code. The default is
13917 @option{-mno-apcs-reentrant}.
13920 @item -mthumb-interwork
13921 @opindex mthumb-interwork
13922 Generate code that supports calling between the ARM and Thumb
13923 instruction sets. Without this option, on pre-v5 architectures, the
13924 two instruction sets cannot be reliably used inside one program. The
13925 default is @option{-mno-thumb-interwork}, since slightly larger code
13926 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13927 configurations this option is meaningless.
13929 @item -mno-sched-prolog
13930 @opindex mno-sched-prolog
13931 Prevent the reordering of instructions in the function prologue, or the
13932 merging of those instruction with the instructions in the function's
13933 body. This means that all functions start with a recognizable set
13934 of instructions (or in fact one of a choice from a small set of
13935 different function prologues), and this information can be used to
13936 locate the start of functions inside an executable piece of code. The
13937 default is @option{-msched-prolog}.
13939 @item -mfloat-abi=@var{name}
13940 @opindex mfloat-abi
13941 Specifies which floating-point ABI to use. Permissible values
13942 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13944 Specifying @samp{soft} causes GCC to generate output containing
13945 library calls for floating-point operations.
13946 @samp{softfp} allows the generation of code using hardware floating-point
13947 instructions, but still uses the soft-float calling conventions.
13948 @samp{hard} allows generation of floating-point instructions
13949 and uses FPU-specific calling conventions.
13951 The default depends on the specific target configuration. Note that
13952 the hard-float and soft-float ABIs are not link-compatible; you must
13953 compile your entire program with the same ABI, and link with a
13954 compatible set of libraries.
13956 @item -mlittle-endian
13957 @opindex mlittle-endian
13958 Generate code for a processor running in little-endian mode. This is
13959 the default for all standard configurations.
13962 @opindex mbig-endian
13963 Generate code for a processor running in big-endian mode; the default is
13964 to compile code for a little-endian processor.
13966 @item -march=@var{name}
13968 This specifies the name of the target ARM architecture. GCC uses this
13969 name to determine what kind of instructions it can emit when generating
13970 assembly code. This option can be used in conjunction with or instead
13971 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13972 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13973 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13974 @samp{armv6}, @samp{armv6j},
13975 @samp{armv6t2}, @samp{armv6z}, @samp{armv6kz}, @samp{armv6-m},
13976 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13977 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
13978 @samp{armv8.1-a+crc}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13980 Architecture revisions older than @option{armv4t} are deprecated.
13982 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13985 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13986 architecture together with the optional CRC32 extensions.
13988 @option{-march=native} causes the compiler to auto-detect the architecture
13989 of the build computer. At present, this feature is only supported on
13990 GNU/Linux, and not all architectures are recognized. If the auto-detect
13991 is unsuccessful the option has no effect.
13993 @item -mtune=@var{name}
13995 This option specifies the name of the target ARM processor for
13996 which GCC should tune the performance of the code.
13997 For some ARM implementations better performance can be obtained by using
13999 Permissible names are: @samp{arm2}, @samp{arm250},
14000 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
14001 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
14002 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
14003 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
14005 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
14006 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
14007 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
14008 @samp{strongarm1110},
14009 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
14010 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
14011 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
14012 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
14013 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
14014 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
14015 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
14016 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
14017 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
14018 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14019 @samp{cortex-a72}, @samp{cortex-r4},
14020 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
14026 @samp{cortex-m0plus},
14027 @samp{cortex-m1.small-multiply},
14028 @samp{cortex-m0.small-multiply},
14029 @samp{cortex-m0plus.small-multiply},
14032 @samp{marvell-pj4},
14033 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
14034 @samp{fa526}, @samp{fa626},
14035 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
14038 Additionally, this option can specify that GCC should tune the performance
14039 of the code for a big.LITTLE system. Permissible names are:
14040 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
14041 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
14043 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
14044 performance for a blend of processors within architecture @var{arch}.
14045 The aim is to generate code that run well on the current most popular
14046 processors, balancing between optimizations that benefit some CPUs in the
14047 range, and avoiding performance pitfalls of other CPUs. The effects of
14048 this option may change in future GCC versions as CPU models come and go.
14050 @option{-mtune=native} causes the compiler to auto-detect the CPU
14051 of the build computer. At present, this feature is only supported on
14052 GNU/Linux, and not all architectures are recognized. If the auto-detect is
14053 unsuccessful the option has no effect.
14055 @item -mcpu=@var{name}
14057 This specifies the name of the target ARM processor. GCC uses this name
14058 to derive the name of the target ARM architecture (as if specified
14059 by @option{-march}) and the ARM processor type for which to tune for
14060 performance (as if specified by @option{-mtune}). Where this option
14061 is used in conjunction with @option{-march} or @option{-mtune},
14062 those options take precedence over the appropriate part of this option.
14064 Permissible names for this option are the same as those for
14067 @option{-mcpu=generic-@var{arch}} is also permissible, and is
14068 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
14069 See @option{-mtune} for more information.
14071 @option{-mcpu=native} causes the compiler to auto-detect the CPU
14072 of the build computer. At present, this feature is only supported on
14073 GNU/Linux, and not all architectures are recognized. If the auto-detect
14074 is unsuccessful the option has no effect.
14076 @item -mfpu=@var{name}
14078 This specifies what floating-point hardware (or hardware emulation) is
14079 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
14080 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
14081 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
14082 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
14083 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
14084 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
14086 If @option{-msoft-float} is specified this specifies the format of
14087 floating-point values.
14089 If the selected floating-point hardware includes the NEON extension
14090 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
14091 operations are not generated by GCC's auto-vectorization pass unless
14092 @option{-funsafe-math-optimizations} is also specified. This is
14093 because NEON hardware does not fully implement the IEEE 754 standard for
14094 floating-point arithmetic (in particular denormal values are treated as
14095 zero), so the use of NEON instructions may lead to a loss of precision.
14097 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}).
14099 @item -mfp16-format=@var{name}
14100 @opindex mfp16-format
14101 Specify the format of the @code{__fp16} half-precision floating-point type.
14102 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
14103 the default is @samp{none}, in which case the @code{__fp16} type is not
14104 defined. @xref{Half-Precision}, for more information.
14106 @item -mstructure-size-boundary=@var{n}
14107 @opindex mstructure-size-boundary
14108 The sizes of all structures and unions are rounded up to a multiple
14109 of the number of bits set by this option. Permissible values are 8, 32
14110 and 64. The default value varies for different toolchains. For the COFF
14111 targeted toolchain the default value is 8. A value of 64 is only allowed
14112 if the underlying ABI supports it.
14114 Specifying a larger number can produce faster, more efficient code, but
14115 can also increase the size of the program. Different values are potentially
14116 incompatible. Code compiled with one value cannot necessarily expect to
14117 work with code or libraries compiled with another value, if they exchange
14118 information using structures or unions.
14120 @item -mabort-on-noreturn
14121 @opindex mabort-on-noreturn
14122 Generate a call to the function @code{abort} at the end of a
14123 @code{noreturn} function. It is executed if the function tries to
14127 @itemx -mno-long-calls
14128 @opindex mlong-calls
14129 @opindex mno-long-calls
14130 Tells the compiler to perform function calls by first loading the
14131 address of the function into a register and then performing a subroutine
14132 call on this register. This switch is needed if the target function
14133 lies outside of the 64-megabyte addressing range of the offset-based
14134 version of subroutine call instruction.
14136 Even if this switch is enabled, not all function calls are turned
14137 into long calls. The heuristic is that static functions, functions
14138 that have the @code{short_call} attribute, functions that are inside
14139 the scope of a @code{#pragma no_long_calls} directive, and functions whose
14140 definitions have already been compiled within the current compilation
14141 unit are not turned into long calls. The exceptions to this rule are
14142 that weak function definitions, functions with the @code{long_call}
14143 attribute or the @code{section} attribute, and functions that are within
14144 the scope of a @code{#pragma long_calls} directive are always
14145 turned into long calls.
14147 This feature is not enabled by default. Specifying
14148 @option{-mno-long-calls} restores the default behavior, as does
14149 placing the function calls within the scope of a @code{#pragma
14150 long_calls_off} directive. Note these switches have no effect on how
14151 the compiler generates code to handle function calls via function
14154 @item -msingle-pic-base
14155 @opindex msingle-pic-base
14156 Treat the register used for PIC addressing as read-only, rather than
14157 loading it in the prologue for each function. The runtime system is
14158 responsible for initializing this register with an appropriate value
14159 before execution begins.
14161 @item -mpic-register=@var{reg}
14162 @opindex mpic-register
14163 Specify the register to be used for PIC addressing.
14164 For standard PIC base case, the default is any suitable register
14165 determined by compiler. For single PIC base case, the default is
14166 @samp{R9} if target is EABI based or stack-checking is enabled,
14167 otherwise the default is @samp{R10}.
14169 @item -mpic-data-is-text-relative
14170 @opindex mpic-data-is-text-relative
14171 Assume that each data segments are relative to text segment at load time.
14172 Therefore, it permits addressing data using PC-relative operations.
14173 This option is on by default for targets other than VxWorks RTP.
14175 @item -mpoke-function-name
14176 @opindex mpoke-function-name
14177 Write the name of each function into the text section, directly
14178 preceding the function prologue. The generated code is similar to this:
14182 .ascii "arm_poke_function_name", 0
14185 .word 0xff000000 + (t1 - t0)
14186 arm_poke_function_name
14188 stmfd sp!, @{fp, ip, lr, pc@}
14192 When performing a stack backtrace, code can inspect the value of
14193 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
14194 location @code{pc - 12} and the top 8 bits are set, then we know that
14195 there is a function name embedded immediately preceding this location
14196 and has length @code{((pc[-3]) & 0xff000000)}.
14203 Select between generating code that executes in ARM and Thumb
14204 states. The default for most configurations is to generate code
14205 that executes in ARM state, but the default can be changed by
14206 configuring GCC with the @option{--with-mode=}@var{state}
14209 You can also override the ARM and Thumb mode for each function
14210 by using the @code{target("thumb")} and @code{target("arm")} function attributes
14211 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
14214 @opindex mtpcs-frame
14215 Generate a stack frame that is compliant with the Thumb Procedure Call
14216 Standard for all non-leaf functions. (A leaf function is one that does
14217 not call any other functions.) The default is @option{-mno-tpcs-frame}.
14219 @item -mtpcs-leaf-frame
14220 @opindex mtpcs-leaf-frame
14221 Generate a stack frame that is compliant with the Thumb Procedure Call
14222 Standard for all leaf functions. (A leaf function is one that does
14223 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
14225 @item -mcallee-super-interworking
14226 @opindex mcallee-super-interworking
14227 Gives all externally visible functions in the file being compiled an ARM
14228 instruction set header which switches to Thumb mode before executing the
14229 rest of the function. This allows these functions to be called from
14230 non-interworking code. This option is not valid in AAPCS configurations
14231 because interworking is enabled by default.
14233 @item -mcaller-super-interworking
14234 @opindex mcaller-super-interworking
14235 Allows calls via function pointers (including virtual functions) to
14236 execute correctly regardless of whether the target code has been
14237 compiled for interworking or not. There is a small overhead in the cost
14238 of executing a function pointer if this option is enabled. This option
14239 is not valid in AAPCS configurations because interworking is enabled
14242 @item -mtp=@var{name}
14244 Specify the access model for the thread local storage pointer. The valid
14245 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
14246 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
14247 (supported in the arm6k architecture), and @samp{auto}, which uses the
14248 best available method for the selected processor. The default setting is
14251 @item -mtls-dialect=@var{dialect}
14252 @opindex mtls-dialect
14253 Specify the dialect to use for accessing thread local storage. Two
14254 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
14255 @samp{gnu} dialect selects the original GNU scheme for supporting
14256 local and global dynamic TLS models. The @samp{gnu2} dialect
14257 selects the GNU descriptor scheme, which provides better performance
14258 for shared libraries. The GNU descriptor scheme is compatible with
14259 the original scheme, but does require new assembler, linker and
14260 library support. Initial and local exec TLS models are unaffected by
14261 this option and always use the original scheme.
14263 @item -mword-relocations
14264 @opindex mword-relocations
14265 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
14266 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
14267 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
14270 @item -mfix-cortex-m3-ldrd
14271 @opindex mfix-cortex-m3-ldrd
14272 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
14273 with overlapping destination and base registers are used. This option avoids
14274 generating these instructions. This option is enabled by default when
14275 @option{-mcpu=cortex-m3} is specified.
14277 @item -munaligned-access
14278 @itemx -mno-unaligned-access
14279 @opindex munaligned-access
14280 @opindex mno-unaligned-access
14281 Enables (or disables) reading and writing of 16- and 32- bit values
14282 from addresses that are not 16- or 32- bit aligned. By default
14283 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
14284 architectures, and enabled for all other architectures. If unaligned
14285 access is not enabled then words in packed data structures are
14286 accessed a byte at a time.
14288 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
14289 generated object file to either true or false, depending upon the
14290 setting of this option. If unaligned access is enabled then the
14291 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
14294 @item -mneon-for-64bits
14295 @opindex mneon-for-64bits
14296 Enables using Neon to handle scalar 64-bits operations. This is
14297 disabled by default since the cost of moving data from core registers
14300 @item -mslow-flash-data
14301 @opindex mslow-flash-data
14302 Assume loading data from flash is slower than fetching instruction.
14303 Therefore literal load is minimized for better performance.
14304 This option is only supported when compiling for ARMv7 M-profile and
14307 @item -masm-syntax-unified
14308 @opindex masm-syntax-unified
14309 Assume inline assembler is using unified asm syntax. The default is
14310 currently off which implies divided syntax. This option has no impact
14311 on Thumb2. However, this may change in future releases of GCC.
14312 Divided syntax should be considered deprecated.
14314 @item -mrestrict-it
14315 @opindex mrestrict-it
14316 Restricts generation of IT blocks to conform to the rules of ARMv8.
14317 IT blocks can only contain a single 16-bit instruction from a select
14318 set of instructions. This option is on by default for ARMv8 Thumb mode.
14320 @item -mprint-tune-info
14321 @opindex mprint-tune-info
14322 Print CPU tuning information as comment in assembler file. This is
14323 an option used only for regression testing of the compiler and not
14324 intended for ordinary use in compiling code. This option is disabled
14329 @subsection AVR Options
14330 @cindex AVR Options
14332 These options are defined for AVR implementations:
14335 @item -mmcu=@var{mcu}
14337 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
14339 The default for this option is@tie{}@samp{avr2}.
14341 GCC supports the following AVR devices and ISAs:
14343 @include avr-mmcu.texi
14345 @item -maccumulate-args
14346 @opindex maccumulate-args
14347 Accumulate outgoing function arguments and acquire/release the needed
14348 stack space for outgoing function arguments once in function
14349 prologue/epilogue. Without this option, outgoing arguments are pushed
14350 before calling a function and popped afterwards.
14352 Popping the arguments after the function call can be expensive on
14353 AVR so that accumulating the stack space might lead to smaller
14354 executables because arguments need not to be removed from the
14355 stack after such a function call.
14357 This option can lead to reduced code size for functions that perform
14358 several calls to functions that get their arguments on the stack like
14359 calls to printf-like functions.
14361 @item -mbranch-cost=@var{cost}
14362 @opindex mbranch-cost
14363 Set the branch costs for conditional branch instructions to
14364 @var{cost}. Reasonable values for @var{cost} are small, non-negative
14365 integers. The default branch cost is 0.
14367 @item -mcall-prologues
14368 @opindex mcall-prologues
14369 Functions prologues/epilogues are expanded as calls to appropriate
14370 subroutines. Code size is smaller.
14374 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
14375 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
14376 and @code{long long} is 4 bytes. Please note that this option does not
14377 conform to the C standards, but it results in smaller code
14380 @item -mn-flash=@var{num}
14382 Assume that the flash memory has a size of
14383 @var{num} times 64@tie{}KiB.
14385 @item -mno-interrupts
14386 @opindex mno-interrupts
14387 Generated code is not compatible with hardware interrupts.
14388 Code size is smaller.
14392 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
14393 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
14394 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
14395 the assembler's command line and the @option{--relax} option to the
14396 linker's command line.
14398 Jump relaxing is performed by the linker because jump offsets are not
14399 known before code is located. Therefore, the assembler code generated by the
14400 compiler is the same, but the instructions in the executable may
14401 differ from instructions in the assembler code.
14403 Relaxing must be turned on if linker stubs are needed, see the
14404 section on @code{EIND} and linker stubs below.
14408 Assume that the device supports the Read-Modify-Write
14409 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
14413 Treat the stack pointer register as an 8-bit register,
14414 i.e.@: assume the high byte of the stack pointer is zero.
14415 In general, you don't need to set this option by hand.
14417 This option is used internally by the compiler to select and
14418 build multilibs for architectures @code{avr2} and @code{avr25}.
14419 These architectures mix devices with and without @code{SPH}.
14420 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
14421 the compiler driver adds or removes this option from the compiler
14422 proper's command line, because the compiler then knows if the device
14423 or architecture has an 8-bit stack pointer and thus no @code{SPH}
14428 Use address register @code{X} in a way proposed by the hardware. This means
14429 that @code{X} is only used in indirect, post-increment or
14430 pre-decrement addressing.
14432 Without this option, the @code{X} register may be used in the same way
14433 as @code{Y} or @code{Z} which then is emulated by additional
14435 For example, loading a value with @code{X+const} addressing with a
14436 small non-negative @code{const < 64} to a register @var{Rn} is
14440 adiw r26, const ; X += const
14441 ld @var{Rn}, X ; @var{Rn} = *X
14442 sbiw r26, const ; X -= const
14446 @opindex mtiny-stack
14447 Only change the lower 8@tie{}bits of the stack pointer.
14450 @opindex nodevicelib
14451 Don't link against AVR-LibC's device specific library @code{libdev.a}.
14453 @item -Waddr-space-convert
14454 @opindex Waddr-space-convert
14455 Warn about conversions between address spaces in the case where the
14456 resulting address space is not contained in the incoming address space.
14459 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
14460 @cindex @code{EIND}
14461 Pointers in the implementation are 16@tie{}bits wide.
14462 The address of a function or label is represented as word address so
14463 that indirect jumps and calls can target any code address in the
14464 range of 64@tie{}Ki words.
14466 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
14467 bytes of program memory space, there is a special function register called
14468 @code{EIND} that serves as most significant part of the target address
14469 when @code{EICALL} or @code{EIJMP} instructions are used.
14471 Indirect jumps and calls on these devices are handled as follows by
14472 the compiler and are subject to some limitations:
14477 The compiler never sets @code{EIND}.
14480 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
14481 instructions or might read @code{EIND} directly in order to emulate an
14482 indirect call/jump by means of a @code{RET} instruction.
14485 The compiler assumes that @code{EIND} never changes during the startup
14486 code or during the application. In particular, @code{EIND} is not
14487 saved/restored in function or interrupt service routine
14491 For indirect calls to functions and computed goto, the linker
14492 generates @emph{stubs}. Stubs are jump pads sometimes also called
14493 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
14494 The stub contains a direct jump to the desired address.
14497 Linker relaxation must be turned on so that the linker generates
14498 the stubs correctly in all situations. See the compiler option
14499 @option{-mrelax} and the linker option @option{--relax}.
14500 There are corner cases where the linker is supposed to generate stubs
14501 but aborts without relaxation and without a helpful error message.
14504 The default linker script is arranged for code with @code{EIND = 0}.
14505 If code is supposed to work for a setup with @code{EIND != 0}, a custom
14506 linker script has to be used in order to place the sections whose
14507 name start with @code{.trampolines} into the segment where @code{EIND}
14511 The startup code from libgcc never sets @code{EIND}.
14512 Notice that startup code is a blend of code from libgcc and AVR-LibC.
14513 For the impact of AVR-LibC on @code{EIND}, see the
14514 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
14517 It is legitimate for user-specific startup code to set up @code{EIND}
14518 early, for example by means of initialization code located in
14519 section @code{.init3}. Such code runs prior to general startup code
14520 that initializes RAM and calls constructors, but after the bit
14521 of startup code from AVR-LibC that sets @code{EIND} to the segment
14522 where the vector table is located.
14524 #include <avr/io.h>
14527 __attribute__((section(".init3"),naked,used,no_instrument_function))
14528 init3_set_eind (void)
14530 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
14531 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
14536 The @code{__trampolines_start} symbol is defined in the linker script.
14539 Stubs are generated automatically by the linker if
14540 the following two conditions are met:
14543 @item The address of a label is taken by means of the @code{gs} modifier
14544 (short for @emph{generate stubs}) like so:
14546 LDI r24, lo8(gs(@var{func}))
14547 LDI r25, hi8(gs(@var{func}))
14549 @item The final location of that label is in a code segment
14550 @emph{outside} the segment where the stubs are located.
14554 The compiler emits such @code{gs} modifiers for code labels in the
14555 following situations:
14557 @item Taking address of a function or code label.
14558 @item Computed goto.
14559 @item If prologue-save function is used, see @option{-mcall-prologues}
14560 command-line option.
14561 @item Switch/case dispatch tables. If you do not want such dispatch
14562 tables you can specify the @option{-fno-jump-tables} command-line option.
14563 @item C and C++ constructors/destructors called during startup/shutdown.
14564 @item If the tools hit a @code{gs()} modifier explained above.
14568 Jumping to non-symbolic addresses like so is @emph{not} supported:
14573 /* Call function at word address 0x2 */
14574 return ((int(*)(void)) 0x2)();
14578 Instead, a stub has to be set up, i.e.@: the function has to be called
14579 through a symbol (@code{func_4} in the example):
14584 extern int func_4 (void);
14586 /* Call function at byte address 0x4 */
14591 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
14592 Alternatively, @code{func_4} can be defined in the linker script.
14595 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14596 @cindex @code{RAMPD}
14597 @cindex @code{RAMPX}
14598 @cindex @code{RAMPY}
14599 @cindex @code{RAMPZ}
14600 Some AVR devices support memories larger than the 64@tie{}KiB range
14601 that can be accessed with 16-bit pointers. To access memory locations
14602 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14603 register is used as high part of the address:
14604 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14605 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14606 register, respectively, to get a wide address. Similarly,
14607 @code{RAMPD} is used together with direct addressing.
14611 The startup code initializes the @code{RAMP} special function
14612 registers with zero.
14615 If a @ref{AVR Named Address Spaces,named address space} other than
14616 generic or @code{__flash} is used, then @code{RAMPZ} is set
14617 as needed before the operation.
14620 If the device supports RAM larger than 64@tie{}KiB and the compiler
14621 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14622 is reset to zero after the operation.
14625 If the device comes with a specific @code{RAMP} register, the ISR
14626 prologue/epilogue saves/restores that SFR and initializes it with
14627 zero in case the ISR code might (implicitly) use it.
14630 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14631 If you use inline assembler to read from locations outside the
14632 16-bit address range and change one of the @code{RAMP} registers,
14633 you must reset it to zero after the access.
14637 @subsubsection AVR Built-in Macros
14639 GCC defines several built-in macros so that the user code can test
14640 for the presence or absence of features. Almost any of the following
14641 built-in macros are deduced from device capabilities and thus
14642 triggered by the @option{-mmcu=} command-line option.
14644 For even more AVR-specific built-in macros see
14645 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14650 Build-in macro that resolves to a decimal number that identifies the
14651 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14652 Possible values are:
14654 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14655 @code{4}, @code{5}, @code{51}, @code{6}
14657 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14658 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14662 @code{100}, @code{102}, @code{104},
14663 @code{105}, @code{106}, @code{107}
14665 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14666 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14667 If @var{mcu} specifies a device, this built-in macro is set
14668 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14669 defined to @code{4}.
14671 @item __AVR_@var{Device}__
14672 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14673 the device's name. For example, @option{-mmcu=atmega8} defines the
14674 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14675 @code{__AVR_ATtiny261A__}, etc.
14677 The built-in macros' names follow
14678 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14679 the device name as from the AVR user manual. The difference between
14680 @var{Device} in the built-in macro and @var{device} in
14681 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14683 If @var{device} is not a device but only a core architecture like
14684 @samp{avr51}, this macro is not defined.
14686 @item __AVR_DEVICE_NAME__
14687 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14688 the device's name. For example, with @option{-mmcu=atmega8} the macro
14689 is defined to @code{atmega8}.
14691 If @var{device} is not a device but only a core architecture like
14692 @samp{avr51}, this macro is not defined.
14694 @item __AVR_XMEGA__
14695 The device / architecture belongs to the XMEGA family of devices.
14697 @item __AVR_HAVE_ELPM__
14698 The device has the @code{ELPM} instruction.
14700 @item __AVR_HAVE_ELPMX__
14701 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14702 R@var{n},Z+} instructions.
14704 @item __AVR_HAVE_MOVW__
14705 The device has the @code{MOVW} instruction to perform 16-bit
14706 register-register moves.
14708 @item __AVR_HAVE_LPMX__
14709 The device has the @code{LPM R@var{n},Z} and
14710 @code{LPM R@var{n},Z+} instructions.
14712 @item __AVR_HAVE_MUL__
14713 The device has a hardware multiplier.
14715 @item __AVR_HAVE_JMP_CALL__
14716 The device has the @code{JMP} and @code{CALL} instructions.
14717 This is the case for devices with at least 16@tie{}KiB of program
14720 @item __AVR_HAVE_EIJMP_EICALL__
14721 @itemx __AVR_3_BYTE_PC__
14722 The device has the @code{EIJMP} and @code{EICALL} instructions.
14723 This is the case for devices with more than 128@tie{}KiB of program memory.
14724 This also means that the program counter
14725 (PC) is 3@tie{}bytes wide.
14727 @item __AVR_2_BYTE_PC__
14728 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14729 with up to 128@tie{}KiB of program memory.
14731 @item __AVR_HAVE_8BIT_SP__
14732 @itemx __AVR_HAVE_16BIT_SP__
14733 The stack pointer (SP) register is treated as 8-bit respectively
14734 16-bit register by the compiler.
14735 The definition of these macros is affected by @option{-mtiny-stack}.
14737 @item __AVR_HAVE_SPH__
14739 The device has the SPH (high part of stack pointer) special function
14740 register or has an 8-bit stack pointer, respectively.
14741 The definition of these macros is affected by @option{-mmcu=} and
14742 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14745 @item __AVR_HAVE_RAMPD__
14746 @itemx __AVR_HAVE_RAMPX__
14747 @itemx __AVR_HAVE_RAMPY__
14748 @itemx __AVR_HAVE_RAMPZ__
14749 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14750 @code{RAMPZ} special function register, respectively.
14752 @item __NO_INTERRUPTS__
14753 This macro reflects the @option{-mno-interrupts} command-line option.
14755 @item __AVR_ERRATA_SKIP__
14756 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14757 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14758 instructions because of a hardware erratum. Skip instructions are
14759 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14760 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14763 @item __AVR_ISA_RMW__
14764 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14766 @item __AVR_SFR_OFFSET__=@var{offset}
14767 Instructions that can address I/O special function registers directly
14768 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14769 address as if addressed by an instruction to access RAM like @code{LD}
14770 or @code{STS}. This offset depends on the device architecture and has
14771 to be subtracted from the RAM address in order to get the
14772 respective I/O@tie{}address.
14774 @item __WITH_AVRLIBC__
14775 The compiler is configured to be used together with AVR-Libc.
14776 See the @option{--with-avrlibc} configure option.
14780 @node Blackfin Options
14781 @subsection Blackfin Options
14782 @cindex Blackfin Options
14785 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14787 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14788 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14789 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14790 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14791 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14792 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14793 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14794 @samp{bf561}, @samp{bf592}.
14796 The optional @var{sirevision} specifies the silicon revision of the target
14797 Blackfin processor. Any workarounds available for the targeted silicon revision
14798 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14799 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14800 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14801 hexadecimal digits representing the major and minor numbers in the silicon
14802 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14803 is not defined. If @var{sirevision} is @samp{any}, the
14804 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14805 If this optional @var{sirevision} is not used, GCC assumes the latest known
14806 silicon revision of the targeted Blackfin processor.
14808 GCC defines a preprocessor macro for the specified @var{cpu}.
14809 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14810 provided by libgloss to be linked in if @option{-msim} is not given.
14812 Without this option, @samp{bf532} is used as the processor by default.
14814 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14815 only the preprocessor macro is defined.
14819 Specifies that the program will be run on the simulator. This causes
14820 the simulator BSP provided by libgloss to be linked in. This option
14821 has effect only for @samp{bfin-elf} toolchain.
14822 Certain other options, such as @option{-mid-shared-library} and
14823 @option{-mfdpic}, imply @option{-msim}.
14825 @item -momit-leaf-frame-pointer
14826 @opindex momit-leaf-frame-pointer
14827 Don't keep the frame pointer in a register for leaf functions. This
14828 avoids the instructions to save, set up and restore frame pointers and
14829 makes an extra register available in leaf functions. The option
14830 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14831 which might make debugging harder.
14833 @item -mspecld-anomaly
14834 @opindex mspecld-anomaly
14835 When enabled, the compiler ensures that the generated code does not
14836 contain speculative loads after jump instructions. If this option is used,
14837 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14839 @item -mno-specld-anomaly
14840 @opindex mno-specld-anomaly
14841 Don't generate extra code to prevent speculative loads from occurring.
14843 @item -mcsync-anomaly
14844 @opindex mcsync-anomaly
14845 When enabled, the compiler ensures that the generated code does not
14846 contain CSYNC or SSYNC instructions too soon after conditional branches.
14847 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14849 @item -mno-csync-anomaly
14850 @opindex mno-csync-anomaly
14851 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14852 occurring too soon after a conditional branch.
14856 When enabled, the compiler is free to take advantage of the knowledge that
14857 the entire program fits into the low 64k of memory.
14860 @opindex mno-low-64k
14861 Assume that the program is arbitrarily large. This is the default.
14863 @item -mstack-check-l1
14864 @opindex mstack-check-l1
14865 Do stack checking using information placed into L1 scratchpad memory by the
14868 @item -mid-shared-library
14869 @opindex mid-shared-library
14870 Generate code that supports shared libraries via the library ID method.
14871 This allows for execute in place and shared libraries in an environment
14872 without virtual memory management. This option implies @option{-fPIC}.
14873 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14875 @item -mno-id-shared-library
14876 @opindex mno-id-shared-library
14877 Generate code that doesn't assume ID-based shared libraries are being used.
14878 This is the default.
14880 @item -mleaf-id-shared-library
14881 @opindex mleaf-id-shared-library
14882 Generate code that supports shared libraries via the library ID method,
14883 but assumes that this library or executable won't link against any other
14884 ID shared libraries. That allows the compiler to use faster code for jumps
14887 @item -mno-leaf-id-shared-library
14888 @opindex mno-leaf-id-shared-library
14889 Do not assume that the code being compiled won't link against any ID shared
14890 libraries. Slower code is generated for jump and call insns.
14892 @item -mshared-library-id=n
14893 @opindex mshared-library-id
14894 Specifies the identification number of the ID-based shared library being
14895 compiled. Specifying a value of 0 generates more compact code; specifying
14896 other values forces the allocation of that number to the current
14897 library but is no more space- or time-efficient than omitting this option.
14901 Generate code that allows the data segment to be located in a different
14902 area of memory from the text segment. This allows for execute in place in
14903 an environment without virtual memory management by eliminating relocations
14904 against the text section.
14906 @item -mno-sep-data
14907 @opindex mno-sep-data
14908 Generate code that assumes that the data segment follows the text segment.
14909 This is the default.
14912 @itemx -mno-long-calls
14913 @opindex mlong-calls
14914 @opindex mno-long-calls
14915 Tells the compiler to perform function calls by first loading the
14916 address of the function into a register and then performing a subroutine
14917 call on this register. This switch is needed if the target function
14918 lies outside of the 24-bit addressing range of the offset-based
14919 version of subroutine call instruction.
14921 This feature is not enabled by default. Specifying
14922 @option{-mno-long-calls} restores the default behavior. Note these
14923 switches have no effect on how the compiler generates code to handle
14924 function calls via function pointers.
14928 Link with the fast floating-point library. This library relaxes some of
14929 the IEEE floating-point standard's rules for checking inputs against
14930 Not-a-Number (NAN), in the interest of performance.
14933 @opindex minline-plt
14934 Enable inlining of PLT entries in function calls to functions that are
14935 not known to bind locally. It has no effect without @option{-mfdpic}.
14938 @opindex mmulticore
14939 Build a standalone application for multicore Blackfin processors.
14940 This option causes proper start files and link scripts supporting
14941 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14942 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14944 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14945 selects the one-application-per-core programming model. Without
14946 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14947 programming model is used. In this model, the main function of Core B
14948 should be named as @code{coreb_main}.
14950 If this option is not used, the single-core application programming
14955 Build a standalone application for Core A of BF561 when using
14956 the one-application-per-core programming model. Proper start files
14957 and link scripts are used to support Core A, and the macro
14958 @code{__BFIN_COREA} is defined.
14959 This option can only be used in conjunction with @option{-mmulticore}.
14963 Build a standalone application for Core B of BF561 when using
14964 the one-application-per-core programming model. Proper start files
14965 and link scripts are used to support Core B, and the macro
14966 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14967 should be used instead of @code{main}.
14968 This option can only be used in conjunction with @option{-mmulticore}.
14972 Build a standalone application for SDRAM. Proper start files and
14973 link scripts are used to put the application into SDRAM, and the macro
14974 @code{__BFIN_SDRAM} is defined.
14975 The loader should initialize SDRAM before loading the application.
14979 Assume that ICPLBs are enabled at run time. This has an effect on certain
14980 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14981 are enabled; for standalone applications the default is off.
14985 @subsection C6X Options
14986 @cindex C6X Options
14989 @item -march=@var{name}
14991 This specifies the name of the target architecture. GCC uses this
14992 name to determine what kind of instructions it can emit when generating
14993 assembly code. Permissible names are: @samp{c62x},
14994 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14997 @opindex mbig-endian
14998 Generate code for a big-endian target.
15000 @item -mlittle-endian
15001 @opindex mlittle-endian
15002 Generate code for a little-endian target. This is the default.
15006 Choose startup files and linker script suitable for the simulator.
15008 @item -msdata=default
15009 @opindex msdata=default
15010 Put small global and static data in the @code{.neardata} section,
15011 which is pointed to by register @code{B14}. Put small uninitialized
15012 global and static data in the @code{.bss} section, which is adjacent
15013 to the @code{.neardata} section. Put small read-only data into the
15014 @code{.rodata} section. The corresponding sections used for large
15015 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
15018 @opindex msdata=all
15019 Put all data, not just small objects, into the sections reserved for
15020 small data, and use addressing relative to the @code{B14} register to
15024 @opindex msdata=none
15025 Make no use of the sections reserved for small data, and use absolute
15026 addresses to access all data. Put all initialized global and static
15027 data in the @code{.fardata} section, and all uninitialized data in the
15028 @code{.far} section. Put all constant data into the @code{.const}
15033 @subsection CRIS Options
15034 @cindex CRIS Options
15036 These options are defined specifically for the CRIS ports.
15039 @item -march=@var{architecture-type}
15040 @itemx -mcpu=@var{architecture-type}
15043 Generate code for the specified architecture. The choices for
15044 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
15045 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
15046 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
15049 @item -mtune=@var{architecture-type}
15051 Tune to @var{architecture-type} everything applicable about the generated
15052 code, except for the ABI and the set of available instructions. The
15053 choices for @var{architecture-type} are the same as for
15054 @option{-march=@var{architecture-type}}.
15056 @item -mmax-stack-frame=@var{n}
15057 @opindex mmax-stack-frame
15058 Warn when the stack frame of a function exceeds @var{n} bytes.
15064 The options @option{-metrax4} and @option{-metrax100} are synonyms for
15065 @option{-march=v3} and @option{-march=v8} respectively.
15067 @item -mmul-bug-workaround
15068 @itemx -mno-mul-bug-workaround
15069 @opindex mmul-bug-workaround
15070 @opindex mno-mul-bug-workaround
15071 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
15072 models where it applies. This option is active by default.
15076 Enable CRIS-specific verbose debug-related information in the assembly
15077 code. This option also has the effect of turning off the @samp{#NO_APP}
15078 formatted-code indicator to the assembler at the beginning of the
15083 Do not use condition-code results from previous instruction; always emit
15084 compare and test instructions before use of condition codes.
15086 @item -mno-side-effects
15087 @opindex mno-side-effects
15088 Do not emit instructions with side effects in addressing modes other than
15091 @item -mstack-align
15092 @itemx -mno-stack-align
15093 @itemx -mdata-align
15094 @itemx -mno-data-align
15095 @itemx -mconst-align
15096 @itemx -mno-const-align
15097 @opindex mstack-align
15098 @opindex mno-stack-align
15099 @opindex mdata-align
15100 @opindex mno-data-align
15101 @opindex mconst-align
15102 @opindex mno-const-align
15103 These options (@samp{no-} options) arrange (eliminate arrangements) for the
15104 stack frame, individual data and constants to be aligned for the maximum
15105 single data access size for the chosen CPU model. The default is to
15106 arrange for 32-bit alignment. ABI details such as structure layout are
15107 not affected by these options.
15115 Similar to the stack- data- and const-align options above, these options
15116 arrange for stack frame, writable data and constants to all be 32-bit,
15117 16-bit or 8-bit aligned. The default is 32-bit alignment.
15119 @item -mno-prologue-epilogue
15120 @itemx -mprologue-epilogue
15121 @opindex mno-prologue-epilogue
15122 @opindex mprologue-epilogue
15123 With @option{-mno-prologue-epilogue}, the normal function prologue and
15124 epilogue which set up the stack frame are omitted and no return
15125 instructions or return sequences are generated in the code. Use this
15126 option only together with visual inspection of the compiled code: no
15127 warnings or errors are generated when call-saved registers must be saved,
15128 or storage for local variables needs to be allocated.
15132 @opindex mno-gotplt
15134 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
15135 instruction sequences that load addresses for functions from the PLT part
15136 of the GOT rather than (traditional on other architectures) calls to the
15137 PLT@. The default is @option{-mgotplt}.
15141 Legacy no-op option only recognized with the cris-axis-elf and
15142 cris-axis-linux-gnu targets.
15146 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
15150 This option, recognized for the cris-axis-elf, arranges
15151 to link with input-output functions from a simulator library. Code,
15152 initialized data and zero-initialized data are allocated consecutively.
15156 Like @option{-sim}, but pass linker options to locate initialized data at
15157 0x40000000 and zero-initialized data at 0x80000000.
15161 @subsection CR16 Options
15162 @cindex CR16 Options
15164 These options are defined specifically for the CR16 ports.
15170 Enable the use of multiply-accumulate instructions. Disabled by default.
15174 @opindex mcr16cplus
15176 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
15181 Links the library libsim.a which is in compatible with simulator. Applicable
15182 to ELF compiler only.
15186 Choose integer type as 32-bit wide.
15190 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
15192 @item -mdata-model=@var{model}
15193 @opindex mdata-model
15194 Choose a data model. The choices for @var{model} are @samp{near},
15195 @samp{far} or @samp{medium}. @samp{medium} is default.
15196 However, @samp{far} is not valid with @option{-mcr16c}, as the
15197 CR16C architecture does not support the far data model.
15200 @node Darwin Options
15201 @subsection Darwin Options
15202 @cindex Darwin options
15204 These options are defined for all architectures running the Darwin operating
15207 FSF GCC on Darwin does not create ``fat'' object files; it creates
15208 an object file for the single architecture that GCC was built to
15209 target. Apple's GCC on Darwin does create ``fat'' files if multiple
15210 @option{-arch} options are used; it does so by running the compiler or
15211 linker multiple times and joining the results together with
15214 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
15215 @samp{i686}) is determined by the flags that specify the ISA
15216 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
15217 @option{-force_cpusubtype_ALL} option can be used to override this.
15219 The Darwin tools vary in their behavior when presented with an ISA
15220 mismatch. The assembler, @file{as}, only permits instructions to
15221 be used that are valid for the subtype of the file it is generating,
15222 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
15223 The linker for shared libraries, @file{/usr/bin/libtool}, fails
15224 and prints an error if asked to create a shared library with a less
15225 restrictive subtype than its input files (for instance, trying to put
15226 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
15227 for executables, @command{ld}, quietly gives the executable the most
15228 restrictive subtype of any of its input files.
15233 Add the framework directory @var{dir} to the head of the list of
15234 directories to be searched for header files. These directories are
15235 interleaved with those specified by @option{-I} options and are
15236 scanned in a left-to-right order.
15238 A framework directory is a directory with frameworks in it. A
15239 framework is a directory with a @file{Headers} and/or
15240 @file{PrivateHeaders} directory contained directly in it that ends
15241 in @file{.framework}. The name of a framework is the name of this
15242 directory excluding the @file{.framework}. Headers associated with
15243 the framework are found in one of those two directories, with
15244 @file{Headers} being searched first. A subframework is a framework
15245 directory that is in a framework's @file{Frameworks} directory.
