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 @gol
193 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
194 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
195 -ffriend-injection @gol
196 -fno-elide-constructors @gol
197 -fno-enforce-eh-specs @gol
198 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
199 -fno-implicit-templates @gol
200 -fno-implicit-inline-templates @gol
201 -fno-implement-inlines -fms-extensions @gol
202 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
203 -fno-optional-diags -fpermissive @gol
204 -fno-pretty-templates @gol
205 -frepo -fno-rtti -fsized-deallocation @gol
206 -ftemplate-backtrace-limit=@var{n} @gol
207 -ftemplate-depth=@var{n} @gol
208 -fno-threadsafe-statics -fuse-cxa-atexit @gol
209 -fno-weak -nostdinc++ @gol
210 -fvisibility-inlines-hidden @gol
211 -fvisibility-ms-compat @gol
212 -fext-numeric-literals @gol
213 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
214 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
215 -Wnamespaces -Wnarrowing @gol
216 -Wnoexcept -Wnon-virtual-dtor -Wreorder -Wregister @gol
217 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
218 -Wno-non-template-friend -Wold-style-cast @gol
219 -Woverloaded-virtual -Wno-pmf-conversions @gol
220 -Wsign-promo -Wvirtual-inheritance}
222 @item Objective-C and Objective-C++ Language Options
223 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
224 Objective-C and Objective-C++ Dialects}.
225 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
226 -fgnu-runtime -fnext-runtime @gol
227 -fno-nil-receivers @gol
228 -fobjc-abi-version=@var{n} @gol
229 -fobjc-call-cxx-cdtors @gol
230 -fobjc-direct-dispatch @gol
231 -fobjc-exceptions @gol
234 -fobjc-std=objc1 @gol
235 -fno-local-ivars @gol
236 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
237 -freplace-objc-classes @gol
240 -Wassign-intercept @gol
241 -Wno-protocol -Wselector @gol
242 -Wstrict-selector-match @gol
243 -Wundeclared-selector}
245 @item Diagnostic Message Formatting Options
246 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
247 @gccoptlist{-fmessage-length=@var{n} @gol
248 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
249 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
250 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
251 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch}
253 @item Warning Options
254 @xref{Warning Options,,Options to Request or Suppress Warnings}.
255 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
256 -pedantic-errors @gol
257 -w -Wextra -Wall -Waddress -Waggregate-return @gol
258 -Walloca -Walloca-larger-than=@var{n} @gol
259 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
260 -Wno-attributes -Wbool-compare -Wbool-operation @gol
261 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
262 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
263 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
264 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
265 -Wdelete-incomplete @gol
266 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
267 -Wdisabled-optimization @gol
268 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
269 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
270 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
271 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
272 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-length=@var{n} @gol
273 -Wformat-nonliteral @gol
274 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
275 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
276 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
277 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
278 -Wimplicit-function-declaration -Wimplicit-int @gol
279 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
280 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
281 -Winvalid-pch -Wlarger-than=@var{len} @gol
282 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
283 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
284 -Wmisleading-indentation -Wmissing-braces @gol
285 -Wmissing-field-initializers -Wmissing-include-dirs @gol
286 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
287 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
288 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
289 -Woverride-init-side-effects -Woverlength-strings @gol
290 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
291 -Wparentheses -Wno-pedantic-ms-format @gol
292 -Wplacement-new -Wplacement-new=@var{n} @gol
293 -Wpointer-arith -Wno-pointer-to-int-cast @gol
294 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
295 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
296 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
297 -Wshift-overflow -Wshift-overflow=@var{n} @gol
298 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
299 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
300 -Wno-scalar-storage-order @gol
301 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
302 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
303 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
304 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
305 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
306 -Wmissing-format-attribute -Wsubobject-linkage @gol
307 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
308 -Wswitch-unreachable -Wsync-nand @gol
309 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
310 -Wtype-limits -Wundef @gol
311 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
312 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
313 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
314 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
315 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
316 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
317 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
318 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
319 -Wzero-as-null-pointer-constant -Whsa}
321 @item C and Objective-C-only Warning Options
322 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
323 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
324 -Wold-style-declaration -Wold-style-definition @gol
325 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
326 -Wdeclaration-after-statement -Wpointer-sign}
328 @item Debugging Options
329 @xref{Debugging Options,,Options for Debugging Your Program}.
330 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
331 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
332 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
333 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
334 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
335 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
336 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
337 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
338 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
339 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
340 -fvar-tracking -fvar-tracking-assignments}
342 @item Optimization Options
343 @xref{Optimize Options,,Options that Control Optimization}.
344 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
345 -falign-jumps[=@var{n}] @gol
346 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
347 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
348 -fauto-inc-dec -fbranch-probabilities @gol
349 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
350 -fbtr-bb-exclusive -fcaller-saves @gol
351 -fcombine-stack-adjustments -fconserve-stack @gol
352 -fcompare-elim -fcprop-registers -fcrossjumping @gol
353 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
354 -fcx-limited-range @gol
355 -fdata-sections -fdce -fdelayed-branch @gol
356 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
357 -fdevirtualize-at-ltrans -fdse @gol
358 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
359 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
360 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
361 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
362 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
363 -fif-conversion2 -findirect-inlining @gol
364 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
365 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment -fipa-bit-cp @gol
366 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
367 -fira-algorithm=@var{algorithm} @gol
368 -fira-region=@var{region} -fira-hoist-pressure @gol
369 -fira-loop-pressure -fno-ira-share-save-slots @gol
370 -fno-ira-share-spill-slots @gol
371 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
372 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
373 -fkeep-static-consts -flive-range-shrinkage @gol
374 -floop-block -floop-interchange -floop-strip-mine @gol
375 -floop-unroll-and-jam -floop-nest-optimize @gol
376 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
377 -flto-partition=@var{alg} -fmerge-all-constants @gol
378 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
379 -fmove-loop-invariants -fno-branch-count-reg @gol
380 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
381 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
382 -fno-peephole2 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
383 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
384 -fomit-frame-pointer -foptimize-sibling-calls @gol
385 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
386 -fprefetch-loop-arrays -fprintf-return-value @gol
387 -fprofile-correction @gol
388 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
389 -fprofile-reorder-functions @gol
390 -freciprocal-math -free -frename-registers -freorder-blocks @gol
391 -freorder-blocks-algorithm=@var{algorithm} @gol
392 -freorder-blocks-and-partition -freorder-functions @gol
393 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
394 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
395 -fsched-spec-load -fsched-spec-load-dangerous @gol
396 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
397 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
398 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
399 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
400 -fschedule-fusion @gol
401 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
402 -fselective-scheduling -fselective-scheduling2 @gol
403 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
404 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
405 -fsignaling-nans @gol
406 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
408 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
409 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
410 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
411 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
412 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
413 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting -ftree-loop-if-convert @gol
414 -ftree-loop-if-convert-stores -ftree-loop-im @gol
415 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
416 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
417 -ftree-loop-vectorize @gol
418 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
419 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
420 -ftree-switch-conversion -ftree-tail-merge @gol
421 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
422 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
423 -funsafe-math-optimizations -funswitch-loops @gol
424 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
425 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
426 --param @var{name}=@var{value}
427 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
429 @item Program Instrumentation Options
430 @xref{Instrumentation Options,,Program Instrumentation Options}.
431 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
432 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
433 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
434 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
435 -fsanitize-undefined-trap-on-error -fbounds-check @gol
436 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
437 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
438 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
439 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
440 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
441 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
442 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
443 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
444 -fchkp-use-wrappers @gol
445 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
446 -fstack-protector-explicit -fstack-check @gol
447 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
448 -fno-stack-limit -fsplit-stack @gol
449 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
450 -fvtv-counts -fvtv-debug @gol
451 -finstrument-functions @gol
452 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
453 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
455 @item Preprocessor Options
456 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
457 @gccoptlist{-A@var{question}=@var{answer} @gol
458 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
459 -C -dD -dI -dM -dN @gol
460 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
461 -idirafter @var{dir} @gol
462 -include @var{file} -imacros @var{file} @gol
463 -iprefix @var{file} -iwithprefix @var{dir} @gol
464 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
465 -imultilib @var{dir} -isysroot @var{dir} @gol
466 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
467 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
468 -remap -trigraphs -undef -U@var{macro} @gol
469 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
471 @item Assembler Option
472 @xref{Assembler Options,,Passing Options to the Assembler}.
473 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
476 @xref{Link Options,,Options for Linking}.
477 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
478 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
479 -s -static -static-libgcc -static-libstdc++ @gol
480 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
481 -static-libmpx -static-libmpxwrappers @gol
482 -shared -shared-libgcc -symbolic @gol
483 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
484 -u @var{symbol} -z @var{keyword}}
486 @item Directory Options
487 @xref{Directory Options,,Options for Directory Search}.
488 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
489 -iquote@var{dir} -L@var{dir} -no-canonical-prefixes -I- @gol
490 --sysroot=@var{dir} --no-sysroot-suffix}
492 @item Code Generation Options
493 @xref{Code Gen Options,,Options for Code Generation Conventions}.
494 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
495 -ffixed-@var{reg} -fexceptions @gol
496 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
497 -fasynchronous-unwind-tables @gol
499 -finhibit-size-directive -fno-common -fno-ident @gol
500 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
501 -fno-jump-tables @gol
502 -frecord-gcc-switches @gol
503 -freg-struct-return -fshort-enums -fshort-wchar @gol
504 -fverbose-asm -fpack-struct[=@var{n}] @gol
505 -fleading-underscore -ftls-model=@var{model} @gol
506 -fstack-reuse=@var{reuse_level} @gol
507 -ftrampolines -ftrapv -fwrapv @gol
508 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
509 -fstrict-volatile-bitfields -fsync-libcalls}
511 @item Developer Options
512 @xref{Developer Options,,GCC Developer Options}.
513 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
514 -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
515 -fdbg-cnt=@var{counter-value-list} @gol
516 -fdisable-ipa-@var{pass_name} @gol
517 -fdisable-rtl-@var{pass_name} @gol
518 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
519 -fdisable-tree-@var{pass_name} @gol
520 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
521 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
522 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
523 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
524 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
526 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
527 -fdump-statistics @gol
529 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
530 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
531 -fdump-tree-cfg -fdump-tree-alias @gol
533 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
534 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
535 -fdump-tree-gimple@r{[}-raw@r{]} @gol
536 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
537 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
538 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
539 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
540 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
541 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
542 -fdump-tree-nrv -fdump-tree-vect @gol
543 -fdump-tree-sink @gol
544 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
545 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
546 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
547 -fdump-tree-vtable-verify @gol
548 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
549 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
550 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
551 -fdump-final-insns=@var{file} @gol
552 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
553 -fenable-@var{kind}-@var{pass} @gol
554 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
555 -fira-verbose=@var{n} @gol
556 -flto-report -flto-report-wpa -fmem-report-wpa @gol
557 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
558 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
559 -fprofile-report @gol
560 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
561 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
562 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
563 -fvar-tracking-assignments-toggle -gtoggle @gol
564 -print-file-name=@var{library} -print-libgcc-file-name @gol
565 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
566 -print-prog-name=@var{program} -print-search-dirs -Q @gol
567 -print-sysroot -print-sysroot-headers-suffix @gol
568 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
570 @item Machine-Dependent Options
571 @xref{Submodel Options,,Machine-Dependent Options}.
572 @c This list is ordered alphanumerically by subsection name.
573 @c Try and put the significant identifier (CPU or system) first,
574 @c so users have a clue at guessing where the ones they want will be.