15246 Includes of subframework headers can only appear in a header of a
15247 framework that contains the subframework, or in a sibling subframework
15248 header. Two subframeworks are siblings if they occur in the same
15249 framework. A subframework should not have the same name as a
15250 framework; a warning is issued if this is violated. Currently a
15251 subframework cannot have subframeworks; in the future, the mechanism
15252 may be extended to support this. The standard frameworks can be found
15253 in @file{/System/Library/Frameworks} and
15254 @file{/Library/Frameworks}. An example include looks like
15255 @code{#include <Framework/header.h>}, where @file{Framework} denotes
15256 the name of the framework and @file{header.h} is found in the
15257 @file{PrivateHeaders} or @file{Headers} directory.
15259 @item -iframework@var{dir}
15260 @opindex iframework
15261 Like @option{-F} except the directory is a treated as a system
15262 directory. The main difference between this @option{-iframework} and
15263 @option{-F} is that with @option{-iframework} the compiler does not
15264 warn about constructs contained within header files found via
15265 @var{dir}. This option is valid only for the C family of languages.
15269 Emit debugging information for symbols that are used. For stabs
15270 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
15271 This is by default ON@.
15275 Emit debugging information for all symbols and types.
15277 @item -mmacosx-version-min=@var{version}
15278 The earliest version of MacOS X that this executable will run on
15279 is @var{version}. Typical values of @var{version} include @code{10.1},
15280 @code{10.2}, and @code{10.3.9}.
15282 If the compiler was built to use the system's headers by default,
15283 then the default for this option is the system version on which the
15284 compiler is running, otherwise the default is to make choices that
15285 are compatible with as many systems and code bases as possible.
15289 Enable kernel development mode. The @option{-mkernel} option sets
15290 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
15291 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
15292 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
15293 applicable. This mode also sets @option{-mno-altivec},
15294 @option{-msoft-float}, @option{-fno-builtin} and
15295 @option{-mlong-branch} for PowerPC targets.
15297 @item -mone-byte-bool
15298 @opindex mone-byte-bool
15299 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
15300 By default @code{sizeof(bool)} is @code{4} when compiling for
15301 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
15302 option has no effect on x86.
15304 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
15305 to generate code that is not binary compatible with code generated
15306 without that switch. Using this switch may require recompiling all
15307 other modules in a program, including system libraries. Use this
15308 switch to conform to a non-default data model.
15310 @item -mfix-and-continue
15311 @itemx -ffix-and-continue
15312 @itemx -findirect-data
15313 @opindex mfix-and-continue
15314 @opindex ffix-and-continue
15315 @opindex findirect-data
15316 Generate code suitable for fast turnaround development, such as to
15317 allow GDB to dynamically load @file{.o} files into already-running
15318 programs. @option{-findirect-data} and @option{-ffix-and-continue}
15319 are provided for backwards compatibility.
15323 Loads all members of static archive libraries.
15324 See man ld(1) for more information.
15326 @item -arch_errors_fatal
15327 @opindex arch_errors_fatal
15328 Cause the errors having to do with files that have the wrong architecture
15331 @item -bind_at_load
15332 @opindex bind_at_load
15333 Causes the output file to be marked such that the dynamic linker will
15334 bind all undefined references when the file is loaded or launched.
15338 Produce a Mach-o bundle format file.
15339 See man ld(1) for more information.
15341 @item -bundle_loader @var{executable}
15342 @opindex bundle_loader
15343 This option specifies the @var{executable} that will load the build
15344 output file being linked. See man ld(1) for more information.
15347 @opindex dynamiclib
15348 When passed this option, GCC produces a dynamic library instead of
15349 an executable when linking, using the Darwin @file{libtool} command.
15351 @item -force_cpusubtype_ALL
15352 @opindex force_cpusubtype_ALL
15353 This causes GCC's output file to have the @samp{ALL} subtype, instead of
15354 one controlled by the @option{-mcpu} or @option{-march} option.
15356 @item -allowable_client @var{client_name}
15357 @itemx -client_name
15358 @itemx -compatibility_version
15359 @itemx -current_version
15361 @itemx -dependency-file
15363 @itemx -dylinker_install_name
15365 @itemx -exported_symbols_list
15368 @itemx -flat_namespace
15369 @itemx -force_flat_namespace
15370 @itemx -headerpad_max_install_names
15373 @itemx -install_name
15374 @itemx -keep_private_externs
15375 @itemx -multi_module
15376 @itemx -multiply_defined
15377 @itemx -multiply_defined_unused
15380 @itemx -no_dead_strip_inits_and_terms
15381 @itemx -nofixprebinding
15382 @itemx -nomultidefs
15384 @itemx -noseglinkedit
15385 @itemx -pagezero_size
15387 @itemx -prebind_all_twolevel_modules
15388 @itemx -private_bundle
15390 @itemx -read_only_relocs
15392 @itemx -sectobjectsymbols
15396 @itemx -sectobjectsymbols
15399 @itemx -segs_read_only_addr
15401 @itemx -segs_read_write_addr
15402 @itemx -seg_addr_table
15403 @itemx -seg_addr_table_filename
15404 @itemx -seglinkedit
15406 @itemx -segs_read_only_addr
15407 @itemx -segs_read_write_addr
15408 @itemx -single_module
15410 @itemx -sub_library
15412 @itemx -sub_umbrella
15413 @itemx -twolevel_namespace
15416 @itemx -unexported_symbols_list
15417 @itemx -weak_reference_mismatches
15418 @itemx -whatsloaded
15419 @opindex allowable_client
15420 @opindex client_name
15421 @opindex compatibility_version
15422 @opindex current_version
15423 @opindex dead_strip
15424 @opindex dependency-file
15425 @opindex dylib_file
15426 @opindex dylinker_install_name
15428 @opindex exported_symbols_list
15430 @opindex flat_namespace
15431 @opindex force_flat_namespace
15432 @opindex headerpad_max_install_names
15433 @opindex image_base
15435 @opindex install_name
15436 @opindex keep_private_externs
15437 @opindex multi_module
15438 @opindex multiply_defined
15439 @opindex multiply_defined_unused
15440 @opindex noall_load
15441 @opindex no_dead_strip_inits_and_terms
15442 @opindex nofixprebinding
15443 @opindex nomultidefs
15445 @opindex noseglinkedit
15446 @opindex pagezero_size
15448 @opindex prebind_all_twolevel_modules
15449 @opindex private_bundle
15450 @opindex read_only_relocs
15452 @opindex sectobjectsymbols
15455 @opindex sectcreate
15456 @opindex sectobjectsymbols
15459 @opindex segs_read_only_addr
15460 @opindex segs_read_write_addr
15461 @opindex seg_addr_table
15462 @opindex seg_addr_table_filename
15463 @opindex seglinkedit
15465 @opindex segs_read_only_addr
15466 @opindex segs_read_write_addr
15467 @opindex single_module
15469 @opindex sub_library
15470 @opindex sub_umbrella
15471 @opindex twolevel_namespace
15474 @opindex unexported_symbols_list
15475 @opindex weak_reference_mismatches
15476 @opindex whatsloaded
15477 These options are passed to the Darwin linker. The Darwin linker man page
15478 describes them in detail.
15481 @node DEC Alpha Options
15482 @subsection DEC Alpha Options
15484 These @samp{-m} options are defined for the DEC Alpha implementations:
15487 @item -mno-soft-float
15488 @itemx -msoft-float
15489 @opindex mno-soft-float
15490 @opindex msoft-float
15491 Use (do not use) the hardware floating-point instructions for
15492 floating-point operations. When @option{-msoft-float} is specified,
15493 functions in @file{libgcc.a} are used to perform floating-point
15494 operations. Unless they are replaced by routines that emulate the
15495 floating-point operations, or compiled in such a way as to call such
15496 emulations routines, these routines issue floating-point
15497 operations. If you are compiling for an Alpha without floating-point
15498 operations, you must ensure that the library is built so as not to call
15501 Note that Alpha implementations without floating-point operations are
15502 required to have floating-point registers.
15505 @itemx -mno-fp-regs
15507 @opindex mno-fp-regs
15508 Generate code that uses (does not use) the floating-point register set.
15509 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
15510 register set is not used, floating-point operands are passed in integer
15511 registers as if they were integers and floating-point results are passed
15512 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
15513 so any function with a floating-point argument or return value called by code
15514 compiled with @option{-mno-fp-regs} must also be compiled with that
15517 A typical use of this option is building a kernel that does not use,
15518 and hence need not save and restore, any floating-point registers.
15522 The Alpha architecture implements floating-point hardware optimized for
15523 maximum performance. It is mostly compliant with the IEEE floating-point
15524 standard. However, for full compliance, software assistance is
15525 required. This option generates code fully IEEE-compliant code
15526 @emph{except} that the @var{inexact-flag} is not maintained (see below).
15527 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
15528 defined during compilation. The resulting code is less efficient but is
15529 able to correctly support denormalized numbers and exceptional IEEE
15530 values such as not-a-number and plus/minus infinity. Other Alpha
15531 compilers call this option @option{-ieee_with_no_inexact}.
15533 @item -mieee-with-inexact
15534 @opindex mieee-with-inexact
15535 This is like @option{-mieee} except the generated code also maintains
15536 the IEEE @var{inexact-flag}. Turning on this option causes the
15537 generated code to implement fully-compliant IEEE math. In addition to
15538 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
15539 macro. On some Alpha implementations the resulting code may execute
15540 significantly slower than the code generated by default. Since there is
15541 very little code that depends on the @var{inexact-flag}, you should
15542 normally not specify this option. Other Alpha compilers call this
15543 option @option{-ieee_with_inexact}.
15545 @item -mfp-trap-mode=@var{trap-mode}
15546 @opindex mfp-trap-mode
15547 This option controls what floating-point related traps are enabled.
15548 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
15549 The trap mode can be set to one of four values:
15553 This is the default (normal) setting. The only traps that are enabled
15554 are the ones that cannot be disabled in software (e.g., division by zero
15558 In addition to the traps enabled by @samp{n}, underflow traps are enabled
15562 Like @samp{u}, but the instructions are marked to be safe for software
15563 completion (see Alpha architecture manual for details).
15566 Like @samp{su}, but inexact traps are enabled as well.
15569 @item -mfp-rounding-mode=@var{rounding-mode}
15570 @opindex mfp-rounding-mode
15571 Selects the IEEE rounding mode. Other Alpha compilers call this option
15572 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
15577 Normal IEEE rounding mode. Floating-point numbers are rounded towards
15578 the nearest machine number or towards the even machine number in case
15582 Round towards minus infinity.
15585 Chopped rounding mode. Floating-point numbers are rounded towards zero.
15588 Dynamic rounding mode. A field in the floating-point control register
15589 (@var{fpcr}, see Alpha architecture reference manual) controls the
15590 rounding mode in effect. The C library initializes this register for
15591 rounding towards plus infinity. Thus, unless your program modifies the
15592 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15595 @item -mtrap-precision=@var{trap-precision}
15596 @opindex mtrap-precision
15597 In the Alpha architecture, floating-point traps are imprecise. This
15598 means without software assistance it is impossible to recover from a
15599 floating trap and program execution normally needs to be terminated.
15600 GCC can generate code that can assist operating system trap handlers
15601 in determining the exact location that caused a floating-point trap.
15602 Depending on the requirements of an application, different levels of
15603 precisions can be selected:
15607 Program precision. This option is the default and means a trap handler
15608 can only identify which program caused a floating-point exception.
15611 Function precision. The trap handler can determine the function that
15612 caused a floating-point exception.
15615 Instruction precision. The trap handler can determine the exact
15616 instruction that caused a floating-point exception.
15619 Other Alpha compilers provide the equivalent options called
15620 @option{-scope_safe} and @option{-resumption_safe}.
15622 @item -mieee-conformant
15623 @opindex mieee-conformant
15624 This option marks the generated code as IEEE conformant. You must not
15625 use this option unless you also specify @option{-mtrap-precision=i} and either
15626 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15627 is to emit the line @samp{.eflag 48} in the function prologue of the
15628 generated assembly file.
15630 @item -mbuild-constants
15631 @opindex mbuild-constants
15632 Normally GCC examines a 32- or 64-bit integer constant to
15633 see if it can construct it from smaller constants in two or three
15634 instructions. If it cannot, it outputs the constant as a literal and
15635 generates code to load it from the data segment at run time.
15637 Use this option to require GCC to construct @emph{all} integer constants
15638 using code, even if it takes more instructions (the maximum is six).
15640 You typically use this option to build a shared library dynamic
15641 loader. Itself a shared library, it must relocate itself in memory
15642 before it can find the variables and constants in its own data segment.
15660 Indicate whether GCC should generate code to use the optional BWX,
15661 CIX, FIX and MAX instruction sets. The default is to use the instruction
15662 sets supported by the CPU type specified via @option{-mcpu=} option or that
15663 of the CPU on which GCC was built if none is specified.
15666 @itemx -mfloat-ieee
15667 @opindex mfloat-vax
15668 @opindex mfloat-ieee
15669 Generate code that uses (does not use) VAX F and G floating-point
15670 arithmetic instead of IEEE single and double precision.
15672 @item -mexplicit-relocs
15673 @itemx -mno-explicit-relocs
15674 @opindex mexplicit-relocs
15675 @opindex mno-explicit-relocs
15676 Older Alpha assemblers provided no way to generate symbol relocations
15677 except via assembler macros. Use of these macros does not allow
15678 optimal instruction scheduling. GNU binutils as of version 2.12
15679 supports a new syntax that allows the compiler to explicitly mark
15680 which relocations should apply to which instructions. This option
15681 is mostly useful for debugging, as GCC detects the capabilities of
15682 the assembler when it is built and sets the default accordingly.
15685 @itemx -mlarge-data
15686 @opindex msmall-data
15687 @opindex mlarge-data
15688 When @option{-mexplicit-relocs} is in effect, static data is
15689 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15690 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15691 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15692 16-bit relocations off of the @code{$gp} register. This limits the
15693 size of the small data area to 64KB, but allows the variables to be
15694 directly accessed via a single instruction.
15696 The default is @option{-mlarge-data}. With this option the data area
15697 is limited to just below 2GB@. Programs that require more than 2GB of
15698 data must use @code{malloc} or @code{mmap} to allocate the data in the
15699 heap instead of in the program's data segment.
15701 When generating code for shared libraries, @option{-fpic} implies
15702 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15705 @itemx -mlarge-text
15706 @opindex msmall-text
15707 @opindex mlarge-text
15708 When @option{-msmall-text} is used, the compiler assumes that the
15709 code of the entire program (or shared library) fits in 4MB, and is
15710 thus reachable with a branch instruction. When @option{-msmall-data}
15711 is used, the compiler can assume that all local symbols share the
15712 same @code{$gp} value, and thus reduce the number of instructions
15713 required for a function call from 4 to 1.
15715 The default is @option{-mlarge-text}.
15717 @item -mcpu=@var{cpu_type}
15719 Set the instruction set and instruction scheduling parameters for
15720 machine type @var{cpu_type}. You can specify either the @samp{EV}
15721 style name or the corresponding chip number. GCC supports scheduling
15722 parameters for the EV4, EV5 and EV6 family of processors and
15723 chooses the default values for the instruction set from the processor
15724 you specify. If you do not specify a processor type, GCC defaults
15725 to the processor on which the compiler was built.
15727 Supported values for @var{cpu_type} are
15733 Schedules as an EV4 and has no instruction set extensions.
15737 Schedules as an EV5 and has no instruction set extensions.
15741 Schedules as an EV5 and supports the BWX extension.
15746 Schedules as an EV5 and supports the BWX and MAX extensions.
15750 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15754 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15757 Native toolchains also support the value @samp{native},
15758 which selects the best architecture option for the host processor.
15759 @option{-mcpu=native} has no effect if GCC does not recognize
15762 @item -mtune=@var{cpu_type}
15764 Set only the instruction scheduling parameters for machine type
15765 @var{cpu_type}. The instruction set is not changed.
15767 Native toolchains also support the value @samp{native},
15768 which selects the best architecture option for the host processor.
15769 @option{-mtune=native} has no effect if GCC does not recognize
15772 @item -mmemory-latency=@var{time}
15773 @opindex mmemory-latency
15774 Sets the latency the scheduler should assume for typical memory
15775 references as seen by the application. This number is highly
15776 dependent on the memory access patterns used by the application
15777 and the size of the external cache on the machine.
15779 Valid options for @var{time} are
15783 A decimal number representing clock cycles.
15789 The compiler contains estimates of the number of clock cycles for
15790 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15791 (also called Dcache, Scache, and Bcache), as well as to main memory.
15792 Note that L3 is only valid for EV5.
15798 @subsection FR30 Options
15799 @cindex FR30 Options
15801 These options are defined specifically for the FR30 port.
15805 @item -msmall-model
15806 @opindex msmall-model
15807 Use the small address space model. This can produce smaller code, but
15808 it does assume that all symbolic values and addresses fit into a
15813 Assume that runtime support has been provided and so there is no need
15814 to include the simulator library (@file{libsim.a}) on the linker
15820 @subsection FT32 Options
15821 @cindex FT32 Options
15823 These options are defined specifically for the FT32 port.
15829 Specifies that the program will be run on the simulator. This causes
15830 an alternate runtime startup and library to be linked.
15831 You must not use this option when generating programs that will run on
15832 real hardware; you must provide your own runtime library for whatever
15833 I/O functions are needed.
15837 Enable Local Register Allocation. This is still experimental for FT32,
15838 so by default the compiler uses standard reload.
15842 Do not use div and mod instructions.
15847 @subsection FRV Options
15848 @cindex FRV Options
15854 Only use the first 32 general-purpose registers.
15859 Use all 64 general-purpose registers.
15864 Use only the first 32 floating-point registers.
15869 Use all 64 floating-point registers.
15872 @opindex mhard-float
15874 Use hardware instructions for floating-point operations.
15877 @opindex msoft-float
15879 Use library routines for floating-point operations.
15884 Dynamically allocate condition code registers.
15889 Do not try to dynamically allocate condition code registers, only
15890 use @code{icc0} and @code{fcc0}.
15895 Change ABI to use double word insns.
15900 Do not use double word instructions.
15905 Use floating-point double instructions.
15908 @opindex mno-double
15910 Do not use floating-point double instructions.
15915 Use media instructions.
15920 Do not use media instructions.
15925 Use multiply and add/subtract instructions.
15928 @opindex mno-muladd
15930 Do not use multiply and add/subtract instructions.
15935 Select the FDPIC ABI, which uses function descriptors to represent
15936 pointers to functions. Without any PIC/PIE-related options, it
15937 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15938 assumes GOT entries and small data are within a 12-bit range from the
15939 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15940 are computed with 32 bits.
15941 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15944 @opindex minline-plt
15946 Enable inlining of PLT entries in function calls to functions that are
15947 not known to bind locally. It has no effect without @option{-mfdpic}.
15948 It's enabled by default if optimizing for speed and compiling for
15949 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15950 optimization option such as @option{-O3} or above is present in the
15956 Assume a large TLS segment when generating thread-local code.
15961 Do not assume a large TLS segment when generating thread-local code.
15966 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15967 that is known to be in read-only sections. It's enabled by default,
15968 except for @option{-fpic} or @option{-fpie}: even though it may help
15969 make the global offset table smaller, it trades 1 instruction for 4.
15970 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15971 one of which may be shared by multiple symbols, and it avoids the need
15972 for a GOT entry for the referenced symbol, so it's more likely to be a
15973 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15975 @item -multilib-library-pic
15976 @opindex multilib-library-pic
15978 Link with the (library, not FD) pic libraries. It's implied by
15979 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15980 @option{-fpic} without @option{-mfdpic}. You should never have to use
15984 @opindex mlinked-fp
15986 Follow the EABI requirement of always creating a frame pointer whenever
15987 a stack frame is allocated. This option is enabled by default and can
15988 be disabled with @option{-mno-linked-fp}.
15991 @opindex mlong-calls
15993 Use indirect addressing to call functions outside the current
15994 compilation unit. This allows the functions to be placed anywhere
15995 within the 32-bit address space.
15997 @item -malign-labels
15998 @opindex malign-labels
16000 Try to align labels to an 8-byte boundary by inserting NOPs into the
16001 previous packet. This option only has an effect when VLIW packing
16002 is enabled. It doesn't create new packets; it merely adds NOPs to
16005 @item -mlibrary-pic
16006 @opindex mlibrary-pic
16008 Generate position-independent EABI code.
16013 Use only the first four media accumulator registers.
16018 Use all eight media accumulator registers.
16023 Pack VLIW instructions.
16028 Do not pack VLIW instructions.
16031 @opindex mno-eflags
16033 Do not mark ABI switches in e_flags.
16036 @opindex mcond-move
16038 Enable the use of conditional-move instructions (default).
16040 This switch is mainly for debugging the compiler and will likely be removed
16041 in a future version.
16043 @item -mno-cond-move
16044 @opindex mno-cond-move
16046 Disable the use of conditional-move instructions.
16048 This switch is mainly for debugging the compiler and will likely be removed
16049 in a future version.
16054 Enable the use of conditional set instructions (default).
16056 This switch is mainly for debugging the compiler and will likely be removed
16057 in a future version.
16062 Disable the use of conditional set instructions.
16064 This switch is mainly for debugging the compiler and will likely be removed
16065 in a future version.
16068 @opindex mcond-exec
16070 Enable the use of conditional execution (default).
16072 This switch is mainly for debugging the compiler and will likely be removed
16073 in a future version.
16075 @item -mno-cond-exec
16076 @opindex mno-cond-exec
16078 Disable the use of conditional execution.
16080 This switch is mainly for debugging the compiler and will likely be removed
16081 in a future version.
16083 @item -mvliw-branch
16084 @opindex mvliw-branch
16086 Run a pass to pack branches into VLIW instructions (default).
16088 This switch is mainly for debugging the compiler and will likely be removed
16089 in a future version.
16091 @item -mno-vliw-branch
16092 @opindex mno-vliw-branch
16094 Do not run a pass to pack branches into VLIW instructions.
16096 This switch is mainly for debugging the compiler and will likely be removed
16097 in a future version.
16099 @item -mmulti-cond-exec
16100 @opindex mmulti-cond-exec
16102 Enable optimization of @code{&&} and @code{||} in conditional execution
16105 This switch is mainly for debugging the compiler and will likely be removed
16106 in a future version.
16108 @item -mno-multi-cond-exec
16109 @opindex mno-multi-cond-exec
16111 Disable optimization of @code{&&} and @code{||} in conditional execution.
16113 This switch is mainly for debugging the compiler and will likely be removed
16114 in a future version.
16116 @item -mnested-cond-exec
16117 @opindex mnested-cond-exec
16119 Enable nested conditional execution optimizations (default).
16121 This switch is mainly for debugging the compiler and will likely be removed
16122 in a future version.
16124 @item -mno-nested-cond-exec
16125 @opindex mno-nested-cond-exec
16127 Disable nested conditional execution optimizations.
16129 This switch is mainly for debugging the compiler and will likely be removed
16130 in a future version.
16132 @item -moptimize-membar
16133 @opindex moptimize-membar
16135 This switch removes redundant @code{membar} instructions from the
16136 compiler-generated code. It is enabled by default.
16138 @item -mno-optimize-membar
16139 @opindex mno-optimize-membar
16141 This switch disables the automatic removal of redundant @code{membar}
16142 instructions from the generated code.
16144 @item -mtomcat-stats
16145 @opindex mtomcat-stats
16147 Cause gas to print out tomcat statistics.
16149 @item -mcpu=@var{cpu}
16152 Select the processor type for which to generate code. Possible values are
16153 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
16154 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
16158 @node GNU/Linux Options
16159 @subsection GNU/Linux Options
16161 These @samp{-m} options are defined for GNU/Linux targets:
16166 Use the GNU C library. This is the default except
16167 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
16168 @samp{*-*-linux-*android*} targets.
16172 Use uClibc C library. This is the default on
16173 @samp{*-*-linux-*uclibc*} targets.
16177 Use the musl C library. This is the default on
16178 @samp{*-*-linux-*musl*} targets.
16182 Use Bionic C library. This is the default on
16183 @samp{*-*-linux-*android*} targets.
16187 Compile code compatible with Android platform. This is the default on
16188 @samp{*-*-linux-*android*} targets.
16190 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
16191 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
16192 this option makes the GCC driver pass Android-specific options to the linker.
16193 Finally, this option causes the preprocessor macro @code{__ANDROID__}
16196 @item -tno-android-cc
16197 @opindex tno-android-cc
16198 Disable compilation effects of @option{-mandroid}, i.e., do not enable
16199 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
16200 @option{-fno-rtti} by default.
16202 @item -tno-android-ld
16203 @opindex tno-android-ld
16204 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
16205 linking options to the linker.
16209 @node H8/300 Options
16210 @subsection H8/300 Options
16212 These @samp{-m} options are defined for the H8/300 implementations:
16217 Shorten some address references at link time, when possible; uses the
16218 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
16219 ld, Using ld}, for a fuller description.
16223 Generate code for the H8/300H@.
16227 Generate code for the H8S@.
16231 Generate code for the H8S and H8/300H in the normal mode. This switch
16232 must be used either with @option{-mh} or @option{-ms}.
16236 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
16240 Extended registers are stored on stack before execution of function
16241 with monitor attribute. Default option is @option{-mexr}.
16242 This option is valid only for H8S targets.
16246 Extended registers are not stored on stack before execution of function
16247 with monitor attribute. Default option is @option{-mno-exr}.
16248 This option is valid only for H8S targets.
16252 Make @code{int} data 32 bits by default.
16255 @opindex malign-300
16256 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
16257 The default for the H8/300H and H8S is to align longs and floats on
16259 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
16260 This option has no effect on the H8/300.
16264 @subsection HPPA Options
16265 @cindex HPPA Options
16267 These @samp{-m} options are defined for the HPPA family of computers:
16270 @item -march=@var{architecture-type}
16272 Generate code for the specified architecture. The choices for
16273 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
16274 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
16275 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
16276 architecture option for your machine. Code compiled for lower numbered
16277 architectures runs on higher numbered architectures, but not the
16280 @item -mpa-risc-1-0
16281 @itemx -mpa-risc-1-1
16282 @itemx -mpa-risc-2-0
16283 @opindex mpa-risc-1-0
16284 @opindex mpa-risc-1-1
16285 @opindex mpa-risc-2-0
16286 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
16288 @item -mjump-in-delay
16289 @opindex mjump-in-delay
16290 This option is ignored and provided for compatibility purposes only.
16292 @item -mdisable-fpregs
16293 @opindex mdisable-fpregs
16294 Prevent floating-point registers from being used in any manner. This is
16295 necessary for compiling kernels that perform lazy context switching of
16296 floating-point registers. If you use this option and attempt to perform
16297 floating-point operations, the compiler aborts.
16299 @item -mdisable-indexing
16300 @opindex mdisable-indexing
16301 Prevent the compiler from using indexing address modes. This avoids some
16302 rather obscure problems when compiling MIG generated code under MACH@.
16304 @item -mno-space-regs
16305 @opindex mno-space-regs
16306 Generate code that assumes the target has no space registers. This allows
16307 GCC to generate faster indirect calls and use unscaled index address modes.
16309 Such code is suitable for level 0 PA systems and kernels.
16311 @item -mfast-indirect-calls
16312 @opindex mfast-indirect-calls
16313 Generate code that assumes calls never cross space boundaries. This
16314 allows GCC to emit code that performs faster indirect calls.
16316 This option does not work in the presence of shared libraries or nested
16319 @item -mfixed-range=@var{register-range}
16320 @opindex mfixed-range
16321 Generate code treating the given register range as fixed registers.
16322 A fixed register is one that the register allocator cannot use. This is
16323 useful when compiling kernel code. A register range is specified as
16324 two registers separated by a dash. Multiple register ranges can be
16325 specified separated by a comma.
16327 @item -mlong-load-store
16328 @opindex mlong-load-store
16329 Generate 3-instruction load and store sequences as sometimes required by
16330 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
16333 @item -mportable-runtime
16334 @opindex mportable-runtime
16335 Use the portable calling conventions proposed by HP for ELF systems.
16339 Enable the use of assembler directives only GAS understands.
16341 @item -mschedule=@var{cpu-type}
16343 Schedule code according to the constraints for the machine type
16344 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
16345 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
16346 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
16347 proper scheduling option for your machine. The default scheduling is
16351 @opindex mlinker-opt
16352 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
16353 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
16354 linkers in which they give bogus error messages when linking some programs.
16357 @opindex msoft-float
16358 Generate output containing library calls for floating point.
16359 @strong{Warning:} the requisite libraries are not available for all HPPA
16360 targets. Normally the facilities of the machine's usual C compiler are
16361 used, but this cannot be done directly in cross-compilation. You must make
16362 your own arrangements to provide suitable library functions for
16365 @option{-msoft-float} changes the calling convention in the output file;
16366 therefore, it is only useful if you compile @emph{all} of a program with
16367 this option. In particular, you need to compile @file{libgcc.a}, the
16368 library that comes with GCC, with @option{-msoft-float} in order for
16373 Generate the predefine, @code{_SIO}, for server IO@. The default is
16374 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
16375 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
16376 options are available under HP-UX and HI-UX@.
16380 Use options specific to GNU @command{ld}.
16381 This passes @option{-shared} to @command{ld} when
16382 building a shared library. It is the default when GCC is configured,
16383 explicitly or implicitly, with the GNU linker. This option does not
16384 affect which @command{ld} is called; it only changes what parameters
16385 are passed to that @command{ld}.
16386 The @command{ld} that is called is determined by the
16387 @option{--with-ld} configure option, GCC's program search path, and
16388 finally by the user's @env{PATH}. The linker used by GCC can be printed
16389 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
16390 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16394 Use options specific to HP @command{ld}.
16395 This passes @option{-b} to @command{ld} when building
16396 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
16397 links. It is the default when GCC is configured, explicitly or
16398 implicitly, with the HP linker. This option does not affect
16399 which @command{ld} is called; it only changes what parameters are passed to that
16401 The @command{ld} that is called is determined by the @option{--with-ld}
16402 configure option, GCC's program search path, and finally by the user's
16403 @env{PATH}. The linker used by GCC can be printed using @samp{which
16404 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
16405 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16408 @opindex mno-long-calls
16409 Generate code that uses long call sequences. This ensures that a call
16410 is always able to reach linker generated stubs. The default is to generate
16411 long calls only when the distance from the call site to the beginning
16412 of the function or translation unit, as the case may be, exceeds a
16413 predefined limit set by the branch type being used. The limits for
16414 normal calls are 7,600,000 and 240,000 bytes, respectively for the
16415 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
16418 Distances are measured from the beginning of functions when using the
16419 @option{-ffunction-sections} option, or when using the @option{-mgas}
16420 and @option{-mno-portable-runtime} options together under HP-UX with
16423 It is normally not desirable to use this option as it degrades
16424 performance. However, it may be useful in large applications,
16425 particularly when partial linking is used to build the application.
16427 The types of long calls used depends on the capabilities of the
16428 assembler and linker, and the type of code being generated. The
16429 impact on systems that support long absolute calls, and long pic
16430 symbol-difference or pc-relative calls should be relatively small.
16431 However, an indirect call is used on 32-bit ELF systems in pic code
16432 and it is quite long.
16434 @item -munix=@var{unix-std}
16436 Generate compiler predefines and select a startfile for the specified
16437 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
16438 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
16439 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
16440 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
16441 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
16444 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
16445 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
16446 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
16447 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
16448 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
16449 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
16451 It is @emph{important} to note that this option changes the interfaces
16452 for various library routines. It also affects the operational behavior
16453 of the C library. Thus, @emph{extreme} care is needed in using this
16456 Library code that is intended to operate with more than one UNIX
16457 standard must test, set and restore the variable @code{__xpg4_extended_mask}
16458 as appropriate. Most GNU software doesn't provide this capability.
16462 Suppress the generation of link options to search libdld.sl when the
16463 @option{-static} option is specified on HP-UX 10 and later.
16467 The HP-UX implementation of setlocale in libc has a dependency on
16468 libdld.sl. There isn't an archive version of libdld.sl. Thus,
16469 when the @option{-static} option is specified, special link options
16470 are needed to resolve this dependency.
16472 On HP-UX 10 and later, the GCC driver adds the necessary options to
16473 link with libdld.sl when the @option{-static} option is specified.
16474 This causes the resulting binary to be dynamic. On the 64-bit port,
16475 the linkers generate dynamic binaries by default in any case. The
16476 @option{-nolibdld} option can be used to prevent the GCC driver from
16477 adding these link options.
16481 Add support for multithreading with the @dfn{dce thread} library
16482 under HP-UX@. This option sets flags for both the preprocessor and
16486 @node IA-64 Options
16487 @subsection IA-64 Options
16488 @cindex IA-64 Options
16490 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16494 @opindex mbig-endian
16495 Generate code for a big-endian target. This is the default for HP-UX@.
16497 @item -mlittle-endian
16498 @opindex mlittle-endian
16499 Generate code for a little-endian target. This is the default for AIX5
16505 @opindex mno-gnu-as
16506 Generate (or don't) code for the GNU assembler. This is the default.
16507 @c Also, this is the default if the configure option @option{--with-gnu-as}
16513 @opindex mno-gnu-ld
16514 Generate (or don't) code for the GNU linker. This is the default.
16515 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16520 Generate code that does not use a global pointer register. The result
16521 is not position independent code, and violates the IA-64 ABI@.
16523 @item -mvolatile-asm-stop
16524 @itemx -mno-volatile-asm-stop
16525 @opindex mvolatile-asm-stop
16526 @opindex mno-volatile-asm-stop
16527 Generate (or don't) a stop bit immediately before and after volatile asm
16530 @item -mregister-names
16531 @itemx -mno-register-names
16532 @opindex mregister-names
16533 @opindex mno-register-names
16534 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16535 the stacked registers. This may make assembler output more readable.
16541 Disable (or enable) optimizations that use the small data section. This may
16542 be useful for working around optimizer bugs.
16544 @item -mconstant-gp
16545 @opindex mconstant-gp
16546 Generate code that uses a single constant global pointer value. This is
16547 useful when compiling kernel code.
16551 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16552 This is useful when compiling firmware code.
16554 @item -minline-float-divide-min-latency
16555 @opindex minline-float-divide-min-latency
16556 Generate code for inline divides of floating-point values
16557 using the minimum latency algorithm.
16559 @item -minline-float-divide-max-throughput
16560 @opindex minline-float-divide-max-throughput
16561 Generate code for inline divides of floating-point values
16562 using the maximum throughput algorithm.
16564 @item -mno-inline-float-divide
16565 @opindex mno-inline-float-divide
16566 Do not generate inline code for divides of floating-point values.
16568 @item -minline-int-divide-min-latency
16569 @opindex minline-int-divide-min-latency
16570 Generate code for inline divides of integer values
16571 using the minimum latency algorithm.
16573 @item -minline-int-divide-max-throughput
16574 @opindex minline-int-divide-max-throughput
16575 Generate code for inline divides of integer values
16576 using the maximum throughput algorithm.
16578 @item -mno-inline-int-divide
16579 @opindex mno-inline-int-divide
16580 Do not generate inline code for divides of integer values.
16582 @item -minline-sqrt-min-latency
16583 @opindex minline-sqrt-min-latency
16584 Generate code for inline square roots
16585 using the minimum latency algorithm.
16587 @item -minline-sqrt-max-throughput
16588 @opindex minline-sqrt-max-throughput
16589 Generate code for inline square roots
16590 using the maximum throughput algorithm.
16592 @item -mno-inline-sqrt
16593 @opindex mno-inline-sqrt
16594 Do not generate inline code for @code{sqrt}.
16597 @itemx -mno-fused-madd
16598 @opindex mfused-madd
16599 @opindex mno-fused-madd
16600 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16601 instructions. The default is to use these instructions.
16603 @item -mno-dwarf2-asm
16604 @itemx -mdwarf2-asm
16605 @opindex mno-dwarf2-asm
16606 @opindex mdwarf2-asm
16607 Don't (or do) generate assembler code for the DWARF line number debugging
16608 info. This may be useful when not using the GNU assembler.
16610 @item -mearly-stop-bits
16611 @itemx -mno-early-stop-bits
16612 @opindex mearly-stop-bits
16613 @opindex mno-early-stop-bits
16614 Allow stop bits to be placed earlier than immediately preceding the
16615 instruction that triggered the stop bit. This can improve instruction
16616 scheduling, but does not always do so.
16618 @item -mfixed-range=@var{register-range}
16619 @opindex mfixed-range
16620 Generate code treating the given register range as fixed registers.
16621 A fixed register is one that the register allocator cannot use. This is
16622 useful when compiling kernel code. A register range is specified as
16623 two registers separated by a dash. Multiple register ranges can be
16624 specified separated by a comma.
16626 @item -mtls-size=@var{tls-size}
16628 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16631 @item -mtune=@var{cpu-type}
16633 Tune the instruction scheduling for a particular CPU, Valid values are
16634 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16635 and @samp{mckinley}.
16641 Generate code for a 32-bit or 64-bit environment.
16642 The 32-bit environment sets int, long and pointer to 32 bits.
16643 The 64-bit environment sets int to 32 bits and long and pointer
16644 to 64 bits. These are HP-UX specific flags.
16646 @item -mno-sched-br-data-spec
16647 @itemx -msched-br-data-spec
16648 @opindex mno-sched-br-data-spec
16649 @opindex msched-br-data-spec
16650 (Dis/En)able data speculative scheduling before reload.
16651 This results in generation of @code{ld.a} instructions and
16652 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16653 The default setting is disabled.
16655 @item -msched-ar-data-spec
16656 @itemx -mno-sched-ar-data-spec
16657 @opindex msched-ar-data-spec
16658 @opindex mno-sched-ar-data-spec
16659 (En/Dis)able data speculative scheduling after reload.
16660 This results in generation of @code{ld.a} instructions and
16661 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16662 The default setting is enabled.
16664 @item -mno-sched-control-spec
16665 @itemx -msched-control-spec
16666 @opindex mno-sched-control-spec
16667 @opindex msched-control-spec
16668 (Dis/En)able control speculative scheduling. This feature is
16669 available only during region scheduling (i.e.@: before reload).
16670 This results in generation of the @code{ld.s} instructions and
16671 the corresponding check instructions @code{chk.s}.
16672 The default setting is disabled.
16674 @item -msched-br-in-data-spec
16675 @itemx -mno-sched-br-in-data-spec
16676 @opindex msched-br-in-data-spec
16677 @opindex mno-sched-br-in-data-spec
16678 (En/Dis)able speculative scheduling of the instructions that
16679 are dependent on the data speculative loads before reload.
16680 This is effective only with @option{-msched-br-data-spec} enabled.
16681 The default setting is enabled.
16683 @item -msched-ar-in-data-spec
16684 @itemx -mno-sched-ar-in-data-spec
16685 @opindex msched-ar-in-data-spec
16686 @opindex mno-sched-ar-in-data-spec
16687 (En/Dis)able speculative scheduling of the instructions that
16688 are dependent on the data speculative loads after reload.
16689 This is effective only with @option{-msched-ar-data-spec} enabled.
16690 The default setting is enabled.
16692 @item -msched-in-control-spec
16693 @itemx -mno-sched-in-control-spec
16694 @opindex msched-in-control-spec
16695 @opindex mno-sched-in-control-spec
16696 (En/Dis)able speculative scheduling of the instructions that
16697 are dependent on the control speculative loads.
16698 This is effective only with @option{-msched-control-spec} enabled.
16699 The default setting is enabled.
16701 @item -mno-sched-prefer-non-data-spec-insns
16702 @itemx -msched-prefer-non-data-spec-insns
16703 @opindex mno-sched-prefer-non-data-spec-insns
16704 @opindex msched-prefer-non-data-spec-insns
16705 If enabled, data-speculative instructions are chosen for schedule
16706 only if there are no other choices at the moment. This makes
16707 the use of the data speculation much more conservative.
16708 The default setting is disabled.
16710 @item -mno-sched-prefer-non-control-spec-insns
16711 @itemx -msched-prefer-non-control-spec-insns
16712 @opindex mno-sched-prefer-non-control-spec-insns
16713 @opindex msched-prefer-non-control-spec-insns
16714 If enabled, control-speculative instructions are chosen for schedule
16715 only if there are no other choices at the moment. This makes
16716 the use of the control speculation much more conservative.
16717 The default setting is disabled.
16719 @item -mno-sched-count-spec-in-critical-path
16720 @itemx -msched-count-spec-in-critical-path
16721 @opindex mno-sched-count-spec-in-critical-path
16722 @opindex msched-count-spec-in-critical-path
16723 If enabled, speculative dependencies are considered during
16724 computation of the instructions priorities. This makes the use of the
16725 speculation a bit more conservative.