576 @emph{AArch64 Options}
577 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
578 -mgeneral-regs-only @gol
579 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
581 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
582 -mtls-dialect=desc -mtls-dialect=traditional @gol
583 -mtls-size=@var{size} @gol
584 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
585 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
586 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
587 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
588 -mlow-precision-div -mno-low-precision-div @gol
589 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
591 @emph{Adapteva Epiphany Options}
592 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
593 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
594 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
595 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
596 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
597 -msplit-vecmove-early -m1reg-@var{reg}}
600 @gccoptlist{-mbarrel-shifter @gol
601 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
602 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
603 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
604 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
605 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
606 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
607 -mlong-calls -mmedium-calls -msdata @gol
608 -mucb-mcount -mvolatile-cache -mtp-regno=@var{regno} @gol
609 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
610 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
611 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
612 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
613 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
614 -mtune=@var{cpu} -mmultcost=@var{num} @gol
615 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
616 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
619 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
620 -mabi=@var{name} @gol
621 -mapcs-stack-check -mno-apcs-stack-check @gol
622 -mapcs-float -mno-apcs-float @gol
623 -mapcs-reentrant -mno-apcs-reentrant @gol
624 -msched-prolog -mno-sched-prolog @gol
625 -mlittle-endian -mbig-endian @gol
626 -mfloat-abi=@var{name} @gol
627 -mfp16-format=@var{name}
628 -mthumb-interwork -mno-thumb-interwork @gol
629 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
630 -mtune=@var{name} -mprint-tune-info @gol
631 -mstructure-size-boundary=@var{n} @gol
632 -mabort-on-noreturn @gol
633 -mlong-calls -mno-long-calls @gol
634 -msingle-pic-base -mno-single-pic-base @gol
635 -mpic-register=@var{reg} @gol
636 -mnop-fun-dllimport @gol
637 -mpoke-function-name @gol
639 -mtpcs-frame -mtpcs-leaf-frame @gol
640 -mcaller-super-interworking -mcallee-super-interworking @gol
641 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
642 -mword-relocations @gol
643 -mfix-cortex-m3-ldrd @gol
644 -munaligned-access @gol
645 -mneon-for-64bits @gol
646 -mslow-flash-data @gol
647 -masm-syntax-unified @gol
652 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
653 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
654 -mrelax -mrmw -mstrict-X -mtiny-stack -mfract-convert-truncate -nodevicelib @gol
655 -Waddr-space-convert -Wmisspelled-isr}
657 @emph{Blackfin Options}
658 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
659 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
660 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
661 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
662 -mno-id-shared-library -mshared-library-id=@var{n} @gol
663 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
664 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
665 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
669 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
670 -msim -msdata=@var{sdata-type}}
673 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
674 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
675 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
676 -mstack-align -mdata-align -mconst-align @gol
677 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
678 -melf -maout -melinux -mlinux -sim -sim2 @gol
679 -mmul-bug-workaround -mno-mul-bug-workaround}
682 @gccoptlist{-mmac @gol
683 -mcr16cplus -mcr16c @gol
684 -msim -mint32 -mbit-ops
685 -mdata-model=@var{model}}
687 @emph{Darwin Options}
688 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
689 -arch_only -bind_at_load -bundle -bundle_loader @gol
690 -client_name -compatibility_version -current_version @gol
692 -dependency-file -dylib_file -dylinker_install_name @gol
693 -dynamic -dynamiclib -exported_symbols_list @gol
694 -filelist -flat_namespace -force_cpusubtype_ALL @gol
695 -force_flat_namespace -headerpad_max_install_names @gol
697 -image_base -init -install_name -keep_private_externs @gol
698 -multi_module -multiply_defined -multiply_defined_unused @gol
699 -noall_load -no_dead_strip_inits_and_terms @gol
700 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
701 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
702 -private_bundle -read_only_relocs -sectalign @gol
703 -sectobjectsymbols -whyload -seg1addr @gol
704 -sectcreate -sectobjectsymbols -sectorder @gol
705 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
706 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
707 -segprot -segs_read_only_addr -segs_read_write_addr @gol
708 -single_module -static -sub_library -sub_umbrella @gol
709 -twolevel_namespace -umbrella -undefined @gol
710 -unexported_symbols_list -weak_reference_mismatches @gol
711 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
712 -mkernel -mone-byte-bool}
714 @emph{DEC Alpha Options}
715 @gccoptlist{-mno-fp-regs -msoft-float @gol
716 -mieee -mieee-with-inexact -mieee-conformant @gol
717 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
718 -mtrap-precision=@var{mode} -mbuild-constants @gol
719 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
720 -mbwx -mmax -mfix -mcix @gol
721 -mfloat-vax -mfloat-ieee @gol
722 -mexplicit-relocs -msmall-data -mlarge-data @gol
723 -msmall-text -mlarge-text @gol
724 -mmemory-latency=@var{time}}
727 @gccoptlist{-msmall-model -mno-lsim}
730 @gccoptlist{-msim -mlra -mnodiv}
733 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
734 -mhard-float -msoft-float @gol
735 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
736 -mdouble -mno-double @gol
737 -mmedia -mno-media -mmuladd -mno-muladd @gol
738 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
739 -mlinked-fp -mlong-calls -malign-labels @gol
740 -mlibrary-pic -macc-4 -macc-8 @gol
741 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
742 -moptimize-membar -mno-optimize-membar @gol
743 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
744 -mvliw-branch -mno-vliw-branch @gol
745 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
746 -mno-nested-cond-exec -mtomcat-stats @gol
750 @emph{GNU/Linux Options}
751 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
752 -tno-android-cc -tno-android-ld}
754 @emph{H8/300 Options}
755 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
758 @gccoptlist{-march=@var{architecture-type} @gol
759 -mdisable-fpregs -mdisable-indexing @gol
760 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
761 -mfixed-range=@var{register-range} @gol
762 -mjump-in-delay -mlinker-opt -mlong-calls @gol
763 -mlong-load-store -mno-disable-fpregs @gol
764 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
765 -mno-jump-in-delay -mno-long-load-store @gol
766 -mno-portable-runtime -mno-soft-float @gol
767 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
768 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
769 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
770 -munix=@var{unix-std} -nolibdld -static -threads}
773 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
774 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
775 -mconstant-gp -mauto-pic -mfused-madd @gol
776 -minline-float-divide-min-latency @gol
777 -minline-float-divide-max-throughput @gol
778 -mno-inline-float-divide @gol
779 -minline-int-divide-min-latency @gol
780 -minline-int-divide-max-throughput @gol
781 -mno-inline-int-divide @gol
782 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
783 -mno-inline-sqrt @gol
784 -mdwarf2-asm -mearly-stop-bits @gol
785 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
786 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
787 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
788 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
789 -msched-spec-ldc -msched-spec-control-ldc @gol
790 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
791 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
792 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
793 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
796 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
797 -msign-extend-enabled -muser-enabled}
799 @emph{M32R/D Options}
800 @gccoptlist{-m32r2 -m32rx -m32r @gol
802 -malign-loops -mno-align-loops @gol
803 -missue-rate=@var{number} @gol
804 -mbranch-cost=@var{number} @gol
805 -mmodel=@var{code-size-model-type} @gol
806 -msdata=@var{sdata-type} @gol
807 -mno-flush-func -mflush-func=@var{name} @gol
808 -mno-flush-trap -mflush-trap=@var{number} @gol
812 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
814 @emph{M680x0 Options}
815 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
816 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
817 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
818 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
819 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
820 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
821 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
822 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
826 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
827 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
828 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
829 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
830 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
833 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
834 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
835 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
836 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
839 @emph{MicroBlaze Options}
840 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
841 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
842 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
843 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
844 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
847 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
848 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
849 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
850 -mips16 -mno-mips16 -mflip-mips16 @gol
851 -minterlink-compressed -mno-interlink-compressed @gol
852 -minterlink-mips16 -mno-interlink-mips16 @gol
853 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
854 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
855 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
856 -mno-float -msingle-float -mdouble-float @gol
857 -modd-spreg -mno-odd-spreg @gol
858 -mabs=@var{mode} -mnan=@var{encoding} @gol
859 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
862 -mvirt -mno-virt @gol
864 -mmicromips -mno-micromips @gol
866 -mfpu=@var{fpu-type} @gol
867 -msmartmips -mno-smartmips @gol
868 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
869 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
870 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
871 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
872 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
873 -membedded-data -mno-embedded-data @gol
874 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
875 -mcode-readable=@var{setting} @gol
876 -msplit-addresses -mno-split-addresses @gol
877 -mexplicit-relocs -mno-explicit-relocs @gol
878 -mcheck-zero-division -mno-check-zero-division @gol
879 -mdivide-traps -mdivide-breaks @gol
880 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
881 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
882 -mfix-24k -mno-fix-24k @gol
883 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
884 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
885 -mfix-vr4120 -mno-fix-vr4120 @gol
886 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
887 -mflush-func=@var{func} -mno-flush-func @gol
888 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
889 -mcompact-branches=@var{policy} @gol
890 -mfp-exceptions -mno-fp-exceptions @gol
891 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
892 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
893 -mframe-header-opt -mno-frame-header-opt}
896 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
897 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
898 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
899 -mno-base-addresses -msingle-exit -mno-single-exit}
901 @emph{MN10300 Options}
902 @gccoptlist{-mmult-bug -mno-mult-bug @gol
903 -mno-am33 -mam33 -mam33-2 -mam34 @gol
904 -mtune=@var{cpu-type} @gol
905 -mreturn-pointer-on-d0 @gol
906 -mno-crt0 -mrelax -mliw -msetlb}
909 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
911 @emph{MSP430 Options}
912 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
914 -mcode-region= -mdata-region= @gol
915 -msilicon-errata= -msilicon-errata-warn= @gol
919 @gccoptlist{-mbig-endian -mlittle-endian @gol
920 -mreduced-regs -mfull-regs @gol
921 -mcmov -mno-cmov @gol
922 -mperf-ext -mno-perf-ext @gol
923 -mv3push -mno-v3push @gol
924 -m16bit -mno-16bit @gol
925 -misr-vector-size=@var{num} @gol
926 -mcache-block-size=@var{num} @gol
927 -march=@var{arch} @gol
928 -mcmodel=@var{code-model} @gol
931 @emph{Nios II Options}
932 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
934 -mno-bypass-cache -mbypass-cache @gol
935 -mno-cache-volatile -mcache-volatile @gol
936 -mno-fast-sw-div -mfast-sw-div @gol
937 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
938 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
939 -mcustom-fpu-cfg=@var{name} @gol
940 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
941 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
943 @emph{Nvidia PTX Options}
944 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
946 @emph{PDP-11 Options}
947 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
948 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
949 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
950 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
951 -mbranch-expensive -mbranch-cheap @gol
952 -munix-asm -mdec-asm}
954 @emph{picoChip Options}
955 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
956 -msymbol-as-address -mno-inefficient-warnings}
958 @emph{PowerPC Options}
959 See RS/6000 and PowerPC Options.
962 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
963 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
964 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
966 @emph{RS/6000 and PowerPC Options}
967 @gccoptlist{-mcpu=@var{cpu-type} @gol
968 -mtune=@var{cpu-type} @gol
969 -mcmodel=@var{code-model} @gol
971 -maltivec -mno-altivec @gol
972 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
973 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
974 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
975 -mfprnd -mno-fprnd @gol
976 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
977 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
978 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
979 -malign-power -malign-natural @gol
980 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
981 -msingle-float -mdouble-float -msimple-fpu @gol
982 -mstring -mno-string -mupdate -mno-update @gol
983 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
984 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
985 -mstrict-align -mno-strict-align -mrelocatable @gol
986 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
987 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
988 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
989 -mprioritize-restricted-insns=@var{priority} @gol
990 -msched-costly-dep=@var{dependence_type} @gol
991 -minsert-sched-nops=@var{scheme} @gol
992 -mcall-sysv -mcall-netbsd @gol
993 -maix-struct-return -msvr4-struct-return @gol
994 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
995 -mblock-move-inline-limit=@var{num} @gol
996 -misel -mno-isel @gol
997 -misel=yes -misel=no @gol
999 -mspe=yes -mspe=no @gol
1001 -mgen-cell-microcode -mwarn-cell-microcode @gol
1002 -mvrsave -mno-vrsave @gol
1003 -mmulhw -mno-mulhw @gol
1004 -mdlmzb -mno-dlmzb @gol
1005 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1006 -mprototype -mno-prototype @gol
1007 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1008 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
1009 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1010 -mno-recip-precision @gol
1011 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1012 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1013 -msave-toc-indirect -mno-save-toc-indirect @gol
1014 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1015 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1016 -mquad-memory -mno-quad-memory @gol
1017 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1018 -mcompat-align-parm -mno-compat-align-parm @gol
1019 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1020 -mupper-regs-di -mno-upper-regs-di @gol
1021 -mupper-regs -mno-upper-regs @gol
1022 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1023 -mgnu-attribute -mno-gnu-attribute @gol
1027 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1029 -mbig-endian-data -mlittle-endian-data @gol
1032 -mas100-syntax -mno-as100-syntax@gol
1034 -mmax-constant-size=@gol
1037 -mallow-string-insns -mno-allow-string-insns@gol
1039 -mno-warn-multiple-fast-interrupts@gol
1040 -msave-acc-in-interrupts}
1042 @emph{S/390 and zSeries Options}
1043 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1044 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1045 -mlong-double-64 -mlong-double-128 @gol
1046 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1047 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1048 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1049 -mhtm -mvx -mzvector @gol
1050 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1051 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1052 -mhotpatch=@var{halfwords},@var{halfwords}}
1054 @emph{Score Options}
1055 @gccoptlist{-meb -mel @gol
1059 -mscore5 -mscore5u -mscore7 -mscore7d}
1062 @gccoptlist{-m1 -m2 -m2e @gol
1063 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1065 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1066 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1067 -mb -ml -mdalign -mrelax @gol
1068 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1069 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1070 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1071 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1072 -maccumulate-outgoing-args @gol
1073 -matomic-model=@var{atomic-model} @gol
1074 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1075 -mcbranch-force-delay-slot @gol
1076 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1077 -mpretend-cmove -mtas}
1079 @emph{Solaris 2 Options}
1080 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1083 @emph{SPARC Options}
1084 @gccoptlist{-mcpu=@var{cpu-type} @gol
1085 -mtune=@var{cpu-type} @gol
1086 -mcmodel=@var{code-model} @gol
1087 -mmemory-model=@var{mem-model} @gol
1088 -m32 -m64 -mapp-regs -mno-app-regs @gol
1089 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1090 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1091 -mhard-quad-float -msoft-quad-float @gol
1092 -mstack-bias -mno-stack-bias @gol
1093 -mstd-struct-return -mno-std-struct-return @gol
1094 -munaligned-doubles -mno-unaligned-doubles @gol
1095 -muser-mode -mno-user-mode @gol
1096 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1097 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1098 -mcbcond -mno-cbcond -mfmaf -mno-fmaf @gol
1099 -mpopc -mno-popc -msubxc -mno-subxc@gol
1100 -mfix-at697f -mfix-ut699}
1103 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1104 -msafe-dma -munsafe-dma @gol
1106 -msmall-mem -mlarge-mem -mstdmain @gol
1107 -mfixed-range=@var{register-range} @gol
1109 -maddress-space-conversion -mno-address-space-conversion @gol
1110 -mcache-size=@var{cache-size} @gol
1111 -matomic-updates -mno-atomic-updates}
1113 @emph{System V Options}
1114 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1116 @emph{TILE-Gx Options}
1117 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1118 -mcmodel=@var{code-model}}
1120 @emph{TILEPro Options}
1121 @gccoptlist{-mcpu=@var{cpu} -m32}
1124 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1125 -mprolog-function -mno-prolog-function -mspace @gol
1126 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1127 -mapp-regs -mno-app-regs @gol
1128 -mdisable-callt -mno-disable-callt @gol
1129 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1130 -mv850e -mv850 -mv850e3v5 @gol
1141 @gccoptlist{-mg -mgnu -munix}
1143 @emph{Visium Options}
1144 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1145 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1148 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1149 -mpointer-size=@var{size}}
1151 @emph{VxWorks Options}
1152 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1153 -Xbind-lazy -Xbind-now}
1156 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1157 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1158 -mfpmath=@var{unit} @gol
1159 -masm=@var{dialect} -mno-fancy-math-387 @gol
1160 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1161 -mno-wide-multiply -mrtd -malign-double @gol
1162 -mpreferred-stack-boundary=@var{num} @gol
1163 -mincoming-stack-boundary=@var{num} @gol
1164 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1165 -mrecip -mrecip=@var{opt} @gol
1166 -mvzeroupper -mprefer-avx128 @gol
1167 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1168 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1169 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1170 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1171 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1172 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1173 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1174 -mpku -mthreads @gol
1175 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1176 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1177 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1178 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1179 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1180 -mregparm=@var{num} -msseregparm @gol
1181 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1182 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1183 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1184 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1185 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1186 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1187 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1188 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1189 -mmitigate-rop -mgeneral-regs-only}
1191 @emph{x86 Windows Options}
1192 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1193 -mnop-fun-dllimport -mthread @gol
1194 -municode -mwin32 -mwindows -fno-set-stack-executable}
1196 @emph{Xstormy16 Options}
1199 @emph{Xtensa Options}
1200 @gccoptlist{-mconst16 -mno-const16 @gol
1201 -mfused-madd -mno-fused-madd @gol
1203 -mserialize-volatile -mno-serialize-volatile @gol
1204 -mtext-section-literals -mno-text-section-literals @gol
1205 -mauto-litpools -mno-auto-litpools @gol
1206 -mtarget-align -mno-target-align @gol
1207 -mlongcalls -mno-longcalls}
1209 @emph{zSeries Options}
1210 See S/390 and zSeries Options.
1214 @node Overall Options
1215 @section Options Controlling the Kind of Output
1217 Compilation can involve up to four stages: preprocessing, compilation
1218 proper, assembly and linking, always in that order. GCC is capable of
1219 preprocessing and compiling several files either into several
1220 assembler input files, or into one assembler input file; then each
1221 assembler input file produces an object file, and linking combines all
1222 the object files (those newly compiled, and those specified as input)
1223 into an executable file.
1225 @cindex file name suffix
1226 For any given input file, the file name suffix determines what kind of
1227 compilation is done:
1231 C source code that must be preprocessed.
1234 C source code that should not be preprocessed.
1237 C++ source code that should not be preprocessed.
1240 Objective-C source code. Note that you must link with the @file{libobjc}
1241 library to make an Objective-C program work.
1244 Objective-C source code that should not be preprocessed.
1248 Objective-C++ source code. Note that you must link with the @file{libobjc}
1249 library to make an Objective-C++ program work. Note that @samp{.M} refers
1250 to a literal capital M@.
1252 @item @var{file}.mii
1253 Objective-C++ source code that should not be preprocessed.
1256 C, C++, Objective-C or Objective-C++ header file to be turned into a
1257 precompiled header (default), or C, C++ header file to be turned into an
1258 Ada spec (via the @option{-fdump-ada-spec} switch).
1261 @itemx @var{file}.cp
1262 @itemx @var{file}.cxx
1263 @itemx @var{file}.cpp
1264 @itemx @var{file}.CPP
1265 @itemx @var{file}.c++
1267 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1268 the last two letters must both be literally @samp{x}. Likewise,
1269 @samp{.C} refers to a literal capital C@.
1273 Objective-C++ source code that must be preprocessed.
1275 @item @var{file}.mii
1276 Objective-C++ source code that should not be preprocessed.
1280 @itemx @var{file}.hp
1281 @itemx @var{file}.hxx
1282 @itemx @var{file}.hpp
1283 @itemx @var{file}.HPP
1284 @itemx @var{file}.h++
1285 @itemx @var{file}.tcc
1286 C++ header file to be turned into a precompiled header or Ada spec.
1289 @itemx @var{file}.for
1290 @itemx @var{file}.ftn
1291 Fixed form Fortran source code that should not be preprocessed.
1294 @itemx @var{file}.FOR
1295 @itemx @var{file}.fpp
1296 @itemx @var{file}.FPP
1297 @itemx @var{file}.FTN
1298 Fixed form Fortran source code that must be preprocessed (with the traditional
1301 @item @var{file}.f90
1302 @itemx @var{file}.f95
1303 @itemx @var{file}.f03
1304 @itemx @var{file}.f08
1305 Free form Fortran source code that should not be preprocessed.
1307 @item @var{file}.F90
1308 @itemx @var{file}.F95
1309 @itemx @var{file}.F03
1310 @itemx @var{file}.F08
1311 Free form Fortran source code that must be preprocessed (with the
1312 traditional preprocessor).
1317 @c FIXME: Descriptions of Java file types.
1323 @item @var{file}.ads
1324 Ada source code file that contains a library unit declaration (a
1325 declaration of a package, subprogram, or generic, or a generic
1326 instantiation), or a library unit renaming declaration (a package,
1327 generic, or subprogram renaming declaration). Such files are also
1330 @item @var{file}.adb
1331 Ada source code file containing a library unit body (a subprogram or
1332 package body). Such files are also called @dfn{bodies}.
1334 @c GCC also knows about some suffixes for languages not yet included:
1345 @itemx @var{file}.sx
1346 Assembler code that must be preprocessed.
1349 An object file to be fed straight into linking.
1350 Any file name with no recognized suffix is treated this way.
1354 You can specify the input language explicitly with the @option{-x} option:
1357 @item -x @var{language}
1358 Specify explicitly the @var{language} for the following input files
1359 (rather than letting the compiler choose a default based on the file
1360 name suffix). This option applies to all following input files until
1361 the next @option{-x} option. Possible values for @var{language} are:
1363 c c-header cpp-output
1364 c++ c++-header c++-cpp-output
1365 objective-c objective-c-header objective-c-cpp-output
1366 objective-c++ objective-c++-header objective-c++-cpp-output
1367 assembler assembler-with-cpp
1369 f77 f77-cpp-input f95 f95-cpp-input
1375 Turn off any specification of a language, so that subsequent files are
1376 handled according to their file name suffixes (as they are if @option{-x}
1377 has not been used at all).
1380 If you only want some of the stages of compilation, you can use
1381 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1382 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1383 @command{gcc} is to stop. Note that some combinations (for example,
1384 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1389 Compile or assemble the source files, but do not link. The linking
1390 stage simply is not done. The ultimate output is in the form of an
1391 object file for each source file.
1393 By default, the object file name for a source file is made by replacing
1394 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1396 Unrecognized input files, not requiring compilation or assembly, are
1401 Stop after the stage of compilation proper; do not assemble. The output
1402 is in the form of an assembler code file for each non-assembler input
1405 By default, the assembler file name for a source file is made by
1406 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1408 Input files that don't require compilation are ignored.
1412 Stop after the preprocessing stage; do not run the compiler proper. The
1413 output is in the form of preprocessed source code, which is sent to the
1416 Input files that don't require preprocessing are ignored.
1418 @cindex output file option
1421 Place output in file @var{file}. This applies to whatever
1422 sort of output is being produced, whether it be an executable file,
1423 an object file, an assembler file or preprocessed C code.
1425 If @option{-o} is not specified, the default is to put an executable
1426 file in @file{a.out}, the object file for
1427 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1428 assembler file in @file{@var{source}.s}, a precompiled header file in
1429 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1434 Print (on standard error output) the commands executed to run the stages
1435 of compilation. Also print the version number of the compiler driver
1436 program and of the preprocessor and the compiler proper.
1440 Like @option{-v} except the commands are not executed and arguments
1441 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1442 This is useful for shell scripts to capture the driver-generated command lines.
1446 Print (on the standard output) a description of the command-line options
1447 understood by @command{gcc}. If the @option{-v} option is also specified
1448 then @option{--help} is also passed on to the various processes
1449 invoked by @command{gcc}, so that they can display the command-line options
1450 they accept. If the @option{-Wextra} option has also been specified
1451 (prior to the @option{--help} option), then command-line options that
1452 have no documentation associated with them are also displayed.
1455 @opindex target-help
1456 Print (on the standard output) a description of target-specific command-line
1457 options for each tool. For some targets extra target-specific
1458 information may also be printed.
1460 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1461 Print (on the standard output) a description of the command-line
1462 options understood by the compiler that fit into all specified classes
1463 and qualifiers. These are the supported classes:
1466 @item @samp{optimizers}
1467 Display all of the optimization options supported by the
1470 @item @samp{warnings}
1471 Display all of the options controlling warning messages
1472 produced by the compiler.
1475 Display target-specific options. Unlike the
1476 @option{--target-help} option however, target-specific options of the
1477 linker and assembler are not displayed. This is because those
1478 tools do not currently support the extended @option{--help=} syntax.
1481 Display the values recognized by the @option{--param}
1484 @item @var{language}
1485 Display the options supported for @var{language}, where
1486 @var{language} is the name of one of the languages supported in this
1490 Display the options that are common to all languages.
1493 These are the supported qualifiers:
1496 @item @samp{undocumented}
1497 Display only those options that are undocumented.
1500 Display options taking an argument that appears after an equal
1501 sign in the same continuous piece of text, such as:
1502 @samp{--help=target}.
1504 @item @samp{separate}
1505 Display options taking an argument that appears as a separate word
1506 following the original option, such as: @samp{-o output-file}.
1509 Thus for example to display all the undocumented target-specific
1510 switches supported by the compiler, use:
1513 --help=target,undocumented
1516 The sense of a qualifier can be inverted by prefixing it with the
1517 @samp{^} character, so for example to display all binary warning
1518 options (i.e., ones that are either on or off and that do not take an
1519 argument) that have a description, use:
1522 --help=warnings,^joined,^undocumented
1525 The argument to @option{--help=} should not consist solely of inverted
1528 Combining several classes is possible, although this usually
1529 restricts the output so much that there is nothing to display. One
1530 case where it does work, however, is when one of the classes is
1531 @var{target}. For example, to display all the target-specific
1532 optimization options, use:
1535 --help=target,optimizers
1538 The @option{--help=} option can be repeated on the command line. Each
1539 successive use displays its requested class of options, skipping
1540 those that have already been displayed.
1542 If the @option{-Q} option appears on the command line before the
1543 @option{--help=} option, then the descriptive text displayed by
1544 @option{--help=} is changed. Instead of describing the displayed
1545 options, an indication is given as to whether the option is enabled,
1546 disabled or set to a specific value (assuming that the compiler
1547 knows this at the point where the @option{--help=} option is used).
1549 Here is a truncated example from the ARM port of @command{gcc}:
1552 % gcc -Q -mabi=2 --help=target -c
1553 The following options are target specific:
1555 -mabort-on-noreturn [disabled]
1559 The output is sensitive to the effects of previous command-line
1560 options, so for example it is possible to find out which optimizations
1561 are enabled at @option{-O2} by using:
1564 -Q -O2 --help=optimizers
1567 Alternatively you can discover which binary optimizations are enabled
1568 by @option{-O3} by using:
1571 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1572 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1573 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1578 Display the version number and copyrights of the invoked GCC@.
1580 @item -pass-exit-codes
1581 @opindex pass-exit-codes
1582 Normally the @command{gcc} program exits with the code of 1 if any
1583 phase of the compiler returns a non-success return code. If you specify
1584 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1585 the numerically highest error produced by any phase returning an error
1586 indication. The C, C++, and Fortran front ends return 4 if an internal
1587 compiler error is encountered.
1591 Use pipes rather than temporary files for communication between the
1592 various stages of compilation. This fails to work on some systems where
1593 the assembler is unable to read from a pipe; but the GNU assembler has
1596 @item -specs=@var{file}
1598 Process @var{file} after the compiler reads in the standard @file{specs}
1599 file, in order to override the defaults which the @command{gcc} driver
1600 program uses when determining what switches to pass to @command{cc1},
1601 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1602 @option{-specs=@var{file}} can be specified on the command line, and they
1603 are processed in order, from left to right. @xref{Spec Files}, for
1604 information about the format of the @var{file}.