16726 The default setting is disabled.
16728 @item -msched-spec-ldc
16729 @opindex msched-spec-ldc
16730 Use a simple data speculation check. This option is on by default.
16732 @item -msched-control-spec-ldc
16733 @opindex msched-spec-ldc
16734 Use a simple check for control speculation. This option is on by default.
16736 @item -msched-stop-bits-after-every-cycle
16737 @opindex msched-stop-bits-after-every-cycle
16738 Place a stop bit after every cycle when scheduling. This option is on
16741 @item -msched-fp-mem-deps-zero-cost
16742 @opindex msched-fp-mem-deps-zero-cost
16743 Assume that floating-point stores and loads are not likely to cause a conflict
16744 when placed into the same instruction group. This option is disabled by
16747 @item -msel-sched-dont-check-control-spec
16748 @opindex msel-sched-dont-check-control-spec
16749 Generate checks for control speculation in selective scheduling.
16750 This flag is disabled by default.
16752 @item -msched-max-memory-insns=@var{max-insns}
16753 @opindex msched-max-memory-insns
16754 Limit on the number of memory insns per instruction group, giving lower
16755 priority to subsequent memory insns attempting to schedule in the same
16756 instruction group. Frequently useful to prevent cache bank conflicts.
16757 The default value is 1.
16759 @item -msched-max-memory-insns-hard-limit
16760 @opindex msched-max-memory-insns-hard-limit
16761 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16762 disallowing more than that number in an instruction group.
16763 Otherwise, the limit is ``soft'', meaning that non-memory operations
16764 are preferred when the limit is reached, but memory operations may still
16770 @subsection LM32 Options
16771 @cindex LM32 options
16773 These @option{-m} options are defined for the LatticeMico32 architecture:
16776 @item -mbarrel-shift-enabled
16777 @opindex mbarrel-shift-enabled
16778 Enable barrel-shift instructions.
16780 @item -mdivide-enabled
16781 @opindex mdivide-enabled
16782 Enable divide and modulus instructions.
16784 @item -mmultiply-enabled
16785 @opindex multiply-enabled
16786 Enable multiply instructions.
16788 @item -msign-extend-enabled
16789 @opindex msign-extend-enabled
16790 Enable sign extend instructions.
16792 @item -muser-enabled
16793 @opindex muser-enabled
16794 Enable user-defined instructions.
16799 @subsection M32C Options
16800 @cindex M32C options
16803 @item -mcpu=@var{name}
16805 Select the CPU for which code is generated. @var{name} may be one of
16806 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16807 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16808 the M32C/80 series.
16812 Specifies that the program will be run on the simulator. This causes
16813 an alternate runtime library to be linked in which supports, for
16814 example, file I/O@. You must not use this option when generating
16815 programs that will run on real hardware; you must provide your own
16816 runtime library for whatever I/O functions are needed.
16818 @item -memregs=@var{number}
16820 Specifies the number of memory-based pseudo-registers GCC uses
16821 during code generation. These pseudo-registers are used like real
16822 registers, so there is a tradeoff between GCC's ability to fit the
16823 code into available registers, and the performance penalty of using
16824 memory instead of registers. Note that all modules in a program must
16825 be compiled with the same value for this option. Because of that, you
16826 must not use this option with GCC's default runtime libraries.
16830 @node M32R/D Options
16831 @subsection M32R/D Options
16832 @cindex M32R/D options
16834 These @option{-m} options are defined for Renesas M32R/D architectures:
16839 Generate code for the M32R/2@.
16843 Generate code for the M32R/X@.
16847 Generate code for the M32R@. This is the default.
16849 @item -mmodel=small
16850 @opindex mmodel=small
16851 Assume all objects live in the lower 16MB of memory (so that their addresses
16852 can be loaded with the @code{ld24} instruction), and assume all subroutines
16853 are reachable with the @code{bl} instruction.
16854 This is the default.
16856 The addressability of a particular object can be set with the
16857 @code{model} attribute.
16859 @item -mmodel=medium
16860 @opindex mmodel=medium
16861 Assume objects may be anywhere in the 32-bit address space (the compiler
16862 generates @code{seth/add3} instructions to load their addresses), and
16863 assume all subroutines are reachable with the @code{bl} instruction.
16865 @item -mmodel=large
16866 @opindex mmodel=large
16867 Assume objects may be anywhere in the 32-bit address space (the compiler
16868 generates @code{seth/add3} instructions to load their addresses), and
16869 assume subroutines may not be reachable with the @code{bl} instruction
16870 (the compiler generates the much slower @code{seth/add3/jl}
16871 instruction sequence).
16874 @opindex msdata=none
16875 Disable use of the small data area. Variables are put into
16876 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16877 @code{section} attribute has been specified).
16878 This is the default.
16880 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16881 Objects may be explicitly put in the small data area with the
16882 @code{section} attribute using one of these sections.
16884 @item -msdata=sdata
16885 @opindex msdata=sdata
16886 Put small global and static data in the small data area, but do not
16887 generate special code to reference them.
16890 @opindex msdata=use
16891 Put small global and static data in the small data area, and generate
16892 special instructions to reference them.
16896 @cindex smaller data references
16897 Put global and static objects less than or equal to @var{num} bytes
16898 into the small data or BSS sections instead of the normal data or BSS
16899 sections. The default value of @var{num} is 8.
16900 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16901 for this option to have any effect.
16903 All modules should be compiled with the same @option{-G @var{num}} value.
16904 Compiling with different values of @var{num} may or may not work; if it
16905 doesn't the linker gives an error message---incorrect code is not
16910 Makes the M32R-specific code in the compiler display some statistics
16911 that might help in debugging programs.
16913 @item -malign-loops
16914 @opindex malign-loops
16915 Align all loops to a 32-byte boundary.
16917 @item -mno-align-loops
16918 @opindex mno-align-loops
16919 Do not enforce a 32-byte alignment for loops. This is the default.
16921 @item -missue-rate=@var{number}
16922 @opindex missue-rate=@var{number}
16923 Issue @var{number} instructions per cycle. @var{number} can only be 1
16926 @item -mbranch-cost=@var{number}
16927 @opindex mbranch-cost=@var{number}
16928 @var{number} can only be 1 or 2. If it is 1 then branches are
16929 preferred over conditional code, if it is 2, then the opposite applies.
16931 @item -mflush-trap=@var{number}
16932 @opindex mflush-trap=@var{number}
16933 Specifies the trap number to use to flush the cache. The default is
16934 12. Valid numbers are between 0 and 15 inclusive.
16936 @item -mno-flush-trap
16937 @opindex mno-flush-trap
16938 Specifies that the cache cannot be flushed by using a trap.
16940 @item -mflush-func=@var{name}
16941 @opindex mflush-func=@var{name}
16942 Specifies the name of the operating system function to call to flush
16943 the cache. The default is @samp{_flush_cache}, but a function call
16944 is only used if a trap is not available.
16946 @item -mno-flush-func
16947 @opindex mno-flush-func
16948 Indicates that there is no OS function for flushing the cache.
16952 @node M680x0 Options
16953 @subsection M680x0 Options
16954 @cindex M680x0 options
16956 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16957 The default settings depend on which architecture was selected when
16958 the compiler was configured; the defaults for the most common choices
16962 @item -march=@var{arch}
16964 Generate code for a specific M680x0 or ColdFire instruction set
16965 architecture. Permissible values of @var{arch} for M680x0
16966 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16967 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16968 architectures are selected according to Freescale's ISA classification
16969 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16970 @samp{isab} and @samp{isac}.
16972 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16973 code for a ColdFire target. The @var{arch} in this macro is one of the
16974 @option{-march} arguments given above.
16976 When used together, @option{-march} and @option{-mtune} select code
16977 that runs on a family of similar processors but that is optimized
16978 for a particular microarchitecture.
16980 @item -mcpu=@var{cpu}
16982 Generate code for a specific M680x0 or ColdFire processor.
16983 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16984 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16985 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16986 below, which also classifies the CPUs into families:
16988 @multitable @columnfractions 0.20 0.80
16989 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16990 @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}
16991 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16992 @item @samp{5206e} @tab @samp{5206e}
16993 @item @samp{5208} @tab @samp{5207} @samp{5208}
16994 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16995 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16996 @item @samp{5216} @tab @samp{5214} @samp{5216}
16997 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16998 @item @samp{5225} @tab @samp{5224} @samp{5225}
16999 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17000 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17001 @item @samp{5249} @tab @samp{5249}
17002 @item @samp{5250} @tab @samp{5250}
17003 @item @samp{5271} @tab @samp{5270} @samp{5271}
17004 @item @samp{5272} @tab @samp{5272}
17005 @item @samp{5275} @tab @samp{5274} @samp{5275}
17006 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17007 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17008 @item @samp{5307} @tab @samp{5307}
17009 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17010 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17011 @item @samp{5407} @tab @samp{5407}
17012 @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}
17015 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17016 @var{arch} is compatible with @var{cpu}. Other combinations of
17017 @option{-mcpu} and @option{-march} are rejected.
17019 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17020 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17021 where the value of @var{family} is given by the table above.
17023 @item -mtune=@var{tune}
17025 Tune the code for a particular microarchitecture within the
17026 constraints set by @option{-march} and @option{-mcpu}.
17027 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17028 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17029 and @samp{cpu32}. The ColdFire microarchitectures
17030 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17032 You can also use @option{-mtune=68020-40} for code that needs
17033 to run relatively well on 68020, 68030 and 68040 targets.
17034 @option{-mtune=68020-60} is similar but includes 68060 targets
17035 as well. These two options select the same tuning decisions as
17036 @option{-m68020-40} and @option{-m68020-60} respectively.
17038 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17039 when tuning for 680x0 architecture @var{arch}. It also defines
17040 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17041 option is used. If GCC is tuning for a range of architectures,
17042 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17043 it defines the macros for every architecture in the range.
17045 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17046 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17047 of the arguments given above.
17053 Generate output for a 68000. This is the default
17054 when the compiler is configured for 68000-based systems.
17055 It is equivalent to @option{-march=68000}.
17057 Use this option for microcontrollers with a 68000 or EC000 core,
17058 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17062 Generate output for a 68010. This is the default
17063 when the compiler is configured for 68010-based systems.
17064 It is equivalent to @option{-march=68010}.
17070 Generate output for a 68020. This is the default
17071 when the compiler is configured for 68020-based systems.
17072 It is equivalent to @option{-march=68020}.
17076 Generate output for a 68030. This is the default when the compiler is
17077 configured for 68030-based systems. It is equivalent to
17078 @option{-march=68030}.
17082 Generate output for a 68040. This is the default when the compiler is
17083 configured for 68040-based systems. It is equivalent to
17084 @option{-march=68040}.
17086 This option inhibits the use of 68881/68882 instructions that have to be
17087 emulated by software on the 68040. Use this option if your 68040 does not
17088 have code to emulate those instructions.
17092 Generate output for a 68060. This is the default when the compiler is
17093 configured for 68060-based systems. It is equivalent to
17094 @option{-march=68060}.
17096 This option inhibits the use of 68020 and 68881/68882 instructions that
17097 have to be emulated by software on the 68060. Use this option if your 68060
17098 does not have code to emulate those instructions.
17102 Generate output for a CPU32. This is the default
17103 when the compiler is configured for CPU32-based systems.
17104 It is equivalent to @option{-march=cpu32}.
17106 Use this option for microcontrollers with a
17107 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
17108 68336, 68340, 68341, 68349 and 68360.
17112 Generate output for a 520X ColdFire CPU@. This is the default
17113 when the compiler is configured for 520X-based systems.
17114 It is equivalent to @option{-mcpu=5206}, and is now deprecated
17115 in favor of that option.
17117 Use this option for microcontroller with a 5200 core, including
17118 the MCF5202, MCF5203, MCF5204 and MCF5206.
17122 Generate output for a 5206e ColdFire CPU@. The option is now
17123 deprecated in favor of the equivalent @option{-mcpu=5206e}.
17127 Generate output for a member of the ColdFire 528X family.
17128 The option is now deprecated in favor of the equivalent
17129 @option{-mcpu=528x}.
17133 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
17134 in favor of the equivalent @option{-mcpu=5307}.
17138 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
17139 in favor of the equivalent @option{-mcpu=5407}.
17143 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
17144 This includes use of hardware floating-point instructions.
17145 The option is equivalent to @option{-mcpu=547x}, and is now
17146 deprecated in favor of that option.
17150 Generate output for a 68040, without using any of the new instructions.
17151 This results in code that can run relatively efficiently on either a
17152 68020/68881 or a 68030 or a 68040. The generated code does use the
17153 68881 instructions that are emulated on the 68040.
17155 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17159 Generate output for a 68060, without using any of the new instructions.
17160 This results in code that can run relatively efficiently on either a
17161 68020/68881 or a 68030 or a 68040. The generated code does use the
17162 68881 instructions that are emulated on the 68060.
17164 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17168 @opindex mhard-float
17170 Generate floating-point instructions. This is the default for 68020
17171 and above, and for ColdFire devices that have an FPU@. It defines the
17172 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
17173 on ColdFire targets.
17176 @opindex msoft-float
17177 Do not generate floating-point instructions; use library calls instead.
17178 This is the default for 68000, 68010, and 68832 targets. It is also
17179 the default for ColdFire devices that have no FPU.
17185 Generate (do not generate) ColdFire hardware divide and remainder
17186 instructions. If @option{-march} is used without @option{-mcpu},
17187 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17188 architectures. Otherwise, the default is taken from the target CPU
17189 (either the default CPU, or the one specified by @option{-mcpu}). For
17190 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17191 @option{-mcpu=5206e}.
17193 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
17197 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17198 Additionally, parameters passed on the stack are also aligned to a
17199 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17203 Do not consider type @code{int} to be 16 bits wide. This is the default.
17206 @itemx -mno-bitfield
17207 @opindex mnobitfield
17208 @opindex mno-bitfield
17209 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17210 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17214 Do use the bit-field instructions. The @option{-m68020} option implies
17215 @option{-mbitfield}. This is the default if you use a configuration
17216 designed for a 68020.
17220 Use a different function-calling convention, in which functions
17221 that take a fixed number of arguments return with the @code{rtd}
17222 instruction, which pops their arguments while returning. This
17223 saves one instruction in the caller since there is no need to pop
17224 the arguments there.
17226 This calling convention is incompatible with the one normally
17227 used on Unix, so you cannot use it if you need to call libraries
17228 compiled with the Unix compiler.
17230 Also, you must provide function prototypes for all functions that
17231 take variable numbers of arguments (including @code{printf});
17232 otherwise incorrect code is generated for calls to those
17235 In addition, seriously incorrect code results if you call a
17236 function with too many arguments. (Normally, extra arguments are
17237 harmlessly ignored.)
17239 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17240 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17244 Do not use the calling conventions selected by @option{-mrtd}.
17245 This is the default.
17248 @itemx -mno-align-int
17249 @opindex malign-int
17250 @opindex mno-align-int
17251 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17252 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17253 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17254 Aligning variables on 32-bit boundaries produces code that runs somewhat
17255 faster on processors with 32-bit busses at the expense of more memory.
17257 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17258 aligns structures containing the above types differently than
17259 most published application binary interface specifications for the m68k.
17263 Use the pc-relative addressing mode of the 68000 directly, instead of
17264 using a global offset table. At present, this option implies @option{-fpic},
17265 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17266 not presently supported with @option{-mpcrel}, though this could be supported for
17267 68020 and higher processors.
17269 @item -mno-strict-align
17270 @itemx -mstrict-align
17271 @opindex mno-strict-align
17272 @opindex mstrict-align
17273 Do not (do) assume that unaligned memory references are handled by
17277 Generate code that allows the data segment to be located in a different
17278 area of memory from the text segment. This allows for execute-in-place in
17279 an environment without virtual memory management. This option implies
17282 @item -mno-sep-data
17283 Generate code that assumes that the data segment follows the text segment.
17284 This is the default.
17286 @item -mid-shared-library
17287 Generate code that supports shared libraries via the library ID method.
17288 This allows for execute-in-place and shared libraries in an environment
17289 without virtual memory management. This option implies @option{-fPIC}.
17291 @item -mno-id-shared-library
17292 Generate code that doesn't assume ID-based shared libraries are being used.
17293 This is the default.
17295 @item -mshared-library-id=n
17296 Specifies the identification number of the ID-based shared library being
17297 compiled. Specifying a value of 0 generates more compact code; specifying
17298 other values forces the allocation of that number to the current
17299 library, but is no more space- or time-efficient than omitting this option.
17305 When generating position-independent code for ColdFire, generate code
17306 that works if the GOT has more than 8192 entries. This code is
17307 larger and slower than code generated without this option. On M680x0
17308 processors, this option is not needed; @option{-fPIC} suffices.
17310 GCC normally uses a single instruction to load values from the GOT@.
17311 While this is relatively efficient, it only works if the GOT
17312 is smaller than about 64k. Anything larger causes the linker
17313 to report an error such as:
17315 @cindex relocation truncated to fit (ColdFire)
17317 relocation truncated to fit: R_68K_GOT16O foobar
17320 If this happens, you should recompile your code with @option{-mxgot}.
17321 It should then work with very large GOTs. However, code generated with
17322 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17323 the value of a global symbol.
17325 Note that some linkers, including newer versions of the GNU linker,
17326 can create multiple GOTs and sort GOT entries. If you have such a linker,
17327 you should only need to use @option{-mxgot} when compiling a single
17328 object file that accesses more than 8192 GOT entries. Very few do.
17330 These options have no effect unless GCC is generating
17331 position-independent code.
17335 @node MCore Options
17336 @subsection MCore Options
17337 @cindex MCore options
17339 These are the @samp{-m} options defined for the Motorola M*Core
17345 @itemx -mno-hardlit
17347 @opindex mno-hardlit
17348 Inline constants into the code stream if it can be done in two
17349 instructions or less.
17355 Use the divide instruction. (Enabled by default).
17357 @item -mrelax-immediate
17358 @itemx -mno-relax-immediate
17359 @opindex mrelax-immediate
17360 @opindex mno-relax-immediate
17361 Allow arbitrary-sized immediates in bit operations.
17363 @item -mwide-bitfields
17364 @itemx -mno-wide-bitfields
17365 @opindex mwide-bitfields
17366 @opindex mno-wide-bitfields
17367 Always treat bit-fields as @code{int}-sized.
17369 @item -m4byte-functions
17370 @itemx -mno-4byte-functions
17371 @opindex m4byte-functions
17372 @opindex mno-4byte-functions
17373 Force all functions to be aligned to a 4-byte boundary.
17375 @item -mcallgraph-data
17376 @itemx -mno-callgraph-data
17377 @opindex mcallgraph-data
17378 @opindex mno-callgraph-data
17379 Emit callgraph information.
17382 @itemx -mno-slow-bytes
17383 @opindex mslow-bytes
17384 @opindex mno-slow-bytes
17385 Prefer word access when reading byte quantities.
17387 @item -mlittle-endian
17388 @itemx -mbig-endian
17389 @opindex mlittle-endian
17390 @opindex mbig-endian
17391 Generate code for a little-endian target.
17397 Generate code for the 210 processor.
17401 Assume that runtime support has been provided and so omit the
17402 simulator library (@file{libsim.a)} from the linker command line.
17404 @item -mstack-increment=@var{size}
17405 @opindex mstack-increment
17406 Set the maximum amount for a single stack increment operation. Large
17407 values can increase the speed of programs that contain functions
17408 that need a large amount of stack space, but they can also trigger a
17409 segmentation fault if the stack is extended too much. The default
17415 @subsection MeP Options
17416 @cindex MeP options
17422 Enables the @code{abs} instruction, which is the absolute difference
17423 between two registers.
17427 Enables all the optional instructions---average, multiply, divide, bit
17428 operations, leading zero, absolute difference, min/max, clip, and
17434 Enables the @code{ave} instruction, which computes the average of two
17437 @item -mbased=@var{n}
17439 Variables of size @var{n} bytes or smaller are placed in the
17440 @code{.based} section by default. Based variables use the @code{$tp}
17441 register as a base register, and there is a 128-byte limit to the
17442 @code{.based} section.
17446 Enables the bit operation instructions---bit test (@code{btstm}), set
17447 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17448 test-and-set (@code{tas}).
17450 @item -mc=@var{name}
17452 Selects which section constant data is placed in. @var{name} may
17453 be @samp{tiny}, @samp{near}, or @samp{far}.
17457 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17458 useful unless you also provide @option{-mminmax}.
17460 @item -mconfig=@var{name}
17462 Selects one of the built-in core configurations. Each MeP chip has
17463 one or more modules in it; each module has a core CPU and a variety of
17464 coprocessors, optional instructions, and peripherals. The
17465 @code{MeP-Integrator} tool, not part of GCC, provides these
17466 configurations through this option; using this option is the same as
17467 using all the corresponding command-line options. The default
17468 configuration is @samp{default}.
17472 Enables the coprocessor instructions. By default, this is a 32-bit
17473 coprocessor. Note that the coprocessor is normally enabled via the
17474 @option{-mconfig=} option.
17478 Enables the 32-bit coprocessor's instructions.
17482 Enables the 64-bit coprocessor's instructions.
17486 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17490 Causes constant variables to be placed in the @code{.near} section.
17494 Enables the @code{div} and @code{divu} instructions.
17498 Generate big-endian code.
17502 Generate little-endian code.
17504 @item -mio-volatile
17505 @opindex mio-volatile
17506 Tells the compiler that any variable marked with the @code{io}
17507 attribute is to be considered volatile.
17511 Causes variables to be assigned to the @code{.far} section by default.
17515 Enables the @code{leadz} (leading zero) instruction.
17519 Causes variables to be assigned to the @code{.near} section by default.
17523 Enables the @code{min} and @code{max} instructions.
17527 Enables the multiplication and multiply-accumulate instructions.
17531 Disables all the optional instructions enabled by @option{-mall-opts}.
17535 Enables the @code{repeat} and @code{erepeat} instructions, used for
17536 low-overhead looping.
17540 Causes all variables to default to the @code{.tiny} section. Note
17541 that there is a 65536-byte limit to this section. Accesses to these
17542 variables use the @code{%gp} base register.
17546 Enables the saturation instructions. Note that the compiler does not
17547 currently generate these itself, but this option is included for
17548 compatibility with other tools, like @code{as}.
17552 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17556 Link the simulator run-time libraries.
17560 Link the simulator runtime libraries, excluding built-in support
17561 for reset and exception vectors and tables.
17565 Causes all functions to default to the @code{.far} section. Without
17566 this option, functions default to the @code{.near} section.
17568 @item -mtiny=@var{n}
17570 Variables that are @var{n} bytes or smaller are allocated to the
17571 @code{.tiny} section. These variables use the @code{$gp} base
17572 register. The default for this option is 4, but note that there's a
17573 65536-byte limit to the @code{.tiny} section.
17577 @node MicroBlaze Options
17578 @subsection MicroBlaze Options
17579 @cindex MicroBlaze Options
17584 @opindex msoft-float
17585 Use software emulation for floating point (default).
17588 @opindex mhard-float
17589 Use hardware floating-point instructions.
17593 Do not optimize block moves, use @code{memcpy}.
17595 @item -mno-clearbss
17596 @opindex mno-clearbss
17597 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17599 @item -mcpu=@var{cpu-type}
17601 Use features of, and schedule code for, the given CPU.
17602 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17603 where @var{X} is a major version, @var{YY} is the minor version, and
17604 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17605 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17607 @item -mxl-soft-mul
17608 @opindex mxl-soft-mul
17609 Use software multiply emulation (default).
17611 @item -mxl-soft-div
17612 @opindex mxl-soft-div
17613 Use software emulation for divides (default).
17615 @item -mxl-barrel-shift
17616 @opindex mxl-barrel-shift
17617 Use the hardware barrel shifter.
17619 @item -mxl-pattern-compare
17620 @opindex mxl-pattern-compare
17621 Use pattern compare instructions.
17623 @item -msmall-divides
17624 @opindex msmall-divides
17625 Use table lookup optimization for small signed integer divisions.
17627 @item -mxl-stack-check
17628 @opindex mxl-stack-check
17629 This option is deprecated. Use @option{-fstack-check} instead.
17632 @opindex mxl-gp-opt
17633 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17635 @item -mxl-multiply-high
17636 @opindex mxl-multiply-high
17637 Use multiply high instructions for high part of 32x32 multiply.
17639 @item -mxl-float-convert
17640 @opindex mxl-float-convert
17641 Use hardware floating-point conversion instructions.
17643 @item -mxl-float-sqrt
17644 @opindex mxl-float-sqrt
17645 Use hardware floating-point square root instruction.
17648 @opindex mbig-endian
17649 Generate code for a big-endian target.
17651 @item -mlittle-endian
17652 @opindex mlittle-endian
17653 Generate code for a little-endian target.
17656 @opindex mxl-reorder
17657 Use reorder instructions (swap and byte reversed load/store).
17659 @item -mxl-mode-@var{app-model}
17660 Select application model @var{app-model}. Valid models are
17663 normal executable (default), uses startup code @file{crt0.o}.
17666 for use with Xilinx Microprocessor Debugger (XMD) based
17667 software intrusive debug agent called xmdstub. This uses startup file
17668 @file{crt1.o} and sets the start address of the program to 0x800.
17671 for applications that are loaded using a bootloader.
17672 This model uses startup file @file{crt2.o} which does not contain a processor
17673 reset vector handler. This is suitable for transferring control on a
17674 processor reset to the bootloader rather than the application.
17677 for applications that do not require any of the
17678 MicroBlaze vectors. This option may be useful for applications running
17679 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17682 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17683 @option{-mxl-mode-@var{app-model}}.
17688 @subsection MIPS Options
17689 @cindex MIPS options
17695 Generate big-endian code.
17699 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17702 @item -march=@var{arch}
17704 Generate code that runs on @var{arch}, which can be the name of a
17705 generic MIPS ISA, or the name of a particular processor.
17707 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17708 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17709 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17710 @samp{mips64r5} and @samp{mips64r6}.
17711 The processor names are:
17712 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17713 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17714 @samp{5kc}, @samp{5kf},
17716 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17717 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17718 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17719 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17720 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17723 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17725 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17726 @samp{m5100}, @samp{m5101},
17727 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17730 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17731 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17732 @samp{rm7000}, @samp{rm9000},
17733 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17736 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17737 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17738 @samp{xlr} and @samp{xlp}.
17739 The special value @samp{from-abi} selects the
17740 most compatible architecture for the selected ABI (that is,
17741 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17743 The native Linux/GNU toolchain also supports the value @samp{native},
17744 which selects the best architecture option for the host processor.
17745 @option{-march=native} has no effect if GCC does not recognize
17748 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17749 (for example, @option{-march=r2k}). Prefixes are optional, and
17750 @samp{vr} may be written @samp{r}.
17752 Names of the form @samp{@var{n}f2_1} refer to processors with
17753 FPUs clocked at half the rate of the core, names of the form
17754 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17755 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17756 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17757 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17758 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17759 accepted as synonyms for @samp{@var{n}f1_1}.
17761 GCC defines two macros based on the value of this option. The first
17762 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17763 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17764 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17765 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17766 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17768 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17769 above. In other words, it has the full prefix and does not
17770 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17771 the macro names the resolved architecture (either @code{"mips1"} or
17772 @code{"mips3"}). It names the default architecture when no
17773 @option{-march} option is given.
17775 @item -mtune=@var{arch}
17777 Optimize for @var{arch}. Among other things, this option controls
17778 the way instructions are scheduled, and the perceived cost of arithmetic
17779 operations. The list of @var{arch} values is the same as for
17782 When this option is not used, GCC optimizes for the processor
17783 specified by @option{-march}. By using @option{-march} and
17784 @option{-mtune} together, it is possible to generate code that
17785 runs on a family of processors, but optimize the code for one
17786 particular member of that family.
17788 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17789 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17790 @option{-march} ones described above.
17794 Equivalent to @option{-march=mips1}.
17798 Equivalent to @option{-march=mips2}.
17802 Equivalent to @option{-march=mips3}.
17806 Equivalent to @option{-march=mips4}.
17810 Equivalent to @option{-march=mips32}.
17814 Equivalent to @option{-march=mips32r3}.
17818 Equivalent to @option{-march=mips32r5}.
17822 Equivalent to @option{-march=mips32r6}.
17826 Equivalent to @option{-march=mips64}.
17830 Equivalent to @option{-march=mips64r2}.
17834 Equivalent to @option{-march=mips64r3}.
17838 Equivalent to @option{-march=mips64r5}.
17842 Equivalent to @option{-march=mips64r6}.
17847 @opindex mno-mips16
17848 Generate (do not generate) MIPS16 code. If GCC is targeting a
17849 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17851 MIPS16 code generation can also be controlled on a per-function basis
17852 by means of @code{mips16} and @code{nomips16} attributes.
17853 @xref{Function Attributes}, for more information.
17855 @item -mflip-mips16
17856 @opindex mflip-mips16
17857 Generate MIPS16 code on alternating functions. This option is provided
17858 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17859 not intended for ordinary use in compiling user code.
17861 @item -minterlink-compressed
17862 @item -mno-interlink-compressed
17863 @opindex minterlink-compressed
17864 @opindex mno-interlink-compressed
17865 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17866 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17868 For example, code using the standard ISA encoding cannot jump directly
17869 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17870 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17871 knows that the target of the jump is not compressed.
17873 @item -minterlink-mips16
17874 @itemx -mno-interlink-mips16
17875 @opindex minterlink-mips16
17876 @opindex mno-interlink-mips16
17877 Aliases of @option{-minterlink-compressed} and
17878 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17879 and are retained for backwards compatibility.
17891 Generate code for the given ABI@.
17893 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17894 generates 64-bit code when you select a 64-bit architecture, but you
17895 can use @option{-mgp32} to get 32-bit code instead.
17897 For information about the O64 ABI, see
17898 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17900 GCC supports a variant of the o32 ABI in which floating-point registers
17901 are 64 rather than 32 bits wide. You can select this combination with
17902 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17903 and @code{mfhc1} instructions and is therefore only supported for
17904 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17906 The register assignments for arguments and return values remain the
17907 same, but each scalar value is passed in a single 64-bit register
17908 rather than a pair of 32-bit registers. For example, scalar
17909 floating-point values are returned in @samp{$f0} only, not a
17910 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17911 remains the same in that the even-numbered double-precision registers
17914 Two additional variants of the o32 ABI are supported to enable
17915 a transition from 32-bit to 64-bit registers. These are FPXX
17916 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17917 The FPXX extension mandates that all code must execute correctly
17918 when run using 32-bit or 64-bit registers. The code can be interlinked
17919 with either FP32 or FP64, but not both.
17920 The FP64A extension is similar to the FP64 extension but forbids the
17921 use of odd-numbered single-precision registers. This can be used
17922 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17923 processors and allows both FP32 and FP64A code to interlink and
17924 run in the same process without changing FPU modes.
17927 @itemx -mno-abicalls
17929 @opindex mno-abicalls
17930 Generate (do not generate) code that is suitable for SVR4-style
17931 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17936 Generate (do not generate) code that is fully position-independent,
17937 and that can therefore be linked into shared libraries. This option
17938 only affects @option{-mabicalls}.
17940 All @option{-mabicalls} code has traditionally been position-independent,
17941 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17942 as an extension, the GNU toolchain allows executables to use absolute
17943 accesses for locally-binding symbols. It can also use shorter GP
17944 initialization sequences and generate direct calls to locally-defined
17945 functions. This mode is selected by @option{-mno-shared}.
17947 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17948 objects that can only be linked by the GNU linker. However, the option
17949 does not affect the ABI of the final executable; it only affects the ABI
17950 of relocatable objects. Using @option{-mno-shared} generally makes
17951 executables both smaller and quicker.
17953 @option{-mshared} is the default.
17959 Assume (do not assume) that the static and dynamic linkers
17960 support PLTs and copy relocations. This option only affects
17961 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17962 has no effect without @option{-msym32}.
17964 You can make @option{-mplt} the default by configuring
17965 GCC with @option{--with-mips-plt}. The default is
17966 @option{-mno-plt} otherwise.
17972 Lift (do not lift) the usual restrictions on the size of the global
17975 GCC normally uses a single instruction to load values from the GOT@.
17976 While this is relatively efficient, it only works if the GOT
17977 is smaller than about 64k. Anything larger causes the linker
17978 to report an error such as:
17980 @cindex relocation truncated to fit (MIPS)
17982 relocation truncated to fit: R_MIPS_GOT16 foobar
17985 If this happens, you should recompile your code with @option{-mxgot}.
17986 This works with very large GOTs, although the code is also
17987 less efficient, since it takes three instructions to fetch the
17988 value of a global symbol.
17990 Note that some linkers can create multiple GOTs. If you have such a
17991 linker, you should only need to use @option{-mxgot} when a single object
17992 file accesses more than 64k's worth of GOT entries. Very few do.
17994 These options have no effect unless GCC is generating position
17999 Assume that general-purpose registers are 32 bits wide.
18003 Assume that general-purpose registers are 64 bits wide.
18007 Assume that floating-point registers are 32 bits wide.
18011 Assume that floating-point registers are 64 bits wide.
18015 Do not assume the width of floating-point registers.
18018 @opindex mhard-float
18019 Use floating-point coprocessor instructions.
18022 @opindex msoft-float
18023 Do not use floating-point coprocessor instructions. Implement
18024 floating-point calculations using library calls instead.
18028 Equivalent to @option{-msoft-float}, but additionally asserts that the
18029 program being compiled does not perform any floating-point operations.
18030 This option is presently supported only by some bare-metal MIPS
18031 configurations, where it may select a special set of libraries
18032 that lack all floating-point support (including, for example, the
18033 floating-point @code{printf} formats).
18034 If code compiled with @option{-mno-float} accidentally contains
18035 floating-point operations, it is likely to suffer a link-time
18036 or run-time failure.
18038 @item -msingle-float
18039 @opindex msingle-float
18040 Assume that the floating-point coprocessor only supports single-precision
18043 @item -mdouble-float
18044 @opindex mdouble-float
18045 Assume that the floating-point coprocessor supports double-precision
18046 operations. This is the default.
18049 @itemx -mno-odd-spreg
18050 @opindex modd-spreg
18051 @opindex mno-odd-spreg
18052 Enable the use of odd-numbered single-precision floating-point registers
18053 for the o32 ABI. This is the default for processors that are known to
18054 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18058 @itemx -mabs=legacy
18060 @opindex mabs=legacy
18061 These options control the treatment of the special not-a-number (NaN)
18062 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18063 @code{neg.@i{fmt}} machine instructions.
18065 By default or when @option{-mabs=legacy} is used the legacy
18066 treatment is selected. In this case these instructions are considered
18067 arithmetic and avoided where correct operation is required and the
18068 input operand might be a NaN. A longer sequence of instructions that
18069 manipulate the sign bit of floating-point datum manually is used
18070 instead unless the @option{-ffinite-math-only} option has also been
18073 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
18074 this case these instructions are considered non-arithmetic and therefore
18075 operating correctly in all cases, including in particular where the
18076 input operand is a NaN. These instructions are therefore always used
18077 for the respective operations.
18080 @itemx -mnan=legacy
18082 @opindex mnan=legacy
18083 These options control the encoding of the special not-a-number (NaN)
18084 IEEE 754 floating-point data.
18086 The @option{-mnan=legacy} option selects the legacy encoding. In this
18087 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
18088 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
18089 by the first bit of their trailing significand field being 1.
18091 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
18092 this case qNaNs are denoted by the first bit of their trailing
18093 significand field being 1, whereas sNaNs are denoted by the first bit of
18094 their trailing significand field being 0.
18096 The default is @option{-mnan=legacy} unless GCC has been configured with
18097 @option{--with-nan=2008}.
18103 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
18104 implement atomic memory built-in functions. When neither option is
18105 specified, GCC uses the instructions if the target architecture
18108 @option{-mllsc} is useful if the runtime environment can emulate the
18109 instructions and @option{-mno-llsc} can be useful when compiling for
18110 nonstandard ISAs. You can make either option the default by
18111 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
18112 respectively. @option{--with-llsc} is the default for some
18113 configurations; see the installation documentation for details.
18119 Use (do not use) revision 1 of the MIPS DSP ASE@.
18120 @xref{MIPS DSP Built-in Functions}. This option defines the
18121 preprocessor macro @code{__mips_dsp}. It also defines
18122 @code{__mips_dsp_rev} to 1.
18128 Use (do not use) revision 2 of the MIPS DSP ASE@.
18129 @xref{MIPS DSP Built-in Functions}. This option defines the
18130 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
18131 It also defines @code{__mips_dsp_rev} to 2.
18134 @itemx -mno-smartmips
18135 @opindex msmartmips
18136 @opindex mno-smartmips
18137 Use (do not use) the MIPS SmartMIPS ASE.
18139 @item -mpaired-single
18140 @itemx -mno-paired-single
18141 @opindex mpaired-single
18142 @opindex mno-paired-single
18143 Use (do not use) paired-single floating-point instructions.
18144 @xref{MIPS Paired-Single Support}. This option requires
18145 hardware floating-point support to be enabled.
18151 Use (do not use) MIPS Digital Media Extension instructions.
18152 This option can only be used when generating 64-bit code and requires
18153 hardware floating-point support to be enabled.
18158 @opindex mno-mips3d
18159 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
18160 The option @option{-mips3d} implies @option{-mpaired-single}.
18163 @itemx -mno-micromips
18164 @opindex mmicromips
18165 @opindex mno-mmicromips
18166 Generate (do not generate) microMIPS code.
18168 MicroMIPS code generation can also be controlled on a per-function basis
18169 by means of @code{micromips} and @code{nomicromips} attributes.
18170 @xref{Function Attributes}, for more information.
18176 Use (do not use) MT Multithreading instructions.
18182 Use (do not use) the MIPS MCU ASE instructions.
18188 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18194 Use (do not use) the MIPS Virtualization Application Specific instructions.
18200 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18204 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18205 an explanation of the default and the way that the pointer size is
18210 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18212 The default size of @code{int}s, @code{long}s and pointers depends on
18213 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18214 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18215 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18216 or the same size as integer registers, whichever is smaller.
18222 Assume (do not assume) that all symbols have 32-bit values, regardless
18223 of the selected ABI@. This option is useful in combination with
18224 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18225 to generate shorter and faster references to symbolic addresses.
18229 Put definitions of externally-visible data in a small data section
18230 if that data is no bigger than @var{num} bytes. GCC can then generate
18231 more efficient accesses to the data; see @option{-mgpopt} for details.
18233 The default @option{-G} option depends on the configuration.
18235 @item -mlocal-sdata
18236 @itemx -mno-local-sdata
18237 @opindex mlocal-sdata
18238 @opindex mno-local-sdata
18239 Extend (do not extend) the @option{-G} behavior to local data too,
18240 such as to static variables in C@. @option{-mlocal-sdata} is the
18241 default for all configurations.
18243 If the linker complains that an application is using too much small data,
18244 you might want to try rebuilding the less performance-critical parts with
18245 @option{-mno-local-sdata}. You might also want to build large
18246 libraries with @option{-mno-local-sdata}, so that the libraries leave
18247 more room for the main program.
18249 @item -mextern-sdata
18250 @itemx -mno-extern-sdata
18251 @opindex mextern-sdata
18252 @opindex mno-extern-sdata
18253 Assume (do not assume) that externally-defined data is in
18254 a small data section if the size of that data is within the @option{-G} limit.
18255 @option{-mextern-sdata} is the default for all configurations.
18257 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18258 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18259 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18260 is placed in a small data section. If @var{Var} is defined by another
18261 module, you must either compile that module with a high-enough
18262 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18263 definition. If @var{Var} is common, you must link the application
18264 with a high-enough @option{-G} setting.
18266 The easiest way of satisfying these restrictions is to compile
18267 and link every module with the same @option{-G} option. However,
18268 you may wish to build a library that supports several different
18269 small data limits. You can do this by compiling the library with
18270 the highest supported @option{-G} setting and additionally using
18271 @option{-mno-extern-sdata} to stop the library from making assumptions
18272 about externally-defined data.
18278 Use (do not use) GP-relative accesses for symbols that are known to be
18279 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18280 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18283 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18284 might not hold the value of @code{_gp}. For example, if the code is
18285 part of a library that might be used in a boot monitor, programs that
18286 call boot monitor routines pass an unknown value in @code{$gp}.
18287 (In such situations, the boot monitor itself is usually compiled
18288 with @option{-G0}.)
18290 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18291 @option{-mno-extern-sdata}.