1608 Invoke all subcommands under a wrapper program. The name of the
1609 wrapper program and its parameters are passed as a comma separated
1613 gcc -c t.c -wrapper gdb,--args
1617 This invokes all subprograms of @command{gcc} under
1618 @samp{gdb --args}, thus the invocation of @command{cc1} is
1619 @samp{gdb --args cc1 @dots{}}.
1621 @item -fplugin=@var{name}.so
1623 Load the plugin code in file @var{name}.so, assumed to be a
1624 shared object to be dlopen'd by the compiler. The base name of
1625 the shared object file is used to identify the plugin for the
1626 purposes of argument parsing (See
1627 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1628 Each plugin should define the callback functions specified in the
1631 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1632 @opindex fplugin-arg
1633 Define an argument called @var{key} with a value of @var{value}
1634 for the plugin called @var{name}.
1636 @item -fdump-ada-spec@r{[}-slim@r{]}
1637 @opindex fdump-ada-spec
1638 For C and C++ source and include files, generate corresponding Ada specs.
1639 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1640 GNAT User's Guide}, which provides detailed documentation on this feature.
1642 @item -fada-spec-parent=@var{unit}
1643 @opindex fada-spec-parent
1644 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1645 Ada specs as child units of parent @var{unit}.
1647 @item -fdump-go-spec=@var{file}
1648 @opindex fdump-go-spec
1649 For input files in any language, generate corresponding Go
1650 declarations in @var{file}. This generates Go @code{const},
1651 @code{type}, @code{var}, and @code{func} declarations which may be a
1652 useful way to start writing a Go interface to code written in some
1655 @include @value{srcdir}/../libiberty/at-file.texi
1659 @section Compiling C++ Programs
1661 @cindex suffixes for C++ source
1662 @cindex C++ source file suffixes
1663 C++ source files conventionally use one of the suffixes @samp{.C},
1664 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1665 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1666 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1667 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1668 files with these names and compiles them as C++ programs even if you
1669 call the compiler the same way as for compiling C programs (usually
1670 with the name @command{gcc}).
1674 However, the use of @command{gcc} does not add the C++ library.
1675 @command{g++} is a program that calls GCC and automatically specifies linking
1676 against the C++ library. It treats @samp{.c},
1677 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1678 files unless @option{-x} is used. This program is also useful when
1679 precompiling a C header file with a @samp{.h} extension for use in C++
1680 compilations. On many systems, @command{g++} is also installed with
1681 the name @command{c++}.
1683 @cindex invoking @command{g++}
1684 When you compile C++ programs, you may specify many of the same
1685 command-line options that you use for compiling programs in any
1686 language; or command-line options meaningful for C and related
1687 languages; or options that are meaningful only for C++ programs.
1688 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1689 explanations of options for languages related to C@.
1690 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1691 explanations of options that are meaningful only for C++ programs.
1693 @node C Dialect Options
1694 @section Options Controlling C Dialect
1695 @cindex dialect options
1696 @cindex language dialect options
1697 @cindex options, dialect
1699 The following options control the dialect of C (or languages derived
1700 from C, such as C++, Objective-C and Objective-C++) that the compiler
1704 @cindex ANSI support
1708 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1709 equivalent to @option{-std=c++98}.
1711 This turns off certain features of GCC that are incompatible with ISO
1712 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1713 such as the @code{asm} and @code{typeof} keywords, and
1714 predefined macros such as @code{unix} and @code{vax} that identify the
1715 type of system you are using. It also enables the undesirable and
1716 rarely used ISO trigraph feature. For the C compiler,
1717 it disables recognition of C++ style @samp{//} comments as well as
1718 the @code{inline} keyword.
1720 The alternate keywords @code{__asm__}, @code{__extension__},
1721 @code{__inline__} and @code{__typeof__} continue to work despite
1722 @option{-ansi}. You would not want to use them in an ISO C program, of
1723 course, but it is useful to put them in header files that might be included
1724 in compilations done with @option{-ansi}. Alternate predefined macros
1725 such as @code{__unix__} and @code{__vax__} are also available, with or
1726 without @option{-ansi}.
1728 The @option{-ansi} option does not cause non-ISO programs to be
1729 rejected gratuitously. For that, @option{-Wpedantic} is required in
1730 addition to @option{-ansi}. @xref{Warning Options}.
1732 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1733 option is used. Some header files may notice this macro and refrain
1734 from declaring certain functions or defining certain macros that the
1735 ISO standard doesn't call for; this is to avoid interfering with any
1736 programs that might use these names for other things.
1738 Functions that are normally built in but do not have semantics
1739 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1740 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1741 built-in functions provided by GCC}, for details of the functions
1746 Determine the language standard. @xref{Standards,,Language Standards
1747 Supported by GCC}, for details of these standard versions. This option
1748 is currently only supported when compiling C or C++.
1750 The compiler can accept several base standards, such as @samp{c90} or
1751 @samp{c++98}, and GNU dialects of those standards, such as
1752 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1753 compiler accepts all programs following that standard plus those
1754 using GNU extensions that do not contradict it. For example,
1755 @option{-std=c90} turns off certain features of GCC that are
1756 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1757 keywords, but not other GNU extensions that do not have a meaning in
1758 ISO C90, such as omitting the middle term of a @code{?:}
1759 expression. On the other hand, when a GNU dialect of a standard is
1760 specified, all features supported by the compiler are enabled, even when
1761 those features change the meaning of the base standard. As a result, some
1762 strict-conforming programs may be rejected. The particular standard
1763 is used by @option{-Wpedantic} to identify which features are GNU
1764 extensions given that version of the standard. For example
1765 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1766 comments, while @option{-std=gnu99 -Wpedantic} does not.
1768 A value for this option must be provided; possible values are
1774 Support all ISO C90 programs (certain GNU extensions that conflict
1775 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1777 @item iso9899:199409
1778 ISO C90 as modified in amendment 1.
1784 ISO C99. This standard is substantially completely supported, modulo
1785 bugs and floating-point issues
1786 (mainly but not entirely relating to optional C99 features from
1787 Annexes F and G). See
1788 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1789 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1794 ISO C11, the 2011 revision of the ISO C standard. This standard is
1795 substantially completely supported, modulo bugs, floating-point issues
1796 (mainly but not entirely relating to optional C11 features from
1797 Annexes F and G) and the optional Annexes K (Bounds-checking
1798 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1802 GNU dialect of ISO C90 (including some C99 features).
1806 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1810 GNU dialect of ISO C11. This is the default for C code.
1811 The name @samp{gnu1x} is deprecated.
1815 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1816 additional defect reports. Same as @option{-ansi} for C++ code.
1820 GNU dialect of @option{-std=c++98}.
1824 The 2011 ISO C++ standard plus amendments.
1825 The name @samp{c++0x} is deprecated.
1829 GNU dialect of @option{-std=c++11}.
1830 The name @samp{gnu++0x} is deprecated.
1834 The 2014 ISO C++ standard plus amendments.
1835 The name @samp{c++1y} is deprecated.
1839 GNU dialect of @option{-std=c++14}.
1840 This is the default for C++ code.
1841 The name @samp{gnu++1y} is deprecated.
1844 The next revision of the ISO C++ standard, tentatively planned for
1845 2017. Support is highly experimental, and will almost certainly
1846 change in incompatible ways in future releases.
1849 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1850 and will almost certainly change in incompatible ways in future
1854 @item -fgnu89-inline
1855 @opindex fgnu89-inline
1856 The option @option{-fgnu89-inline} tells GCC to use the traditional
1857 GNU semantics for @code{inline} functions when in C99 mode.
1858 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1859 Using this option is roughly equivalent to adding the
1860 @code{gnu_inline} function attribute to all inline functions
1861 (@pxref{Function Attributes}).
1863 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1864 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1865 specifies the default behavior).
1866 This option is not supported in @option{-std=c90} or
1867 @option{-std=gnu90} mode.
1869 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1870 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1871 in effect for @code{inline} functions. @xref{Common Predefined
1872 Macros,,,cpp,The C Preprocessor}.
1874 @item -aux-info @var{filename}
1876 Output to the given filename prototyped declarations for all functions
1877 declared and/or defined in a translation unit, including those in header
1878 files. This option is silently ignored in any language other than C@.
1880 Besides declarations, the file indicates, in comments, the origin of
1881 each declaration (source file and line), whether the declaration was
1882 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1883 @samp{O} for old, respectively, in the first character after the line
1884 number and the colon), and whether it came from a declaration or a
1885 definition (@samp{C} or @samp{F}, respectively, in the following
1886 character). In the case of function definitions, a K&R-style list of
1887 arguments followed by their declarations is also provided, inside
1888 comments, after the declaration.
1890 @item -fallow-parameterless-variadic-functions
1891 @opindex fallow-parameterless-variadic-functions
1892 Accept variadic functions without named parameters.
1894 Although it is possible to define such a function, this is not very
1895 useful as it is not possible to read the arguments. This is only
1896 supported for C as this construct is allowed by C++.
1900 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1901 keyword, so that code can use these words as identifiers. You can use
1902 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1903 instead. @option{-ansi} implies @option{-fno-asm}.
1905 In C++, this switch only affects the @code{typeof} keyword, since
1906 @code{asm} and @code{inline} are standard keywords. You may want to
1907 use the @option{-fno-gnu-keywords} flag instead, which has the same
1908 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1909 switch only affects the @code{asm} and @code{typeof} keywords, since
1910 @code{inline} is a standard keyword in ISO C99.
1913 @itemx -fno-builtin-@var{function}
1914 @opindex fno-builtin
1915 @cindex built-in functions
1916 Don't recognize built-in functions that do not begin with
1917 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1918 functions provided by GCC}, for details of the functions affected,
1919 including those which are not built-in functions when @option{-ansi} or
1920 @option{-std} options for strict ISO C conformance are used because they
1921 do not have an ISO standard meaning.
1923 GCC normally generates special code to handle certain built-in functions
1924 more efficiently; for instance, calls to @code{alloca} may become single
1925 instructions which adjust the stack directly, and calls to @code{memcpy}
1926 may become inline copy loops. The resulting code is often both smaller
1927 and faster, but since the function calls no longer appear as such, you
1928 cannot set a breakpoint on those calls, nor can you change the behavior
1929 of the functions by linking with a different library. In addition,
1930 when a function is recognized as a built-in function, GCC may use
1931 information about that function to warn about problems with calls to
1932 that function, or to generate more efficient code, even if the
1933 resulting code still contains calls to that function. For example,
1934 warnings are given with @option{-Wformat} for bad calls to
1935 @code{printf} when @code{printf} is built in and @code{strlen} is
1936 known not to modify global memory.
1938 With the @option{-fno-builtin-@var{function}} option
1939 only the built-in function @var{function} is
1940 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1941 function is named that is not built-in in this version of GCC, this
1942 option is ignored. There is no corresponding
1943 @option{-fbuiltin-@var{function}} option; if you wish to enable
1944 built-in functions selectively when using @option{-fno-builtin} or
1945 @option{-ffreestanding}, you may define macros such as:
1948 #define abs(n) __builtin_abs ((n))
1949 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1954 @cindex hosted environment
1956 Assert that compilation targets a hosted environment. This implies
1957 @option{-fbuiltin}. A hosted environment is one in which the
1958 entire standard library is available, and in which @code{main} has a return
1959 type of @code{int}. Examples are nearly everything except a kernel.
1960 This is equivalent to @option{-fno-freestanding}.
1962 @item -ffreestanding
1963 @opindex ffreestanding
1964 @cindex hosted environment
1966 Assert that compilation targets a freestanding environment. This
1967 implies @option{-fno-builtin}. A freestanding environment
1968 is one in which the standard library may not exist, and program startup may
1969 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1970 This is equivalent to @option{-fno-hosted}.
1972 @xref{Standards,,Language Standards Supported by GCC}, for details of
1973 freestanding and hosted environments.
1977 @cindex OpenACC accelerator programming
1978 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1979 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1980 compiler generates accelerated code according to the OpenACC Application
1981 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1982 implies @option{-pthread}, and thus is only supported on targets that
1983 have support for @option{-pthread}.
1985 @item -fopenacc-dim=@var{geom}
1986 @opindex fopenacc-dim
1987 @cindex OpenACC accelerator programming
1988 Specify default compute dimensions for parallel offload regions that do
1989 not explicitly specify. The @var{geom} value is a triple of
1990 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
1991 can be omitted, to use a target-specific default value.
1995 @cindex OpenMP parallel
1996 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1997 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1998 compiler generates parallel code according to the OpenMP Application
1999 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
2000 implies @option{-pthread}, and thus is only supported on targets that
2001 have support for @option{-pthread}. @option{-fopenmp} implies
2002 @option{-fopenmp-simd}.