18293 @item -membedded-data
18294 @itemx -mno-embedded-data
18295 @opindex membedded-data
18296 @opindex mno-embedded-data
18297 Allocate variables to the read-only data section first if possible, then
18298 next in the small data section if possible, otherwise in data. This gives
18299 slightly slower code than the default, but reduces the amount of RAM required
18300 when executing, and thus may be preferred for some embedded systems.
18302 @item -muninit-const-in-rodata
18303 @itemx -mno-uninit-const-in-rodata
18304 @opindex muninit-const-in-rodata
18305 @opindex mno-uninit-const-in-rodata
18306 Put uninitialized @code{const} variables in the read-only data section.
18307 This option is only meaningful in conjunction with @option{-membedded-data}.
18309 @item -mcode-readable=@var{setting}
18310 @opindex mcode-readable
18311 Specify whether GCC may generate code that reads from executable sections.
18312 There are three possible settings:
18315 @item -mcode-readable=yes
18316 Instructions may freely access executable sections. This is the
18319 @item -mcode-readable=pcrel
18320 MIPS16 PC-relative load instructions can access executable sections,
18321 but other instructions must not do so. This option is useful on 4KSc
18322 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18323 It is also useful on processors that can be configured to have a dual
18324 instruction/data SRAM interface and that, like the M4K, automatically
18325 redirect PC-relative loads to the instruction RAM.
18327 @item -mcode-readable=no
18328 Instructions must not access executable sections. This option can be
18329 useful on targets that are configured to have a dual instruction/data
18330 SRAM interface but that (unlike the M4K) do not automatically redirect
18331 PC-relative loads to the instruction RAM.
18334 @item -msplit-addresses
18335 @itemx -mno-split-addresses
18336 @opindex msplit-addresses
18337 @opindex mno-split-addresses
18338 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18339 relocation operators. This option has been superseded by
18340 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18342 @item -mexplicit-relocs
18343 @itemx -mno-explicit-relocs
18344 @opindex mexplicit-relocs
18345 @opindex mno-explicit-relocs
18346 Use (do not use) assembler relocation operators when dealing with symbolic
18347 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18348 is to use assembler macros instead.
18350 @option{-mexplicit-relocs} is the default if GCC was configured
18351 to use an assembler that supports relocation operators.
18353 @item -mcheck-zero-division
18354 @itemx -mno-check-zero-division
18355 @opindex mcheck-zero-division
18356 @opindex mno-check-zero-division
18357 Trap (do not trap) on integer division by zero.
18359 The default is @option{-mcheck-zero-division}.
18361 @item -mdivide-traps
18362 @itemx -mdivide-breaks
18363 @opindex mdivide-traps
18364 @opindex mdivide-breaks
18365 MIPS systems check for division by zero by generating either a
18366 conditional trap or a break instruction. Using traps results in
18367 smaller code, but is only supported on MIPS II and later. Also, some
18368 versions of the Linux kernel have a bug that prevents trap from
18369 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18370 allow conditional traps on architectures that support them and
18371 @option{-mdivide-breaks} to force the use of breaks.
18373 The default is usually @option{-mdivide-traps}, but this can be
18374 overridden at configure time using @option{--with-divide=breaks}.
18375 Divide-by-zero checks can be completely disabled using
18376 @option{-mno-check-zero-division}.
18381 @opindex mno-memcpy
18382 Force (do not force) the use of @code{memcpy} for non-trivial block
18383 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18384 most constant-sized copies.
18387 @itemx -mno-long-calls
18388 @opindex mlong-calls
18389 @opindex mno-long-calls
18390 Disable (do not disable) use of the @code{jal} instruction. Calling
18391 functions using @code{jal} is more efficient but requires the caller
18392 and callee to be in the same 256 megabyte segment.
18394 This option has no effect on abicalls code. The default is
18395 @option{-mno-long-calls}.
18401 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18402 instructions, as provided by the R4650 ISA@.
18408 Enable (disable) use of the @code{madd} and @code{msub} integer
18409 instructions. The default is @option{-mimadd} on architectures
18410 that support @code{madd} and @code{msub} except for the 74k
18411 architecture where it was found to generate slower code.
18414 @itemx -mno-fused-madd
18415 @opindex mfused-madd
18416 @opindex mno-fused-madd
18417 Enable (disable) use of the floating-point multiply-accumulate
18418 instructions, when they are available. The default is
18419 @option{-mfused-madd}.
18421 On the R8000 CPU when multiply-accumulate instructions are used,
18422 the intermediate product is calculated to infinite precision
18423 and is not subject to the FCSR Flush to Zero bit. This may be
18424 undesirable in some circumstances. On other processors the result
18425 is numerically identical to the equivalent computation using
18426 separate multiply, add, subtract and negate instructions.
18430 Tell the MIPS assembler to not run its preprocessor over user
18431 assembler files (with a @samp{.s} suffix) when assembling them.
18436 @opindex mno-fix-24k
18437 Work around the 24K E48 (lost data on stores during refill) errata.
18438 The workarounds are implemented by the assembler rather than by GCC@.
18441 @itemx -mno-fix-r4000
18442 @opindex mfix-r4000
18443 @opindex mno-fix-r4000
18444 Work around certain R4000 CPU errata:
18447 A double-word or a variable shift may give an incorrect result if executed
18448 immediately after starting an integer division.
18450 A double-word or a variable shift may give an incorrect result if executed
18451 while an integer multiplication is in progress.
18453 An integer division may give an incorrect result if started in a delay slot
18454 of a taken branch or a jump.
18458 @itemx -mno-fix-r4400
18459 @opindex mfix-r4400
18460 @opindex mno-fix-r4400
18461 Work around certain R4400 CPU errata:
18464 A double-word or a variable shift may give an incorrect result if executed
18465 immediately after starting an integer division.
18469 @itemx -mno-fix-r10000
18470 @opindex mfix-r10000
18471 @opindex mno-fix-r10000
18472 Work around certain R10000 errata:
18475 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18476 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18479 This option can only be used if the target architecture supports
18480 branch-likely instructions. @option{-mfix-r10000} is the default when
18481 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18485 @itemx -mno-fix-rm7000
18486 @opindex mfix-rm7000
18487 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18488 workarounds are implemented by the assembler rather than by GCC@.
18491 @itemx -mno-fix-vr4120
18492 @opindex mfix-vr4120
18493 Work around certain VR4120 errata:
18496 @code{dmultu} does not always produce the correct result.
18498 @code{div} and @code{ddiv} do not always produce the correct result if one
18499 of the operands is negative.
18501 The workarounds for the division errata rely on special functions in
18502 @file{libgcc.a}. At present, these functions are only provided by
18503 the @code{mips64vr*-elf} configurations.
18505 Other VR4120 errata require a NOP to be inserted between certain pairs of
18506 instructions. These errata are handled by the assembler, not by GCC itself.
18509 @opindex mfix-vr4130
18510 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18511 workarounds are implemented by the assembler rather than by GCC,
18512 although GCC avoids using @code{mflo} and @code{mfhi} if the
18513 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18514 instructions are available instead.
18517 @itemx -mno-fix-sb1
18519 Work around certain SB-1 CPU core errata.
18520 (This flag currently works around the SB-1 revision 2
18521 ``F1'' and ``F2'' floating-point errata.)
18523 @item -mr10k-cache-barrier=@var{setting}
18524 @opindex mr10k-cache-barrier
18525 Specify whether GCC should insert cache barriers to avoid the
18526 side-effects of speculation on R10K processors.
18528 In common with many processors, the R10K tries to predict the outcome
18529 of a conditional branch and speculatively executes instructions from
18530 the ``taken'' branch. It later aborts these instructions if the
18531 predicted outcome is wrong. However, on the R10K, even aborted
18532 instructions can have side effects.
18534 This problem only affects kernel stores and, depending on the system,
18535 kernel loads. As an example, a speculatively-executed store may load
18536 the target memory into cache and mark the cache line as dirty, even if
18537 the store itself is later aborted. If a DMA operation writes to the
18538 same area of memory before the ``dirty'' line is flushed, the cached
18539 data overwrites the DMA-ed data. See the R10K processor manual
18540 for a full description, including other potential problems.
18542 One workaround is to insert cache barrier instructions before every memory
18543 access that might be speculatively executed and that might have side
18544 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18545 controls GCC's implementation of this workaround. It assumes that
18546 aborted accesses to any byte in the following regions does not have
18551 the memory occupied by the current function's stack frame;
18554 the memory occupied by an incoming stack argument;
18557 the memory occupied by an object with a link-time-constant address.
18560 It is the kernel's responsibility to ensure that speculative
18561 accesses to these regions are indeed safe.
18563 If the input program contains a function declaration such as:
18569 then the implementation of @code{foo} must allow @code{j foo} and
18570 @code{jal foo} to be executed speculatively. GCC honors this
18571 restriction for functions it compiles itself. It expects non-GCC
18572 functions (such as hand-written assembly code) to do the same.
18574 The option has three forms:
18577 @item -mr10k-cache-barrier=load-store
18578 Insert a cache barrier before a load or store that might be
18579 speculatively executed and that might have side effects even
18582 @item -mr10k-cache-barrier=store
18583 Insert a cache barrier before a store that might be speculatively
18584 executed and that might have side effects even if aborted.
18586 @item -mr10k-cache-barrier=none
18587 Disable the insertion of cache barriers. This is the default setting.
18590 @item -mflush-func=@var{func}
18591 @itemx -mno-flush-func
18592 @opindex mflush-func
18593 Specifies the function to call to flush the I and D caches, or to not
18594 call any such function. If called, the function must take the same
18595 arguments as the common @code{_flush_func}, that is, the address of the
18596 memory range for which the cache is being flushed, the size of the
18597 memory range, and the number 3 (to flush both caches). The default
18598 depends on the target GCC was configured for, but commonly is either
18599 @code{_flush_func} or @code{__cpu_flush}.
18601 @item mbranch-cost=@var{num}
18602 @opindex mbranch-cost
18603 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18604 This cost is only a heuristic and is not guaranteed to produce
18605 consistent results across releases. A zero cost redundantly selects
18606 the default, which is based on the @option{-mtune} setting.
18608 @item -mbranch-likely
18609 @itemx -mno-branch-likely
18610 @opindex mbranch-likely
18611 @opindex mno-branch-likely
18612 Enable or disable use of Branch Likely instructions, regardless of the
18613 default for the selected architecture. By default, Branch Likely
18614 instructions may be generated if they are supported by the selected
18615 architecture. An exception is for the MIPS32 and MIPS64 architectures
18616 and processors that implement those architectures; for those, Branch
18617 Likely instructions are not be generated by default because the MIPS32
18618 and MIPS64 architectures specifically deprecate their use.
18620 @item -mcompact-branches=never
18621 @itemx -mcompact-branches=optimal
18622 @itemx -mcompact-branches=always
18623 @opindex mcompact-branches=never
18624 @opindex mcompact-branches=optimal
18625 @opindex mcompact-branches=always
18626 These options control which form of branches will be generated. The
18627 default is @option{-mcompact-branches=optimal}.
18629 The @option{-mcompact-branches=never} option ensures that compact branch
18630 instructions will never be generated.
18632 The @option{-mcompact-branches=always} option ensures that a compact
18633 branch instruction will be generated if available. If a compact branch
18634 instruction is not available, a delay slot form of the branch will be
18637 This option is supported from MIPS Release 6 onwards.
18639 The @option{-mcompact-branches=optimal} option will cause a delay slot
18640 branch to be used if one is available in the current ISA and the delay
18641 slot is successfully filled. If the delay slot is not filled, a compact
18642 branch will be chosen if one is available.
18644 @item -mfp-exceptions
18645 @itemx -mno-fp-exceptions
18646 @opindex mfp-exceptions
18647 Specifies whether FP exceptions are enabled. This affects how
18648 FP instructions are scheduled for some processors.
18649 The default is that FP exceptions are
18652 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18653 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18656 @item -mvr4130-align
18657 @itemx -mno-vr4130-align
18658 @opindex mvr4130-align
18659 The VR4130 pipeline is two-way superscalar, but can only issue two
18660 instructions together if the first one is 8-byte aligned. When this
18661 option is enabled, GCC aligns pairs of instructions that it
18662 thinks should execute in parallel.
18664 This option only has an effect when optimizing for the VR4130.
18665 It normally makes code faster, but at the expense of making it bigger.
18666 It is enabled by default at optimization level @option{-O3}.
18671 Enable (disable) generation of @code{synci} instructions on
18672 architectures that support it. The @code{synci} instructions (if
18673 enabled) are generated when @code{__builtin___clear_cache} is
18676 This option defaults to @option{-mno-synci}, but the default can be
18677 overridden by configuring GCC with @option{--with-synci}.
18679 When compiling code for single processor systems, it is generally safe
18680 to use @code{synci}. However, on many multi-core (SMP) systems, it
18681 does not invalidate the instruction caches on all cores and may lead
18682 to undefined behavior.
18684 @item -mrelax-pic-calls
18685 @itemx -mno-relax-pic-calls
18686 @opindex mrelax-pic-calls
18687 Try to turn PIC calls that are normally dispatched via register
18688 @code{$25} into direct calls. This is only possible if the linker can
18689 resolve the destination at link time and if the destination is within
18690 range for a direct call.
18692 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18693 an assembler and a linker that support the @code{.reloc} assembly
18694 directive and @option{-mexplicit-relocs} is in effect. With
18695 @option{-mno-explicit-relocs}, this optimization can be performed by the
18696 assembler and the linker alone without help from the compiler.
18698 @item -mmcount-ra-address
18699 @itemx -mno-mcount-ra-address
18700 @opindex mmcount-ra-address
18701 @opindex mno-mcount-ra-address
18702 Emit (do not emit) code that allows @code{_mcount} to modify the
18703 calling function's return address. When enabled, this option extends
18704 the usual @code{_mcount} interface with a new @var{ra-address}
18705 parameter, which has type @code{intptr_t *} and is passed in register
18706 @code{$12}. @code{_mcount} can then modify the return address by
18707 doing both of the following:
18710 Returning the new address in register @code{$31}.
18712 Storing the new address in @code{*@var{ra-address}},
18713 if @var{ra-address} is nonnull.
18716 The default is @option{-mno-mcount-ra-address}.
18718 @item -mframe-header-opt
18719 @itemx -mno-frame-header-opt
18720 @opindex mframe-header-opt
18721 Enable (disable) frame header optimization in the o32 ABI. When using the
18722 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18723 function to write out register arguments. When enabled, this optimization
18724 will suppress the allocation of the frame header if it can be determined that
18727 This optimization is off by default at all optimization levels.
18732 @subsection MMIX Options
18733 @cindex MMIX Options
18735 These options are defined for the MMIX:
18739 @itemx -mno-libfuncs
18741 @opindex mno-libfuncs
18742 Specify that intrinsic library functions are being compiled, passing all
18743 values in registers, no matter the size.
18746 @itemx -mno-epsilon
18748 @opindex mno-epsilon
18749 Generate floating-point comparison instructions that compare with respect
18750 to the @code{rE} epsilon register.
18752 @item -mabi=mmixware
18754 @opindex mabi=mmixware
18756 Generate code that passes function parameters and return values that (in
18757 the called function) are seen as registers @code{$0} and up, as opposed to
18758 the GNU ABI which uses global registers @code{$231} and up.
18760 @item -mzero-extend
18761 @itemx -mno-zero-extend
18762 @opindex mzero-extend
18763 @opindex mno-zero-extend
18764 When reading data from memory in sizes shorter than 64 bits, use (do not
18765 use) zero-extending load instructions by default, rather than
18766 sign-extending ones.
18769 @itemx -mno-knuthdiv
18771 @opindex mno-knuthdiv
18772 Make the result of a division yielding a remainder have the same sign as
18773 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18774 remainder follows the sign of the dividend. Both methods are
18775 arithmetically valid, the latter being almost exclusively used.
18777 @item -mtoplevel-symbols
18778 @itemx -mno-toplevel-symbols
18779 @opindex mtoplevel-symbols
18780 @opindex mno-toplevel-symbols
18781 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18782 code can be used with the @code{PREFIX} assembly directive.
18786 Generate an executable in the ELF format, rather than the default
18787 @samp{mmo} format used by the @command{mmix} simulator.
18789 @item -mbranch-predict
18790 @itemx -mno-branch-predict
18791 @opindex mbranch-predict
18792 @opindex mno-branch-predict
18793 Use (do not use) the probable-branch instructions, when static branch
18794 prediction indicates a probable branch.
18796 @item -mbase-addresses
18797 @itemx -mno-base-addresses
18798 @opindex mbase-addresses
18799 @opindex mno-base-addresses
18800 Generate (do not generate) code that uses @emph{base addresses}. Using a
18801 base address automatically generates a request (handled by the assembler
18802 and the linker) for a constant to be set up in a global register. The
18803 register is used for one or more base address requests within the range 0
18804 to 255 from the value held in the register. The generally leads to short
18805 and fast code, but the number of different data items that can be
18806 addressed is limited. This means that a program that uses lots of static
18807 data may require @option{-mno-base-addresses}.
18809 @item -msingle-exit
18810 @itemx -mno-single-exit
18811 @opindex msingle-exit
18812 @opindex mno-single-exit
18813 Force (do not force) generated code to have a single exit point in each
18817 @node MN10300 Options
18818 @subsection MN10300 Options
18819 @cindex MN10300 options
18821 These @option{-m} options are defined for Matsushita MN10300 architectures:
18826 Generate code to avoid bugs in the multiply instructions for the MN10300
18827 processors. This is the default.
18829 @item -mno-mult-bug
18830 @opindex mno-mult-bug
18831 Do not generate code to avoid bugs in the multiply instructions for the
18832 MN10300 processors.
18836 Generate code using features specific to the AM33 processor.
18840 Do not generate code using features specific to the AM33 processor. This
18845 Generate code using features specific to the AM33/2.0 processor.
18849 Generate code using features specific to the AM34 processor.
18851 @item -mtune=@var{cpu-type}
18853 Use the timing characteristics of the indicated CPU type when
18854 scheduling instructions. This does not change the targeted processor
18855 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18856 @samp{am33-2} or @samp{am34}.
18858 @item -mreturn-pointer-on-d0
18859 @opindex mreturn-pointer-on-d0
18860 When generating a function that returns a pointer, return the pointer
18861 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18862 only in @code{a0}, and attempts to call such functions without a prototype
18863 result in errors. Note that this option is on by default; use
18864 @option{-mno-return-pointer-on-d0} to disable it.
18868 Do not link in the C run-time initialization object file.
18872 Indicate to the linker that it should perform a relaxation optimization pass
18873 to shorten branches, calls and absolute memory addresses. This option only
18874 has an effect when used on the command line for the final link step.
18876 This option makes symbolic debugging impossible.
18880 Allow the compiler to generate @emph{Long Instruction Word}
18881 instructions if the target is the @samp{AM33} or later. This is the
18882 default. This option defines the preprocessor macro @code{__LIW__}.
18886 Do not allow the compiler to generate @emph{Long Instruction Word}
18887 instructions. This option defines the preprocessor macro
18892 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18893 instructions if the target is the @samp{AM33} or later. This is the
18894 default. This option defines the preprocessor macro @code{__SETLB__}.
18898 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18899 instructions. This option defines the preprocessor macro
18900 @code{__NO_SETLB__}.
18904 @node Moxie Options
18905 @subsection Moxie Options
18906 @cindex Moxie Options
18912 Generate big-endian code. This is the default for @samp{moxie-*-*}
18917 Generate little-endian code.
18921 Generate mul.x and umul.x instructions. This is the default for
18922 @samp{moxiebox-*-*} configurations.
18926 Do not link in the C run-time initialization object file.
18930 @node MSP430 Options
18931 @subsection MSP430 Options
18932 @cindex MSP430 Options
18934 These options are defined for the MSP430:
18940 Force assembly output to always use hex constants. Normally such
18941 constants are signed decimals, but this option is available for
18942 testsuite and/or aesthetic purposes.
18946 Select the MCU to target. This is used to create a C preprocessor
18947 symbol based upon the MCU name, converted to upper case and pre- and
18948 post-fixed with @samp{__}. This in turn is used by the
18949 @file{msp430.h} header file to select an MCU-specific supplementary
18952 The option also sets the ISA to use. If the MCU name is one that is
18953 known to only support the 430 ISA then that is selected, otherwise the
18954 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18955 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18956 name selects the 430X ISA.
18958 In addition an MCU-specific linker script is added to the linker
18959 command line. The script's name is the name of the MCU with
18960 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18961 command line defines the C preprocessor symbol @code{__XXX__} and
18962 cause the linker to search for a script called @file{xxx.ld}.
18964 This option is also passed on to the assembler.
18967 @itemx -mno-warn-mcu
18969 @opindex mno-warn-mcu
18970 This option enables or disables warnings about conflicts between the
18971 MCU name specified by the @option{-mmcu} option and the ISA set by the
18972 @option{-mcpu} option and/or the hardware multiply support set by the
18973 @option{-mhwmult} option. It also toggles warnings about unrecognized
18974 MCU names. This option is on by default.
18978 Specifies the ISA to use. Accepted values are @samp{msp430},
18979 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18980 @option{-mmcu=} option should be used to select the ISA.
18984 Link to the simulator runtime libraries and linker script. Overrides
18985 any scripts that would be selected by the @option{-mmcu=} option.
18989 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18993 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18997 This option is passed to the assembler and linker, and allows the
18998 linker to perform certain optimizations that cannot be done until
19003 Describes the type of hardware multiply supported by the target.
19004 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19005 for the original 16-bit-only multiply supported by early MCUs.
19006 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19007 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19008 A value of @samp{auto} can also be given. This tells GCC to deduce
19009 the hardware multiply support based upon the MCU name provided by the
19010 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
19011 the MCU name is not recognized then no hardware multiply support is
19012 assumed. @code{auto} is the default setting.
19014 Hardware multiplies are normally performed by calling a library
19015 routine. This saves space in the generated code. When compiling at
19016 @option{-O3} or higher however the hardware multiplier is invoked
19017 inline. This makes for bigger, but faster code.
19019 The hardware multiply routines disable interrupts whilst running and
19020 restore the previous interrupt state when they finish. This makes
19021 them safe to use inside interrupt handlers as well as in normal code.
19025 Enable the use of a minimum runtime environment - no static
19026 initializers or constructors. This is intended for memory-constrained
19027 devices. The compiler includes special symbols in some objects
19028 that tell the linker and runtime which code fragments are required.
19030 @item -mcode-region=
19031 @itemx -mdata-region=
19032 @opindex mcode-region
19033 @opindex mdata-region
19034 These options tell the compiler where to place functions and data that
19035 do not have one of the @code{lower}, @code{upper}, @code{either} or
19036 @code{section} attributes. Possible values are @code{lower},
19037 @code{upper}, @code{either} or @code{any}. The first three behave
19038 like the corresponding attribute. The fourth possible value -
19039 @code{any} - is the default. It leaves placement entirely up to the
19040 linker script and how it assigns the standard sections
19041 (@code{.text}, @code{.data}, etc) to the memory regions.
19043 @item -msilicon-errata=
19044 @opindex msilicon-errata
19045 This option passes on a request to assembler to enable the fixes for
19046 the named silicon errata.
19048 @item -msilicon-errata-warn=
19049 @opindex msilicon-errata-warn
19050 This option passes on a request to the assembler to enable warning
19051 messages when a silicon errata might need to be applied.
19055 @node NDS32 Options
19056 @subsection NDS32 Options
19057 @cindex NDS32 Options
19059 These options are defined for NDS32 implementations:
19064 @opindex mbig-endian
19065 Generate code in big-endian mode.
19067 @item -mlittle-endian
19068 @opindex mlittle-endian
19069 Generate code in little-endian mode.
19071 @item -mreduced-regs
19072 @opindex mreduced-regs
19073 Use reduced-set registers for register allocation.
19076 @opindex mfull-regs
19077 Use full-set registers for register allocation.
19081 Generate conditional move instructions.
19085 Do not generate conditional move instructions.
19089 Generate performance extension instructions.
19091 @item -mno-perf-ext
19092 @opindex mno-perf-ext
19093 Do not generate performance extension instructions.
19097 Generate v3 push25/pop25 instructions.
19100 @opindex mno-v3push
19101 Do not generate v3 push25/pop25 instructions.
19105 Generate 16-bit instructions.
19108 @opindex mno-16-bit
19109 Do not generate 16-bit instructions.
19111 @item -misr-vector-size=@var{num}
19112 @opindex misr-vector-size
19113 Specify the size of each interrupt vector, which must be 4 or 16.
19115 @item -mcache-block-size=@var{num}
19116 @opindex mcache-block-size
19117 Specify the size of each cache block,
19118 which must be a power of 2 between 4 and 512.
19120 @item -march=@var{arch}
19122 Specify the name of the target architecture.
19124 @item -mcmodel=@var{code-model}
19126 Set the code model to one of
19129 All the data and read-only data segments must be within 512KB addressing space.
19130 The text segment must be within 16MB addressing space.
19131 @item @samp{medium}
19132 The data segment must be within 512KB while the read-only data segment can be
19133 within 4GB addressing space. The text segment should be still within 16MB
19136 All the text and data segments can be within 4GB addressing space.
19140 @opindex mctor-dtor
19141 Enable constructor/destructor feature.
19145 Guide linker to relax instructions.
19149 @node Nios II Options
19150 @subsection Nios II Options
19151 @cindex Nios II options
19152 @cindex Altera Nios II options
19154 These are the options defined for the Altera Nios II processor.
19160 @cindex smaller data references
19161 Put global and static objects less than or equal to @var{num} bytes
19162 into the small data or BSS sections instead of the normal data or BSS
19163 sections. The default value of @var{num} is 8.
19165 @item -mgpopt=@var{option}
19170 Generate (do not generate) GP-relative accesses. The following
19171 @var{option} names are recognized:
19176 Do not generate GP-relative accesses.
19179 Generate GP-relative accesses for small data objects that are not
19180 external, weak, or uninitialized common symbols.
19181 Also use GP-relative addressing for objects that
19182 have been explicitly placed in a small data section via a @code{section}
19186 As for @samp{local}, but also generate GP-relative accesses for
19187 small data objects that are external, weak, or common. If you use this option,
19188 you must ensure that all parts of your program (including libraries) are
19189 compiled with the same @option{-G} setting.
19192 Generate GP-relative accesses for all data objects in the program. If you
19193 use this option, the entire data and BSS segments
19194 of your program must fit in 64K of memory and you must use an appropriate
19195 linker script to allocate them within the addressable range of the
19199 Generate GP-relative addresses for function pointers as well as data
19200 pointers. If you use this option, the entire text, data, and BSS segments
19201 of your program must fit in 64K of memory and you must use an appropriate
19202 linker script to allocate them within the addressable range of the
19207 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
19208 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
19210 The default is @option{-mgpopt} except when @option{-fpic} or
19211 @option{-fPIC} is specified to generate position-independent code.
19212 Note that the Nios II ABI does not permit GP-relative accesses from
19215 You may need to specify @option{-mno-gpopt} explicitly when building
19216 programs that include large amounts of small data, including large
19217 GOT data sections. In this case, the 16-bit offset for GP-relative
19218 addressing may not be large enough to allow access to the entire
19219 small data section.
19225 Generate little-endian (default) or big-endian (experimental) code,
19228 @item -march=@var{arch}
19230 This specifies the name of the target Nios II architecture. GCC uses this
19231 name to determine what kind of instructions it can emit when generating
19232 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
19234 The preprocessor macro @code{__nios2_arch__} is available to programs,
19235 with value 1 or 2, indicating the targeted ISA level.
19237 @item -mbypass-cache
19238 @itemx -mno-bypass-cache
19239 @opindex mno-bypass-cache
19240 @opindex mbypass-cache
19241 Force all load and store instructions to always bypass cache by
19242 using I/O variants of the instructions. The default is not to
19245 @item -mno-cache-volatile
19246 @itemx -mcache-volatile
19247 @opindex mcache-volatile
19248 @opindex mno-cache-volatile
19249 Volatile memory access bypass the cache using the I/O variants of
19250 the load and store instructions. The default is not to bypass the cache.
19252 @item -mno-fast-sw-div
19253 @itemx -mfast-sw-div
19254 @opindex mno-fast-sw-div
19255 @opindex mfast-sw-div
19256 Do not use table-based fast divide for small numbers. The default
19257 is to use the fast divide at @option{-O3} and above.
19261 @itemx -mno-hw-mulx
19265 @opindex mno-hw-mul
19267 @opindex mno-hw-mulx
19269 @opindex mno-hw-div
19271 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
19272 instructions by the compiler. The default is to emit @code{mul}
19273 and not emit @code{div} and @code{mulx}.
19279 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
19280 CDX (code density) instructions. Enabling these instructions also
19281 requires @option{-march=r2}. Since these instructions are optional
19282 extensions to the R2 architecture, the default is not to emit them.
19284 @item -mcustom-@var{insn}=@var{N}
19285 @itemx -mno-custom-@var{insn}
19286 @opindex mcustom-@var{insn}
19287 @opindex mno-custom-@var{insn}
19288 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19289 custom instruction with encoding @var{N} when generating code that uses
19290 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
19291 instruction 253 for single-precision floating-point add operations instead
19292 of the default behavior of using a library call.
19294 The following values of @var{insn} are supported. Except as otherwise
19295 noted, floating-point operations are expected to be implemented with
19296 normal IEEE 754 semantics and correspond directly to the C operators or the
19297 equivalent GCC built-in functions (@pxref{Other Builtins}).
19299 Single-precision floating point:
19302 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19303 Binary arithmetic operations.
19309 Unary absolute value.
19311 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19312 Comparison operations.
19314 @item @samp{fmins}, @samp{fmaxs}
19315 Floating-point minimum and maximum. These instructions are only
19316 generated if @option{-ffinite-math-only} is specified.
19318 @item @samp{fsqrts}
19319 Unary square root operation.
19321 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19322 Floating-point trigonometric and exponential functions. These instructions
19323 are only generated if @option{-funsafe-math-optimizations} is also specified.
19327 Double-precision floating point:
19330 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19331 Binary arithmetic operations.
19337 Unary absolute value.
19339 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19340 Comparison operations.
19342 @item @samp{fmind}, @samp{fmaxd}
19343 Double-precision minimum and maximum. These instructions are only
19344 generated if @option{-ffinite-math-only} is specified.
19346 @item @samp{fsqrtd}
19347 Unary square root operation.
19349 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19350 Double-precision trigonometric and exponential functions. These instructions
19351 are only generated if @option{-funsafe-math-optimizations} is also specified.
19357 @item @samp{fextsd}
19358 Conversion from single precision to double precision.
19360 @item @samp{ftruncds}
19361 Conversion from double precision to single precision.
19363 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19364 Conversion from floating point to signed or unsigned integer types, with
19365 truncation towards zero.
19368 Conversion from single-precision floating point to signed integer,
19369 rounding to the nearest integer and ties away from zero.
19370 This corresponds to the @code{__builtin_lroundf} function when
19371 @option{-fno-math-errno} is used.
19373 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19374 Conversion from signed or unsigned integer types to floating-point types.
19378 In addition, all of the following transfer instructions for internal
19379 registers X and Y must be provided to use any of the double-precision
19380 floating-point instructions. Custom instructions taking two
19381 double-precision source operands expect the first operand in the
19382 64-bit register X. The other operand (or only operand of a unary
19383 operation) is given to the custom arithmetic instruction with the
19384 least significant half in source register @var{src1} and the most
19385 significant half in @var{src2}. A custom instruction that returns a
19386 double-precision result returns the most significant 32 bits in the
19387 destination register and the other half in 32-bit register Y.
19388 GCC automatically generates the necessary code sequences to write
19389 register X and/or read register Y when double-precision floating-point
19390 instructions are used.
19395 Write @var{src1} into the least significant half of X and @var{src2} into
19396 the most significant half of X.
19399 Write @var{src1} into Y.
19401 @item @samp{frdxhi}, @samp{frdxlo}
19402 Read the most or least (respectively) significant half of X and store it in
19406 Read the value of Y and store it into @var{dest}.
19409 Note that you can gain more local control over generation of Nios II custom
19410 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19411 and @code{target("no-custom-@var{insn}")} function attributes
19412 (@pxref{Function Attributes})
19413 or pragmas (@pxref{Function Specific Option Pragmas}).
19415 @item -mcustom-fpu-cfg=@var{name}
19416 @opindex mcustom-fpu-cfg
19418 This option enables a predefined, named set of custom instruction encodings
19419 (see @option{-mcustom-@var{insn}} above).
19420 Currently, the following sets are defined:
19422 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19423 @gccoptlist{-mcustom-fmuls=252 @gol
19424 -mcustom-fadds=253 @gol
19425 -mcustom-fsubs=254 @gol
19426 -fsingle-precision-constant}
19428 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19429 @gccoptlist{-mcustom-fmuls=252 @gol
19430 -mcustom-fadds=253 @gol
19431 -mcustom-fsubs=254 @gol
19432 -mcustom-fdivs=255 @gol
19433 -fsingle-precision-constant}
19435 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19436 @gccoptlist{-mcustom-floatus=243 @gol
19437 -mcustom-fixsi=244 @gol
19438 -mcustom-floatis=245 @gol
19439 -mcustom-fcmpgts=246 @gol
19440 -mcustom-fcmples=249 @gol
19441 -mcustom-fcmpeqs=250 @gol
19442 -mcustom-fcmpnes=251 @gol
19443 -mcustom-fmuls=252 @gol
19444 -mcustom-fadds=253 @gol
19445 -mcustom-fsubs=254 @gol
19446 -mcustom-fdivs=255 @gol
19447 -fsingle-precision-constant}
19449 Custom instruction assignments given by individual
19450 @option{-mcustom-@var{insn}=} options override those given by
19451 @option{-mcustom-fpu-cfg=}, regardless of the
19452 order of the options on the command line.
19454 Note that you can gain more local control over selection of a FPU
19455 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19456 function attribute (@pxref{Function Attributes})
19457 or pragma (@pxref{Function Specific Option Pragmas}).
19461 These additional @samp{-m} options are available for the Altera Nios II
19462 ELF (bare-metal) target:
19468 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19469 startup and termination code, and is typically used in conjunction with
19470 @option{-msys-crt0=} to specify the location of the alternate startup code
19471 provided by the HAL BSP.
19475 Link with a limited version of the C library, @option{-lsmallc}, rather than
19478 @item -msys-crt0=@var{startfile}
19480 @var{startfile} is the file name of the startfile (crt0) to use
19481 when linking. This option is only useful in conjunction with @option{-mhal}.
19483 @item -msys-lib=@var{systemlib}
19485 @var{systemlib} is the library name of the library that provides
19486 low-level system calls required by the C library,
19487 e.g. @code{read} and @code{write}.
19488 This option is typically used to link with a library provided by a HAL BSP.
19492 @node Nvidia PTX Options
19493 @subsection Nvidia PTX Options
19494 @cindex Nvidia PTX options
19495 @cindex nvptx options
19497 These options are defined for Nvidia PTX:
19505 Generate code for 32-bit or 64-bit ABI.
19508 @opindex mmainkernel
19509 Link in code for a __main kernel. This is for stand-alone instead of
19510 offloading execution.
19514 Apply partitioned execution optimizations. This is the default when any
19515 level of optimization is selected.
19519 @node PDP-11 Options
19520 @subsection PDP-11 Options
19521 @cindex PDP-11 Options
19523 These options are defined for the PDP-11:
19528 Use hardware FPP floating point. This is the default. (FIS floating
19529 point on the PDP-11/40 is not supported.)
19532 @opindex msoft-float
19533 Do not use hardware floating point.
19537 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19541 Return floating-point results in memory. This is the default.
19545 Generate code for a PDP-11/40.
19549 Generate code for a PDP-11/45. This is the default.
19553 Generate code for a PDP-11/10.
19555 @item -mbcopy-builtin
19556 @opindex mbcopy-builtin
19557 Use inline @code{movmemhi} patterns for copying memory. This is the
19562 Do not use inline @code{movmemhi} patterns for copying memory.
19568 Use 16-bit @code{int}. This is the default.
19574 Use 32-bit @code{int}.
19577 @itemx -mno-float32
19579 @opindex mno-float32
19580 Use 64-bit @code{float}. This is the default.
19583 @itemx -mno-float64
19585 @opindex mno-float64
19586 Use 32-bit @code{float}.
19590 Use @code{abshi2} pattern. This is the default.
19594 Do not use @code{abshi2} pattern.
19596 @item -mbranch-expensive
19597 @opindex mbranch-expensive
19598 Pretend that branches are expensive. This is for experimenting with
19599 code generation only.
19601 @item -mbranch-cheap
19602 @opindex mbranch-cheap
19603 Do not pretend that branches are expensive. This is the default.
19607 Use Unix assembler syntax. This is the default when configured for
19608 @samp{pdp11-*-bsd}.
19612 Use DEC assembler syntax. This is the default when configured for any
19613 PDP-11 target other than @samp{pdp11-*-bsd}.
19616 @node picoChip Options
19617 @subsection picoChip Options
19618 @cindex picoChip options
19620 These @samp{-m} options are defined for picoChip implementations:
19624 @item -mae=@var{ae_type}
19626 Set the instruction set, register set, and instruction scheduling
19627 parameters for array element type @var{ae_type}. Supported values
19628 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19630 @option{-mae=ANY} selects a completely generic AE type. Code
19631 generated with this option runs on any of the other AE types. The
19632 code is not as efficient as it would be if compiled for a specific
19633 AE type, and some types of operation (e.g., multiplication) do not
19634 work properly on all types of AE.
19636 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19637 for compiled code, and is the default.
19639 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19640 option may suffer from poor performance of byte (char) manipulation,
19641 since the DSP AE does not provide hardware support for byte load/stores.
19643 @item -msymbol-as-address
19644 Enable the compiler to directly use a symbol name as an address in a
19645 load/store instruction, without first loading it into a
19646 register. Typically, the use of this option generates larger
19647 programs, which run faster than when the option isn't used. However, the
19648 results vary from program to program, so it is left as a user option,
19649 rather than being permanently enabled.
19651 @item -mno-inefficient-warnings
19652 Disables warnings about the generation of inefficient code. These
19653 warnings can be generated, for example, when compiling code that
19654 performs byte-level memory operations on the MAC AE type. The MAC AE has
19655 no hardware support for byte-level memory operations, so all byte
19656 load/stores must be synthesized from word load/store operations. This is
19657 inefficient and a warning is generated to indicate
19658 that you should rewrite the code to avoid byte operations, or to target
19659 an AE type that has the necessary hardware support. This option disables
19664 @node PowerPC Options
19665 @subsection PowerPC Options
19666 @cindex PowerPC options
19668 These are listed under @xref{RS/6000 and PowerPC Options}.
19671 @subsection RL78 Options
19672 @cindex RL78 Options
19678 Links in additional target libraries to support operation within a
19687 Specifies the type of hardware multiplication and division support to
19688 be used. The simplest is @code{none}, which uses software for both
19689 multiplication and division. This is the default. The @code{g13}
19690 value is for the hardware multiply/divide peripheral found on the
19691 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19692 the multiplication and division instructions supported by the RL78/G14
19693 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19694 the value @code{mg10} is an alias for @code{none}.
19696 In addition a C preprocessor macro is defined, based upon the setting
19697 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19698 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19705 Specifies the RL78 core to target. The default is the G14 core, also
19706 known as an S3 core or just RL78. The G13 or S2 core does not have
19707 multiply or divide instructions, instead it uses a hardware peripheral
19708 for these operations. The G10 or S1 core does not have register
19709 banks, so it uses a different calling convention.
19711 If this option is set it also selects the type of hardware multiply
19712 support to use, unless this is overridden by an explicit
19713 @option{-mmul=none} option on the command line. Thus specifying
19714 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19715 peripheral and specifying @option{-mcpu=g10} disables the use of
19716 hardware multiplications altogether.
19718 Note, although the RL78/G14 core is the default target, specifying
19719 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19720 change the behavior of the toolchain since it also enables G14
19721 hardware multiply support. If these options are not specified on the
19722 command line then software multiplication routines will be used even
19723 though the code targets the RL78 core. This is for backwards
19724 compatibility with older toolchains which did not have hardware
19725 multiply and divide support.
19727 In addition a C preprocessor macro is defined, based upon the setting
19728 of this option. Possible values are: @code{__RL78_G10__},
19729 @code{__RL78_G13__} or @code{__RL78_G14__}.
19739 These are aliases for the corresponding @option{-mcpu=} option. They
19740 are provided for backwards compatibility.
19744 Allow the compiler to use all of the available registers. By default
19745 registers @code{r24..r31} are reserved for use in interrupt handlers.
19746 With this option enabled these registers can be used in ordinary
19749 @item -m64bit-doubles
19750 @itemx -m32bit-doubles
19751 @opindex m64bit-doubles
19752 @opindex m32bit-doubles
19753 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19754 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19755 @option{-m32bit-doubles}.