2005 @opindex fopenmp-simd
2008 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2009 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2014 @cindex Enable Cilk Plus
2015 Enable the usage of Cilk Plus language extension features for C/C++.
2016 When the option @option{-fcilkplus} is specified, enable the usage of
2017 the Cilk Plus Language extension features for C/C++. The present
2018 implementation follows ABI version 1.2. This is an experimental
2019 feature that is only partially complete, and whose interface may
2020 change in future versions of GCC as the official specification
2021 changes. Currently, all features but @code{_Cilk_for} have been
2026 When the option @option{-fgnu-tm} is specified, the compiler
2027 generates code for the Linux variant of Intel's current Transactional
2028 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2029 an experimental feature whose interface may change in future versions
2030 of GCC, as the official specification changes. Please note that not
2031 all architectures are supported for this feature.
2033 For more information on GCC's support for transactional memory,
2034 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2035 Transactional Memory Library}.
2037 Note that the transactional memory feature is not supported with
2038 non-call exceptions (@option{-fnon-call-exceptions}).
2040 @item -fms-extensions
2041 @opindex fms-extensions
2042 Accept some non-standard constructs used in Microsoft header files.
2044 In C++ code, this allows member names in structures to be similar
2045 to previous types declarations.
2054 Some cases of unnamed fields in structures and unions are only
2055 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2056 fields within structs/unions}, for details.
2058 Note that this option is off for all targets but x86
2059 targets using ms-abi.
2061 @item -fplan9-extensions
2062 @opindex fplan9-extensions
2063 Accept some non-standard constructs used in Plan 9 code.
2065 This enables @option{-fms-extensions}, permits passing pointers to
2066 structures with anonymous fields to functions that expect pointers to
2067 elements of the type of the field, and permits referring to anonymous
2068 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2069 struct/union fields within structs/unions}, for details. This is only
2070 supported for C, not C++.
2074 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2075 options for strict ISO C conformance) implies @option{-trigraphs}.
2077 @cindex traditional C language
2078 @cindex C language, traditional
2080 @itemx -traditional-cpp
2081 @opindex traditional-cpp
2082 @opindex traditional
2083 Formerly, these options caused GCC to attempt to emulate a pre-standard
2084 C compiler. They are now only supported with the @option{-E} switch.
2085 The preprocessor continues to support a pre-standard mode. See the GNU
2086 CPP manual for details.
2088 @item -fcond-mismatch
2089 @opindex fcond-mismatch
2090 Allow conditional expressions with mismatched types in the second and
2091 third arguments. The value of such an expression is void. This option
2092 is not supported for C++.
2094 @item -flax-vector-conversions
2095 @opindex flax-vector-conversions
2096 Allow implicit conversions between vectors with differing numbers of
2097 elements and/or incompatible element types. This option should not be
2100 @item -funsigned-char
2101 @opindex funsigned-char
2102 Let the type @code{char} be unsigned, like @code{unsigned char}.
2104 Each kind of machine has a default for what @code{char} should
2105 be. It is either like @code{unsigned char} by default or like
2106 @code{signed char} by default.
2108 Ideally, a portable program should always use @code{signed char} or
2109 @code{unsigned char} when it depends on the signedness of an object.
2110 But many programs have been written to use plain @code{char} and
2111 expect it to be signed, or expect it to be unsigned, depending on the
2112 machines they were written for. This option, and its inverse, let you
2113 make such a program work with the opposite default.
2115 The type @code{char} is always a distinct type from each of
2116 @code{signed char} or @code{unsigned char}, even though its behavior
2117 is always just like one of those two.
2120 @opindex fsigned-char
2121 Let the type @code{char} be signed, like @code{signed char}.
2123 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2124 the negative form of @option{-funsigned-char}. Likewise, the option
2125 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2127 @item -fsigned-bitfields
2128 @itemx -funsigned-bitfields
2129 @itemx -fno-signed-bitfields
2130 @itemx -fno-unsigned-bitfields
2131 @opindex fsigned-bitfields
2132 @opindex funsigned-bitfields
2133 @opindex fno-signed-bitfields
2134 @opindex fno-unsigned-bitfields
2135 These options control whether a bit-field is signed or unsigned, when the
2136 declaration does not use either @code{signed} or @code{unsigned}. By
2137 default, such a bit-field is signed, because this is consistent: the
2138 basic integer types such as @code{int} are signed types.
2140 @item -fsso-struct=@var{endianness}
2141 @opindex fsso-struct
2142 Set the default scalar storage order of structures and unions to the
2143 specified endianness. The accepted values are @samp{big-endian} and
2144 @samp{little-endian}. If the option is not passed, the compiler uses
2145 the native endianness of the target. This option is not supported for C++.
2147 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2148 code that is not binary compatible with code generated without it if the
2149 specified endianness is not the native endianness of the target.
2152 @node C++ Dialect Options
2153 @section Options Controlling C++ Dialect
2155 @cindex compiler options, C++
2156 @cindex C++ options, command-line
2157 @cindex options, C++
2158 This section describes the command-line options that are only meaningful
2159 for C++ programs. You can also use most of the GNU compiler options
2160 regardless of what language your program is in. For example, you
2161 might compile a file @file{firstClass.C} like this:
2164 g++ -g -fstrict-enums -O -c firstClass.C
2168 In this example, only @option{-fstrict-enums} is an option meant
2169 only for C++ programs; you can use the other options with any
2170 language supported by GCC@.
2172 Some options for compiling C programs, such as @option{-std}, are also
2173 relevant for C++ programs.
2174 @xref{C Dialect Options,,Options Controlling C Dialect}.
2176 Here is a list of options that are @emph{only} for compiling C++ programs:
2180 @item -fabi-version=@var{n}
2181 @opindex fabi-version
2182 Use version @var{n} of the C++ ABI@. The default is version 0.
2184 Version 0 refers to the version conforming most closely to
2185 the C++ ABI specification. Therefore, the ABI obtained using version 0
2186 will change in different versions of G++ as ABI bugs are fixed.
2188 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2190 Version 2 is the version of the C++ ABI that first appeared in G++
2191 3.4, and was the default through G++ 4.9.
2193 Version 3 corrects an error in mangling a constant address as a
2196 Version 4, which first appeared in G++ 4.5, implements a standard
2197 mangling for vector types.
2199 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2200 attribute const/volatile on function pointer types, decltype of a
2201 plain decl, and use of a function parameter in the declaration of
2204 Version 6, which first appeared in G++ 4.7, corrects the promotion
2205 behavior of C++11 scoped enums and the mangling of template argument
2206 packs, const/static_cast, prefix ++ and --, and a class scope function
2207 used as a template argument.
2209 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2210 builtin type and corrects the mangling of lambdas in default argument
2213 Version 8, which first appeared in G++ 4.9, corrects the substitution
2214 behavior of function types with function-cv-qualifiers.
2216 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2219 Version 10, which first appeared in G++ 6.1, adds mangling of
2220 attributes that affect type identity, such as ia32 calling convention
2221 attributes (e.g. @samp{stdcall}).
2223 See also @option{-Wabi}.
2225 @item -fabi-compat-version=@var{n}
2226 @opindex fabi-compat-version
2227 On targets that support strong aliases, G++
2228 works around mangling changes by creating an alias with the correct
2229 mangled name when defining a symbol with an incorrect mangled name.
2230 This switch specifies which ABI version to use for the alias.
2232 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2233 compatibility). If another ABI version is explicitly selected, this
2234 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2235 use @option{-fabi-compat-version=2}.
2237 If this option is not provided but @option{-Wabi=@var{n}} is, that
2238 version is used for compatibility aliases. If this option is provided
2239 along with @option{-Wabi} (without the version), the version from this
2240 option is used for the warning.
2242 @item -fno-access-control
2243 @opindex fno-access-control
2244 Turn off all access checking. This switch is mainly useful for working
2245 around bugs in the access control code.
2248 @opindex faligned-new
2249 Enable support for C++17 @code{new} of types that require more
2250 alignment than @code{void* ::operator new(std::size_t)} provides. A
2251 numeric argument such as @code{-faligned-new=32} can be used to
2252 specify how much alignment (in bytes) is provided by that function,
2253 but few users will need to override the default of
2254 @code{alignof(std::max_align_t)}.
2258 Check that the pointer returned by @code{operator new} is non-null
2259 before attempting to modify the storage allocated. This check is
2260 normally unnecessary because the C++ standard specifies that
2261 @code{operator new} only returns @code{0} if it is declared
2262 @code{throw()}, in which case the compiler always checks the
2263 return value even without this option. In all other cases, when
2264 @code{operator new} has a non-empty exception specification, memory
2265 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2266 @samp{new (nothrow)}.
2270 Enable support for the C++ Extensions for Concepts Technical
2271 Specification, ISO 19217 (2015), which allows code like
2274 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2275 template <Addable T> T add (T a, T b) @{ return a + b; @}
2278 @item -fconstexpr-depth=@var{n}
2279 @opindex fconstexpr-depth
2280 Set the maximum nested evaluation depth for C++11 constexpr functions
2281 to @var{n}. A limit is needed to detect endless recursion during
2282 constant expression evaluation. The minimum specified by the standard
2285 @item -fconstexpr-loop-limit=@var{n}
2286 @opindex fconstexpr-loop-limit
2287 Set the maximum number of iterations for a loop in C++14 constexpr functions
2288 to @var{n}. A limit is needed to detect infinite loops during
2289 constant expression evaluation. The default is 262144 (1<<18).
2291 @item -fdeduce-init-list
2292 @opindex fdeduce-init-list
2293 Enable deduction of a template type parameter as
2294 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2297 template <class T> auto forward(T t) -> decltype (realfn (t))
2304 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2308 This deduction was implemented as a possible extension to the
2309 originally proposed semantics for the C++11 standard, but was not part
2310 of the final standard, so it is disabled by default. This option is
2311 deprecated, and may be removed in a future version of G++.
2313 @item -ffriend-injection
2314 @opindex ffriend-injection
2315 Inject friend functions into the enclosing namespace, so that they are
2316 visible outside the scope of the class in which they are declared.
2317 Friend functions were documented to work this way in the old Annotated
2318 C++ Reference Manual.
2319 However, in ISO C++ a friend function that is not declared
2320 in an enclosing scope can only be found using argument dependent
2321 lookup. GCC defaults to the standard behavior.
2323 This option is for compatibility, and may be removed in a future
2326 @item -fno-elide-constructors
2327 @opindex fno-elide-constructors
2328 The C++ standard allows an implementation to omit creating a temporary
2329 that is only used to initialize another object of the same type.
2330 Specifying this option disables that optimization, and forces G++ to
2331 call the copy constructor in all cases. This option also causes G++
2332 to call trivial member functions which otherwise would be expanded inline.
2334 In C++17, the compiler is required to omit these temporaries, but this
2335 option still affects trivial member functions.
2337 @item -fno-enforce-eh-specs
2338 @opindex fno-enforce-eh-specs
2339 Don't generate code to check for violation of exception specifications
2340 at run time. This option violates the C++ standard, but may be useful
2341 for reducing code size in production builds, much like defining
2342 @code{NDEBUG}. This does not give user code permission to throw
2343 exceptions in violation of the exception specifications; the compiler
2344 still optimizes based on the specifications, so throwing an
2345 unexpected exception results in undefined behavior at run time.
2347 @item -fextern-tls-init
2348 @itemx -fno-extern-tls-init
2349 @opindex fextern-tls-init
2350 @opindex fno-extern-tls-init
2351 The C++11 and OpenMP standards allow @code{thread_local} and
2352 @code{threadprivate} variables to have dynamic (runtime)
2353 initialization. To support this, any use of such a variable goes
2354 through a wrapper function that performs any necessary initialization.
2355 When the use and definition of the variable are in the same
2356 translation unit, this overhead can be optimized away, but when the
2357 use is in a different translation unit there is significant overhead
2358 even if the variable doesn't actually need dynamic initialization. If
2359 the programmer can be sure that no use of the variable in a
2360 non-defining TU needs to trigger dynamic initialization (either
2361 because the variable is statically initialized, or a use of the
2362 variable in the defining TU will be executed before any uses in
2363 another TU), they can avoid this overhead with the
2364 @option{-fno-extern-tls-init} option.
2366 On targets that support symbol aliases, the default is
2367 @option{-fextern-tls-init}. On targets that do not support symbol
2368 aliases, the default is @option{-fno-extern-tls-init}.
2371 @itemx -fno-for-scope
2373 @opindex fno-for-scope
2374 If @option{-ffor-scope} is specified, the scope of variables declared in
2375 a @i{for-init-statement} is limited to the @code{for} loop itself,
2376 as specified by the C++ standard.
2377 If @option{-fno-for-scope} is specified, the scope of variables declared in
2378 a @i{for-init-statement} extends to the end of the enclosing scope,
2379 as was the case in old versions of G++, and other (traditional)
2380 implementations of C++.
2382 If neither flag is given, the default is to follow the standard,
2383 but to allow and give a warning for old-style code that would
2384 otherwise be invalid, or have different behavior.
2386 @item -fno-gnu-keywords
2387 @opindex fno-gnu-keywords
2388 Do not recognize @code{typeof} as a keyword, so that code can use this
2389 word as an identifier. You can use the keyword @code{__typeof__} instead.
2390 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2391 @option{-std=c++98}, @option{-std=c++11}, etc.
2393 @item -fno-implicit-templates
2394 @opindex fno-implicit-templates
2395 Never emit code for non-inline templates that are instantiated
2396 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2397 @xref{Template Instantiation}, for more information.
2399 @item -fno-implicit-inline-templates
2400 @opindex fno-implicit-inline-templates
2401 Don't emit code for implicit instantiations of inline templates, either.
2402 The default is to handle inlines differently so that compiles with and
2403 without optimization need the same set of explicit instantiations.
2405 @item -fno-implement-inlines
2406 @opindex fno-implement-inlines
2407 To save space, do not emit out-of-line copies of inline functions
2408 controlled by @code{#pragma implementation}. This causes linker
2409 errors if these functions are not inlined everywhere they are called.
2411 @item -fms-extensions
2412 @opindex fms-extensions
2413 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2414 int and getting a pointer to member function via non-standard syntax.
2416 @item -fno-nonansi-builtins
2417 @opindex fno-nonansi-builtins
2418 Disable built-in declarations of functions that are not mandated by
2419 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2420 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2423 @opindex fnothrow-opt
2424 Treat a @code{throw()} exception specification as if it were a
2425 @code{noexcept} specification to reduce or eliminate the text size
2426 overhead relative to a function with no exception specification. If
2427 the function has local variables of types with non-trivial
2428 destructors, the exception specification actually makes the
2429 function smaller because the EH cleanups for those variables can be
2430 optimized away. The semantic effect is that an exception thrown out of
2431 a function with such an exception specification results in a call
2432 to @code{terminate} rather than @code{unexpected}.
2434 @item -fno-operator-names
2435 @opindex fno-operator-names
2436 Do not treat the operator name keywords @code{and}, @code{bitand},
2437 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2438 synonyms as keywords.
2440 @item -fno-optional-diags
2441 @opindex fno-optional-diags
2442 Disable diagnostics that the standard says a compiler does not need to
2443 issue. Currently, the only such diagnostic issued by G++ is the one for
2444 a name having multiple meanings within a class.
2447 @opindex fpermissive
2448 Downgrade some diagnostics about nonconformant code from errors to
2449 warnings. Thus, using @option{-fpermissive} allows some
2450 nonconforming code to compile.
2452 @item -fno-pretty-templates
2453 @opindex fno-pretty-templates
2454 When an error message refers to a specialization of a function
2455 template, the compiler normally prints the signature of the
2456 template followed by the template arguments and any typedefs or
2457 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2458 rather than @code{void f(int)}) so that it's clear which template is
2459 involved. When an error message refers to a specialization of a class
2460 template, the compiler omits any template arguments that match
2461 the default template arguments for that template. If either of these
2462 behaviors make it harder to understand the error message rather than
2463 easier, you can use @option{-fno-pretty-templates} to disable them.
2467 Enable automatic template instantiation at link time. This option also
2468 implies @option{-fno-implicit-templates}. @xref{Template
2469 Instantiation}, for more information.
2473 Disable generation of information about every class with virtual
2474 functions for use by the C++ run-time type identification features
2475 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2476 of the language, you can save some space by using this flag. Note that
2477 exception handling uses the same information, but G++ generates it as
2478 needed. The @code{dynamic_cast} operator can still be used for casts that
2479 do not require run-time type information, i.e.@: casts to @code{void *} or to
2480 unambiguous base classes.
2482 @item -fsized-deallocation
2483 @opindex fsized-deallocation
2484 Enable the built-in global declarations
2486 void operator delete (void *, std::size_t) noexcept;
2487 void operator delete[] (void *, std::size_t) noexcept;
2489 as introduced in C++14. This is useful for user-defined replacement
2490 deallocation functions that, for example, use the size of the object
2491 to make deallocation faster. Enabled by default under
2492 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2493 warns about places that might want to add a definition.
2495 @item -fstrict-enums
2496 @opindex fstrict-enums
2497 Allow the compiler to optimize using the assumption that a value of
2498 enumerated type can only be one of the values of the enumeration (as
2499 defined in the C++ standard; basically, a value that can be
2500 represented in the minimum number of bits needed to represent all the
2501 enumerators). This assumption may not be valid if the program uses a
2502 cast to convert an arbitrary integer value to the enumerated type.
2504 @item -fstrong-eval-order
2505 @opindex fstrong-eval-order
2506 Evaluate member access, array subscripting, and shift expressions in
2507 left-to-right order, and evaluate assignment in right-to-left order,
2508 as adopted for C++17. Enabled by default with @option{-std=c++1z}.
2509 @option{-fstrong-eval-order=some} enables just the ordering of member
2510 access and shift expressions, and is the default without
2511 @option{-std=c++1z}.
2513 @item -ftemplate-backtrace-limit=@var{n}
2514 @opindex ftemplate-backtrace-limit
2515 Set the maximum number of template instantiation notes for a single
2516 warning or error to @var{n}. The default value is 10.
2518 @item -ftemplate-depth=@var{n}
2519 @opindex ftemplate-depth
2520 Set the maximum instantiation depth for template classes to @var{n}.
2521 A limit on the template instantiation depth is needed to detect
2522 endless recursions during template class instantiation. ANSI/ISO C++
2523 conforming programs must not rely on a maximum depth greater than 17
2524 (changed to 1024 in C++11). The default value is 900, as the compiler
2525 can run out of stack space before hitting 1024 in some situations.
2527 @item -fno-threadsafe-statics
2528 @opindex fno-threadsafe-statics
2529 Do not emit the extra code to use the routines specified in the C++
2530 ABI for thread-safe initialization of local statics. You can use this
2531 option to reduce code size slightly in code that doesn't need to be
2534 @item -fuse-cxa-atexit
2535 @opindex fuse-cxa-atexit
2536 Register destructors for objects with static storage duration with the
2537 @code{__cxa_atexit} function rather than the @code{atexit} function.
2538 This option is required for fully standards-compliant handling of static
2539 destructors, but only works if your C library supports
2540 @code{__cxa_atexit}.
2542 @item -fno-use-cxa-get-exception-ptr
2543 @opindex fno-use-cxa-get-exception-ptr
2544 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2545 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2546 if the runtime routine is not available.
2548 @item -fvisibility-inlines-hidden
2549 @opindex fvisibility-inlines-hidden
2550 This switch declares that the user does not attempt to compare
2551 pointers to inline functions or methods where the addresses of the two functions
2552 are taken in different shared objects.
2554 The effect of this is that GCC may, effectively, mark inline methods with
2555 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2556 appear in the export table of a DSO and do not require a PLT indirection
2557 when used within the DSO@. Enabling this option can have a dramatic effect
2558 on load and link times of a DSO as it massively reduces the size of the
2559 dynamic export table when the library makes heavy use of templates.
2561 The behavior of this switch is not quite the same as marking the
2562 methods as hidden directly, because it does not affect static variables
2563 local to the function or cause the compiler to deduce that
2564 the function is defined in only one shared object.
2566 You may mark a method as having a visibility explicitly to negate the
2567 effect of the switch for that method. For example, if you do want to
2568 compare pointers to a particular inline method, you might mark it as
2569 having default visibility. Marking the enclosing class with explicit
2570 visibility has no effect.
2572 Explicitly instantiated inline methods are unaffected by this option
2573 as their linkage might otherwise cross a shared library boundary.
2574 @xref{Template Instantiation}.
2576 @item -fvisibility-ms-compat
2577 @opindex fvisibility-ms-compat
2578 This flag attempts to use visibility settings to make GCC's C++
2579 linkage model compatible with that of Microsoft Visual Studio.
2581 The flag makes these changes to GCC's linkage model:
2585 It sets the default visibility to @code{hidden}, like
2586 @option{-fvisibility=hidden}.
2589 Types, but not their members, are not hidden by default.
2592 The One Definition Rule is relaxed for types without explicit
2593 visibility specifications that are defined in more than one
2594 shared object: those declarations are permitted if they are
2595 permitted when this option is not used.
2598 In new code it is better to use @option{-fvisibility=hidden} and
2599 export those classes that are intended to be externally visible.
2600 Unfortunately it is possible for code to rely, perhaps accidentally,
2601 on the Visual Studio behavior.
2603 Among the consequences of these changes are that static data members
2604 of the same type with the same name but defined in different shared
2605 objects are different, so changing one does not change the other;
2606 and that pointers to function members defined in different shared
2607 objects may not compare equal. When this flag is given, it is a
2608 violation of the ODR to define types with the same name differently.
2612 Do not use weak symbol support, even if it is provided by the linker.
2613 By default, G++ uses weak symbols if they are available. This
2614 option exists only for testing, and should not be used by end-users;
2615 it results in inferior code and has no benefits. This option may
2616 be removed in a future release of G++.
2620 Do not search for header files in the standard directories specific to
2621 C++, but do still search the other standard directories. (This option
2622 is used when building the C++ library.)
2625 In addition, these optimization, warning, and code generation options
2626 have meanings only for C++ programs:
2629 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2632 Warn when G++ it generates code that is probably not compatible with
2633 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2634 ABI with each major release, normally @option{-Wabi} will warn only if
2635 there is a check added later in a release series for an ABI issue
2636 discovered since the initial release. @option{-Wabi} will warn about
2637 more things if an older ABI version is selected (with
2638 @option{-fabi-version=@var{n}}).
2640 @option{-Wabi} can also be used with an explicit version number to
2641 warn about compatibility with a particular @option{-fabi-version}
2642 level, e.g. @option{-Wabi=2} to warn about changes relative to
2643 @option{-fabi-version=2}.
2645 If an explicit version number is provided and
2646 @option{-fabi-compat-version} is not specified, the version number
2647 from this option is used for compatibility aliases. If no explicit
2648 version number is provided with this option, but
2649 @option{-fabi-compat-version} is specified, that version number is
2650 used for ABI warnings.
2652 Although an effort has been made to warn about
2653 all such cases, there are probably some cases that are not warned about,
2654 even though G++ is generating incompatible code. There may also be
2655 cases where warnings are emitted even though the code that is generated
2658 You should rewrite your code to avoid these warnings if you are
2659 concerned about the fact that code generated by G++ may not be binary
2660 compatible with code generated by other compilers.
2662 Known incompatibilities in @option{-fabi-version=2} (which was the
2663 default from GCC 3.4 to 4.9) include:
2668 A template with a non-type template parameter of reference type was
2669 mangled incorrectly:
2672 template <int &> struct S @{@};
2676 This was fixed in @option{-fabi-version=3}.
2679 SIMD vector types declared using @code{__attribute ((vector_size))} were
2680 mangled in a non-standard way that does not allow for overloading of
2681 functions taking vectors of different sizes.
2683 The mangling was changed in @option{-fabi-version=4}.
2686 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2687 qualifiers, and @code{decltype} of a plain declaration was folded away.
2689 These mangling issues were fixed in @option{-fabi-version=5}.
2692 Scoped enumerators passed as arguments to a variadic function are
2693 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2694 On most targets this does not actually affect the parameter passing
2695 ABI, as there is no way to pass an argument smaller than @code{int}.
2697 Also, the ABI changed the mangling of template argument packs,
2698 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2699 a class scope function used as a template argument.
2701 These issues were corrected in @option{-fabi-version=6}.
2704 Lambdas in default argument scope were mangled incorrectly, and the
2705 ABI changed the mangling of @code{nullptr_t}.
2707 These issues were corrected in @option{-fabi-version=7}.
2710 When mangling a function type with function-cv-qualifiers, the
2711 un-qualified function type was incorrectly treated as a substitution
2714 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2717 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2718 unaligned accesses. Note that this did not affect the ABI of a
2719 function with a @code{nullptr_t} parameter, as parameters have a
2722 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2725 Target-specific attributes that affect the identity of a type, such as
2726 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2727 did not affect the mangled name, leading to name collisions when
2728 function pointers were used as template arguments.
2730 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2734 It also warns about psABI-related changes. The known psABI changes at this
2740 For SysV/x86-64, unions with @code{long double} members are
2741 passed in memory as specified in psABI. For example:
2751 @code{union U} is always passed in memory.
2755 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2758 Warn when a type with an ABI tag is used in a context that does not
2759 have that ABI tag. See @ref{C++ Attributes} for more information
2762 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2763 @opindex Wctor-dtor-privacy
2764 @opindex Wno-ctor-dtor-privacy
2765 Warn when a class seems unusable because all the constructors or
2766 destructors in that class are private, and it has neither friends nor
2767 public static member functions. Also warn if there are no non-private
2768 methods, and there's at least one private member function that isn't
2769 a constructor or destructor.
2771 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2772 @opindex Wdelete-non-virtual-dtor
2773 @opindex Wno-delete-non-virtual-dtor
2774 Warn when @code{delete} is used to destroy an instance of a class that
2775 has virtual functions and non-virtual destructor. It is unsafe to delete
2776 an instance of a derived class through a pointer to a base class if the
2777 base class does not have a virtual destructor. This warning is enabled
2780 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2781 @opindex Wliteral-suffix
2782 @opindex Wno-literal-suffix
2783 Warn when a string or character literal is followed by a ud-suffix which does
2784 not begin with an underscore. As a conforming extension, GCC treats such
2785 suffixes as separate preprocessing tokens in order to maintain backwards
2786 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2790 #define __STDC_FORMAT_MACROS
2791 #include <inttypes.h>
2796 printf("My int64: %" PRId64"\n", i64);
2800 In this case, @code{PRId64} is treated as a separate preprocessing token.
2802 This warning is enabled by default.
2804 @item -Wlto-type-mismatch
2805 @opindex Wlto-type-mismatch
2806 @opindex Wno-lto-type-mismatch
2808 During the link-time optimization warn about type mismatches in
2809 global declarations from different compilation units.