19759 @node RS/6000 and PowerPC Options
19760 @subsection IBM RS/6000 and PowerPC Options
19761 @cindex RS/6000 and PowerPC Options
19762 @cindex IBM RS/6000 and PowerPC Options
19764 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19766 @item -mpowerpc-gpopt
19767 @itemx -mno-powerpc-gpopt
19768 @itemx -mpowerpc-gfxopt
19769 @itemx -mno-powerpc-gfxopt
19772 @itemx -mno-powerpc64
19776 @itemx -mno-popcntb
19778 @itemx -mno-popcntd
19787 @itemx -mno-hard-dfp
19788 @opindex mpowerpc-gpopt
19789 @opindex mno-powerpc-gpopt
19790 @opindex mpowerpc-gfxopt
19791 @opindex mno-powerpc-gfxopt
19792 @opindex mpowerpc64
19793 @opindex mno-powerpc64
19797 @opindex mno-popcntb
19799 @opindex mno-popcntd
19805 @opindex mno-mfpgpr
19807 @opindex mno-hard-dfp
19808 You use these options to specify which instructions are available on the
19809 processor you are using. The default value of these options is
19810 determined when configuring GCC@. Specifying the
19811 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19812 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19813 rather than the options listed above.
19815 Specifying @option{-mpowerpc-gpopt} allows
19816 GCC to use the optional PowerPC architecture instructions in the
19817 General Purpose group, including floating-point square root. Specifying
19818 @option{-mpowerpc-gfxopt} allows GCC to
19819 use the optional PowerPC architecture instructions in the Graphics
19820 group, including floating-point select.
19822 The @option{-mmfcrf} option allows GCC to generate the move from
19823 condition register field instruction implemented on the POWER4
19824 processor and other processors that support the PowerPC V2.01
19826 The @option{-mpopcntb} option allows GCC to generate the popcount and
19827 double-precision FP reciprocal estimate instruction implemented on the
19828 POWER5 processor and other processors that support the PowerPC V2.02
19830 The @option{-mpopcntd} option allows GCC to generate the popcount
19831 instruction implemented on the POWER7 processor and other processors
19832 that support the PowerPC V2.06 architecture.
19833 The @option{-mfprnd} option allows GCC to generate the FP round to
19834 integer instructions implemented on the POWER5+ processor and other
19835 processors that support the PowerPC V2.03 architecture.
19836 The @option{-mcmpb} option allows GCC to generate the compare bytes
19837 instruction implemented on the POWER6 processor and other processors
19838 that support the PowerPC V2.05 architecture.
19839 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19840 general-purpose register instructions implemented on the POWER6X
19841 processor and other processors that support the extended PowerPC V2.05
19843 The @option{-mhard-dfp} option allows GCC to generate the decimal
19844 floating-point instructions implemented on some POWER processors.
19846 The @option{-mpowerpc64} option allows GCC to generate the additional
19847 64-bit instructions that are found in the full PowerPC64 architecture
19848 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19849 @option{-mno-powerpc64}.
19851 @item -mcpu=@var{cpu_type}
19853 Set architecture type, register usage, and
19854 instruction scheduling parameters for machine type @var{cpu_type}.
19855 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19856 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19857 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19858 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19859 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19860 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19861 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19862 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19863 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19864 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
19865 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
19868 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19869 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19870 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19871 architecture machine types, with an appropriate, generic processor
19872 model assumed for scheduling purposes.
19874 The other options specify a specific processor. Code generated under
19875 those options runs best on that processor, and may not run at all on
19878 The @option{-mcpu} options automatically enable or disable the
19881 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19882 -mpopcntb -mpopcntd -mpowerpc64 @gol
19883 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19884 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19885 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19886 -mquad-memory -mquad-memory-atomic -mmodulo -mfloat128 -mfloat128-hardware @gol
19887 -mpower9-fusion -mpower9-vector}
19889 The particular options set for any particular CPU varies between
19890 compiler versions, depending on what setting seems to produce optimal
19891 code for that CPU; it doesn't necessarily reflect the actual hardware's
19892 capabilities. If you wish to set an individual option to a particular
19893 value, you may specify it after the @option{-mcpu} option, like
19894 @option{-mcpu=970 -mno-altivec}.
19896 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19897 not enabled or disabled by the @option{-mcpu} option at present because
19898 AIX does not have full support for these options. You may still
19899 enable or disable them individually if you're sure it'll work in your
19902 @item -mtune=@var{cpu_type}
19904 Set the instruction scheduling parameters for machine type
19905 @var{cpu_type}, but do not set the architecture type or register usage,
19906 as @option{-mcpu=@var{cpu_type}} does. The same
19907 values for @var{cpu_type} are used for @option{-mtune} as for
19908 @option{-mcpu}. If both are specified, the code generated uses the
19909 architecture and registers set by @option{-mcpu}, but the
19910 scheduling parameters set by @option{-mtune}.
19912 @item -mcmodel=small
19913 @opindex mcmodel=small
19914 Generate PowerPC64 code for the small model: The TOC is limited to
19917 @item -mcmodel=medium
19918 @opindex mcmodel=medium
19919 Generate PowerPC64 code for the medium model: The TOC and other static
19920 data may be up to a total of 4G in size.
19922 @item -mcmodel=large
19923 @opindex mcmodel=large
19924 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19925 in size. Other data and code is only limited by the 64-bit address
19929 @itemx -mno-altivec
19931 @opindex mno-altivec
19932 Generate code that uses (does not use) AltiVec instructions, and also
19933 enable the use of built-in functions that allow more direct access to
19934 the AltiVec instruction set. You may also need to set
19935 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19938 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19939 @option{-maltivec=be}, the element order for AltiVec intrinsics such
19940 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19941 match array element order corresponding to the endianness of the
19942 target. That is, element zero identifies the leftmost element in a
19943 vector register when targeting a big-endian platform, and identifies
19944 the rightmost element in a vector register when targeting a
19945 little-endian platform.
19948 @opindex maltivec=be
19949 Generate AltiVec instructions using big-endian element order,
19950 regardless of whether the target is big- or little-endian. This is
19951 the default when targeting a big-endian platform.
19953 The element order is used to interpret element numbers in AltiVec
19954 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19955 @code{vec_insert}. By default, these match array element order
19956 corresponding to the endianness for the target.
19959 @opindex maltivec=le
19960 Generate AltiVec instructions using little-endian element order,
19961 regardless of whether the target is big- or little-endian. This is
19962 the default when targeting a little-endian platform. This option is
19963 currently ignored when targeting a big-endian platform.
19965 The element order is used to interpret element numbers in AltiVec
19966 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19967 @code{vec_insert}. By default, these match array element order
19968 corresponding to the endianness for the target.
19973 @opindex mno-vrsave
19974 Generate VRSAVE instructions when generating AltiVec code.
19976 @item -mgen-cell-microcode
19977 @opindex mgen-cell-microcode
19978 Generate Cell microcode instructions.
19980 @item -mwarn-cell-microcode
19981 @opindex mwarn-cell-microcode
19982 Warn when a Cell microcode instruction is emitted. An example
19983 of a Cell microcode instruction is a variable shift.
19986 @opindex msecure-plt
19987 Generate code that allows @command{ld} and @command{ld.so}
19988 to build executables and shared
19989 libraries with non-executable @code{.plt} and @code{.got} sections.
19991 32-bit SYSV ABI option.
19995 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19997 requires @code{.plt} and @code{.got}
19998 sections that are both writable and executable.
19999 This is a PowerPC 32-bit SYSV ABI option.
20005 This switch enables or disables the generation of ISEL instructions.
20007 @item -misel=@var{yes/no}
20008 This switch has been deprecated. Use @option{-misel} and
20009 @option{-mno-isel} instead.
20015 This switch enables or disables the generation of SPE simd
20021 @opindex mno-paired
20022 This switch enables or disables the generation of PAIRED simd
20025 @item -mspe=@var{yes/no}
20026 This option has been deprecated. Use @option{-mspe} and
20027 @option{-mno-spe} instead.
20033 Generate code that uses (does not use) vector/scalar (VSX)
20034 instructions, and also enable the use of built-in functions that allow
20035 more direct access to the VSX instruction set.
20040 @opindex mno-crypto
20041 Enable the use (disable) of the built-in functions that allow direct
20042 access to the cryptographic instructions that were added in version
20043 2.07 of the PowerPC ISA.
20045 @item -mdirect-move
20046 @itemx -mno-direct-move
20047 @opindex mdirect-move
20048 @opindex mno-direct-move
20049 Generate code that uses (does not use) the instructions to move data
20050 between the general purpose registers and the vector/scalar (VSX)
20051 registers that were added in version 2.07 of the PowerPC ISA.
20053 @item -mpower8-fusion
20054 @itemx -mno-power8-fusion
20055 @opindex mpower8-fusion
20056 @opindex mno-power8-fusion
20057 Generate code that keeps (does not keeps) some integer operations
20058 adjacent so that the instructions can be fused together on power8 and
20061 @item -mpower8-vector
20062 @itemx -mno-power8-vector
20063 @opindex mpower8-vector
20064 @opindex mno-power8-vector
20065 Generate code that uses (does not use) the vector and scalar
20066 instructions that were added in version 2.07 of the PowerPC ISA. Also
20067 enable the use of built-in functions that allow more direct access to
20068 the vector instructions.
20070 @item -mquad-memory
20071 @itemx -mno-quad-memory
20072 @opindex mquad-memory
20073 @opindex mno-quad-memory
20074 Generate code that uses (does not use) the non-atomic quad word memory
20075 instructions. The @option{-mquad-memory} option requires use of
20078 @item -mquad-memory-atomic
20079 @itemx -mno-quad-memory-atomic
20080 @opindex mquad-memory-atomic
20081 @opindex mno-quad-memory-atomic
20082 Generate code that uses (does not use) the atomic quad word memory
20083 instructions. The @option{-mquad-memory-atomic} option requires use of
20086 @item -mupper-regs-df
20087 @itemx -mno-upper-regs-df
20088 @opindex mupper-regs-df
20089 @opindex mno-upper-regs-df
20090 Generate code that uses (does not use) the scalar double precision
20091 instructions that target all 64 registers in the vector/scalar
20092 floating point register set that were added in version 2.06 of the
20093 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
20094 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
20095 @option{-mvsx} options.
20097 @item -mupper-regs-sf
20098 @itemx -mno-upper-regs-sf
20099 @opindex mupper-regs-sf
20100 @opindex mno-upper-regs-sf
20101 Generate code that uses (does not use) the scalar single precision
20102 instructions that target all 64 registers in the vector/scalar
20103 floating point register set that were added in version 2.07 of the
20104 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
20105 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
20109 @itemx -mno-upper-regs
20110 @opindex mupper-regs
20111 @opindex mno-upper-regs
20112 Generate code that uses (does not use) the scalar
20113 instructions that target all 64 registers in the vector/scalar
20114 floating point register set, depending on the model of the machine.
20116 If the @option{-mno-upper-regs} option is used, it turns off both
20117 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
20120 @itemx -mno-float128
20122 @opindex mno-float128
20123 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
20124 and use either software emulation for IEEE 128-bit floating point or
20125 hardware instructions.
20127 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
20128 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
20129 option. The @code{-mfloat128} option only works on PowerPC 64-bit
20132 @item -mfloat128-hardware
20133 @itemx -mno-float128-hardware
20134 @opindex mfloat128-hardware
20135 @opindex mno-float128-hardware
20136 Enable/disable using ISA 3.0 hardware instructions to support the
20137 @var{__float128} data type.
20142 @opindex mno-module
20143 Generate code that uses (does not use) the ISA 3.0 integer modulo
20144 instructions. The @option{-mmodulo} option is enabled by default
20145 with the @option{-mcpu=power9} option.
20147 @item -mpower9-fusion
20148 @itemx -mno-power9-fusion
20149 @opindex mpower9-fusion
20150 @opindex mno-power9-fusion
20151 Generate code that keeps (does not keeps) some operations adjacent so
20152 that the instructions can be fused together on power9 and later
20155 @item -mpower9-vector
20156 @itemx -mno-power9-vector
20157 @opindex mpower9-vector
20158 @opindex mno-power9-vector
20159 Generate code that uses (does not use) the vector and scalar
20160 instructions that were added in version 2.07 of the PowerPC ISA. Also
20161 enable the use of built-in functions that allow more direct access to
20162 the vector instructions.
20164 @item -mfloat-gprs=@var{yes/single/double/no}
20165 @itemx -mfloat-gprs
20166 @opindex mfloat-gprs
20167 This switch enables or disables the generation of floating-point
20168 operations on the general-purpose registers for architectures that
20171 The argument @samp{yes} or @samp{single} enables the use of
20172 single-precision floating-point operations.
20174 The argument @samp{double} enables the use of single and
20175 double-precision floating-point operations.
20177 The argument @samp{no} disables floating-point operations on the
20178 general-purpose registers.
20180 This option is currently only available on the MPC854x.
20186 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
20187 targets (including GNU/Linux). The 32-bit environment sets int, long
20188 and pointer to 32 bits and generates code that runs on any PowerPC
20189 variant. The 64-bit environment sets int to 32 bits and long and
20190 pointer to 64 bits, and generates code for PowerPC64, as for
20191 @option{-mpowerpc64}.
20194 @itemx -mno-fp-in-toc
20195 @itemx -mno-sum-in-toc
20196 @itemx -mminimal-toc
20198 @opindex mno-fp-in-toc
20199 @opindex mno-sum-in-toc
20200 @opindex mminimal-toc
20201 Modify generation of the TOC (Table Of Contents), which is created for
20202 every executable file. The @option{-mfull-toc} option is selected by
20203 default. In that case, GCC allocates at least one TOC entry for
20204 each unique non-automatic variable reference in your program. GCC
20205 also places floating-point constants in the TOC@. However, only
20206 16,384 entries are available in the TOC@.
20208 If you receive a linker error message that saying you have overflowed
20209 the available TOC space, you can reduce the amount of TOC space used
20210 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
20211 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
20212 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
20213 generate code to calculate the sum of an address and a constant at
20214 run time instead of putting that sum into the TOC@. You may specify one
20215 or both of these options. Each causes GCC to produce very slightly
20216 slower and larger code at the expense of conserving TOC space.
20218 If you still run out of space in the TOC even when you specify both of
20219 these options, specify @option{-mminimal-toc} instead. This option causes
20220 GCC to make only one TOC entry for every file. When you specify this
20221 option, GCC produces code that is slower and larger but which
20222 uses extremely little TOC space. You may wish to use this option
20223 only on files that contain less frequently-executed code.
20229 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
20230 @code{long} type, and the infrastructure needed to support them.
20231 Specifying @option{-maix64} implies @option{-mpowerpc64},
20232 while @option{-maix32} disables the 64-bit ABI and
20233 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
20236 @itemx -mno-xl-compat
20237 @opindex mxl-compat
20238 @opindex mno-xl-compat
20239 Produce code that conforms more closely to IBM XL compiler semantics
20240 when using AIX-compatible ABI@. Pass floating-point arguments to
20241 prototyped functions beyond the register save area (RSA) on the stack
20242 in addition to argument FPRs. Do not assume that most significant
20243 double in 128-bit long double value is properly rounded when comparing
20244 values and converting to double. Use XL symbol names for long double
20247 The AIX calling convention was extended but not initially documented to
20248 handle an obscure K&R C case of calling a function that takes the
20249 address of its arguments with fewer arguments than declared. IBM XL
20250 compilers access floating-point arguments that do not fit in the
20251 RSA from the stack when a subroutine is compiled without
20252 optimization. Because always storing floating-point arguments on the
20253 stack is inefficient and rarely needed, this option is not enabled by
20254 default and only is necessary when calling subroutines compiled by IBM
20255 XL compilers without optimization.
20259 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
20260 application written to use message passing with special startup code to
20261 enable the application to run. The system must have PE installed in the
20262 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
20263 must be overridden with the @option{-specs=} option to specify the
20264 appropriate directory location. The Parallel Environment does not
20265 support threads, so the @option{-mpe} option and the @option{-pthread}
20266 option are incompatible.
20268 @item -malign-natural
20269 @itemx -malign-power
20270 @opindex malign-natural
20271 @opindex malign-power
20272 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
20273 @option{-malign-natural} overrides the ABI-defined alignment of larger
20274 types, such as floating-point doubles, on their natural size-based boundary.
20275 The option @option{-malign-power} instructs GCC to follow the ABI-specified
20276 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
20278 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
20282 @itemx -mhard-float
20283 @opindex msoft-float
20284 @opindex mhard-float
20285 Generate code that does not use (uses) the floating-point register set.
20286 Software floating-point emulation is provided if you use the
20287 @option{-msoft-float} option, and pass the option to GCC when linking.
20289 @item -msingle-float
20290 @itemx -mdouble-float
20291 @opindex msingle-float
20292 @opindex mdouble-float
20293 Generate code for single- or double-precision floating-point operations.
20294 @option{-mdouble-float} implies @option{-msingle-float}.
20297 @opindex msimple-fpu
20298 Do not generate @code{sqrt} and @code{div} instructions for hardware
20299 floating-point unit.
20301 @item -mfpu=@var{name}
20303 Specify type of floating-point unit. Valid values for @var{name} are
20304 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
20305 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
20306 @samp{sp_full} (equivalent to @option{-msingle-float}),
20307 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
20310 @opindex mxilinx-fpu
20311 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
20314 @itemx -mno-multiple
20316 @opindex mno-multiple
20317 Generate code that uses (does not use) the load multiple word
20318 instructions and the store multiple word instructions. These
20319 instructions are generated by default on POWER systems, and not
20320 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
20321 PowerPC systems, since those instructions do not work when the
20322 processor is in little-endian mode. The exceptions are PPC740 and
20323 PPC750 which permit these instructions in little-endian mode.
20328 @opindex mno-string
20329 Generate code that uses (does not use) the load string instructions
20330 and the store string word instructions to save multiple registers and
20331 do small block moves. These instructions are generated by default on
20332 POWER systems, and not generated on PowerPC systems. Do not use
20333 @option{-mstring} on little-endian PowerPC systems, since those
20334 instructions do not work when the processor is in little-endian mode.
20335 The exceptions are PPC740 and PPC750 which permit these instructions
20336 in little-endian mode.
20341 @opindex mno-update
20342 Generate code that uses (does not use) the load or store instructions
20343 that update the base register to the address of the calculated memory
20344 location. These instructions are generated by default. If you use
20345 @option{-mno-update}, there is a small window between the time that the
20346 stack pointer is updated and the address of the previous frame is
20347 stored, which means code that walks the stack frame across interrupts or
20348 signals may get corrupted data.
20350 @item -mavoid-indexed-addresses
20351 @itemx -mno-avoid-indexed-addresses
20352 @opindex mavoid-indexed-addresses
20353 @opindex mno-avoid-indexed-addresses
20354 Generate code that tries to avoid (not avoid) the use of indexed load
20355 or store instructions. These instructions can incur a performance
20356 penalty on Power6 processors in certain situations, such as when
20357 stepping through large arrays that cross a 16M boundary. This option
20358 is enabled by default when targeting Power6 and disabled otherwise.
20361 @itemx -mno-fused-madd
20362 @opindex mfused-madd
20363 @opindex mno-fused-madd
20364 Generate code that uses (does not use) the floating-point multiply and
20365 accumulate instructions. These instructions are generated by default
20366 if hardware floating point is used. The machine-dependent
20367 @option{-mfused-madd} option is now mapped to the machine-independent
20368 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20369 mapped to @option{-ffp-contract=off}.
20375 Generate code that uses (does not use) the half-word multiply and
20376 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
20377 These instructions are generated by default when targeting those
20384 Generate code that uses (does not use) the string-search @samp{dlmzb}
20385 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
20386 generated by default when targeting those processors.
20388 @item -mno-bit-align
20390 @opindex mno-bit-align
20391 @opindex mbit-align
20392 On System V.4 and embedded PowerPC systems do not (do) force structures
20393 and unions that contain bit-fields to be aligned to the base type of the
20396 For example, by default a structure containing nothing but 8
20397 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
20398 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
20399 the structure is aligned to a 1-byte boundary and is 1 byte in
20402 @item -mno-strict-align
20403 @itemx -mstrict-align
20404 @opindex mno-strict-align
20405 @opindex mstrict-align
20406 On System V.4 and embedded PowerPC systems do not (do) assume that
20407 unaligned memory references are handled by the system.
20409 @item -mrelocatable
20410 @itemx -mno-relocatable
20411 @opindex mrelocatable
20412 @opindex mno-relocatable
20413 Generate code that allows (does not allow) a static executable to be
20414 relocated to a different address at run time. A simple embedded
20415 PowerPC system loader should relocate the entire contents of
20416 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20417 a table of 32-bit addresses generated by this option. For this to
20418 work, all objects linked together must be compiled with
20419 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20420 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20422 @item -mrelocatable-lib
20423 @itemx -mno-relocatable-lib
20424 @opindex mrelocatable-lib
20425 @opindex mno-relocatable-lib
20426 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20427 @code{.fixup} section to allow static executables to be relocated at
20428 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20429 alignment of @option{-mrelocatable}. Objects compiled with
20430 @option{-mrelocatable-lib} may be linked with objects compiled with
20431 any combination of the @option{-mrelocatable} options.
20437 On System V.4 and embedded PowerPC systems do not (do) assume that
20438 register 2 contains a pointer to a global area pointing to the addresses
20439 used in the program.
20442 @itemx -mlittle-endian
20444 @opindex mlittle-endian
20445 On System V.4 and embedded PowerPC systems compile code for the
20446 processor in little-endian mode. The @option{-mlittle-endian} option is
20447 the same as @option{-mlittle}.
20450 @itemx -mbig-endian
20452 @opindex mbig-endian
20453 On System V.4 and embedded PowerPC systems compile code for the
20454 processor in big-endian mode. The @option{-mbig-endian} option is
20455 the same as @option{-mbig}.
20457 @item -mdynamic-no-pic
20458 @opindex mdynamic-no-pic
20459 On Darwin and Mac OS X systems, compile code so that it is not
20460 relocatable, but that its external references are relocatable. The
20461 resulting code is suitable for applications, but not shared
20464 @item -msingle-pic-base
20465 @opindex msingle-pic-base
20466 Treat the register used for PIC addressing as read-only, rather than
20467 loading it in the prologue for each function. The runtime system is
20468 responsible for initializing this register with an appropriate value
20469 before execution begins.
20471 @item -mprioritize-restricted-insns=@var{priority}
20472 @opindex mprioritize-restricted-insns
20473 This option controls the priority that is assigned to
20474 dispatch-slot restricted instructions during the second scheduling
20475 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20476 or @samp{2} to assign no, highest, or second-highest (respectively)
20477 priority to dispatch-slot restricted
20480 @item -msched-costly-dep=@var{dependence_type}
20481 @opindex msched-costly-dep
20482 This option controls which dependences are considered costly
20483 by the target during instruction scheduling. The argument
20484 @var{dependence_type} takes one of the following values:
20488 No dependence is costly.
20491 All dependences are costly.
20493 @item @samp{true_store_to_load}
20494 A true dependence from store to load is costly.
20496 @item @samp{store_to_load}
20497 Any dependence from store to load is costly.
20500 Any dependence for which the latency is greater than or equal to
20501 @var{number} is costly.
20504 @item -minsert-sched-nops=@var{scheme}
20505 @opindex minsert-sched-nops
20506 This option controls which NOP insertion scheme is used during
20507 the second scheduling pass. The argument @var{scheme} takes one of the
20515 Pad with NOPs any dispatch group that has vacant issue slots,
20516 according to the scheduler's grouping.
20518 @item @samp{regroup_exact}
20519 Insert NOPs to force costly dependent insns into
20520 separate groups. Insert exactly as many NOPs as needed to force an insn
20521 to a new group, according to the estimated processor grouping.
20524 Insert NOPs to force costly dependent insns into
20525 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20529 @opindex mcall-sysv
20530 On System V.4 and embedded PowerPC systems compile code using calling
20531 conventions that adhere to the March 1995 draft of the System V
20532 Application Binary Interface, PowerPC processor supplement. This is the
20533 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20535 @item -mcall-sysv-eabi
20537 @opindex mcall-sysv-eabi
20538 @opindex mcall-eabi
20539 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20541 @item -mcall-sysv-noeabi
20542 @opindex mcall-sysv-noeabi
20543 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20545 @item -mcall-aixdesc
20547 On System V.4 and embedded PowerPC systems compile code for the AIX
20551 @opindex mcall-linux
20552 On System V.4 and embedded PowerPC systems compile code for the
20553 Linux-based GNU system.
20555 @item -mcall-freebsd
20556 @opindex mcall-freebsd
20557 On System V.4 and embedded PowerPC systems compile code for the
20558 FreeBSD operating system.
20560 @item -mcall-netbsd
20561 @opindex mcall-netbsd
20562 On System V.4 and embedded PowerPC systems compile code for the
20563 NetBSD operating system.
20565 @item -mcall-openbsd
20566 @opindex mcall-netbsd
20567 On System V.4 and embedded PowerPC systems compile code for the
20568 OpenBSD operating system.
20570 @item -maix-struct-return
20571 @opindex maix-struct-return
20572 Return all structures in memory (as specified by the AIX ABI)@.
20574 @item -msvr4-struct-return
20575 @opindex msvr4-struct-return
20576 Return structures smaller than 8 bytes in registers (as specified by the
20579 @item -mabi=@var{abi-type}
20581 Extend the current ABI with a particular extension, or remove such extension.
20582 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20583 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20584 @samp{elfv1}, @samp{elfv2}@.
20588 Extend the current ABI with SPE ABI extensions. This does not change
20589 the default ABI, instead it adds the SPE ABI extensions to the current
20593 @opindex mabi=no-spe
20594 Disable Book-E SPE ABI extensions for the current ABI@.
20596 @item -mabi=ibmlongdouble
20597 @opindex mabi=ibmlongdouble
20598 Change the current ABI to use IBM extended-precision long double.
20599 This is a PowerPC 32-bit SYSV ABI option.
20601 @item -mabi=ieeelongdouble
20602 @opindex mabi=ieeelongdouble
20603 Change the current ABI to use IEEE extended-precision long double.
20604 This is a PowerPC 32-bit Linux ABI option.
20607 @opindex mabi=elfv1
20608 Change the current ABI to use the ELFv1 ABI.
20609 This is the default ABI for big-endian PowerPC 64-bit Linux.
20610 Overriding the default ABI requires special system support and is
20611 likely to fail in spectacular ways.
20614 @opindex mabi=elfv2
20615 Change the current ABI to use the ELFv2 ABI.
20616 This is the default ABI for little-endian PowerPC 64-bit Linux.
20617 Overriding the default ABI requires special system support and is
20618 likely to fail in spectacular ways.
20621 @itemx -mno-prototype
20622 @opindex mprototype
20623 @opindex mno-prototype
20624 On System V.4 and embedded PowerPC systems assume that all calls to
20625 variable argument functions are properly prototyped. Otherwise, the
20626 compiler must insert an instruction before every non-prototyped call to
20627 set or clear bit 6 of the condition code register (@code{CR}) to
20628 indicate whether floating-point values are passed in the floating-point
20629 registers in case the function takes variable arguments. With
20630 @option{-mprototype}, only calls to prototyped variable argument functions
20631 set or clear the bit.
20635 On embedded PowerPC systems, assume that the startup module is called
20636 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20637 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20642 On embedded PowerPC systems, assume that the startup module is called
20643 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20648 On embedded PowerPC systems, assume that the startup module is called
20649 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20652 @item -myellowknife
20653 @opindex myellowknife
20654 On embedded PowerPC systems, assume that the startup module is called
20655 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20660 On System V.4 and embedded PowerPC systems, specify that you are
20661 compiling for a VxWorks system.
20665 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20666 header to indicate that @samp{eabi} extended relocations are used.
20672 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20673 Embedded Applications Binary Interface (EABI), which is a set of
20674 modifications to the System V.4 specifications. Selecting @option{-meabi}
20675 means that the stack is aligned to an 8-byte boundary, a function
20676 @code{__eabi} is called from @code{main} to set up the EABI
20677 environment, and the @option{-msdata} option can use both @code{r2} and
20678 @code{r13} to point to two separate small data areas. Selecting
20679 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20680 no EABI initialization function is called from @code{main}, and the
20681 @option{-msdata} option only uses @code{r13} to point to a single
20682 small data area. The @option{-meabi} option is on by default if you
20683 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20686 @opindex msdata=eabi
20687 On System V.4 and embedded PowerPC systems, put small initialized
20688 @code{const} global and static data in the @code{.sdata2} section, which
20689 is pointed to by register @code{r2}. Put small initialized
20690 non-@code{const} global and static data in the @code{.sdata} section,
20691 which is pointed to by register @code{r13}. Put small uninitialized
20692 global and static data in the @code{.sbss} section, which is adjacent to
20693 the @code{.sdata} section. The @option{-msdata=eabi} option is
20694 incompatible with the @option{-mrelocatable} option. The
20695 @option{-msdata=eabi} option also sets the @option{-memb} option.
20698 @opindex msdata=sysv
20699 On System V.4 and embedded PowerPC systems, put small global and static
20700 data in the @code{.sdata} section, which is pointed to by register
20701 @code{r13}. Put small uninitialized global and static data in the
20702 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20703 The @option{-msdata=sysv} option is incompatible with the
20704 @option{-mrelocatable} option.
20706 @item -msdata=default
20708 @opindex msdata=default
20710 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20711 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20712 same as @option{-msdata=sysv}.
20715 @opindex msdata=data
20716 On System V.4 and embedded PowerPC systems, put small global
20717 data in the @code{.sdata} section. Put small uninitialized global
20718 data in the @code{.sbss} section. Do not use register @code{r13}
20719 to address small data however. This is the default behavior unless
20720 other @option{-msdata} options are used.
20724 @opindex msdata=none
20726 On embedded PowerPC systems, put all initialized global and static data
20727 in the @code{.data} section, and all uninitialized data in the
20728 @code{.bss} section.
20730 @item -mblock-move-inline-limit=@var{num}
20731 @opindex mblock-move-inline-limit
20732 Inline all block moves (such as calls to @code{memcpy} or structure
20733 copies) less than or equal to @var{num} bytes. The minimum value for
20734 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20735 targets. The default value is target-specific.
20739 @cindex smaller data references (PowerPC)
20740 @cindex .sdata/.sdata2 references (PowerPC)
20741 On embedded PowerPC systems, put global and static items less than or
20742 equal to @var{num} bytes into the small data or BSS sections instead of
20743 the normal data or BSS section. By default, @var{num} is 8. The
20744 @option{-G @var{num}} switch is also passed to the linker.
20745 All modules should be compiled with the same @option{-G @var{num}} value.
20748 @itemx -mno-regnames
20750 @opindex mno-regnames
20751 On System V.4 and embedded PowerPC systems do (do not) emit register
20752 names in the assembly language output using symbolic forms.
20755 @itemx -mno-longcall
20757 @opindex mno-longcall
20758 By default assume that all calls are far away so that a longer and more
20759 expensive calling sequence is required. This is required for calls
20760 farther than 32 megabytes (33,554,432 bytes) from the current location.
20761 A short call is generated if the compiler knows
20762 the call cannot be that far away. This setting can be overridden by
20763 the @code{shortcall} function attribute, or by @code{#pragma
20766 Some linkers are capable of detecting out-of-range calls and generating
20767 glue code on the fly. On these systems, long calls are unnecessary and
20768 generate slower code. As of this writing, the AIX linker can do this,
20769 as can the GNU linker for PowerPC/64. It is planned to add this feature
20770 to the GNU linker for 32-bit PowerPC systems as well.
20772 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20773 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20774 addresses represent the callee and the branch island. The
20775 Darwin/PPC linker prefers the first address and generates a @code{bl
20776 callee} if the PPC @code{bl} instruction reaches the callee directly;
20777 otherwise, the linker generates @code{bl L42} to call the branch
20778 island. The branch island is appended to the body of the
20779 calling function; it computes the full 32-bit address of the callee
20782 On Mach-O (Darwin) systems, this option directs the compiler emit to
20783 the glue for every direct call, and the Darwin linker decides whether
20784 to use or discard it.
20786 In the future, GCC may ignore all longcall specifications
20787 when the linker is known to generate glue.
20789 @item -mtls-markers
20790 @itemx -mno-tls-markers
20791 @opindex mtls-markers
20792 @opindex mno-tls-markers
20793 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20794 specifying the function argument. The relocation allows the linker to
20795 reliably associate function call with argument setup instructions for
20796 TLS optimization, which in turn allows GCC to better schedule the
20801 Adds support for multithreading with the @dfn{pthreads} library.
20802 This option sets flags for both the preprocessor and linker.
20807 This option enables use of the reciprocal estimate and
20808 reciprocal square root estimate instructions with additional
20809 Newton-Raphson steps to increase precision instead of doing a divide or
20810 square root and divide for floating-point arguments. You should use
20811 the @option{-ffast-math} option when using @option{-mrecip} (or at
20812 least @option{-funsafe-math-optimizations},
20813 @option{-ffinite-math-only}, @option{-freciprocal-math} and
20814 @option{-fno-trapping-math}). Note that while the throughput of the
20815 sequence is generally higher than the throughput of the non-reciprocal
20816 instruction, the precision of the sequence can be decreased by up to 2
20817 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20820 @item -mrecip=@var{opt}
20821 @opindex mrecip=opt
20822 This option controls which reciprocal estimate instructions
20823 may be used. @var{opt} is a comma-separated list of options, which may
20824 be preceded by a @code{!} to invert the option:
20829 Enable all estimate instructions.
20832 Enable the default instructions, equivalent to @option{-mrecip}.
20835 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20838 Enable the reciprocal approximation instructions for both
20839 single and double precision.
20842 Enable the single-precision reciprocal approximation instructions.
20845 Enable the double-precision reciprocal approximation instructions.
20848 Enable the reciprocal square root approximation instructions for both
20849 single and double precision.
20852 Enable the single-precision reciprocal square root approximation instructions.
20855 Enable the double-precision reciprocal square root approximation instructions.
20859 So, for example, @option{-mrecip=all,!rsqrtd} enables
20860 all of the reciprocal estimate instructions, except for the
20861 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20862 which handle the double-precision reciprocal square root calculations.
20864 @item -mrecip-precision
20865 @itemx -mno-recip-precision
20866 @opindex mrecip-precision
20867 Assume (do not assume) that the reciprocal estimate instructions
20868 provide higher-precision estimates than is mandated by the PowerPC
20869 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20870 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20871 The double-precision square root estimate instructions are not generated by
20872 default on low-precision machines, since they do not provide an
20873 estimate that converges after three steps.
20875 @item -mveclibabi=@var{type}
20876 @opindex mveclibabi
20877 Specifies the ABI type to use for vectorizing intrinsics using an
20878 external library. The only type supported at present is @samp{mass},
20879 which specifies to use IBM's Mathematical Acceleration Subsystem
20880 (MASS) libraries for vectorizing intrinsics using external libraries.
20881 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20882 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20883 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20884 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20885 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20886 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20887 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20888 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20889 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20890 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20891 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20892 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20893 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20894 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20895 for power7. Both @option{-ftree-vectorize} and
20896 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20897 libraries must be specified at link time.
20902 Generate (do not generate) the @code{friz} instruction when the
20903 @option{-funsafe-math-optimizations} option is used to optimize
20904 rounding of floating-point values to 64-bit integer and back to floating
20905 point. The @code{friz} instruction does not return the same value if
20906 the floating-point number is too large to fit in an integer.
20908 @item -mpointers-to-nested-functions
20909 @itemx -mno-pointers-to-nested-functions
20910 @opindex mpointers-to-nested-functions
20911 Generate (do not generate) code to load up the static chain register
20912 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20913 systems where a function pointer points to a 3-word descriptor giving
20914 the function address, TOC value to be loaded in register @code{r2}, and
20915 static chain value to be loaded in register @code{r11}. The
20916 @option{-mpointers-to-nested-functions} is on by default. You cannot
20917 call through pointers to nested functions or pointers
20918 to functions compiled in other languages that use the static chain if
20919 you use @option{-mno-pointers-to-nested-functions}.
20921 @item -msave-toc-indirect
20922 @itemx -mno-save-toc-indirect
20923 @opindex msave-toc-indirect
20924 Generate (do not generate) code to save the TOC value in the reserved
20925 stack location in the function prologue if the function calls through
20926 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20927 saved in the prologue, it is saved just before the call through the
20928 pointer. The @option{-mno-save-toc-indirect} option is the default.
20930 @item -mcompat-align-parm
20931 @itemx -mno-compat-align-parm
20932 @opindex mcompat-align-parm
20933 Generate (do not generate) code to pass structure parameters with a
20934 maximum alignment of 64 bits, for compatibility with older versions
20937 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20938 structure parameter on a 128-bit boundary when that structure contained
20939 a member requiring 128-bit alignment. This is corrected in more
20940 recent versions of GCC. This option may be used to generate code
20941 that is compatible with functions compiled with older versions of
20944 The @option{-mno-compat-align-parm} option is the default.
20948 @subsection RX Options
20951 These command-line options are defined for RX targets:
20954 @item -m64bit-doubles
20955 @itemx -m32bit-doubles
20956 @opindex m64bit-doubles
20957 @opindex m32bit-doubles
20958 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20959 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20960 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20961 works on 32-bit values, which is why the default is
20962 @option{-m32bit-doubles}.
20968 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20969 floating-point hardware. The default is enabled for the RX600
20970 series and disabled for the RX200 series.
20972 Floating-point instructions are only generated for 32-bit floating-point
20973 values, however, so the FPU hardware is not used for doubles if the
20974 @option{-m64bit-doubles} option is used.
20976 @emph{Note} If the @option{-fpu} option is enabled then
20977 @option{-funsafe-math-optimizations} is also enabled automatically.
20978 This is because the RX FPU instructions are themselves unsafe.
20980 @item -mcpu=@var{name}
20982 Selects the type of RX CPU to be targeted. Currently three types are
20983 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20984 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20986 The only difference between @samp{RX600} and @samp{RX610} is that the
20987 @samp{RX610} does not support the @code{MVTIPL} instruction.
20989 The @samp{RX200} series does not have a hardware floating-point unit
20990 and so @option{-nofpu} is enabled by default when this type is
20993 @item -mbig-endian-data
20994 @itemx -mlittle-endian-data
20995 @opindex mbig-endian-data
20996 @opindex mlittle-endian-data
20997 Store data (but not code) in the big-endian format. The default is
20998 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
21001 @item -msmall-data-limit=@var{N}
21002 @opindex msmall-data-limit
21003 Specifies the maximum size in bytes of global and static variables
21004 which can be placed into the small data area. Using the small data
21005 area can lead to smaller and faster code, but the size of area is
21006 limited and it is up to the programmer to ensure that the area does
21007 not overflow. Also when the small data area is used one of the RX's
21008 registers (usually @code{r13}) is reserved for use pointing to this
21009 area, so it is no longer available for use by the compiler. This
21010 could result in slower and/or larger code if variables are pushed onto
21011 the stack instead of being held in this register.
21013 Note, common variables (variables that have not been initialized) and
21014 constants are not placed into the small data area as they are assigned
21015 to other sections in the output executable.
21017 The default value is zero, which disables this feature. Note, this
21018 feature is not enabled by default with higher optimization levels
21019 (@option{-O2} etc) because of the potentially detrimental effects of
21020 reserving a register. It is up to the programmer to experiment and
21021 discover whether this feature is of benefit to their program. See the
21022 description of the @option{-mpid} option for a description of how the
21023 actual register to hold the small data area pointer is chosen.
21029 Use the simulator runtime. The default is to use the libgloss
21030 board-specific runtime.
21032 @item -mas100-syntax
21033 @itemx -mno-as100-syntax
21034 @opindex mas100-syntax
21035 @opindex mno-as100-syntax
21036 When generating assembler output use a syntax that is compatible with
21037 Renesas's AS100 assembler. This syntax can also be handled by the GAS
21038 assembler, but it has some restrictions so it is not generated by default.
21040 @item -mmax-constant-size=@var{N}
21041 @opindex mmax-constant-size
21042 Specifies the maximum size, in bytes, of a constant that can be used as
21043 an operand in a RX instruction. Although the RX instruction set does
21044 allow constants of up to 4 bytes in length to be used in instructions,
21045 a longer value equates to a longer instruction. Thus in some
21046 circumstances it can be beneficial to restrict the size of constants
21047 that are used in instructions. Constants that are too big are instead
21048 placed into a constant pool and referenced via register indirection.
21050 The value @var{N} can be between 0 and 4. A value of 0 (the default)
21051 or 4 means that constants of any size are allowed.
21055 Enable linker relaxation. Linker relaxation is a process whereby the
21056 linker attempts to reduce the size of a program by finding shorter
21057 versions of various instructions. Disabled by default.
21059 @item -mint-register=@var{N}
21060 @opindex mint-register
21061 Specify the number of registers to reserve for fast interrupt handler
21062 functions. The value @var{N} can be between 0 and 4. A value of 1
21063 means that register @code{r13} is reserved for the exclusive use
21064 of fast interrupt handlers. A value of 2 reserves @code{r13} and
21065 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
21066 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
21067 A value of 0, the default, does not reserve any registers.
21069 @item -msave-acc-in-interrupts
21070 @opindex msave-acc-in-interrupts
21071 Specifies that interrupt handler functions should preserve the
21072 accumulator register. This is only necessary if normal code might use
21073 the accumulator register, for example because it performs 64-bit
21074 multiplications. The default is to ignore the accumulator as this
21075 makes the interrupt handlers faster.
21081 Enables the generation of position independent data. When enabled any
21082 access to constant data is done via an offset from a base address
21083 held in a register. This allows the location of constant data to be
21084 determined at run time without requiring the executable to be
21085 relocated, which is a benefit to embedded applications with tight
21086 memory constraints. Data that can be modified is not affected by this
21089 Note, using this feature reserves a register, usually @code{r13}, for
21090 the constant data base address. This can result in slower and/or
21091 larger code, especially in complicated functions.
21093 The actual register chosen to hold the constant data base address
21094 depends upon whether the @option{-msmall-data-limit} and/or the
21095 @option{-mint-register} command-line options are enabled. Starting
21096 with register @code{r13} and proceeding downwards, registers are
21097 allocated first to satisfy the requirements of @option{-mint-register},
21098 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
21099 is possible for the small data area register to be @code{r8} if both
21100 @option{-mint-register=4} and @option{-mpid} are specified on the
21103 By default this feature is not enabled. The default can be restored
21104 via the @option{-mno-pid} command-line option.
21106 @item -mno-warn-multiple-fast-interrupts
21107 @itemx -mwarn-multiple-fast-interrupts
21108 @opindex mno-warn-multiple-fast-interrupts
21109 @opindex mwarn-multiple-fast-interrupts
21110 Prevents GCC from issuing a warning message if it finds more than one
21111 fast interrupt handler when it is compiling a file. The default is to
21112 issue a warning for each extra fast interrupt handler found, as the RX
21113 only supports one such interrupt.
21115 @item -mallow-string-insns
21116 @itemx -mno-allow-string-insns
21117 @opindex mallow-string-insns
21118 @opindex mno-allow-string-insns
21119 Enables or disables the use of the string manipulation instructions
21120 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
21121 @code{SWHILE} and also the @code{RMPA} instruction. These
21122 instructions may prefetch data, which is not safe to do if accessing
21123 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
21124 for more information).
21126 The default is to allow these instructions, but it is not possible for
21127 GCC to reliably detect all circumstances where a string instruction
21128 might be used to access an I/O register, so their use cannot be
21129 disabled automatically. Instead it is reliant upon the programmer to
21130 use the @option{-mno-allow-string-insns} option if their program
21131 accesses I/O space.
21133 When the instructions are enabled GCC defines the C preprocessor
21134 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
21135 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
21141 Use only (or not only) @code{JSR} instructions to access functions.
21142 This option can be used when code size exceeds the range of @code{BSR}
21143 instructions. Note that @option{-mno-jsr} does not mean to not use
21144 @code{JSR} but instead means that any type of branch may be used.
21147 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
21148 has special significance to the RX port when used with the
21149 @code{interrupt} function attribute. This attribute indicates a
21150 function intended to process fast interrupts. GCC ensures
21151 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
21152 and/or @code{r13} and only provided that the normal use of the
21153 corresponding registers have been restricted via the
21154 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
21157 @node S/390 and zSeries Options
21158 @subsection S/390 and zSeries Options
21159 @cindex S/390 and zSeries Options
21161 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
21165 @itemx -msoft-float
21166 @opindex mhard-float
21167 @opindex msoft-float
21168 Use (do not use) the hardware floating-point instructions and registers
21169 for floating-point operations. When @option{-msoft-float} is specified,
21170 functions in @file{libgcc.a} are used to perform floating-point
21171 operations. When @option{-mhard-float} is specified, the compiler
21172 generates IEEE floating-point instructions. This is the default.
21175 @itemx -mno-hard-dfp
21177 @opindex mno-hard-dfp
21178 Use (do not use) the hardware decimal-floating-point instructions for
21179 decimal-floating-point operations. When @option{-mno-hard-dfp} is
21180 specified, functions in @file{libgcc.a} are used to perform
21181 decimal-floating-point operations. When @option{-mhard-dfp} is
21182 specified, the compiler generates decimal-floating-point hardware
21183 instructions. This is the default for @option{-march=z9-ec} or higher.
21185 @item -mlong-double-64
21186 @itemx -mlong-double-128
21187 @opindex mlong-double-64
21188 @opindex mlong-double-128
21189 These switches control the size of @code{long double} type. A size
21190 of 64 bits makes the @code{long double} type equivalent to the @code{double}
21191 type. This is the default.
21194 @itemx -mno-backchain
21195 @opindex mbackchain
21196 @opindex mno-backchain
21197 Store (do not store) the address of the caller's frame as backchain pointer
21198 into the callee's stack frame.
21199 A backchain may be needed to allow debugging using tools that do not understand
21200 DWARF call frame information.
21201 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
21202 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
21203 the backchain is placed into the topmost word of the 96/160 byte register
21206 In general, code compiled with @option{-mbackchain} is call-compatible with
21207 code compiled with @option{-mmo-backchain}; however, use of the backchain
21208 for debugging purposes usually requires that the whole binary is built with
21209 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
21210 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21211 to build a linux kernel use @option{-msoft-float}.
21213 The default is to not maintain the backchain.
21215 @item -mpacked-stack
21216 @itemx -mno-packed-stack
21217 @opindex mpacked-stack
21218 @opindex mno-packed-stack
21219 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
21220 specified, the compiler uses the all fields of the 96/160 byte register save
21221 area only for their default purpose; unused fields still take up stack space.
21222 When @option{-mpacked-stack} is specified, register save slots are densely
21223 packed at the top of the register save area; unused space is reused for other
21224 purposes, allowing for more efficient use of the available stack space.
21225 However, when @option{-mbackchain} is also in effect, the topmost word of
21226 the save area is always used to store the backchain, and the return address
21227 register is always saved two words below the backchain.
21229 As long as the stack frame backchain is not used, code generated with
21230 @option{-mpacked-stack} is call-compatible with code generated with
21231 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
21232 S/390 or zSeries generated code that uses the stack frame backchain at run
21233 time, not just for debugging purposes. Such code is not call-compatible
21234 with code compiled with @option{-mpacked-stack}. Also, note that the
21235 combination of @option{-mbackchain},
21236 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21237 to build a linux kernel use @option{-msoft-float}.
21239 The default is to not use the packed stack layout.
21242 @itemx -mno-small-exec
21243 @opindex msmall-exec
21244 @opindex mno-small-exec
21245 Generate (or do not generate) code using the @code{bras} instruction
21246 to do subroutine calls.
21247 This only works reliably if the total executable size does not
21248 exceed 64k. The default is to use the @code{basr} instruction instead,
21249 which does not have this limitation.
21255 When @option{-m31} is specified, generate code compliant to the
21256 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
21257 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
21258 particular to generate 64-bit instructions. For the @samp{s390}
21259 targets, the default is @option{-m31}, while the @samp{s390x}
21260 targets default to @option{-m64}.
21266 When @option{-mzarch} is specified, generate code using the
21267 instructions available on z/Architecture.
21268 When @option{-mesa} is specified, generate code using the
21269 instructions available on ESA/390. Note that @option{-mesa} is
21270 not possible with @option{-m64}.
21271 When generating code compliant to the GNU/Linux for S/390 ABI,
21272 the default is @option{-mesa}. When generating code compliant
21273 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
21279 The @option{-mhtm} option enables a set of builtins making use of
21280 instructions available with the transactional execution facility
21281 introduced with the IBM zEnterprise EC12 machine generation
21282 @ref{S/390 System z Built-in Functions}.
21283 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
21289 When @option{-mvx} is specified, generate code using the instructions
21290 available with the vector extension facility introduced with the IBM
21291 z13 machine generation.
21292 This option changes the ABI for some vector type values with regard to
21293 alignment and calling conventions. In case vector type values are
21294 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
21295 command will be added to mark the resulting binary with the ABI used.
21296 @option{-mvx} is enabled by default when using @option{-march=z13}.
21299 @itemx -mno-zvector
21301 @opindex mno-zvector
21302 The @option{-mzvector} option enables vector language extensions and
21303 builtins using instructions available with the vector extension
21304 facility introduced with the IBM z13 machine generation.
21305 This option adds support for @samp{vector} to be used as a keyword to
21306 define vector type variables and arguments. @samp{vector} is only
21307 available when GNU extensions are enabled. It will not be expanded
21308 when requesting strict standard compliance e.g. with @option{-std=c99}.
21309 In addition to the GCC low-level builtins @option{-mzvector} enables
21310 a set of builtins added for compatibility with AltiVec-style
21311 implementations like Power and Cell. In order to make use of these
21312 builtins the header file @file{vecintrin.h} needs to be included.
21313 @option{-mzvector} is disabled by default.
21319 Generate (or do not generate) code using the @code{mvcle} instruction
21320 to perform block moves. When @option{-mno-mvcle} is specified,
21321 use a @code{mvc} loop instead. This is the default unless optimizing for
21328 Print (or do not print) additional debug information when compiling.
21329 The default is to not print debug information.
21331 @item -march=@var{cpu-type}
21333 Generate code that runs on @var{cpu-type}, which is the name of a
21334 system representing a certain processor type. Possible values for
21335 @var{cpu-type} are @samp{z900}, @samp{z990}, @samp{z9-109},
21336 @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12}, and @samp{z13}.
21337 The default is @option{-march=z900}. @samp{g5} and @samp{g6} are
21338 deprecated and will be removed with future releases.
21340 @item -mtune=@var{cpu-type}
21342 Tune to @var{cpu-type} everything applicable about the generated code,
21343 except for the ABI and the set of available instructions.
21344 The list of @var{cpu-type} values is the same as for @option{-march}.
21345 The default is the value used for @option{-march}.
21348 @itemx -mno-tpf-trace
21349 @opindex mtpf-trace
21350 @opindex mno-tpf-trace
21351 Generate code that adds (does not add) in TPF OS specific branches to trace
21352 routines in the operating system. This option is off by default, even
21353 when compiling for the TPF OS@.
21356 @itemx -mno-fused-madd
21357 @opindex mfused-madd
21358 @opindex mno-fused-madd
21359 Generate code that uses (does not use) the floating-point multiply and
21360 accumulate instructions. These instructions are generated by default if
21361 hardware floating point is used.
21363 @item -mwarn-framesize=@var{framesize}
21364 @opindex mwarn-framesize
21365 Emit a warning if the current function exceeds the given frame size. Because
21366 this is a compile-time check it doesn't need to be a real problem when the program
21367 runs. It is intended to identify functions that most probably cause
21368 a stack overflow. It is useful to be used in an environment with limited stack
21369 size e.g.@: the linux kernel.
21371 @item -mwarn-dynamicstack
21372 @opindex mwarn-dynamicstack
21373 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
21374 arrays. This is generally a bad idea with a limited stack size.
21376 @item -mstack-guard=@var{stack-guard}
21377 @itemx -mstack-size=@var{stack-size}
21378 @opindex mstack-guard
21379 @opindex mstack-size
21380 If these options are provided the S/390 back end emits additional instructions in
21381 the function prologue that trigger a trap if the stack size is @var{stack-guard}
21382 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
21383 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
21384 the frame size of the compiled function is chosen.
21385 These options are intended to be used to help debugging stack overflow problems.
21386 The additionally emitted code causes only little overhead and hence can also be
21387 used in production-like systems without greater performance degradation. The given
21388 values have to be exact powers of 2 and @var{stack-size} has to be greater than
21389 @var{stack-guard} without exceeding 64k.
21390 In order to be efficient the extra code makes the assumption that the stack starts
21391 at an address aligned to the value given by @var{stack-size}.
21392 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
21394 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
21396 If the hotpatch option is enabled, a ``hot-patching'' function
21397 prologue is generated for all functions in the compilation unit.
21398 The funtion label is prepended with the given number of two-byte
21399 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
21400 the label, 2 * @var{post-halfwords} bytes are appended, using the
21401 largest NOP like instructions the architecture allows (maximum
21404 If both arguments are zero, hotpatching is disabled.
21406 This option can be overridden for individual functions with the
21407 @code{hotpatch} attribute.
21410 @node Score Options
21411 @subsection Score Options
21412 @cindex Score Options
21414 These options are defined for Score implementations:
21419 Compile code for big-endian mode. This is the default.
21423 Compile code for little-endian mode.
21427 Disable generation of @code{bcnz} instructions.
21431 Enable generation of unaligned load and store instructions.
21435 Enable the use of multiply-accumulate instructions. Disabled by default.
21439 Specify the SCORE5 as the target architecture.
21443 Specify the SCORE5U of the target architecture.
21447 Specify the SCORE7 as the target architecture. This is the default.
21451 Specify the SCORE7D as the target architecture.
21455 @subsection SH Options
21457 These @samp{-m} options are defined for the SH implementations:
21462 Generate code for the SH1.
21466 Generate code for the SH2.
21469 Generate code for the SH2e.
21473 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21474 that the floating-point unit is not used.
21476 @item -m2a-single-only
21477 @opindex m2a-single-only
21478 Generate code for the SH2a-FPU, in such a way that no double-precision
21479 floating-point operations are used.
21482 @opindex m2a-single
21483 Generate code for the SH2a-FPU assuming the floating-point unit is in
21484 single-precision mode by default.
21488 Generate code for the SH2a-FPU assuming the floating-point unit is in
21489 double-precision mode by default.
21493 Generate code for the SH3.
21497 Generate code for the SH3e.
21501 Generate code for the SH4 without a floating-point unit.
21503 @item -m4-single-only
21504 @opindex m4-single-only
21505 Generate code for the SH4 with a floating-point unit that only
21506 supports single-precision arithmetic.
21510 Generate code for the SH4 assuming the floating-point unit is in
21511 single-precision mode by default.
21515 Generate code for the SH4.
21519 Generate code for SH4-100.
21521 @item -m4-100-nofpu
21522 @opindex m4-100-nofpu
21523 Generate code for SH4-100 in such a way that the
21524 floating-point unit is not used.
21526 @item -m4-100-single
21527 @opindex m4-100-single
21528 Generate code for SH4-100 assuming the floating-point unit is in
21529 single-precision mode by default.
21531 @item -m4-100-single-only
21532 @opindex m4-100-single-only
21533 Generate code for SH4-100 in such a way that no double-precision
21534 floating-point operations are used.
21538 Generate code for SH4-200.
21540 @item -m4-200-nofpu
21541 @opindex m4-200-nofpu
21542 Generate code for SH4-200 without in such a way that the
21543 floating-point unit is not used.
21545 @item -m4-200-single
21546 @opindex m4-200-single
21547 Generate code for SH4-200 assuming the floating-point unit is in
21548 single-precision mode by default.
21550 @item -m4-200-single-only
21551 @opindex m4-200-single-only
21552 Generate code for SH4-200 in such a way that no double-precision
21553 floating-point operations are used.
21557 Generate code for SH4-300.
21559 @item -m4-300-nofpu
21560 @opindex m4-300-nofpu
21561 Generate code for SH4-300 without in such a way that the
21562 floating-point unit is not used.
21564 @item -m4-300-single
21565 @opindex m4-300-single
21566 Generate code for SH4-300 in such a way that no double-precision
21567 floating-point operations are used.
21569 @item -m4-300-single-only
21570 @opindex m4-300-single-only
21571 Generate code for SH4-300 in such a way that no double-precision
21572 floating-point operations are used.
21576 Generate code for SH4-340 (no MMU, no FPU).
21580 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21585 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21586 floating-point unit is not used.
21588 @item -m4a-single-only
21589 @opindex m4a-single-only
21590 Generate code for the SH4a, in such a way that no double-precision
21591 floating-point operations are used.
21594 @opindex m4a-single
21595 Generate code for the SH4a assuming the floating-point unit is in
21596 single-precision mode by default.
21600 Generate code for the SH4a.
21604 Same as @option{-m4a-nofpu}, except that it implicitly passes
21605 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21606 instructions at the moment.
21610 Compile code for the processor in big-endian mode.
21614 Compile code for the processor in little-endian mode.
21618 Align doubles at 64-bit boundaries. Note that this changes the calling
21619 conventions, and thus some functions from the standard C library do
21620 not work unless you recompile it first with @option{-mdalign}.
21624 Shorten some address references at link time, when possible; uses the
21625 linker option @option{-relax}.
21629 Use 32-bit offsets in @code{switch} tables. The default is to use
21634 Enable the use of bit manipulation instructions on SH2A.
21638 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21639 alignment constraints.
21643 Comply with the calling conventions defined by Renesas.
21646 @opindex mno-renesas
21647 Comply with the calling conventions defined for GCC before the Renesas
21648 conventions were available. This option is the default for all
21649 targets of the SH toolchain.
21652 @opindex mnomacsave
21653 Mark the @code{MAC} register as call-clobbered, even if
21654 @option{-mrenesas} is given.
21660 Control the IEEE compliance of floating-point comparisons, which affects the
21661 handling of cases where the result of a comparison is unordered. By default
21662 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21663 enabled @option{-mno-ieee} is implicitly set, which results in faster
21664 floating-point greater-equal and less-equal comparisons. The implicit settings
21665 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21667 @item -minline-ic_invalidate
21668 @opindex minline-ic_invalidate
21669 Inline code to invalidate instruction cache entries after setting up
21670 nested function trampolines.
21671 This option has no effect if @option{-musermode} is in effect and the selected
21672 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21674 If the selected code generation option does not allow the use of the @code{icbi}
21675 instruction, and @option{-musermode} is not in effect, the inlined code
21676 manipulates the instruction cache address array directly with an associative
21677 write. This not only requires privileged mode at run time, but it also
21678 fails if the cache line had been mapped via the TLB and has become unmapped.
21682 Dump instruction size and location in the assembly code.
21685 @opindex mpadstruct
21686 This option is deprecated. It pads structures to multiple of 4 bytes,
21687 which is incompatible with the SH ABI@.
21689 @item -matomic-model=@var{model}
21690 @opindex matomic-model=@var{model}
21691 Sets the model of atomic operations and additional parameters as a comma
21692 separated list. For details on the atomic built-in functions see
21693 @ref{__atomic Builtins}. The following models and parameters are supported:
21698 Disable compiler generated atomic sequences and emit library calls for atomic
21699 operations. This is the default if the target is not @code{sh*-*-linux*}.
21702 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21703 built-in functions. The generated atomic sequences require additional support
21704 from the interrupt/exception handling code of the system and are only suitable
21705 for SH3* and SH4* single-core systems. This option is enabled by default when
21706 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21707 this option also partially utilizes the hardware atomic instructions
21708 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21709 @samp{strict} is specified.
21712 Generate software atomic sequences that use a variable in the thread control
21713 block. This is a variation of the gUSA sequences which can also be used on
21714 SH1* and SH2* targets. The generated atomic sequences require additional
21715 support from the interrupt/exception handling code of the system and are only
21716 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21717 parameter has to be specified as well.
21720 Generate software atomic sequences that temporarily disable interrupts by
21721 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21722 in privileged mode and is only suitable for single-core systems. Additional
21723 support from the interrupt/exception handling code of the system is not
21724 required. This model is enabled by default when the target is
21725 @code{sh*-*-linux*} and SH1* or SH2*.
21728 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21729 instructions only. This is only available on SH4A and is suitable for
21730 multi-core systems. Since the hardware instructions support only 32 bit atomic
21731 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21732 Code compiled with this option is also compatible with other software
21733 atomic model interrupt/exception handling systems if executed on an SH4A
21734 system. Additional support from the interrupt/exception handling code of the
21735 system is not required for this model.
21738 This parameter specifies the offset in bytes of the variable in the thread
21739 control block structure that should be used by the generated atomic sequences
21740 when the @samp{soft-tcb} model has been selected. For other models this
21741 parameter is ignored. The specified value must be an integer multiple of four
21742 and in the range 0-1020.
21745 This parameter prevents mixed usage of multiple atomic models, even if they
21746 are compatible, and makes the compiler generate atomic sequences of the
21747 specified model only.
21753 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21754 Notice that depending on the particular hardware and software configuration
21755 this can degrade overall performance due to the operand cache line flushes
21756 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21757 processors the @code{tas.b} instruction must be used with caution since it
21758 can result in data corruption for certain cache configurations.
21761 @opindex mprefergot
21762 When generating position-independent code, emit function calls using
21763 the Global Offset Table instead of the Procedure Linkage Table.
21766 @itemx -mno-usermode
21768 @opindex mno-usermode
21769 Don't allow (allow) the compiler generating privileged mode code. Specifying
21770 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21771 inlined code would not work in user mode. @option{-musermode} is the default
21772 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21773 @option{-musermode} has no effect, since there is no user mode.
21775 @item -multcost=@var{number}
21776 @opindex multcost=@var{number}
21777 Set the cost to assume for a multiply insn.
21779 @item -mdiv=@var{strategy}
21780 @opindex mdiv=@var{strategy}
21781 Set the division strategy to be used for integer division operations.
21782 @var{strategy} can be one of:
21787 Calls a library function that uses the single-step division instruction
21788 @code{div1} to perform the operation. Division by zero calculates an
21789 unspecified result and does not trap. This is the default except for SH4,
21790 SH2A and SHcompact.
21793 Calls a library function that performs the operation in double precision
21794 floating point. Division by zero causes a floating-point exception. This is
21795 the default for SHcompact with FPU. Specifying this for targets that do not
21796 have a double precision FPU defaults to @code{call-div1}.
21799 Calls a library function that uses a lookup table for small divisors and
21800 the @code{div1} instruction with case distinction for larger divisors. Division
21801 by zero calculates an unspecified result and does not trap. This is the default
21802 for SH4. Specifying this for targets that do not have dynamic shift
21803 instructions defaults to @code{call-div1}.
21807 When a division strategy has not been specified the default strategy is
21808 selected based on the current target. For SH2A the default strategy is to
21809 use the @code{divs} and @code{divu} instructions instead of library function
21812 @item -maccumulate-outgoing-args
21813 @opindex maccumulate-outgoing-args
21814 Reserve space once for outgoing arguments in the function prologue rather
21815 than around each call. Generally beneficial for performance and size. Also
21816 needed for unwinding to avoid changing the stack frame around conditional code.
21818 @item -mdivsi3_libfunc=@var{name}
21819 @opindex mdivsi3_libfunc=@var{name}
21820 Set the name of the library function used for 32-bit signed division to
21822 This only affects the name used in the @samp{call} division strategies, and
21823 the compiler still expects the same sets of input/output/clobbered registers as
21824 if this option were not present.
21826 @item -mfixed-range=@var{register-range}
21827 @opindex mfixed-range
21828 Generate code treating the given register range as fixed registers.
21829 A fixed register is one that the register allocator can not use. This is
21830 useful when compiling kernel code. A register range is specified as
21831 two registers separated by a dash. Multiple register ranges can be
21832 specified separated by a comma.
21834 @item -mbranch-cost=@var{num}
21835 @opindex mbranch-cost=@var{num}
21836 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21837 make the compiler try to generate more branch-free code if possible.
21838 If not specified the value is selected depending on the processor type that
21839 is being compiled for.
21842 @itemx -mno-zdcbranch
21843 @opindex mzdcbranch
21844 @opindex mno-zdcbranch
21845 Assume (do not assume) that zero displacement conditional branch instructions
21846 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21847 compiler prefers zero displacement branch code sequences. This is
21848 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21849 disabled by specifying @option{-mno-zdcbranch}.
21851 @item -mcbranch-force-delay-slot
21852 @opindex mcbranch-force-delay-slot
21853 Force the usage of delay slots for conditional branches, which stuffs the delay
21854 slot with a @code{nop} if a suitable instruction can't be found. By default
21855 this option is disabled. It can be enabled to work around hardware bugs as
21856 found in the original SH7055.
21859 @itemx -mno-fused-madd
21860 @opindex mfused-madd
21861 @opindex mno-fused-madd
21862 Generate code that uses (does not use) the floating-point multiply and
21863 accumulate instructions. These instructions are generated by default
21864 if hardware floating point is used. The machine-dependent
21865 @option{-mfused-madd} option is now mapped to the machine-independent
21866 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21867 mapped to @option{-ffp-contract=off}.
21873 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21874 and cosine approximations. The option @option{-mfsca} must be used in
21875 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21876 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21877 approximations even if @option{-funsafe-math-optimizations} is in effect.
21883 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21884 reciprocal square root approximations. The option @option{-mfsrra} must be used
21885 in combination with @option{-funsafe-math-optimizations} and
21886 @option{-ffinite-math-only}. It is enabled by default when generating code for
21887 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21888 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21891 @item -mpretend-cmove
21892 @opindex mpretend-cmove
21893 Prefer zero-displacement conditional branches for conditional move instruction
21894 patterns. This can result in faster code on the SH4 processor.
21898 Generate code using the FDPIC ABI.
21902 @node Solaris 2 Options
21903 @subsection Solaris 2 Options
21904 @cindex Solaris 2 options
21906 These @samp{-m} options are supported on Solaris 2:
21909 @item -mclear-hwcap
21910 @opindex mclear-hwcap
21911 @option{-mclear-hwcap} tells the compiler to remove the hardware
21912 capabilities generated by the Solaris assembler. This is only necessary
21913 when object files use ISA extensions not supported by the current
21914 machine, but check at runtime whether or not to use them.
21916 @item -mimpure-text
21917 @opindex mimpure-text
21918 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21919 the compiler to not pass @option{-z text} to the linker when linking a
21920 shared object. Using this option, you can link position-dependent
21921 code into a shared object.
21923 @option{-mimpure-text} suppresses the ``relocations remain against
21924 allocatable but non-writable sections'' linker error message.
21925 However, the necessary relocations trigger copy-on-write, and the
21926 shared object is not actually shared across processes. Instead of
21927 using @option{-mimpure-text}, you should compile all source code with
21928 @option{-fpic} or @option{-fPIC}.
21932 These switches are supported in addition to the above on Solaris 2:
21937 Add support for multithreading using the POSIX threads library. This
21938 option sets flags for both the preprocessor and linker. This option does
21939 not affect the thread safety of object code produced by the compiler or
21940 that of libraries supplied with it.
21944 This is a synonym for @option{-pthreads}.
21947 @node SPARC Options
21948 @subsection SPARC Options
21949 @cindex SPARC options
21951 These @samp{-m} options are supported on the SPARC:
21954 @item -mno-app-regs
21956 @opindex mno-app-regs
21958 Specify @option{-mapp-regs} to generate output using the global registers
21959 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21960 global register 1, each global register 2 through 4 is then treated as an
21961 allocable register that is clobbered by function calls. This is the default.
21963 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21964 specify @option{-mno-app-regs}. You should compile libraries and system
21965 software with this option.
21971 With @option{-mflat}, the compiler does not generate save/restore instructions
21972 and uses a ``flat'' or single register window model. This model is compatible
21973 with the regular register window model. The local registers and the input
21974 registers (0--5) are still treated as ``call-saved'' registers and are
21975 saved on the stack as needed.
21977 With @option{-mno-flat} (the default), the compiler generates save/restore
21978 instructions (except for leaf functions). This is the normal operating mode.
21981 @itemx -mhard-float
21983 @opindex mhard-float
21984 Generate output containing floating-point instructions. This is the
21988 @itemx -msoft-float
21990 @opindex msoft-float
21991 Generate output containing library calls for floating point.
21992 @strong{Warning:} the requisite libraries are not available for all SPARC
21993 targets. Normally the facilities of the machine's usual C compiler are
21994 used, but this cannot be done directly in cross-compilation. You must make
21995 your own arrangements to provide suitable library functions for
21996 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21997 @samp{sparclite-*-*} do provide software floating-point support.
21999 @option{-msoft-float} changes the calling convention in the output file;
22000 therefore, it is only useful if you compile @emph{all} of a program with
22001 this option. In particular, you need to compile @file{libgcc.a}, the
22002 library that comes with GCC, with @option{-msoft-float} in order for
22005 @item -mhard-quad-float
22006 @opindex mhard-quad-float
22007 Generate output containing quad-word (long double) floating-point
22010 @item -msoft-quad-float
22011 @opindex msoft-quad-float
22012 Generate output containing library calls for quad-word (long double)
22013 floating-point instructions. The functions called are those specified
22014 in the SPARC ABI@. This is the default.
22016 As of this writing, there are no SPARC implementations that have hardware
22017 support for the quad-word floating-point instructions. They all invoke
22018 a trap handler for one of these instructions, and then the trap handler
22019 emulates the effect of the instruction. Because of the trap handler overhead,
22020 this is much slower than calling the ABI library routines. Thus the
22021 @option{-msoft-quad-float} option is the default.
22023 @item -mno-unaligned-doubles
22024 @itemx -munaligned-doubles
22025 @opindex mno-unaligned-doubles
22026 @opindex munaligned-doubles
22027 Assume that doubles have 8-byte alignment. This is the default.
22029 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
22030 alignment only if they are contained in another type, or if they have an
22031 absolute address. Otherwise, it assumes they have 4-byte alignment.
22032 Specifying this option avoids some rare compatibility problems with code
22033 generated by other compilers. It is not the default because it results
22034 in a performance loss, especially for floating-point code.
22037 @itemx -mno-user-mode
22038 @opindex muser-mode
22039 @opindex mno-user-mode
22040 Do not generate code that can only run in supervisor mode. This is relevant
22041 only for the @code{casa} instruction emitted for the LEON3 processor. This
22044 @item -mfaster-structs
22045 @itemx -mno-faster-structs
22046 @opindex mfaster-structs
22047 @opindex mno-faster-structs
22048 With @option{-mfaster-structs}, the compiler assumes that structures
22049 should have 8-byte alignment. This enables the use of pairs of
22050 @code{ldd} and @code{std} instructions for copies in structure
22051 assignment, in place of twice as many @code{ld} and @code{st} pairs.
22052 However, the use of this changed alignment directly violates the SPARC
22053 ABI@. Thus, it's intended only for use on targets where the developer
22054 acknowledges that their resulting code is not directly in line with
22055 the rules of the ABI@.
22057 @item -mstd-struct-return
22058 @itemx -mno-std-struct-return
22059 @opindex mstd-struct-return
22060 @opindex mno-std-struct-return
22061 With @option{-mstd-struct-return}, the compiler generates checking code
22062 in functions returning structures or unions to detect size mismatches
22063 between the two sides of function calls, as per the 32-bit ABI@.
22065 The default is @option{-mno-std-struct-return}. This option has no effect
22068 @item -mcpu=@var{cpu_type}
22070 Set the instruction set, register set, and instruction scheduling parameters
22071 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22072 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
22073 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
22074 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
22075 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
22076 @samp{niagara3} and @samp{niagara4}.
22078 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
22079 which selects the best architecture option for the host processor.
22080 @option{-mcpu=native} has no effect if GCC does not recognize
22083 Default instruction scheduling parameters are used for values that select
22084 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
22085 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
22087 Here is a list of each supported architecture and their supported
22095 supersparc, hypersparc, leon, leon3
22098 f930, f934, sparclite86x
22104 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
22107 By default (unless configured otherwise), GCC generates code for the V7
22108 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
22109 additionally optimizes it for the Cypress CY7C602 chip, as used in the
22110 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
22111 SPARCStation 1, 2, IPX etc.
22113 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
22114 architecture. The only difference from V7 code is that the compiler emits
22115 the integer multiply and integer divide instructions which exist in SPARC-V8
22116 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
22117 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
22120 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
22121 the SPARC architecture. This adds the integer multiply, integer divide step
22122 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
22123 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
22124 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
22125 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
22126 MB86934 chip, which is the more recent SPARClite with FPU@.
22128 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
22129 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
22130 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
22131 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
22132 optimizes it for the TEMIC SPARClet chip.
22134 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
22135 architecture. This adds 64-bit integer and floating-point move instructions,
22136 3 additional floating-point condition code registers and conditional move
22137 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
22138 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
22139 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
22140 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
22141 @option{-mcpu=niagara}, the compiler additionally optimizes it for
22142 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
22143 additionally optimizes it for Sun UltraSPARC T2 chips. With
22144 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
22145 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
22146 additionally optimizes it for Sun UltraSPARC T4 chips.
22148 @item -mtune=@var{cpu_type}
22150 Set the instruction scheduling parameters for machine type
22151 @var{cpu_type}, but do not set the instruction set or register set that the
22152 option @option{-mcpu=@var{cpu_type}} does.
22154 The same values for @option{-mcpu=@var{cpu_type}} can be used for
22155 @option{-mtune=@var{cpu_type}}, but the only useful values are those
22156 that select a particular CPU implementation. Those are @samp{cypress},
22157 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
22158 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
22159 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
22160 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
22161 toolchains, @samp{native} can also be used.
22166 @opindex mno-v8plus
22167 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
22168 difference from the V8 ABI is that the global and out registers are
22169 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
22170 mode for all SPARC-V9 processors.
22176 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
22177 Visual Instruction Set extensions. The default is @option{-mno-vis}.
22183 With @option{-mvis2}, GCC generates code that takes advantage of
22184 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
22185 default is @option{-mvis2} when targeting a cpu that supports such
22186 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
22187 also sets @option{-mvis}.
22193 With @option{-mvis3}, GCC generates code that takes advantage of
22194 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
22195 default is @option{-mvis3} when targeting a cpu that supports such
22196 instructions, such as niagara-3 and later. Setting @option{-mvis3}
22197 also sets @option{-mvis2} and @option{-mvis}.
22202 @opindex mno-cbcond
22203 With @option{-mcbcond}, GCC generates code that takes advantage of
22204 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
22205 The default is @option{-mcbcond} when targeting a cpu that supports such
22206 instructions, such as niagara-4 and later.
22212 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
22213 population count instruction. The default is @option{-mpopc}
22214 when targeting a cpu that supports such instructions, such as Niagara-2 and
22221 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
22222 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
22223 when targeting a cpu that supports such instructions, such as Niagara-3 and
22227 @opindex mfix-at697f
22228 Enable the documented workaround for the single erratum of the Atmel AT697F
22229 processor (which corresponds to erratum #13 of the AT697E processor).
22232 @opindex mfix-ut699
22233 Enable the documented workarounds for the floating-point errata and the data
22234 cache nullify errata of the UT699 processor.
22237 These @samp{-m} options are supported in addition to the above
22238 on SPARC-V9 processors in 64-bit environments:
22245 Generate code for a 32-bit or 64-bit environment.
22246 The 32-bit environment sets int, long and pointer to 32 bits.
22247 The 64-bit environment sets int to 32 bits and long and pointer
22250 @item -mcmodel=@var{which}
22252 Set the code model to one of
22256 The Medium/Low code model: 64-bit addresses, programs
22257 must be linked in the low 32 bits of memory. Programs can be statically
22258 or dynamically linked.
22261 The Medium/Middle code model: 64-bit addresses, programs
22262 must be linked in the low 44 bits of memory, the text and data segments must
22263 be less than 2GB in size and the data segment must be located within 2GB of
22267 The Medium/Anywhere code model: 64-bit addresses, programs
22268 may be linked anywhere in memory, the text and data segments must be less
22269 than 2GB in size and the data segment must be located within 2GB of the
22273 The Medium/Anywhere code model for embedded systems:
22274 64-bit addresses, the text and data segments must be less than 2GB in
22275 size, both starting anywhere in memory (determined at link time). The
22276 global register %g4 points to the base of the data segment. Programs
22277 are statically linked and PIC is not supported.
22280 @item -mmemory-model=@var{mem-model}
22281 @opindex mmemory-model
22282 Set the memory model in force on the processor to one of
22286 The default memory model for the processor and operating system.
22289 Relaxed Memory Order
22292 Partial Store Order
22298 Sequential Consistency
22301 These memory models are formally defined in Appendix D of the Sparc V9
22302 architecture manual, as set in the processor's @code{PSTATE.MM} field.
22305 @itemx -mno-stack-bias
22306 @opindex mstack-bias
22307 @opindex mno-stack-bias
22308 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
22309 frame pointer if present, are offset by @minus{}2047 which must be added back
22310 when making stack frame references. This is the default in 64-bit mode.
22311 Otherwise, assume no such offset is present.
22315 @subsection SPU Options
22316 @cindex SPU options
22318 These @samp{-m} options are supported on the SPU:
22322 @itemx -merror-reloc
22323 @opindex mwarn-reloc
22324 @opindex merror-reloc
22326 The loader for SPU does not handle dynamic relocations. By default, GCC
22327 gives an error when it generates code that requires a dynamic
22328 relocation. @option{-mno-error-reloc} disables the error,
22329 @option{-mwarn-reloc} generates a warning instead.
22332 @itemx -munsafe-dma
22334 @opindex munsafe-dma
22336 Instructions that initiate or test completion of DMA must not be
22337 reordered with respect to loads and stores of the memory that is being
22339 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
22340 memory accesses, but that can lead to inefficient code in places where the
22341 memory is known to not change. Rather than mark the memory as volatile,
22342 you can use @option{-msafe-dma} to tell the compiler to treat
22343 the DMA instructions as potentially affecting all memory.
22345 @item -mbranch-hints
22346 @opindex mbranch-hints
22348 By default, GCC generates a branch hint instruction to avoid
22349 pipeline stalls for always-taken or probably-taken branches. A hint
22350 is not generated closer than 8 instructions away from its branch.
22351 There is little reason to disable them, except for debugging purposes,
22352 or to make an object a little bit smaller.
22356 @opindex msmall-mem
22357 @opindex mlarge-mem
22359 By default, GCC generates code assuming that addresses are never larger
22360 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
22361 a full 32-bit address.
22366 By default, GCC links against startup code that assumes the SPU-style
22367 main function interface (which has an unconventional parameter list).
22368 With @option{-mstdmain}, GCC links your program against startup
22369 code that assumes a C99-style interface to @code{main}, including a
22370 local copy of @code{argv} strings.
22372 @item -mfixed-range=@var{register-range}
22373 @opindex mfixed-range
22374 Generate code treating the given register range as fixed registers.
22375 A fixed register is one that the register allocator cannot use. This is
22376 useful when compiling kernel code. A register range is specified as
22377 two registers separated by a dash. Multiple register ranges can be
22378 specified separated by a comma.
22384 Compile code assuming that pointers to the PPU address space accessed
22385 via the @code{__ea} named address space qualifier are either 32 or 64
22386 bits wide. The default is 32 bits. As this is an ABI-changing option,
22387 all object code in an executable must be compiled with the same setting.
22389 @item -maddress-space-conversion
22390 @itemx -mno-address-space-conversion
22391 @opindex maddress-space-conversion
22392 @opindex mno-address-space-conversion
22393 Allow/disallow treating the @code{__ea} address space as superset
22394 of the generic address space. This enables explicit type casts
22395 between @code{__ea} and generic pointer as well as implicit
22396 conversions of generic pointers to @code{__ea} pointers. The
22397 default is to allow address space pointer conversions.
22399 @item -mcache-size=@var{cache-size}
22400 @opindex mcache-size
22401 This option controls the version of libgcc that the compiler links to an
22402 executable and selects a software-managed cache for accessing variables
22403 in the @code{__ea} address space with a particular cache size. Possible
22404 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
22405 and @samp{128}. The default cache size is 64KB.
22407 @item -matomic-updates
22408 @itemx -mno-atomic-updates
22409 @opindex matomic-updates
22410 @opindex mno-atomic-updates
22411 This option controls the version of libgcc that the compiler links to an
22412 executable and selects whether atomic updates to the software-managed
22413 cache of PPU-side variables are used. If you use atomic updates, changes
22414 to a PPU variable from SPU code using the @code{__ea} named address space
22415 qualifier do not interfere with changes to other PPU variables residing
22416 in the same cache line from PPU code. If you do not use atomic updates,
22417 such interference may occur; however, writing back cache lines is
22418 more efficient. The default behavior is to use atomic updates.
22421 @itemx -mdual-nops=@var{n}
22422 @opindex mdual-nops
22423 By default, GCC inserts nops to increase dual issue when it expects
22424 it to increase performance. @var{n} can be a value from 0 to 10. A
22425 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22426 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22428 @item -mhint-max-nops=@var{n}
22429 @opindex mhint-max-nops
22430 Maximum number of nops to insert for a branch hint. A branch hint must
22431 be at least 8 instructions away from the branch it is affecting. GCC
22432 inserts up to @var{n} nops to enforce this, otherwise it does not
22433 generate the branch hint.