2810 Requires @option{-flto} to be enabled. Enabled by default.
2812 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2814 @opindex Wno-narrowing
2815 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2816 conversion prohibited by C++11 occurs within
2820 int i = @{ 2.2 @}; // error: narrowing from double to int
2823 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2825 When a later standard is in effect, e.g. when using @option{-std=c++11},
2826 narrowing conversions are diagnosed by default, as required by the standard.
2827 A narrowing conversion from a constant produces an error,
2828 and a narrowing conversion from a non-constant produces a warning,
2829 but @option{-Wno-narrowing} suppresses the diagnostic.
2830 Note that this does not affect the meaning of well-formed code;
2831 narrowing conversions are still considered ill-formed in SFINAE contexts.
2833 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2835 @opindex Wno-noexcept
2836 Warn when a noexcept-expression evaluates to false because of a call
2837 to a function that does not have a non-throwing exception
2838 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2839 the compiler to never throw an exception.
2841 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2842 @opindex Wnon-virtual-dtor
2843 @opindex Wno-non-virtual-dtor
2844 Warn when a class has virtual functions and an accessible non-virtual
2845 destructor itself or in an accessible polymorphic base class, in which
2846 case it is possible but unsafe to delete an instance of a derived
2847 class through a pointer to the class itself or base class. This
2848 warning is automatically enabled if @option{-Weffc++} is specified.
2850 @item -Wregister @r{(C++ and Objective-C++ only)}
2852 @opindex Wno-register
2853 Warn on uses of the @code{register} storage class specifier, except
2854 when it is part of the GNU @ref{Explicit Register Variables} extension.
2855 The use of the @code{register} keyword as storage class specifier has
2856 been deprecated in C++11 and removed in C++17.
2857 Enabled by default with @option{-std=c++1z}.
2859 @item -Wreorder @r{(C++ and Objective-C++ only)}
2861 @opindex Wno-reorder
2862 @cindex reordering, warning
2863 @cindex warning for reordering of member initializers
2864 Warn when the order of member initializers given in the code does not
2865 match the order in which they must be executed. For instance:
2871 A(): j (0), i (1) @{ @}
2876 The compiler rearranges the member initializers for @code{i}
2877 and @code{j} to match the declaration order of the members, emitting
2878 a warning to that effect. This warning is enabled by @option{-Wall}.
2880 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2881 @opindex fext-numeric-literals
2882 @opindex fno-ext-numeric-literals
2883 Accept imaginary, fixed-point, or machine-defined
2884 literal number suffixes as GNU extensions.
2885 When this option is turned off these suffixes are treated
2886 as C++11 user-defined literal numeric suffixes.
2887 This is on by default for all pre-C++11 dialects and all GNU dialects:
2888 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2889 @option{-std=gnu++14}.
2890 This option is off by default
2891 for ISO C++11 onwards (@option{-std=c++11}, ...).
2894 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2897 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2900 Warn about violations of the following style guidelines from Scott Meyers'
2901 @cite{Effective C++} series of books:
2905 Define a copy constructor and an assignment operator for classes
2906 with dynamically-allocated memory.
2909 Prefer initialization to assignment in constructors.
2912 Have @code{operator=} return a reference to @code{*this}.
2915 Don't try to return a reference when you must return an object.
2918 Distinguish between prefix and postfix forms of increment and
2919 decrement operators.
2922 Never overload @code{&&}, @code{||}, or @code{,}.
2926 This option also enables @option{-Wnon-virtual-dtor}, which is also
2927 one of the effective C++ recommendations. However, the check is
2928 extended to warn about the lack of virtual destructor in accessible
2929 non-polymorphic bases classes too.
2931 When selecting this option, be aware that the standard library
2932 headers do not obey all of these guidelines; use @samp{grep -v}
2933 to filter out those warnings.
2935 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2936 @opindex Wstrict-null-sentinel
2937 @opindex Wno-strict-null-sentinel
2938 Warn about the use of an uncasted @code{NULL} as sentinel. When
2939 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2940 to @code{__null}. Although it is a null pointer constant rather than a
2941 null pointer, it is guaranteed to be of the same size as a pointer.
2942 But this use is not portable across different compilers.
2944 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2945 @opindex Wno-non-template-friend
2946 @opindex Wnon-template-friend
2947 Disable warnings when non-templatized friend functions are declared
2948 within a template. Since the advent of explicit template specification
2949 support in G++, if the name of the friend is an unqualified-id (i.e.,
2950 @samp{friend foo(int)}), the C++ language specification demands that the
2951 friend declare or define an ordinary, nontemplate function. (Section
2952 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2953 could be interpreted as a particular specialization of a templatized
2954 function. Because this non-conforming behavior is no longer the default
2955 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2956 check existing code for potential trouble spots and is on by default.
2957 This new compiler behavior can be turned off with
2958 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2959 but disables the helpful warning.
2961 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2962 @opindex Wold-style-cast
2963 @opindex Wno-old-style-cast
2964 Warn if an old-style (C-style) cast to a non-void type is used within
2965 a C++ program. The new-style casts (@code{dynamic_cast},
2966 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2967 less vulnerable to unintended effects and much easier to search for.
2969 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2970 @opindex Woverloaded-virtual
2971 @opindex Wno-overloaded-virtual
2972 @cindex overloaded virtual function, warning
2973 @cindex warning for overloaded virtual function
2974 Warn when a function declaration hides virtual functions from a
2975 base class. For example, in:
2982 struct B: public A @{
2987 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2998 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2999 @opindex Wno-pmf-conversions
3000 @opindex Wpmf-conversions
3001 Disable the diagnostic for converting a bound pointer to member function
3004 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3005 @opindex Wsign-promo
3006 @opindex Wno-sign-promo
3007 Warn when overload resolution chooses a promotion from unsigned or
3008 enumerated type to a signed type, over a conversion to an unsigned type of
3009 the same size. Previous versions of G++ tried to preserve
3010 unsignedness, but the standard mandates the current behavior.
3012 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3014 Warn when a primary template declaration is encountered. Some coding
3015 rules disallow templates, and this may be used to enforce that rule.
3016 The warning is inactive inside a system header file, such as the STL, so
3017 one can still use the STL. One may also instantiate or specialize
3020 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3021 @opindex Wmultiple-inheritance
3022 Warn when a class is defined with multiple direct base classes. Some
3023 coding rules disallow multiple inheritance, and this may be used to
3024 enforce that rule. The warning is inactive inside a system header file,
3025 such as the STL, so one can still use the STL. One may also define
3026 classes that indirectly use multiple inheritance.
3028 @item -Wvirtual-inheritance
3029 @opindex Wvirtual-inheritance
3030 Warn when a class is defined with a virtual direct base classe. Some
3031 coding rules disallow multiple inheritance, and this may be used to
3032 enforce that rule. The warning is inactive inside a system header file,
3033 such as the STL, so one can still use the STL. One may also define
3034 classes that indirectly use virtual inheritance.
3037 @opindex Wnamespaces
3038 Warn when a namespace definition is opened. Some coding rules disallow
3039 namespaces, and this may be used to enforce that rule. The warning is
3040 inactive inside a system header file, such as the STL, so one can still
3041 use the STL. One may also use using directives and qualified names.
3043 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3045 @opindex Wno-terminate
3046 Disable the warning about a throw-expression that will immediately
3047 result in a call to @code{terminate}.
3050 @node Objective-C and Objective-C++ Dialect Options
3051 @section Options Controlling Objective-C and Objective-C++ Dialects
3053 @cindex compiler options, Objective-C and Objective-C++
3054 @cindex Objective-C and Objective-C++ options, command-line
3055 @cindex options, Objective-C and Objective-C++
3056 (NOTE: This manual does not describe the Objective-C and Objective-C++
3057 languages themselves. @xref{Standards,,Language Standards
3058 Supported by GCC}, for references.)
3060 This section describes the command-line options that are only meaningful
3061 for Objective-C and Objective-C++ programs. You can also use most of
3062 the language-independent GNU compiler options.
3063 For example, you might compile a file @file{some_class.m} like this:
3066 gcc -g -fgnu-runtime -O -c some_class.m
3070 In this example, @option{-fgnu-runtime} is an option meant only for
3071 Objective-C and Objective-C++ programs; you can use the other options with
3072 any language supported by GCC@.
3074 Note that since Objective-C is an extension of the C language, Objective-C
3075 compilations may also use options specific to the C front-end (e.g.,
3076 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3077 C++-specific options (e.g., @option{-Wabi}).
3079 Here is a list of options that are @emph{only} for compiling Objective-C
3080 and Objective-C++ programs:
3083 @item -fconstant-string-class=@var{class-name}
3084 @opindex fconstant-string-class
3085 Use @var{class-name} as the name of the class to instantiate for each
3086 literal string specified with the syntax @code{@@"@dots{}"}. The default
3087 class name is @code{NXConstantString} if the GNU runtime is being used, and
3088 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3089 @option{-fconstant-cfstrings} option, if also present, overrides the
3090 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3091 to be laid out as constant CoreFoundation strings.
3094 @opindex fgnu-runtime
3095 Generate object code compatible with the standard GNU Objective-C
3096 runtime. This is the default for most types of systems.
3098 @item -fnext-runtime
3099 @opindex fnext-runtime
3100 Generate output compatible with the NeXT runtime. This is the default
3101 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3102 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3105 @item -fno-nil-receivers
3106 @opindex fno-nil-receivers
3107 Assume that all Objective-C message dispatches (@code{[receiver
3108 message:arg]}) in this translation unit ensure that the receiver is
3109 not @code{nil}. This allows for more efficient entry points in the
3110 runtime to be used. This option is only available in conjunction with
3111 the NeXT runtime and ABI version 0 or 1.
3113 @item -fobjc-abi-version=@var{n}
3114 @opindex fobjc-abi-version
3115 Use version @var{n} of the Objective-C ABI for the selected runtime.
3116 This option is currently supported only for the NeXT runtime. In that
3117 case, Version 0 is the traditional (32-bit) ABI without support for
3118 properties and other Objective-C 2.0 additions. Version 1 is the
3119 traditional (32-bit) ABI with support for properties and other
3120 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3121 nothing is specified, the default is Version 0 on 32-bit target
3122 machines, and Version 2 on 64-bit target machines.
3124 @item -fobjc-call-cxx-cdtors
3125 @opindex fobjc-call-cxx-cdtors
3126 For each Objective-C class, check if any of its instance variables is a
3127 C++ object with a non-trivial default constructor. If so, synthesize a
3128 special @code{- (id) .cxx_construct} instance method which runs
3129 non-trivial default constructors on any such instance variables, in order,
3130 and then return @code{self}. Similarly, check if any instance variable
3131 is a C++ object with a non-trivial destructor, and if so, synthesize a
3132 special @code{- (void) .cxx_destruct} method which runs
3133 all such default destructors, in reverse order.
3135 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3136 methods thusly generated only operate on instance variables
3137 declared in the current Objective-C class, and not those inherited
3138 from superclasses. It is the responsibility of the Objective-C
3139 runtime to invoke all such methods in an object's inheritance
3140 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3141 by the runtime immediately after a new object instance is allocated;
3142 the @code{- (void) .cxx_destruct} methods are invoked immediately
3143 before the runtime deallocates an object instance.
3145 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3146 support for invoking the @code{- (id) .cxx_construct} and
3147 @code{- (void) .cxx_destruct} methods.
3149 @item -fobjc-direct-dispatch
3150 @opindex fobjc-direct-dispatch
3151 Allow fast jumps to the message dispatcher. On Darwin this is
3152 accomplished via the comm page.
3154 @item -fobjc-exceptions
3155 @opindex fobjc-exceptions
3156 Enable syntactic support for structured exception handling in
3157 Objective-C, similar to what is offered by C++ and Java. This option
3158 is required to use the Objective-C keywords @code{@@try},
3159 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3160 @code{@@synchronized}. This option is available with both the GNU
3161 runtime and the NeXT runtime (but not available in conjunction with
3162 the NeXT runtime on Mac OS X 10.2 and earlier).
3166 Enable garbage collection (GC) in Objective-C and Objective-C++
3167 programs. This option is only available with the NeXT runtime; the
3168 GNU runtime has a different garbage collection implementation that
3169 does not require special compiler flags.
3171 @item -fobjc-nilcheck
3172 @opindex fobjc-nilcheck
3173 For the NeXT runtime with version 2 of the ABI, check for a nil
3174 receiver in method invocations before doing the actual method call.
3175 This is the default and can be disabled using
3176 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3177 checked for nil in this way no matter what this flag is set to.
3178 Currently this flag does nothing when the GNU runtime, or an older
3179 version of the NeXT runtime ABI, is used.