22435 @item -mhint-max-distance=@var{n}
22436 @opindex mhint-max-distance
22437 The encoding of the branch hint instruction limits the hint to be within
22438 256 instructions of the branch it is affecting. By default, GCC makes
22439 sure it is within 125.
22442 @opindex msafe-hints
22443 Work around a hardware bug that causes the SPU to stall indefinitely.
22444 By default, GCC inserts the @code{hbrp} instruction to make sure
22445 this stall won't happen.
22449 @node System V Options
22450 @subsection Options for System V
22452 These additional options are available on System V Release 4 for
22453 compatibility with other compilers on those systems:
22458 Create a shared object.
22459 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22463 Identify the versions of each tool used by the compiler, in a
22464 @code{.ident} assembler directive in the output.
22468 Refrain from adding @code{.ident} directives to the output file (this is
22471 @item -YP,@var{dirs}
22473 Search the directories @var{dirs}, and no others, for libraries
22474 specified with @option{-l}.
22476 @item -Ym,@var{dir}
22478 Look in the directory @var{dir} to find the M4 preprocessor.
22479 The assembler uses this option.
22480 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22481 @c the generic assembler that comes with Solaris takes just -Ym.
22484 @node TILE-Gx Options
22485 @subsection TILE-Gx Options
22486 @cindex TILE-Gx options
22488 These @samp{-m} options are supported on the TILE-Gx:
22491 @item -mcmodel=small
22492 @opindex mcmodel=small
22493 Generate code for the small model. The distance for direct calls is
22494 limited to 500M in either direction. PC-relative addresses are 32
22495 bits. Absolute addresses support the full address range.
22497 @item -mcmodel=large
22498 @opindex mcmodel=large
22499 Generate code for the large model. There is no limitation on call
22500 distance, pc-relative addresses, or absolute addresses.
22502 @item -mcpu=@var{name}
22504 Selects the type of CPU to be targeted. Currently the only supported
22505 type is @samp{tilegx}.
22511 Generate code for a 32-bit or 64-bit environment. The 32-bit
22512 environment sets int, long, and pointer to 32 bits. The 64-bit
22513 environment sets int to 32 bits and long and pointer to 64 bits.
22516 @itemx -mlittle-endian
22517 @opindex mbig-endian
22518 @opindex mlittle-endian
22519 Generate code in big/little endian mode, respectively.
22522 @node TILEPro Options
22523 @subsection TILEPro Options
22524 @cindex TILEPro options
22526 These @samp{-m} options are supported on the TILEPro:
22529 @item -mcpu=@var{name}
22531 Selects the type of CPU to be targeted. Currently the only supported
22532 type is @samp{tilepro}.
22536 Generate code for a 32-bit environment, which sets int, long, and
22537 pointer to 32 bits. This is the only supported behavior so the flag
22538 is essentially ignored.
22542 @subsection V850 Options
22543 @cindex V850 Options
22545 These @samp{-m} options are defined for V850 implementations:
22549 @itemx -mno-long-calls
22550 @opindex mlong-calls
22551 @opindex mno-long-calls
22552 Treat all calls as being far away (near). If calls are assumed to be
22553 far away, the compiler always loads the function's address into a
22554 register, and calls indirect through the pointer.
22560 Do not optimize (do optimize) basic blocks that use the same index
22561 pointer 4 or more times to copy pointer into the @code{ep} register, and
22562 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22563 option is on by default if you optimize.
22565 @item -mno-prolog-function
22566 @itemx -mprolog-function
22567 @opindex mno-prolog-function
22568 @opindex mprolog-function
22569 Do not use (do use) external functions to save and restore registers
22570 at the prologue and epilogue of a function. The external functions
22571 are slower, but use less code space if more than one function saves
22572 the same number of registers. The @option{-mprolog-function} option
22573 is on by default if you optimize.
22577 Try to make the code as small as possible. At present, this just turns
22578 on the @option{-mep} and @option{-mprolog-function} options.
22580 @item -mtda=@var{n}
22582 Put static or global variables whose size is @var{n} bytes or less into
22583 the tiny data area that register @code{ep} points to. The tiny data
22584 area can hold up to 256 bytes in total (128 bytes for byte references).
22586 @item -msda=@var{n}
22588 Put static or global variables whose size is @var{n} bytes or less into
22589 the small data area that register @code{gp} points to. The small data
22590 area can hold up to 64 kilobytes.
22592 @item -mzda=@var{n}
22594 Put static or global variables whose size is @var{n} bytes or less into
22595 the first 32 kilobytes of memory.
22599 Specify that the target processor is the V850.
22603 Specify that the target processor is the V850E3V5. The preprocessor
22604 constant @code{__v850e3v5__} is defined if this option is used.
22608 Specify that the target processor is the V850E3V5. This is an alias for
22609 the @option{-mv850e3v5} option.
22613 Specify that the target processor is the V850E2V3. The preprocessor
22614 constant @code{__v850e2v3__} is defined if this option is used.
22618 Specify that the target processor is the V850E2. The preprocessor
22619 constant @code{__v850e2__} is defined if this option is used.
22623 Specify that the target processor is the V850E1. The preprocessor
22624 constants @code{__v850e1__} and @code{__v850e__} are defined if
22625 this option is used.
22629 Specify that the target processor is the V850ES. This is an alias for
22630 the @option{-mv850e1} option.
22634 Specify that the target processor is the V850E@. The preprocessor
22635 constant @code{__v850e__} is defined if this option is used.
22637 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22638 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22639 are defined then a default target processor is chosen and the
22640 relevant @samp{__v850*__} preprocessor constant is defined.
22642 The preprocessor constants @code{__v850} and @code{__v851__} are always
22643 defined, regardless of which processor variant is the target.
22645 @item -mdisable-callt
22646 @itemx -mno-disable-callt
22647 @opindex mdisable-callt
22648 @opindex mno-disable-callt
22649 This option suppresses generation of the @code{CALLT} instruction for the
22650 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22653 This option is enabled by default when the RH850 ABI is
22654 in use (see @option{-mrh850-abi}), and disabled by default when the
22655 GCC ABI is in use. If @code{CALLT} instructions are being generated
22656 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22662 Pass on (or do not pass on) the @option{-mrelax} command-line option
22666 @itemx -mno-long-jumps
22667 @opindex mlong-jumps
22668 @opindex mno-long-jumps
22669 Disable (or re-enable) the generation of PC-relative jump instructions.
22672 @itemx -mhard-float
22673 @opindex msoft-float
22674 @opindex mhard-float
22675 Disable (or re-enable) the generation of hardware floating point
22676 instructions. This option is only significant when the target
22677 architecture is @samp{V850E2V3} or higher. If hardware floating point
22678 instructions are being generated then the C preprocessor symbol
22679 @code{__FPU_OK__} is defined, otherwise the symbol
22680 @code{__NO_FPU__} is defined.
22684 Enables the use of the e3v5 LOOP instruction. The use of this
22685 instruction is not enabled by default when the e3v5 architecture is
22686 selected because its use is still experimental.
22690 @opindex mrh850-abi
22692 Enables support for the RH850 version of the V850 ABI. This is the
22693 default. With this version of the ABI the following rules apply:
22697 Integer sized structures and unions are returned via a memory pointer
22698 rather than a register.
22701 Large structures and unions (more than 8 bytes in size) are passed by
22705 Functions are aligned to 16-bit boundaries.
22708 The @option{-m8byte-align} command-line option is supported.
22711 The @option{-mdisable-callt} command-line option is enabled by
22712 default. The @option{-mno-disable-callt} command-line option is not
22716 When this version of the ABI is enabled the C preprocessor symbol
22717 @code{__V850_RH850_ABI__} is defined.
22721 Enables support for the old GCC version of the V850 ABI. With this
22722 version of the ABI the following rules apply:
22726 Integer sized structures and unions are returned in register @code{r10}.
22729 Large structures and unions (more than 8 bytes in size) are passed by
22733 Functions are aligned to 32-bit boundaries, unless optimizing for
22737 The @option{-m8byte-align} command-line option is not supported.
22740 The @option{-mdisable-callt} command-line option is supported but not
22741 enabled by default.
22744 When this version of the ABI is enabled the C preprocessor symbol
22745 @code{__V850_GCC_ABI__} is defined.
22747 @item -m8byte-align
22748 @itemx -mno-8byte-align
22749 @opindex m8byte-align
22750 @opindex mno-8byte-align
22751 Enables support for @code{double} and @code{long long} types to be
22752 aligned on 8-byte boundaries. The default is to restrict the
22753 alignment of all objects to at most 4-bytes. When
22754 @option{-m8byte-align} is in effect the C preprocessor symbol
22755 @code{__V850_8BYTE_ALIGN__} is defined.
22758 @opindex mbig-switch
22759 Generate code suitable for big switch tables. Use this option only if
22760 the assembler/linker complain about out of range branches within a switch
22765 This option causes r2 and r5 to be used in the code generated by
22766 the compiler. This setting is the default.
22768 @item -mno-app-regs
22769 @opindex mno-app-regs
22770 This option causes r2 and r5 to be treated as fixed registers.
22775 @subsection VAX Options
22776 @cindex VAX options
22778 These @samp{-m} options are defined for the VAX:
22783 Do not output certain jump instructions (@code{aobleq} and so on)
22784 that the Unix assembler for the VAX cannot handle across long
22789 Do output those jump instructions, on the assumption that the
22790 GNU assembler is being used.
22794 Output code for G-format floating-point numbers instead of D-format.
22797 @node Visium Options
22798 @subsection Visium Options
22799 @cindex Visium options
22805 A program which performs file I/O and is destined to run on an MCM target
22806 should be linked with this option. It causes the libraries libc.a and
22807 libdebug.a to be linked. The program should be run on the target under
22808 the control of the GDB remote debugging stub.
22812 A program which performs file I/O and is destined to run on the simulator
22813 should be linked with option. This causes libraries libc.a and libsim.a to
22817 @itemx -mhard-float
22819 @opindex mhard-float
22820 Generate code containing floating-point instructions. This is the
22824 @itemx -msoft-float
22826 @opindex msoft-float
22827 Generate code containing library calls for floating-point.
22829 @option{-msoft-float} changes the calling convention in the output file;
22830 therefore, it is only useful if you compile @emph{all} of a program with
22831 this option. In particular, you need to compile @file{libgcc.a}, the
22832 library that comes with GCC, with @option{-msoft-float} in order for
22835 @item -mcpu=@var{cpu_type}
22837 Set the instruction set, register set, and instruction scheduling parameters
22838 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22839 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22841 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22843 By default (unless configured otherwise), GCC generates code for the GR5
22844 variant of the Visium architecture.
22846 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22847 architecture. The only difference from GR5 code is that the compiler will
22848 generate block move instructions.
22850 @item -mtune=@var{cpu_type}
22852 Set the instruction scheduling parameters for machine type @var{cpu_type},
22853 but do not set the instruction set or register set that the option
22854 @option{-mcpu=@var{cpu_type}} would.
22858 Generate code for the supervisor mode, where there are no restrictions on
22859 the access to general registers. This is the default.
22862 @opindex muser-mode
22863 Generate code for the user mode, where the access to some general registers
22864 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22865 mode; on the GR6, only registers r29 to r31 are affected.
22869 @subsection VMS Options
22871 These @samp{-m} options are defined for the VMS implementations:
22874 @item -mvms-return-codes
22875 @opindex mvms-return-codes
22876 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22877 condition (e.g.@ error) codes.
22879 @item -mdebug-main=@var{prefix}
22880 @opindex mdebug-main=@var{prefix}
22881 Flag the first routine whose name starts with @var{prefix} as the main
22882 routine for the debugger.
22886 Default to 64-bit memory allocation routines.
22888 @item -mpointer-size=@var{size}
22889 @opindex mpointer-size=@var{size}
22890 Set the default size of pointers. Possible options for @var{size} are
22891 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22892 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22893 The later option disables @code{pragma pointer_size}.
22896 @node VxWorks Options
22897 @subsection VxWorks Options
22898 @cindex VxWorks Options
22900 The options in this section are defined for all VxWorks targets.
22901 Options specific to the target hardware are listed with the other
22902 options for that target.
22907 GCC can generate code for both VxWorks kernels and real time processes
22908 (RTPs). This option switches from the former to the latter. It also
22909 defines the preprocessor macro @code{__RTP__}.
22912 @opindex non-static
22913 Link an RTP executable against shared libraries rather than static
22914 libraries. The options @option{-static} and @option{-shared} can
22915 also be used for RTPs (@pxref{Link Options}); @option{-static}
22922 These options are passed down to the linker. They are defined for
22923 compatibility with Diab.
22926 @opindex Xbind-lazy
22927 Enable lazy binding of function calls. This option is equivalent to
22928 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22932 Disable lazy binding of function calls. This option is the default and
22933 is defined for compatibility with Diab.
22937 @subsection x86 Options
22938 @cindex x86 Options
22940 These @samp{-m} options are defined for the x86 family of computers.
22944 @item -march=@var{cpu-type}
22946 Generate instructions for the machine type @var{cpu-type}. In contrast to
22947 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22948 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22949 to generate code that may not run at all on processors other than the one
22950 indicated. Specifying @option{-march=@var{cpu-type}} implies
22951 @option{-mtune=@var{cpu-type}}.
22953 The choices for @var{cpu-type} are:
22957 This selects the CPU to generate code for at compilation time by determining
22958 the processor type of the compiling machine. Using @option{-march=native}
22959 enables all instruction subsets supported by the local machine (hence
22960 the result might not run on different machines). Using @option{-mtune=native}
22961 produces code optimized for the local machine under the constraints
22962 of the selected instruction set.
22965 Original Intel i386 CPU@.
22968 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22972 Intel Pentium CPU with no MMX support.
22975 Intel Lakemont MCU, based on Intel Pentium CPU.
22978 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22981 Intel Pentium Pro CPU@.
22984 When used with @option{-march}, the Pentium Pro
22985 instruction set is used, so the code runs on all i686 family chips.
22986 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22989 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22994 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22998 Intel Pentium M; low-power version of Intel Pentium III CPU
22999 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
23003 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
23006 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
23010 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
23011 SSE2 and SSE3 instruction set support.
23014 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
23015 instruction set support.
23018 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23019 SSE4.1, SSE4.2 and POPCNT instruction set support.
23022 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23023 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
23026 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23027 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
23030 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23031 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
23032 instruction set support.
23035 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23036 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23037 BMI, BMI2 and F16C instruction set support.
23040 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23041 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23042 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
23045 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23046 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23047 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
23048 XSAVES instruction set support.
23051 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
23052 instruction set support.
23055 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23056 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
23059 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23060 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23061 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
23062 AVX512CD instruction set support.
23064 @item skylake-avx512
23065 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23066 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23067 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
23068 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
23071 AMD K6 CPU with MMX instruction set support.
23075 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
23078 @itemx athlon-tbird
23079 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
23085 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
23086 instruction set support.
23092 Processors based on the AMD K8 core with x86-64 instruction set support,
23093 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
23094 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
23095 instruction set extensions.)
23098 @itemx opteron-sse3
23099 @itemx athlon64-sse3
23100 Improved versions of AMD K8 cores with SSE3 instruction set support.
23104 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
23105 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
23106 instruction set extensions.)
23109 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
23110 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
23111 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
23113 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23114 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
23115 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
23118 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23119 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
23120 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
23121 64-bit instruction set extensions.
23123 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23124 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
23125 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
23126 SSE4.2, ABM and 64-bit instruction set extensions.
23129 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
23130 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
23131 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
23132 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
23133 instruction set extensions.
23136 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
23137 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
23138 instruction set extensions.)
23141 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
23142 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
23143 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
23146 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
23150 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
23151 instruction set support.
23154 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
23155 implemented for this chip.)
23158 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
23160 implemented for this chip.)
23163 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
23166 @item -mtune=@var{cpu-type}
23168 Tune to @var{cpu-type} everything applicable about the generated code, except
23169 for the ABI and the set of available instructions.
23170 While picking a specific @var{cpu-type} schedules things appropriately
23171 for that particular chip, the compiler does not generate any code that
23172 cannot run on the default machine type unless you use a
23173 @option{-march=@var{cpu-type}} option.
23174 For example, if GCC is configured for i686-pc-linux-gnu
23175 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
23176 but still runs on i686 machines.
23178 The choices for @var{cpu-type} are the same as for @option{-march}.
23179 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
23183 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
23184 If you know the CPU on which your code will run, then you should use
23185 the corresponding @option{-mtune} or @option{-march} option instead of
23186 @option{-mtune=generic}. But, if you do not know exactly what CPU users
23187 of your application will have, then you should use this option.
23189 As new processors are deployed in the marketplace, the behavior of this
23190 option will change. Therefore, if you upgrade to a newer version of
23191 GCC, code generation controlled by this option will change to reflect
23193 that are most common at the time that version of GCC is released.
23195 There is no @option{-march=generic} option because @option{-march}
23196 indicates the instruction set the compiler can use, and there is no
23197 generic instruction set applicable to all processors. In contrast,
23198 @option{-mtune} indicates the processor (or, in this case, collection of
23199 processors) for which the code is optimized.
23202 Produce code optimized for the most current Intel processors, which are
23203 Haswell and Silvermont for this version of GCC. If you know the CPU
23204 on which your code will run, then you should use the corresponding
23205 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
23206 But, if you want your application performs better on both Haswell and
23207 Silvermont, then you should use this option.
23209 As new Intel processors are deployed in the marketplace, the behavior of
23210 this option will change. Therefore, if you upgrade to a newer version of
23211 GCC, code generation controlled by this option will change to reflect
23212 the most current Intel processors at the time that version of GCC is
23215 There is no @option{-march=intel} option because @option{-march} indicates
23216 the instruction set the compiler can use, and there is no common
23217 instruction set applicable to all processors. In contrast,
23218 @option{-mtune} indicates the processor (or, in this case, collection of
23219 processors) for which the code is optimized.
23222 @item -mcpu=@var{cpu-type}
23224 A deprecated synonym for @option{-mtune}.
23226 @item -mfpmath=@var{unit}
23228 Generate floating-point arithmetic for selected unit @var{unit}. The choices
23229 for @var{unit} are:
23233 Use the standard 387 floating-point coprocessor present on the majority of chips and
23234 emulated otherwise. Code compiled with this option runs almost everywhere.
23235 The temporary results are computed in 80-bit precision instead of the precision
23236 specified by the type, resulting in slightly different results compared to most
23237 of other chips. See @option{-ffloat-store} for more detailed description.
23239 This is the default choice for x86-32 targets.
23242 Use scalar floating-point instructions present in the SSE instruction set.
23243 This instruction set is supported by Pentium III and newer chips,
23244 and in the AMD line
23245 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
23246 instruction set supports only single-precision arithmetic, thus the double and
23247 extended-precision arithmetic are still done using 387. A later version, present
23248 only in Pentium 4 and AMD x86-64 chips, supports double-precision
23251 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
23252 or @option{-msse2} switches to enable SSE extensions and make this option
23253 effective. For the x86-64 compiler, these extensions are enabled by default.
23255 The resulting code should be considerably faster in the majority of cases and avoid
23256 the numerical instability problems of 387 code, but may break some existing
23257 code that expects temporaries to be 80 bits.
23259 This is the default choice for the x86-64 compiler.
23264 Attempt to utilize both instruction sets at once. This effectively doubles the
23265 amount of available registers, and on chips with separate execution units for
23266 387 and SSE the execution resources too. Use this option with care, as it is
23267 still experimental, because the GCC register allocator does not model separate
23268 functional units well, resulting in unstable performance.
23271 @item -masm=@var{dialect}
23272 @opindex masm=@var{dialect}
23273 Output assembly instructions using selected @var{dialect}. Also affects
23274 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
23275 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
23276 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
23277 not support @samp{intel}.
23280 @itemx -mno-ieee-fp
23282 @opindex mno-ieee-fp
23283 Control whether or not the compiler uses IEEE floating-point
23284 comparisons. These correctly handle the case where the result of a
23285 comparison is unordered.
23288 @opindex msoft-float
23289 Generate output containing library calls for floating point.
23291 @strong{Warning:} the requisite libraries are not part of GCC@.
23292 Normally the facilities of the machine's usual C compiler are used, but
23293 this can't be done directly in cross-compilation. You must make your
23294 own arrangements to provide suitable library functions for
23297 On machines where a function returns floating-point results in the 80387
23298 register stack, some floating-point opcodes may be emitted even if
23299 @option{-msoft-float} is used.
23301 @item -mno-fp-ret-in-387
23302 @opindex mno-fp-ret-in-387
23303 Do not use the FPU registers for return values of functions.
23305 The usual calling convention has functions return values of types
23306 @code{float} and @code{double} in an FPU register, even if there
23307 is no FPU@. The idea is that the operating system should emulate
23310 The option @option{-mno-fp-ret-in-387} causes such values to be returned
23311 in ordinary CPU registers instead.
23313 @item -mno-fancy-math-387
23314 @opindex mno-fancy-math-387
23315 Some 387 emulators do not support the @code{sin}, @code{cos} and
23316 @code{sqrt} instructions for the 387. Specify this option to avoid
23317 generating those instructions. This option is the default on
23318 OpenBSD and NetBSD@. This option is overridden when @option{-march}
23319 indicates that the target CPU always has an FPU and so the
23320 instruction does not need emulation. These
23321 instructions are not generated unless you also use the
23322 @option{-funsafe-math-optimizations} switch.
23324 @item -malign-double
23325 @itemx -mno-align-double
23326 @opindex malign-double
23327 @opindex mno-align-double
23328 Control whether GCC aligns @code{double}, @code{long double}, and
23329 @code{long long} variables on a two-word boundary or a one-word
23330 boundary. Aligning @code{double} variables on a two-word boundary
23331 produces code that runs somewhat faster on a Pentium at the
23332 expense of more memory.
23334 On x86-64, @option{-malign-double} is enabled by default.
23336 @strong{Warning:} if you use the @option{-malign-double} switch,
23337 structures containing the above types are aligned differently than
23338 the published application binary interface specifications for the x86-32
23339 and are not binary compatible with structures in code compiled
23340 without that switch.
23342 @item -m96bit-long-double
23343 @itemx -m128bit-long-double
23344 @opindex m96bit-long-double
23345 @opindex m128bit-long-double
23346 These switches control the size of @code{long double} type. The x86-32
23347 application binary interface specifies the size to be 96 bits,
23348 so @option{-m96bit-long-double} is the default in 32-bit mode.
23350 Modern architectures (Pentium and newer) prefer @code{long double}
23351 to be aligned to an 8- or 16-byte boundary. In arrays or structures
23352 conforming to the ABI, this is not possible. So specifying
23353 @option{-m128bit-long-double} aligns @code{long double}
23354 to a 16-byte boundary by padding the @code{long double} with an additional
23357 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
23358 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
23360 Notice that neither of these options enable any extra precision over the x87
23361 standard of 80 bits for a @code{long double}.
23363 @strong{Warning:} if you override the default value for your target ABI, this
23364 changes the size of
23365 structures and arrays containing @code{long double} variables,
23366 as well as modifying the function calling convention for functions taking
23367 @code{long double}. Hence they are not binary-compatible
23368 with code compiled without that switch.
23370 @item -mlong-double-64
23371 @itemx -mlong-double-80
23372 @itemx -mlong-double-128
23373 @opindex mlong-double-64
23374 @opindex mlong-double-80
23375 @opindex mlong-double-128
23376 These switches control the size of @code{long double} type. A size
23377 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23378 type. This is the default for 32-bit Bionic C library. A size
23379 of 128 bits makes the @code{long double} type equivalent to the
23380 @code{__float128} type. This is the default for 64-bit Bionic C library.
23382 @strong{Warning:} if you override the default value for your target ABI, this
23383 changes the size of
23384 structures and arrays containing @code{long double} variables,
23385 as well as modifying the function calling convention for functions taking
23386 @code{long double}. Hence they are not binary-compatible
23387 with code compiled without that switch.
23389 @item -malign-data=@var{type}
23390 @opindex malign-data
23391 Control how GCC aligns variables. Supported values for @var{type} are
23392 @samp{compat} uses increased alignment value compatible uses GCC 4.8
23393 and earlier, @samp{abi} uses alignment value as specified by the
23394 psABI, and @samp{cacheline} uses increased alignment value to match
23395 the cache line size. @samp{compat} is the default.
23397 @item -mlarge-data-threshold=@var{threshold}
23398 @opindex mlarge-data-threshold
23399 When @option{-mcmodel=medium} is specified, data objects larger than
23400 @var{threshold} are placed in the large data section. This value must be the
23401 same across all objects linked into the binary, and defaults to 65535.
23405 Use a different function-calling convention, in which functions that
23406 take a fixed number of arguments return with the @code{ret @var{num}}
23407 instruction, which pops their arguments while returning. This saves one
23408 instruction in the caller since there is no need to pop the arguments
23411 You can specify that an individual function is called with this calling
23412 sequence with the function attribute @code{stdcall}. You can also
23413 override the @option{-mrtd} option by using the function attribute
23414 @code{cdecl}. @xref{Function Attributes}.
23416 @strong{Warning:} this calling convention is incompatible with the one
23417 normally used on Unix, so you cannot use it if you need to call
23418 libraries compiled with the Unix compiler.
23420 Also, you must provide function prototypes for all functions that
23421 take variable numbers of arguments (including @code{printf});
23422 otherwise incorrect code is generated for calls to those
23425 In addition, seriously incorrect code results if you call a
23426 function with too many arguments. (Normally, extra arguments are
23427 harmlessly ignored.)
23429 @item -mregparm=@var{num}
23431 Control how many registers are used to pass integer arguments. By
23432 default, no registers are used to pass arguments, and at most 3
23433 registers can be used. You can control this behavior for a specific
23434 function by using the function attribute @code{regparm}.
23435 @xref{Function Attributes}.
23437 @strong{Warning:} if you use this switch, and
23438 @var{num} is nonzero, then you must build all modules with the same
23439 value, including any libraries. This includes the system libraries and
23443 @opindex msseregparm
23444 Use SSE register passing conventions for float and double arguments
23445 and return values. You can control this behavior for a specific
23446 function by using the function attribute @code{sseregparm}.
23447 @xref{Function Attributes}.
23449 @strong{Warning:} if you use this switch then you must build all
23450 modules with the same value, including any libraries. This includes
23451 the system libraries and startup modules.
23453 @item -mvect8-ret-in-mem
23454 @opindex mvect8-ret-in-mem
23455 Return 8-byte vectors in memory instead of MMX registers. This is the
23456 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
23457 Studio compilers until version 12. Later compiler versions (starting
23458 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
23459 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
23460 you need to remain compatible with existing code produced by those
23461 previous compiler versions or older versions of GCC@.
23470 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
23471 is specified, the significands of results of floating-point operations are
23472 rounded to 24 bits (single precision); @option{-mpc64} rounds the
23473 significands of results of floating-point operations to 53 bits (double
23474 precision) and @option{-mpc80} rounds the significands of results of
23475 floating-point operations to 64 bits (extended double precision), which is
23476 the default. When this option is used, floating-point operations in higher
23477 precisions are not available to the programmer without setting the FPU
23478 control word explicitly.
23480 Setting the rounding of floating-point operations to less than the default
23481 80 bits can speed some programs by 2% or more. Note that some mathematical
23482 libraries assume that extended-precision (80-bit) floating-point operations
23483 are enabled by default; routines in such libraries could suffer significant
23484 loss of accuracy, typically through so-called ``catastrophic cancellation'',
23485 when this option is used to set the precision to less than extended precision.
23487 @item -mstackrealign
23488 @opindex mstackrealign
23489 Realign the stack at entry. On the x86, the @option{-mstackrealign}
23490 option generates an alternate prologue and epilogue that realigns the
23491 run-time stack if necessary. This supports mixing legacy codes that keep
23492 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
23493 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
23494 applicable to individual functions.
23496 @item -mpreferred-stack-boundary=@var{num}
23497 @opindex mpreferred-stack-boundary
23498 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23499 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23500 the default is 4 (16 bytes or 128 bits).
23502 @strong{Warning:} When generating code for the x86-64 architecture with
23503 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
23504 used to keep the stack boundary aligned to 8 byte boundary. Since
23505 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
23506 intended to be used in controlled environment where stack space is
23507 important limitation. This option leads to wrong code when functions
23508 compiled with 16 byte stack alignment (such as functions from a standard
23509 library) are called with misaligned stack. In this case, SSE
23510 instructions may lead to misaligned memory access traps. In addition,
23511 variable arguments are handled incorrectly for 16 byte aligned
23512 objects (including x87 long double and __int128), leading to wrong
23513 results. You must build all modules with
23514 @option{-mpreferred-stack-boundary=3}, including any libraries. This
23515 includes the system libraries and startup modules.
23517 @item -mincoming-stack-boundary=@var{num}
23518 @opindex mincoming-stack-boundary
23519 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
23520 boundary. If @option{-mincoming-stack-boundary} is not specified,
23521 the one specified by @option{-mpreferred-stack-boundary} is used.
23523 On Pentium and Pentium Pro, @code{double} and @code{long double} values
23524 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
23525 suffer significant run time performance penalties. On Pentium III, the
23526 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
23527 properly if it is not 16-byte aligned.
23529 To ensure proper alignment of this values on the stack, the stack boundary
23530 must be as aligned as that required by any value stored on the stack.
23531 Further, every function must be generated such that it keeps the stack
23532 aligned. Thus calling a function compiled with a higher preferred
23533 stack boundary from a function compiled with a lower preferred stack
23534 boundary most likely misaligns the stack. It is recommended that
23535 libraries that use callbacks always use the default setting.
23537 This extra alignment does consume extra stack space, and generally
23538 increases code size. Code that is sensitive to stack space usage, such
23539 as embedded systems and operating system kernels, may want to reduce the
23540 preferred alignment to @option{-mpreferred-stack-boundary=2}.
23597 @itemx -mavx512ifma
23598 @opindex mavx512ifma
23600 @itemx -mavx512vbmi
23601 @opindex mavx512vbmi
23613 @opindex mclfushopt
23630 @itemx -mprefetchwt1
23631 @opindex mprefetchwt1
23687 These switches enable the use of instructions in the MMX, SSE,
23688 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
23689 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
23690 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
23691 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
23692 extended instruction sets. Each has a corresponding @option{-mno-} option
23693 to disable use of these instructions.
23695 These extensions are also available as built-in functions: see
23696 @ref{x86 Built-in Functions}, for details of the functions enabled and
23697 disabled by these switches.
23699 To generate SSE/SSE2 instructions automatically from floating-point
23700 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
23702 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
23703 generates new AVX instructions or AVX equivalence for all SSEx instructions
23706 These options enable GCC to use these extended instructions in
23707 generated code, even without @option{-mfpmath=sse}. Applications that
23708 perform run-time CPU detection must compile separate files for each
23709 supported architecture, using the appropriate flags. In particular,
23710 the file containing the CPU detection code should be compiled without
23713 @item -mdump-tune-features
23714 @opindex mdump-tune-features
23715 This option instructs GCC to dump the names of the x86 performance
23716 tuning features and default settings. The names can be used in
23717 @option{-mtune-ctrl=@var{feature-list}}.
23719 @item -mtune-ctrl=@var{feature-list}
23720 @opindex mtune-ctrl=@var{feature-list}
23721 This option is used to do fine grain control of x86 code generation features.
23722 @var{feature-list} is a comma separated list of @var{feature} names. See also
23723 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
23724 on if it is not preceded with @samp{^}, otherwise, it is turned off.
23725 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
23726 developers. Using it may lead to code paths not covered by testing and can
23727 potentially result in compiler ICEs or runtime errors.
23730 @opindex mno-default
23731 This option instructs GCC to turn off all tunable features. See also
23732 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23736 This option instructs GCC to emit a @code{cld} instruction in the prologue
23737 of functions that use string instructions. String instructions depend on
23738 the DF flag to select between autoincrement or autodecrement mode. While the
23739 ABI specifies the DF flag to be cleared on function entry, some operating
23740 systems violate this specification by not clearing the DF flag in their
23741 exception dispatchers. The exception handler can be invoked with the DF flag
23742 set, which leads to wrong direction mode when string instructions are used.
23743 This option can be enabled by default on 32-bit x86 targets by configuring
23744 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23745 instructions can be suppressed with the @option{-mno-cld} compiler option
23749 @opindex mvzeroupper
23750 This option instructs GCC to emit a @code{vzeroupper} instruction
23751 before a transfer of control flow out of the function to minimize
23752 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23755 @item -mprefer-avx128
23756 @opindex mprefer-avx128
23757 This option instructs GCC to use 128-bit AVX instructions instead of
23758 256-bit AVX instructions in the auto-vectorizer.
23762 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23763 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23764 (or oword) data types.
23765 This is useful for high-resolution counters that can be updated
23766 by multiple processors (or cores). This instruction is generated as part of
23767 atomic built-in functions: see @ref{__sync Builtins} or
23768 @ref{__atomic Builtins} for details.
23772 This option enables generation of @code{SAHF} instructions in 64-bit code.
23773 Early Intel Pentium 4 CPUs with Intel 64 support,
23774 prior to the introduction of Pentium 4 G1 step in December 2005,
23775 lacked the @code{LAHF} and @code{SAHF} instructions
23776 which are supported by AMD64.
23777 These are load and store instructions, respectively, for certain status flags.
23778 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23779 @code{drem}, and @code{remainder} built-in functions;
23780 see @ref{Other Builtins} for details.
23784 This option enables use of the @code{movbe} instruction to implement
23785 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23789 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23790 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23791 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23795 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23796 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23797 with an additional Newton-Raphson step
23798 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23799 (and their vectorized
23800 variants) for single-precision floating-point arguments. These instructions
23801 are generated only when @option{-funsafe-math-optimizations} is enabled
23802 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
23803 Note that while the throughput of the sequence is higher than the throughput
23804 of the non-reciprocal instruction, the precision of the sequence can be
23805 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23807 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23808 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23809 combination), and doesn't need @option{-mrecip}.
23811 Also note that GCC emits the above sequence with additional Newton-Raphson step
23812 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23813 already with @option{-ffast-math} (or the above option combination), and
23814 doesn't need @option{-mrecip}.
23816 @item -mrecip=@var{opt}
23817 @opindex mrecip=opt
23818 This option controls which reciprocal estimate instructions
23819 may be used. @var{opt} is a comma-separated list of options, which may
23820 be preceded by a @samp{!} to invert the option:
23824 Enable all estimate instructions.
23827 Enable the default instructions, equivalent to @option{-mrecip}.
23830 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23833 Enable the approximation for scalar division.
23836 Enable the approximation for vectorized division.
23839 Enable the approximation for scalar square root.
23842 Enable the approximation for vectorized square root.
23845 So, for example, @option{-mrecip=all,!sqrt} enables
23846 all of the reciprocal approximations, except for square root.
23848 @item -mveclibabi=@var{type}
23849 @opindex mveclibabi
23850 Specifies the ABI type to use for vectorizing intrinsics using an
23851 external library. Supported values for @var{type} are @samp{svml}
23852 for the Intel short
23853 vector math library and @samp{acml} for the AMD math core library.
23854 To use this option, both @option{-ftree-vectorize} and
23855 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23856 ABI-compatible library must be specified at link time.
23858 GCC currently emits calls to @code{vmldExp2},
23859 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23860 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23861 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23862 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23863 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23864 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23865 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23866 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23867 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23868 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23869 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23870 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23871 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23872 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23873 when @option{-mveclibabi=acml} is used.
23875 @item -mabi=@var{name}
23877 Generate code for the specified calling convention. Permissible values
23878 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23879 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23880 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23881 You can control this behavior for specific functions by
23882 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23883 @xref{Function Attributes}.
23885 @item -mtls-dialect=@var{type}
23886 @opindex mtls-dialect
23887 Generate code to access thread-local storage using the @samp{gnu} or
23888 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23889 @samp{gnu2} is more efficient, but it may add compile- and run-time
23890 requirements that cannot be satisfied on all systems.
23893 @itemx -mno-push-args
23894 @opindex mpush-args
23895 @opindex mno-push-args
23896 Use PUSH operations to store outgoing parameters. This method is shorter
23897 and usually equally fast as method using SUB/MOV operations and is enabled
23898 by default. In some cases disabling it may improve performance because of
23899 improved scheduling and reduced dependencies.
23901 @item -maccumulate-outgoing-args
23902 @opindex maccumulate-outgoing-args
23903 If enabled, the maximum amount of space required for outgoing arguments is
23904 computed in the function prologue. This is faster on most modern CPUs
23905 because of reduced dependencies, improved scheduling and reduced stack usage
23906 when the preferred stack boundary is not equal to 2. The drawback is a notable
23907 increase in code size. This switch implies @option{-mno-push-args}.
23911 Support thread-safe exception handling on MinGW. Programs that rely
23912 on thread-safe exception handling must compile and link all code with the
23913 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23914 @option{-D_MT}; when linking, it links in a special thread helper library
23915 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23917 @item -mms-bitfields
23918 @itemx -mno-ms-bitfields
23919 @opindex mms-bitfields
23920 @opindex mno-ms-bitfields
23922 Enable/disable bit-field layout compatible with the native Microsoft
23925 If @code{packed} is used on a structure, or if bit-fields are used,
23926 it may be that the Microsoft ABI lays out the structure differently
23927 than the way GCC normally does. Particularly when moving packed
23928 data between functions compiled with GCC and the native Microsoft compiler
23929 (either via function call or as data in a file), it may be necessary to access
23932 This option is enabled by default for Microsoft Windows
23933 targets. This behavior can also be controlled locally by use of variable
23934 or type attributes. For more information, see @ref{x86 Variable Attributes}
23935 and @ref{x86 Type Attributes}.
23937 The Microsoft structure layout algorithm is fairly simple with the exception
23938 of the bit-field packing.
23939 The padding and alignment of members of structures and whether a bit-field
23940 can straddle a storage-unit boundary are determine by these rules:
23943 @item Structure members are stored sequentially in the order in which they are
23944 declared: the first member has the lowest memory address and the last member
23947 @item Every data object has an alignment requirement. The alignment requirement
23948 for all data except structures, unions, and arrays is either the size of the
23949 object or the current packing size (specified with either the
23950 @code{aligned} attribute or the @code{pack} pragma),
23951 whichever is less. For structures, unions, and arrays,
23952 the alignment requirement is the largest alignment requirement of its members.
23953 Every object is allocated an offset so that:
23956 offset % alignment_requirement == 0
23959 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
23960 unit if the integral types are the same size and if the next bit-field fits
23961 into the current allocation unit without crossing the boundary imposed by the
23962 common alignment requirements of the bit-fields.
23965 MSVC interprets zero-length bit-fields in the following ways:
23968 @item If a zero-length bit-field is inserted between two bit-fields that
23969 are normally coalesced, the bit-fields are not coalesced.
23976 unsigned long bf_1 : 12;
23978 unsigned long bf_2 : 12;
23983 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
23984 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
23986 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
23987 alignment of the zero-length bit-field is greater than the member that follows it,
23988 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
24009 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
24010 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
24011 bit-field does not affect the alignment of @code{bar} or, as a result, the size
24014 Taking this into account, it is important to note the following:
24017 @item If a zero-length bit-field follows a normal bit-field, the type of the
24018 zero-length bit-field may affect the alignment of the structure as whole. For
24019 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
24020 normal bit-field, and is of type short.
24022 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
24023 still affect the alignment of the structure:
24034 Here, @code{t4} takes up 4 bytes.
24037 @item Zero-length bit-fields following non-bit-field members are ignored:
24049 Here, @code{t5} takes up 2 bytes.
24053 @item -mno-align-stringops
24054 @opindex mno-align-stringops
24055 Do not align the destination of inlined string operations. This switch reduces
24056 code size and improves performance in case the destination is already aligned,
24057 but GCC doesn't know about it.
24059 @item -minline-all-stringops
24060 @opindex minline-all-stringops
24061 By default GCC inlines string operations only when the destination is
24062 known to be aligned to least a 4-byte boundary.
24063 This enables more inlining and increases code
24064 size, but may improve performance of code that depends on fast
24065 @code{memcpy}, @code{strlen},
24066 and @code{memset} for short lengths.
24068 @item -minline-stringops-dynamically
24069 @opindex minline-stringops-dynamically
24070 For string operations of unknown size, use run-time checks with
24071 inline code for small blocks and a library call for large blocks.
24073 @item -mstringop-strategy=@var{alg}
24074 @opindex mstringop-strategy=@var{alg}
24075 Override the internal decision heuristic for the particular algorithm to use
24076 for inlining string operations. The allowed values for @var{alg} are:
24082 Expand using i386 @code{rep} prefix of the specified size.
24086 @itemx unrolled_loop
24087 Expand into an inline loop.
24090 Always use a library call.
24093 @item -mmemcpy-strategy=@var{strategy}
24094 @opindex mmemcpy-strategy=@var{strategy}
24095 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
24096 should be inlined and what inline algorithm to use when the expected size
24097 of the copy operation is known. @var{strategy}
24098 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
24099 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
24100 the max byte size with which inline algorithm @var{alg} is allowed. For the last
24101 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
24102 in the list must be specified in increasing order. The minimal byte size for
24103 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
24106 @item -mmemset-strategy=@var{strategy}
24107 @opindex mmemset-strategy=@var{strategy}
24108 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
24109 @code{__builtin_memset} expansion.
24111 @item -momit-leaf-frame-pointer
24112 @opindex momit-leaf-frame-pointer
24113 Don't keep the frame pointer in a register for leaf functions. This
24114 avoids the instructions to save, set up, and restore frame pointers and
24115 makes an extra register available in leaf functions. The option
24116 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
24117 which might make debugging harder.
24119 @item -mtls-direct-seg-refs
24120 @itemx -mno-tls-direct-seg-refs
24121 @opindex mtls-direct-seg-refs
24122 Controls whether TLS variables may be accessed with offsets from the
24123 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
24124 or whether the thread base pointer must be added. Whether or not this
24125 is valid depends on the operating system, and whether it maps the
24126 segment to cover the entire TLS area.
24128 For systems that use the GNU C Library, the default is on.
24131 @itemx -mno-sse2avx
24133 Specify that the assembler should encode SSE instructions with VEX
24134 prefix. The option @option{-mavx} turns this on by default.
24139 If profiling is active (@option{-pg}), put the profiling
24140 counter call before the prologue.
24141 Note: On x86 architectures the attribute @code{ms_hook_prologue}
24142 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
24144 @item -mrecord-mcount
24145 @itemx -mno-record-mcount
24146 @opindex mrecord-mcount
24147 If profiling is active (@option{-pg}), generate a __mcount_loc section
24148 that contains pointers to each profiling call. This is useful for
24149 automatically patching and out calls.
24152 @itemx -mno-nop-mcount
24153 @opindex mnop-mcount
24154 If profiling is active (@option{-pg}), generate the calls to
24155 the profiling functions as nops. This is useful when they
24156 should be patched in later dynamically. This is likely only
24157 useful together with @option{-mrecord-mcount}.
24159 @item -mskip-rax-setup
24160 @itemx -mno-skip-rax-setup
24161 @opindex mskip-rax-setup
24162 When generating code for the x86-64 architecture with SSE extensions
24163 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
24164 register when there are no variable arguments passed in vector registers.
24166 @strong{Warning:} Since RAX register is used to avoid unnecessarily
24167 saving vector registers on stack when passing variable arguments, the
24168 impacts of this option are callees may waste some stack space,
24169 misbehave or jump to a random location. GCC 4.4 or newer don't have
24170 those issues, regardless the RAX register value.
24173 @itemx -mno-8bit-idiv
24174 @opindex m8bit-idiv
24175 On some processors, like Intel Atom, 8-bit unsigned integer divide is
24176 much faster than 32-bit/64-bit integer divide. This option generates a
24177 run-time check. If both dividend and divisor are within range of 0
24178 to 255, 8-bit unsigned integer divide is used instead of
24179 32-bit/64-bit integer divide.
24181 @item -mavx256-split-unaligned-load
24182 @itemx -mavx256-split-unaligned-store
24183 @opindex mavx256-split-unaligned-load
24184 @opindex mavx256-split-unaligned-store
24185 Split 32-byte AVX unaligned load and store.
24187 @item -mstack-protector-guard=@var{guard}
24188 @opindex mstack-protector-guard=@var{guard}
24189 Generate stack protection code using canary at @var{guard}. Supported
24190 locations are @samp{global} for global canary or @samp{tls} for per-thread
24191 canary in the TLS block (the default). This option has effect only when
24192 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
24194 @item -mmitigate-rop
24195 @opindex mmitigate-rop
24196 Try to avoid generating code sequences that contain unintended return
24197 opcodes, to mitigate against certain forms of attack. At the moment,
24198 this option is limited in what it can do and should not be relied
24199 on to provide serious protection.
24203 These @samp{-m} switches are supported in addition to the above
24204 on x86-64 processors in 64-bit environments.
24217 Generate code for a 16-bit, 32-bit or 64-bit environment.
24218 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
24220 generates code that runs on any i386 system.
24222 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
24223 types to 64 bits, and generates code for the x86-64 architecture.
24224 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
24225 and @option{-mdynamic-no-pic} options.
24227 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
24229 generates code for the x86-64 architecture.
24231 The @option{-m16} option is the same as @option{-m32}, except for that
24232 it outputs the @code{.code16gcc} assembly directive at the beginning of
24233 the assembly output so that the binary can run in 16-bit mode.
24235 The @option{-miamcu} option generates code which conforms to Intel MCU
24236 psABI. It requires the @option{-m32} option to be turned on.
24238 @item -mno-red-zone
24239 @opindex mno-red-zone
24240 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
24241 by the x86-64 ABI; it is a 128-byte area beyond the location of the
24242 stack pointer that is not modified by signal or interrupt handlers
24243 and therefore can be used for temporary data without adjusting the stack
24244 pointer. The flag @option{-mno-red-zone} disables this red zone.
24246 @item -mcmodel=small
24247 @opindex mcmodel=small
24248 Generate code for the small code model: the program and its symbols must
24249 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
24250 Programs can be statically or dynamically linked. This is the default
24253 @item -mcmodel=kernel
24254 @opindex mcmodel=kernel
24255 Generate code for the kernel code model. The kernel runs in the
24256 negative 2 GB of the address space.
24257 This model has to be used for Linux kernel code.
24259 @item -mcmodel=medium
24260 @opindex mcmodel=medium
24261 Generate code for the medium model: the program is linked in the lower 2
24262 GB of the address space. Small symbols are also placed there. Symbols
24263 with sizes larger than @option{-mlarge-data-threshold} are put into
24264 large data or BSS sections and can be located above 2GB. Programs can
24265 be statically or dynamically linked.
24267 @item -mcmodel=large
24268 @opindex mcmodel=large
24269 Generate code for the large model. This model makes no assumptions
24270 about addresses and sizes of sections.
24272 @item -maddress-mode=long
24273 @opindex maddress-mode=long
24274 Generate code for long address mode. This is only supported for 64-bit
24275 and x32 environments. It is the default address mode for 64-bit
24278 @item -maddress-mode=short
24279 @opindex maddress-mode=short
24280 Generate code for short address mode. This is only supported for 32-bit
24281 and x32 environments. It is the default address mode for 32-bit and
24285 @node x86 Windows Options
24286 @subsection x86 Windows Options
24287 @cindex x86 Windows Options
24288 @cindex Windows Options for x86
24290 These additional options are available for Microsoft Windows targets:
24296 specifies that a console application is to be generated, by
24297 instructing the linker to set the PE header subsystem type
24298 required for console applications.
24299 This option is available for Cygwin and MinGW targets and is
24300 enabled by default on those targets.
24304 This option is available for Cygwin and MinGW targets. It
24305 specifies that a DLL---a dynamic link library---is to be
24306 generated, enabling the selection of the required runtime
24307 startup object and entry point.
24309 @item -mnop-fun-dllimport
24310 @opindex mnop-fun-dllimport
24311 This option is available for Cygwin and MinGW targets. It
24312 specifies that the @code{dllimport} attribute should be ignored.
24316 This option is available for MinGW targets. It specifies
24317 that MinGW-specific thread support is to be used.
24321 This option is available for MinGW-w64 targets. It causes
24322 the @code{UNICODE} preprocessor macro to be predefined, and
24323 chooses Unicode-capable runtime startup code.
24327 This option is available for Cygwin and MinGW targets. It
24328 specifies that the typical Microsoft Windows predefined macros are to
24329 be set in the pre-processor, but does not influence the choice
24330 of runtime library/startup code.
24334 This option is available for Cygwin and MinGW targets. It
24335 specifies that a GUI application is to be generated by
24336 instructing the linker to set the PE header subsystem type
24339 @item -fno-set-stack-executable
24340 @opindex fno-set-stack-executable
24341 This option is available for MinGW targets. It specifies that
24342 the executable flag for the stack used by nested functions isn't
24343 set. This is necessary for binaries running in kernel mode of
24344 Microsoft Windows, as there the User32 API, which is used to set executable
24345 privileges, isn't available.
24347 @item -fwritable-relocated-rdata
24348 @opindex fno-writable-relocated-rdata
24349 This option is available for MinGW and Cygwin targets. It specifies
24350 that relocated-data in read-only section is put into the @code{.data}
24351 section. This is a necessary for older runtimes not supporting
24352 modification of @code{.rdata} sections for pseudo-relocation.
24354 @item -mpe-aligned-commons
24355 @opindex mpe-aligned-commons
24356 This option is available for Cygwin and MinGW targets. It
24357 specifies that the GNU extension to the PE file format that
24358 permits the correct alignment of COMMON variables should be
24359 used when generating code. It is enabled by default if
24360 GCC detects that the target assembler found during configuration
24361 supports the feature.
24364 See also under @ref{x86 Options} for standard options.
24366 @node Xstormy16 Options
24367 @subsection Xstormy16 Options
24368 @cindex Xstormy16 Options
24370 These options are defined for Xstormy16:
24375 Choose startup files and linker script suitable for the simulator.
24378 @node Xtensa Options
24379 @subsection Xtensa Options
24380 @cindex Xtensa Options
24382 These options are supported for Xtensa targets:
24386 @itemx -mno-const16
24388 @opindex mno-const16
24389 Enable or disable use of @code{CONST16} instructions for loading
24390 constant values. The @code{CONST16} instruction is currently not a
24391 standard option from Tensilica. When enabled, @code{CONST16}
24392 instructions are always used in place of the standard @code{L32R}
24393 instructions. The use of @code{CONST16} is enabled by default only if
24394 the @code{L32R} instruction is not available.
24397 @itemx -mno-fused-madd
24398 @opindex mfused-madd
24399 @opindex mno-fused-madd
24400 Enable or disable use of fused multiply/add and multiply/subtract
24401 instructions in the floating-point option. This has no effect if the
24402 floating-point option is not also enabled. Disabling fused multiply/add
24403 and multiply/subtract instructions forces the compiler to use separate
24404 instructions for the multiply and add/subtract operations. This may be
24405 desirable in some cases where strict IEEE 754-compliant results are
24406 required: the fused multiply add/subtract instructions do not round the
24407 intermediate result, thereby producing results with @emph{more} bits of
24408 precision than specified by the IEEE standard. Disabling fused multiply
24409 add/subtract instructions also ensures that the program output is not
24410 sensitive to the compiler's ability to combine multiply and add/subtract
24413 @item -mserialize-volatile
24414 @itemx -mno-serialize-volatile
24415 @opindex mserialize-volatile
24416 @opindex mno-serialize-volatile
24417 When this option is enabled, GCC inserts @code{MEMW} instructions before
24418 @code{volatile} memory references to guarantee sequential consistency.
24419 The default is @option{-mserialize-volatile}. Use
24420 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
24422 @item -mforce-no-pic
24423 @opindex mforce-no-pic
24424 For targets, like GNU/Linux, where all user-mode Xtensa code must be
24425 position-independent code (PIC), this option disables PIC for compiling
24428 @item -mtext-section-literals
24429 @itemx -mno-text-section-literals
24430 @opindex mtext-section-literals
24431 @opindex mno-text-section-literals
24432 These options control the treatment of literal pools. The default is
24433 @option{-mno-text-section-literals}, which places literals in a separate
24434 section in the output file. This allows the literal pool to be placed
24435 in a data RAM/ROM, and it also allows the linker to combine literal
24436 pools from separate object files to remove redundant literals and
24437 improve code size. With @option{-mtext-section-literals}, the literals
24438 are interspersed in the text section in order to keep them as close as
24439 possible to their references. This may be necessary for large assembly
24440 files. Literals for each function are placed right before that function.
24442 @item -mauto-litpools
24443 @itemx -mno-auto-litpools
24444 @opindex mauto-litpools
24445 @opindex mno-auto-litpools
24446 These options control the treatment of literal pools. The default is
24447 @option{-mno-auto-litpools}, which places literals in a separate
24448 section in the output file unless @option{-mtext-section-literals} is
24449 used. With @option{-mauto-litpools} the literals are interspersed in
24450 the text section by the assembler. Compiler does not produce explicit
24451 @code{.literal} directives and loads literals into registers with
24452 @code{MOVI} instructions instead of @code{L32R} to let the assembler
24453 do relaxation and place literals as necessary. This option allows
24454 assembler to create several literal pools per function and assemble
24455 very big functions, which may not be possible with
24456 @option{-mtext-section-literals}.
24458 @item -mtarget-align
24459 @itemx -mno-target-align
24460 @opindex mtarget-align
24461 @opindex mno-target-align
24462 When this option is enabled, GCC instructs the assembler to
24463 automatically align instructions to reduce branch penalties at the
24464 expense of some code density. The assembler attempts to widen density
24465 instructions to align branch targets and the instructions following call
24466 instructions. If there are not enough preceding safe density
24467 instructions to align a target, no widening is performed. The
24468 default is @option{-mtarget-align}. These options do not affect the
24469 treatment of auto-aligned instructions like @code{LOOP}, which the
24470 assembler always aligns, either by widening density instructions or
24471 by inserting NOP instructions.
24474 @itemx -mno-longcalls
24475 @opindex mlongcalls
24476 @opindex mno-longcalls
24477 When this option is enabled, GCC instructs the assembler to translate
24478 direct calls to indirect calls unless it can determine that the target
24479 of a direct call is in the range allowed by the call instruction. This
24480 translation typically occurs for calls to functions in other source
24481 files. Specifically, the assembler translates a direct @code{CALL}
24482 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
24483 The default is @option{-mno-longcalls}. This option should be used in
24484 programs where the call target can potentially be out of range. This
24485 option is implemented in the assembler, not the compiler, so the
24486 assembly code generated by GCC still shows direct call
24487 instructions---look at the disassembled object code to see the actual
24488 instructions. Note that the assembler uses an indirect call for
24489 every cross-file call, not just those that really are out of range.
24492 @node zSeries Options
24493 @subsection zSeries Options
24494 @cindex zSeries options
24496 These are listed under @xref{S/390 and zSeries Options}.
24502 @section Specifying Subprocesses and the Switches to Pass to Them
24505 @command{gcc} is a driver program. It performs its job by invoking a
24506 sequence of other programs to do the work of compiling, assembling and
24507 linking. GCC interprets its command-line parameters and uses these to
24508 deduce which programs it should invoke, and which command-line options
24509 it ought to place on their command lines. This behavior is controlled
24510 by @dfn{spec strings}. In most cases there is one spec string for each
24511 program that GCC can invoke, but a few programs have multiple spec
24512 strings to control their behavior. The spec strings built into GCC can
24513 be overridden by using the @option{-specs=} command-line switch to specify
24516 @dfn{Spec files} are plain-text files that are used to construct spec
24517 strings. They consist of a sequence of directives separated by blank
24518 lines. The type of directive is determined by the first non-whitespace
24519 character on the line, which can be one of the following:
24522 @item %@var{command}
24523 Issues a @var{command} to the spec file processor. The commands that can
24527 @item %include <@var{file}>
24528 @cindex @code{%include}
24529 Search for @var{file} and insert its text at the current point in the
24532 @item %include_noerr <@var{file}>
24533 @cindex @code{%include_noerr}
24534 Just like @samp{%include}, but do not generate an error message if the include
24535 file cannot be found.
24537 @item %rename @var{old_name} @var{new_name}
24538 @cindex @code{%rename}
24539 Rename the spec string @var{old_name} to @var{new_name}.
24543 @item *[@var{spec_name}]:
24544 This tells the compiler to create, override or delete the named spec
24545 string. All lines after this directive up to the next directive or
24546 blank line are considered to be the text for the spec string. If this
24547 results in an empty string then the spec is deleted. (Or, if the
24548 spec did not exist, then nothing happens.) Otherwise, if the spec
24549 does not currently exist a new spec is created. If the spec does
24550 exist then its contents are overridden by the text of this
24551 directive, unless the first character of that text is the @samp{+}
24552 character, in which case the text is appended to the spec.
24554 @item [@var{suffix}]:
24555 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
24556 and up to the next directive or blank line are considered to make up the
24557 spec string for the indicated suffix. When the compiler encounters an
24558 input file with the named suffix, it processes the spec string in
24559 order to work out how to compile that file. For example:
24563 z-compile -input %i
24566 This says that any input file whose name ends in @samp{.ZZ} should be
24567 passed to the program @samp{z-compile}, which should be invoked with the
24568 command-line switch @option{-input} and with the result of performing the
24569 @samp{%i} substitution. (See below.)
24571 As an alternative to providing a spec string, the text following a
24572 suffix directive can be one of the following:
24575 @item @@@var{language}
24576 This says that the suffix is an alias for a known @var{language}. This is
24577 similar to using the @option{-x} command-line switch to GCC to specify a
24578 language explicitly. For example:
24585 Says that .ZZ files are, in fact, C++ source files.
24588 This causes an error messages saying:
24591 @var{name} compiler not installed on this system.
24595 GCC already has an extensive list of suffixes built into it.
24596 This directive adds an entry to the end of the list of suffixes, but
24597 since the list is searched from the end backwards, it is effectively
24598 possible to override earlier entries using this technique.
24602 GCC has the following spec strings built into it. Spec files can
24603 override these strings or create their own. Note that individual
24604 targets can also add their own spec strings to this list.
24607 asm Options to pass to the assembler
24608 asm_final Options to pass to the assembler post-processor
24609 cpp Options to pass to the C preprocessor
24610 cc1 Options to pass to the C compiler
24611 cc1plus Options to pass to the C++ compiler
24612 endfile Object files to include at the end of the link
24613 link Options to pass to the linker
24614 lib Libraries to include on the command line to the linker
24615 libgcc Decides which GCC support library to pass to the linker
24616 linker Sets the name of the linker
24617 predefines Defines to be passed to the C preprocessor
24618 signed_char Defines to pass to CPP to say whether @code{char} is signed
24620 startfile Object files to include at the start of the link
24623 Here is a small example of a spec file:
24626 %rename lib old_lib
24629 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
24632 This example renames the spec called @samp{lib} to @samp{old_lib} and
24633 then overrides the previous definition of @samp{lib} with a new one.
24634 The new definition adds in some extra command-line options before
24635 including the text of the old definition.
24637 @dfn{Spec strings} are a list of command-line options to be passed to their
24638 corresponding program. In addition, the spec strings can contain
24639 @samp{%}-prefixed sequences to substitute variable text or to
24640 conditionally insert text into the command line. Using these constructs
24641 it is possible to generate quite complex command lines.
24643 Here is a table of all defined @samp{%}-sequences for spec
24644 strings. Note that spaces are not generated automatically around the
24645 results of expanding these sequences. Therefore you can concatenate them
24646 together or combine them with constant text in a single argument.
24650 Substitute one @samp{%} into the program name or argument.
24653 Substitute the name of the input file being processed.
24656 Substitute the basename of the input file being processed.
24657 This is the substring up to (and not including) the last period
24658 and not including the directory.
24661 This is the same as @samp{%b}, but include the file suffix (text after
24665 Marks the argument containing or following the @samp{%d} as a
24666 temporary file name, so that that file is deleted if GCC exits
24667 successfully. Unlike @samp{%g}, this contributes no text to the
24670 @item %g@var{suffix}
24671 Substitute a file name that has suffix @var{suffix} and is chosen
24672 once per compilation, and mark the argument in the same way as
24673 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
24674 name is now chosen in a way that is hard to predict even when previously
24675 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
24676 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
24677 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
24678 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
24679 was simply substituted with a file name chosen once per compilation,
24680 without regard to any appended suffix (which was therefore treated
24681 just like ordinary text), making such attacks more likely to succeed.
24683 @item %u@var{suffix}
24684 Like @samp{%g}, but generates a new temporary file name
24685 each time it appears instead of once per compilation.
24687 @item %U@var{suffix}
24688 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
24689 new one if there is no such last file name. In the absence of any
24690 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
24691 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
24692 involves the generation of two distinct file names, one
24693 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
24694 simply substituted with a file name chosen for the previous @samp{%u},
24695 without regard to any appended suffix.
24697 @item %j@var{suffix}
24698 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
24699 writable, and if @option{-save-temps} is not used;
24700 otherwise, substitute the name
24701 of a temporary file, just like @samp{%u}. This temporary file is not
24702 meant for communication between processes, but rather as a junk
24703 disposal mechanism.
24705 @item %|@var{suffix}
24706 @itemx %m@var{suffix}
24707 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
24708 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
24709 all. These are the two most common ways to instruct a program that it
24710 should read from standard input or write to standard output. If you
24711 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
24712 construct: see for example @file{f/lang-specs.h}.
24714 @item %.@var{SUFFIX}
24715 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
24716 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
24717 terminated by the next space or %.
24720 Marks the argument containing or following the @samp{%w} as the
24721 designated output file of this compilation. This puts the argument
24722 into the sequence of arguments that @samp{%o} substitutes.
24725 Substitutes the names of all the output files, with spaces
24726 automatically placed around them. You should write spaces
24727 around the @samp{%o} as well or the results are undefined.
24728 @samp{%o} is for use in the specs for running the linker.
24729 Input files whose names have no recognized suffix are not compiled
24730 at all, but they are included among the output files, so they are
24734 Substitutes the suffix for object files. Note that this is
24735 handled specially when it immediately follows @samp{%g, %u, or %U},
24736 because of the need for those to form complete file names. The
24737 handling is such that @samp{%O} is treated exactly as if it had already
24738 been substituted, except that @samp{%g, %u, and %U} do not currently
24739 support additional @var{suffix} characters following @samp{%O} as they do
24740 following, for example, @samp{.o}.
24743 Substitutes the standard macro predefinitions for the
24744 current target machine. Use this when running @command{cpp}.
24747 Like @samp{%p}, but puts @samp{__} before and after the name of each
24748 predefined macro, except for macros that start with @samp{__} or with
24749 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
24753 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
24754 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
24755 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
24756 and @option{-imultilib} as necessary.
24759 Current argument is the name of a library or startup file of some sort.
24760 Search for that file in a standard list of directories and substitute
24761 the full name found. The current working directory is included in the
24762 list of directories scanned.
24765 Current argument is the name of a linker script. Search for that file
24766 in the current list of directories to scan for libraries. If the file
24767 is located insert a @option{--script} option into the command line
24768 followed by the full path name found. If the file is not found then
24769 generate an error message. Note: the current working directory is not
24773 Print @var{str} as an error message. @var{str} is terminated by a newline.
24774 Use this when inconsistent options are detected.
24776 @item %(@var{name})
24777 Substitute the contents of spec string @var{name} at this point.
24779 @item %x@{@var{option}@}
24780 Accumulate an option for @samp{%X}.
24783 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
24787 Output the accumulated assembler options specified by @option{-Wa}.
24790 Output the accumulated preprocessor options specified by @option{-Wp}.
24793 Process the @code{asm} spec. This is used to compute the
24794 switches to be passed to the assembler.
24797 Process the @code{asm_final} spec. This is a spec string for
24798 passing switches to an assembler post-processor, if such a program is
24802 Process the @code{link} spec. This is the spec for computing the
24803 command line passed to the linker. Typically it makes use of the
24804 @samp{%L %G %S %D and %E} sequences.
24807 Dump out a @option{-L} option for each directory that GCC believes might
24808 contain startup files. If the target supports multilibs then the
24809 current multilib directory is prepended to each of these paths.
24812 Process the @code{lib} spec. This is a spec string for deciding which
24813 libraries are included on the command line to the linker.
24816 Process the @code{libgcc} spec. This is a spec string for deciding
24817 which GCC support library is included on the command line to the linker.
24820 Process the @code{startfile} spec. This is a spec for deciding which
24821 object files are the first ones passed to the linker. Typically
24822 this might be a file named @file{crt0.o}.
24825 Process the @code{endfile} spec. This is a spec string that specifies
24826 the last object files that are passed to the linker.
24829 Process the @code{cpp} spec. This is used to construct the arguments
24830 to be passed to the C preprocessor.
24833 Process the @code{cc1} spec. This is used to construct the options to be
24834 passed to the actual C compiler (@command{cc1}).
24837 Process the @code{cc1plus} spec. This is used to construct the options to be
24838 passed to the actual C++ compiler (@command{cc1plus}).
24841 Substitute the variable part of a matched option. See below.
24842 Note that each comma in the substituted string is replaced by
24846 Remove all occurrences of @code{-S} from the command line. Note---this
24847 command is position dependent. @samp{%} commands in the spec string
24848 before this one see @code{-S}, @samp{%} commands in the spec string
24849 after this one do not.
24851 @item %:@var{function}(@var{args})
24852 Call the named function @var{function}, passing it @var{args}.
24853 @var{args} is first processed as a nested spec string, then split
24854 into an argument vector in the usual fashion. The function returns
24855 a string which is processed as if it had appeared literally as part
24856 of the current spec.
24858 The following built-in spec functions are provided:
24861 @item @code{getenv}
24862 The @code{getenv} spec function takes two arguments: an environment
24863 variable name and a string. If the environment variable is not
24864 defined, a fatal error is issued. Otherwise, the return value is the
24865 value of the environment variable concatenated with the string. For
24866 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
24869 %:getenv(TOPDIR /include)
24872 expands to @file{/path/to/top/include}.
24874 @item @code{if-exists}
24875 The @code{if-exists} spec function takes one argument, an absolute
24876 pathname to a file. If the file exists, @code{if-exists} returns the
24877 pathname. Here is a small example of its usage:
24881 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
24884 @item @code{if-exists-else}
24885 The @code{if-exists-else} spec function is similar to the @code{if-exists}
24886 spec function, except that it takes two arguments. The first argument is
24887 an absolute pathname to a file. If the file exists, @code{if-exists-else}
24888 returns the pathname. If it does not exist, it returns the second argument.
24889 This way, @code{if-exists-else} can be used to select one file or another,
24890 based on the existence of the first. Here is a small example of its usage:
24894 crt0%O%s %:if-exists(crti%O%s) \
24895 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
24898 @item @code{replace-outfile}
24899 The @code{replace-outfile} spec function takes two arguments. It looks for the
24900 first argument in the outfiles array and replaces it with the second argument. Here
24901 is a small example of its usage:
24904 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
24907 @item @code{remove-outfile}
24908 The @code{remove-outfile} spec function takes one argument. It looks for the
24909 first argument in the outfiles array and removes it. Here is a small example
24913 %:remove-outfile(-lm)
24916 @item @code{pass-through-libs}
24917 The @code{pass-through-libs} spec function takes any number of arguments. It
24918 finds any @option{-l} options and any non-options ending in @file{.a} (which it
24919 assumes are the names of linker input library archive files) and returns a
24920 result containing all the found arguments each prepended by
24921 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
24922 intended to be passed to the LTO linker plugin.
24925 %:pass-through-libs(%G %L %G)
24928 @item @code{print-asm-header}
24929 The @code{print-asm-header} function takes no arguments and simply
24930 prints a banner like:
24936 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
24939 It is used to separate compiler options from assembler options
24940 in the @option{--target-help} output.
24943 @item %@{@code{S}@}
24944 Substitutes the @code{-S} switch, if that switch is given to GCC@.
24945 If that switch is not specified, this substitutes nothing. Note that
24946 the leading dash is omitted when specifying this option, and it is
24947 automatically inserted if the substitution is performed. Thus the spec
24948 string @samp{%@{foo@}} matches the command-line option @option{-foo}
24949 and outputs the command-line option @option{-foo}.
24951 @item %W@{@code{S}@}
24952 Like %@{@code{S}@} but mark last argument supplied within as a file to be
24953 deleted on failure.
24955 @item %@{@code{S}*@}
24956 Substitutes all the switches specified to GCC whose names start
24957 with @code{-S}, but which also take an argument. This is used for
24958 switches like @option{-o}, @option{-D}, @option{-I}, etc.
24959 GCC considers @option{-o foo} as being
24960 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
24961 text, including the space. Thus two arguments are generated.
24963 @item %@{@code{S}*&@code{T}*@}
24964 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
24965 (the order of @code{S} and @code{T} in the spec is not significant).
24966 There can be any number of ampersand-separated variables; for each the
24967 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
24969 @item %@{@code{S}:@code{X}@}
24970 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
24972 @item %@{!@code{S}:@code{X}@}
24973 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
24975 @item %@{@code{S}*:@code{X}@}
24976 Substitutes @code{X} if one or more switches whose names start with
24977 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
24978 once, no matter how many such switches appeared. However, if @code{%*}
24979 appears somewhere in @code{X}, then @code{X} is substituted once
24980 for each matching switch, with the @code{%*} replaced by the part of
24981 that switch matching the @code{*}.
24983 If @code{%*} appears as the last part of a spec sequence then a space
24984 is added after the end of the last substitution. If there is more
24985 text in the sequence, however, then a space is not generated. This
24986 allows the @code{%*} substitution to be used as part of a larger
24987 string. For example, a spec string like this:
24990 %@{mcu=*:--script=%*/memory.ld@}
24994 when matching an option like @option{-mcu=newchip} produces:
24997 --script=newchip/memory.ld
25000 @item %@{.@code{S}:@code{X}@}
25001 Substitutes @code{X}, if processing a file with suffix @code{S}.
25003 @item %@{!.@code{S}:@code{X}@}
25004 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
25006 @item %@{,@code{S}:@code{X}@}
25007 Substitutes @code{X}, if processing a file for language @code{S}.
25009 @item %@{!,@code{S}:@code{X}@}
25010 Substitutes @code{X}, if not processing a file for language @code{S}.
25012 @item %@{@code{S}|@code{P}:@code{X}@}
25013 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
25014 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
25015 @code{*} sequences as well, although they have a stronger binding than
25016 the @samp{|}. If @code{%*} appears in @code{X}, all of the
25017 alternatives must be starred, and only the first matching alternative
25020 For example, a spec string like this:
25023 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
25027 outputs the following command-line options from the following input
25028 command-line options:
25033 -d fred.c -foo -baz -boggle
25034 -d jim.d -bar -baz -boggle
25037 @item %@{S:X; T:Y; :D@}
25039 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
25040 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
25041 be as many clauses as you need. This may be combined with @code{.},
25042 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
25047 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
25048 construct may contain other nested @samp{%} constructs or spaces, or
25049 even newlines. They are processed as usual, as described above.
25050 Trailing white space in @code{X} is ignored. White space may also
25051 appear anywhere on the left side of the colon in these constructs,
25052 except between @code{.} or @code{*} and the corresponding word.
25054 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
25055 handled specifically in these constructs. If another value of
25056 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
25057 @option{-W} switch is found later in the command line, the earlier
25058 switch value is ignored, except with @{@code{S}*@} where @code{S} is
25059 just one letter, which passes all matching options.
25061 The character @samp{|} at the beginning of the predicate text is used to
25062 indicate that a command should be piped to the following command, but
25063 only if @option{-pipe} is specified.
25065 It is built into GCC which switches take arguments and which do not.
25066 (You might think it would be useful to generalize this to allow each
25067 compiler's spec to say which switches take arguments. But this cannot
25068 be done in a consistent fashion. GCC cannot even decide which input
25069 files have been specified without knowing which switches take arguments,
25070 and it must know which input files to compile in order to tell which
25073 GCC also knows implicitly that arguments starting in @option{-l} are to be
25074 treated as compiler output files, and passed to the linker in their
25075 proper position among the other output files.
25077 @node Environment Variables
25078 @section Environment Variables Affecting GCC
25079 @cindex environment variables
25081 @c man begin ENVIRONMENT
25082 This section describes several environment variables that affect how GCC
25083 operates. Some of them work by specifying directories or prefixes to use
25084 when searching for various kinds of files. Some are used to specify other
25085 aspects of the compilation environment.
25087 Note that you can also specify places to search using options such as
25088 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
25089 take precedence over places specified using environment variables, which
25090 in turn take precedence over those specified by the configuration of GCC@.
25091 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
25092 GNU Compiler Collection (GCC) Internals}.
25097 @c @itemx LC_COLLATE
25099 @c @itemx LC_MONETARY
25100 @c @itemx LC_NUMERIC
25105 @c @findex LC_COLLATE
25106 @findex LC_MESSAGES
25107 @c @findex LC_MONETARY
25108 @c @findex LC_NUMERIC
25112 These environment variables control the way that GCC uses
25113 localization information which allows GCC to work with different
25114 national conventions. GCC inspects the locale categories
25115 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
25116 so. These locale categories can be set to any value supported by your
25117 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
25118 Kingdom encoded in UTF-8.
25120 The @env{LC_CTYPE} environment variable specifies character
25121 classification. GCC uses it to determine the character boundaries in
25122 a string; this is needed for some multibyte encodings that contain quote
25123 and escape characters that are otherwise interpreted as a string
25126 The @env{LC_MESSAGES} environment variable specifies the language to
25127 use in diagnostic messages.
25129 If the @env{LC_ALL} environment variable is set, it overrides the value
25130 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
25131 and @env{LC_MESSAGES} default to the value of the @env{LANG}
25132 environment variable. If none of these variables are set, GCC
25133 defaults to traditional C English behavior.
25137 If @env{TMPDIR} is set, it specifies the directory to use for temporary
25138 files. GCC uses temporary files to hold the output of one stage of
25139 compilation which is to be used as input to the next stage: for example,
25140 the output of the preprocessor, which is the input to the compiler
25143 @item GCC_COMPARE_DEBUG
25144 @findex GCC_COMPARE_DEBUG
25145 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
25146 @option{-fcompare-debug} to the compiler driver. See the documentation
25147 of this option for more details.
25149 @item GCC_EXEC_PREFIX
25150 @findex GCC_EXEC_PREFIX
25151 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
25152 names of the subprograms executed by the compiler. No slash is added
25153 when this prefix is combined with the name of a subprogram, but you can
25154 specify a prefix that ends with a slash if you wish.
25156 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
25157 an appropriate prefix to use based on the pathname it is invoked with.
25159 If GCC cannot find the subprogram using the specified prefix, it
25160 tries looking in the usual places for the subprogram.
25162 The default value of @env{GCC_EXEC_PREFIX} is
25163 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
25164 the installed compiler. In many cases @var{prefix} is the value
25165 of @code{prefix} when you ran the @file{configure} script.
25167 Other prefixes specified with @option{-B} take precedence over this prefix.
25169 This prefix is also used for finding files such as @file{crt0.o} that are
25172 In addition, the prefix is used in an unusual way in finding the
25173 directories to search for header files. For each of the standard
25174 directories whose name normally begins with @samp{/usr/local/lib/gcc}
25175 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
25176 replacing that beginning with the specified prefix to produce an
25177 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
25178 @file{foo/bar} just before it searches the standard directory
25179 @file{/usr/local/lib/bar}.
25180 If a standard directory begins with the configured
25181 @var{prefix} then the value of @var{prefix} is replaced by
25182 @env{GCC_EXEC_PREFIX} when looking for header files.
25184 @item COMPILER_PATH
25185 @findex COMPILER_PATH
25186 The value of @env{COMPILER_PATH} is a colon-separated list of
25187 directories, much like @env{PATH}. GCC tries the directories thus
25188 specified when searching for subprograms, if it can't find the
25189 subprograms using @env{GCC_EXEC_PREFIX}.
25192 @findex LIBRARY_PATH
25193 The value of @env{LIBRARY_PATH} is a colon-separated list of
25194 directories, much like @env{PATH}. When configured as a native compiler,
25195 GCC tries the directories thus specified when searching for special
25196 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
25197 using GCC also uses these directories when searching for ordinary
25198 libraries for the @option{-l} option (but directories specified with
25199 @option{-L} come first).
25203 @cindex locale definition
25204 This variable is used to pass locale information to the compiler. One way in
25205 which this information is used is to determine the character set to be used
25206 when character literals, string literals and comments are parsed in C and C++.
25207 When the compiler is configured to allow multibyte characters,
25208 the following values for @env{LANG} are recognized:
25212 Recognize JIS characters.
25214 Recognize SJIS characters.
25216 Recognize EUCJP characters.
25219 If @env{LANG} is not defined, or if it has some other value, then the
25220 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
25221 recognize and translate multibyte characters.
25225 Some additional environment variables affect the behavior of the
25228 @include cppenv.texi
25232 @node Precompiled Headers
25233 @section Using Precompiled Headers
25234 @cindex precompiled headers
25235 @cindex speed of compilation
25237 Often large projects have many header files that are included in every
25238 source file. The time the compiler takes to process these header files
25239 over and over again can account for nearly all of the time required to
25240 build the project. To make builds faster, GCC allows you to
25241 @dfn{precompile} a header file.
25243 To create a precompiled header file, simply compile it as you would any
25244 other file, if necessary using the @option{-x} option to make the driver
25245 treat it as a C or C++ header file. You may want to use a
25246 tool like @command{make} to keep the precompiled header up-to-date when
25247 the headers it contains change.
25249 A precompiled header file is searched for when @code{#include} is
25250 seen in the compilation. As it searches for the included file
25251 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
25252 compiler looks for a precompiled header in each directory just before it
25253 looks for the include file in that directory. The name searched for is
25254 the name specified in the @code{#include} with @samp{.gch} appended. If
25255 the precompiled header file can't be used, it is ignored.
25257 For instance, if you have @code{#include "all.h"}, and you have
25258 @file{all.h.gch} in the same directory as @file{all.h}, then the
25259 precompiled header file is used if possible, and the original
25260 header is used otherwise.
25262 Alternatively, you might decide to put the precompiled header file in a
25263 directory and use @option{-I} to ensure that directory is searched
25264 before (or instead of) the directory containing the original header.
25265 Then, if you want to check that the precompiled header file is always
25266 used, you can put a file of the same name as the original header in this
25267 directory containing an @code{#error} command.
25269 This also works with @option{-include}. So yet another way to use
25270 precompiled headers, good for projects not designed with precompiled
25271 header files in mind, is to simply take most of the header files used by
25272 a project, include them from another header file, precompile that header
25273 file, and @option{-include} the precompiled header. If the header files
25274 have guards against multiple inclusion, they are skipped because
25275 they've already been included (in the precompiled header).
25277 If you need to precompile the same header file for different
25278 languages, targets, or compiler options, you can instead make a
25279 @emph{directory} named like @file{all.h.gch}, and put each precompiled
25280 header in the directory, perhaps using @option{-o}. It doesn't matter
25281 what you call the files in the directory; every precompiled header in
25282 the directory is considered. The first precompiled header
25283 encountered in the directory that is valid for this compilation is
25284 used; they're searched in no particular order.
25286 There are many other possibilities, limited only by your imagination,
25287 good sense, and the constraints of your build system.
25289 A precompiled header file can be used only when these conditions apply:
25293 Only one precompiled header can be used in a particular compilation.
25296 A precompiled header can't be used once the first C token is seen. You
25297 can have preprocessor directives before a precompiled header; you cannot
25298 include a precompiled header from inside another header.
25301 The precompiled header file must be produced for the same language as
25302 the current compilation. You can't use a C precompiled header for a C++
25306 The precompiled header file must have been produced by the same compiler
25307 binary as the current compilation is using.
25310 Any macros defined before the precompiled header is included must
25311 either be defined in the same way as when the precompiled header was
25312 generated, or must not affect the precompiled header, which usually
25313 means that they don't appear in the precompiled header at all.
25315 The @option{-D} option is one way to define a macro before a
25316 precompiled header is included; using a @code{#define} can also do it.
25317 There are also some options that define macros implicitly, like
25318 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
25321 @item If debugging information is output when using the precompiled
25322 header, using @option{-g} or similar, the same kind of debugging information
25323 must have been output when building the precompiled header. However,
25324 a precompiled header built using @option{-g} can be used in a compilation
25325 when no debugging information is being output.
25327 @item The same @option{-m} options must generally be used when building
25328 and using the precompiled header. @xref{Submodel Options},
25329 for any cases where this rule is relaxed.
25331 @item Each of the following options must be the same when building and using
25332 the precompiled header:
25334 @gccoptlist{-fexceptions}
25337 Some other command-line options starting with @option{-f},
25338 @option{-p}, or @option{-O} must be defined in the same way as when
25339 the precompiled header was generated. At present, it's not clear
25340 which options are safe to change and which are not; the safest choice
25341 is to use exactly the same options when generating and using the
25342 precompiled header. The following are known to be safe:
25344 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
25345 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
25346 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
25351 For all of these except the last, the compiler automatically
25352 ignores the precompiled header if the conditions aren't met. If you
25353 find an option combination that doesn't work and doesn't cause the
25354 precompiled header to be ignored, please consider filing a bug report,
25357 If you do use differing options when generating and using the
25358 precompiled header, the actual behavior is a mixture of the
25359 behavior for the options. For instance, if you use @option{-g} to
25360 generate the precompiled header but not when using it, you may or may
25361 not get debugging information for routines in the precompiled header.