3181 @item -fobjc-std=objc1
3183 Conform to the language syntax of Objective-C 1.0, the language
3184 recognized by GCC 4.0. This only affects the Objective-C additions to
3185 the C/C++ language; it does not affect conformance to C/C++ standards,
3186 which is controlled by the separate C/C++ dialect option flags. When
3187 this option is used with the Objective-C or Objective-C++ compiler,
3188 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3189 This is useful if you need to make sure that your Objective-C code can
3190 be compiled with older versions of GCC@.
3192 @item -freplace-objc-classes
3193 @opindex freplace-objc-classes
3194 Emit a special marker instructing @command{ld(1)} not to statically link in
3195 the resulting object file, and allow @command{dyld(1)} to load it in at
3196 run time instead. This is used in conjunction with the Fix-and-Continue
3197 debugging mode, where the object file in question may be recompiled and
3198 dynamically reloaded in the course of program execution, without the need
3199 to restart the program itself. Currently, Fix-and-Continue functionality
3200 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3205 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3206 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3207 compile time) with static class references that get initialized at load time,
3208 which improves run-time performance. Specifying the @option{-fzero-link} flag
3209 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3210 to be retained. This is useful in Zero-Link debugging mode, since it allows
3211 for individual class implementations to be modified during program execution.
3212 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3213 regardless of command-line options.
3215 @item -fno-local-ivars
3216 @opindex fno-local-ivars
3217 @opindex flocal-ivars
3218 By default instance variables in Objective-C can be accessed as if
3219 they were local variables from within the methods of the class they're
3220 declared in. This can lead to shadowing between instance variables
3221 and other variables declared either locally inside a class method or
3222 globally with the same name. Specifying the @option{-fno-local-ivars}
3223 flag disables this behavior thus avoiding variable shadowing issues.
3225 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3226 @opindex fivar-visibility
3227 Set the default instance variable visibility to the specified option
3228 so that instance variables declared outside the scope of any access
3229 modifier directives default to the specified visibility.
3233 Dump interface declarations for all classes seen in the source file to a
3234 file named @file{@var{sourcename}.decl}.
3236 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3237 @opindex Wassign-intercept
3238 @opindex Wno-assign-intercept
3239 Warn whenever an Objective-C assignment is being intercepted by the
3242 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3243 @opindex Wno-protocol
3245 If a class is declared to implement a protocol, a warning is issued for
3246 every method in the protocol that is not implemented by the class. The
3247 default behavior is to issue a warning for every method not explicitly
3248 implemented in the class, even if a method implementation is inherited
3249 from the superclass. If you use the @option{-Wno-protocol} option, then
3250 methods inherited from the superclass are considered to be implemented,
3251 and no warning is issued for them.
3253 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3255 @opindex Wno-selector
3256 Warn if multiple methods of different types for the same selector are
3257 found during compilation. The check is performed on the list of methods
3258 in the final stage of compilation. Additionally, a check is performed
3259 for each selector appearing in a @code{@@selector(@dots{})}
3260 expression, and a corresponding method for that selector has been found
3261 during compilation. Because these checks scan the method table only at
3262 the end of compilation, these warnings are not produced if the final
3263 stage of compilation is not reached, for example because an error is
3264 found during compilation, or because the @option{-fsyntax-only} option is
3267 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3268 @opindex Wstrict-selector-match
3269 @opindex Wno-strict-selector-match
3270 Warn if multiple methods with differing argument and/or return types are
3271 found for a given selector when attempting to send a message using this
3272 selector to a receiver of type @code{id} or @code{Class}. When this flag
3273 is off (which is the default behavior), the compiler omits such warnings
3274 if any differences found are confined to types that share the same size
3277 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3278 @opindex Wundeclared-selector
3279 @opindex Wno-undeclared-selector
3280 Warn if a @code{@@selector(@dots{})} expression referring to an
3281 undeclared selector is found. A selector is considered undeclared if no
3282 method with that name has been declared before the
3283 @code{@@selector(@dots{})} expression, either explicitly in an
3284 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3285 an @code{@@implementation} section. This option always performs its
3286 checks as soon as a @code{@@selector(@dots{})} expression is found,
3287 while @option{-Wselector} only performs its checks in the final stage of
3288 compilation. This also enforces the coding style convention
3289 that methods and selectors must be declared before being used.
3291 @item -print-objc-runtime-info
3292 @opindex print-objc-runtime-info
3293 Generate C header describing the largest structure that is passed by
3298 @node Diagnostic Message Formatting Options
3299 @section Options to Control Diagnostic Messages Formatting
3300 @cindex options to control diagnostics formatting
3301 @cindex diagnostic messages
3302 @cindex message formatting
3304 Traditionally, diagnostic messages have been formatted irrespective of
3305 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3306 options described below
3307 to control the formatting algorithm for diagnostic messages,
3308 e.g.@: how many characters per line, how often source location
3309 information should be reported. Note that some language front ends may not
3310 honor these options.
3313 @item -fmessage-length=@var{n}
3314 @opindex fmessage-length
3315 Try to format error messages so that they fit on lines of about
3316 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3317 done; each error message appears on a single line. This is the
3318 default for all front ends.
3320 @item -fdiagnostics-show-location=once
3321 @opindex fdiagnostics-show-location
3322 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3323 reporter to emit source location information @emph{once}; that is, in
3324 case the message is too long to fit on a single physical line and has to
3325 be wrapped, the source location won't be emitted (as prefix) again,
3326 over and over, in subsequent continuation lines. This is the default
3329 @item -fdiagnostics-show-location=every-line
3330 Only meaningful in line-wrapping mode. Instructs the diagnostic
3331 messages reporter to emit the same source location information (as
3332 prefix) for physical lines that result from the process of breaking
3333 a message which is too long to fit on a single line.
3335 @item -fdiagnostics-color[=@var{WHEN}]
3336 @itemx -fno-diagnostics-color
3337 @opindex fdiagnostics-color
3338 @cindex highlight, color
3339 @vindex GCC_COLORS @r{environment variable}
3340 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3341 or @samp{auto}. The default depends on how the compiler has been configured,
3342 it can be any of the above @var{WHEN} options or also @samp{never}
3343 if @env{GCC_COLORS} environment variable isn't present in the environment,
3344 and @samp{auto} otherwise.
3345 @samp{auto} means to use color only when the standard error is a terminal.
3346 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3347 aliases for @option{-fdiagnostics-color=always} and
3348 @option{-fdiagnostics-color=never}, respectively.
3350 The colors are defined by the environment variable @env{GCC_COLORS}.
3351 Its value is a colon-separated list of capabilities and Select Graphic
3352 Rendition (SGR) substrings. SGR commands are interpreted by the
3353 terminal or terminal emulator. (See the section in the documentation
3354 of your text terminal for permitted values and their meanings as
3355 character attributes.) These substring values are integers in decimal
3356 representation and can be concatenated with semicolons.
3357 Common values to concatenate include
3359 @samp{4} for underline,
3361 @samp{7} for inverse,
3362 @samp{39} for default foreground color,
3363 @samp{30} to @samp{37} for foreground colors,
3364 @samp{90} to @samp{97} for 16-color mode foreground colors,
3365 @samp{38;5;0} to @samp{38;5;255}
3366 for 88-color and 256-color modes foreground colors,
3367 @samp{49} for default background color,
3368 @samp{40} to @samp{47} for background colors,
3369 @samp{100} to @samp{107} for 16-color mode background colors,
3370 and @samp{48;5;0} to @samp{48;5;255}
3371 for 88-color and 256-color modes background colors.
3373 The default @env{GCC_COLORS} is
3375 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:quote=01:\
3376 fixit-insert=32:fixit-delete=31:\
3377 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32
3380 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3381 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3382 @samp{01} is bold, and @samp{31} is red.
3383 Setting @env{GCC_COLORS} to the empty string disables colors.
3384 Supported capabilities are as follows.
3388 @vindex error GCC_COLORS @r{capability}
3389 SGR substring for error: markers.
3392 @vindex warning GCC_COLORS @r{capability}
3393 SGR substring for warning: markers.
3396 @vindex note GCC_COLORS @r{capability}
3397 SGR substring for note: markers.
3400 @vindex range1 GCC_COLORS @r{capability}
3401 SGR substring for first additional range.
3404 @vindex range2 GCC_COLORS @r{capability}
3405 SGR substring for second additional range.
3408 @vindex locus GCC_COLORS @r{capability}
3409 SGR substring for location information, @samp{file:line} or
3410 @samp{file:line:column} etc.
3413 @vindex quote GCC_COLORS @r{capability}
3414 SGR substring for information printed within quotes.
3417 @vindex fixit-insert GCC_COLORS @r{capability}
3418 SGR substring for fix-it hints suggesting text to
3419 be inserted or replaced.
3422 @vindex fixit-delete GCC_COLORS @r{capability}
3423 SGR substring for fix-it hints suggesting text to
3426 @item diff-filename=
3427 @vindex diff-filename GCC_COLORS @r{capability}
3428 SGR substring for filename headers within generated patches.
3431 @vindex diff-hunk GCC_COLORS @r{capability}
3432 SGR substring for the starts of hunks within generated patches.
3435 @vindex diff-delete GCC_COLORS @r{capability}
3436 SGR substring for deleted lines within generated patches.
3439 @vindex diff-insert GCC_COLORS @r{capability}
3440 SGR substring for inserted lines within generated patches.
3443 @item -fno-diagnostics-show-option
3444 @opindex fno-diagnostics-show-option
3445 @opindex fdiagnostics-show-option
3446 By default, each diagnostic emitted includes text indicating the
3447 command-line option that directly controls the diagnostic (if such an
3448 option is known to the diagnostic machinery). Specifying the
3449 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3451 @item -fno-diagnostics-show-caret
3452 @opindex fno-diagnostics-show-caret
3453 @opindex fdiagnostics-show-caret
3454 By default, each diagnostic emitted includes the original source line
3455 and a caret @samp{^} indicating the column. This option suppresses this
3456 information. The source line is truncated to @var{n} characters, if
3457 the @option{-fmessage-length=n} option is given. When the output is done
3458 to the terminal, the width is limited to the width given by the
3459 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3461 @item -fdiagnostics-parseable-fixits
3462 @opindex fdiagnostics-parseable-fixits
3463 Emit fix-it hints in a machine-parseable format, suitable for consumption
3464 by IDEs. For each fix-it, a line will be printed after the relevant
3465 diagnostic, starting with the string ``fix-it:''. For example:
3468 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3471 The location is expressed as a half-open range, expressed as a count of
3472 bytes, starting at byte 1 for the initial column. In the above example,
3473 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3477 00000000011111111112222222222
3478 12345678901234567890123456789
3479 gtk_widget_showall (dlg);
3484 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3485 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3486 (e.g. vertical tab as ``\013'').
3488 An empty replacement string indicates that the given range is to be removed.
3489 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3490 be inserted at the given position.
3492 @item -fdiagnostics-generate-patch
3493 @opindex fdiagnostics-generate-patch
3494 Print fix-it hints to stderr in unified diff format, after any diagnostics
3495 are printed. For example:
3502 void show_cb(GtkDialog *dlg)
3504 - gtk_widget_showall(dlg);
3505 + gtk_widget_show_all(dlg);
3510 The diff may or may not be colorized, following the same rules
3511 as for diagnostics (see @option{-fdiagnostics-color}).
3515 @node Warning Options
3516 @section Options to Request or Suppress Warnings
3517 @cindex options to control warnings
3518 @cindex warning messages
3519 @cindex messages, warning
3520 @cindex suppressing warnings
3522 Warnings are diagnostic messages that report constructions that
3523 are not inherently erroneous but that are risky or suggest there
3524 may have been an error.
3526 The following language-independent options do not enable specific
3527 warnings but control the kinds of diagnostics produced by GCC@.
3530 @cindex syntax checking
3532 @opindex fsyntax-only
3533 Check the code for syntax errors, but don't do anything beyond that.
3535 @item -fmax-errors=@var{n}
3536 @opindex fmax-errors
3537 Limits the maximum number of error messages to @var{n}, at which point
3538 GCC bails out rather than attempting to continue processing the source
3539 code. If @var{n} is 0 (the default), there is no limit on the number
3540 of error messages produced. If @option{-Wfatal-errors} is also
3541 specified, then @option{-Wfatal-errors} takes precedence over this
3546 Inhibit all warning messages.
3551 Make all warnings into errors.
3556 Make the specified warning into an error. The specifier for a warning
3557 is appended; for example @option{-Werror=switch} turns the warnings
3558 controlled by @option{-Wswitch} into errors. This switch takes a
3559 negative form, to be used to negate @option{-Werror} for specific
3560 warnings; for example @option{-Wno-error=switch} makes
3561 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3564 The warning message for each controllable warning includes the
3565 option that controls the warning. That option can then be used with
3566 @option{-Werror=} and @option{-Wno-error=} as described above.
3567 (Printing of the option in the warning message can be disabled using the
3568 @option{-fno-diagnostics-show-option} flag.)
3570 Note that specifying @option{-Werror=}@var{foo} automatically implies
3571 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3574 @item -Wfatal-errors
3575 @opindex Wfatal-errors
3576 @opindex Wno-fatal-errors
3577 This option causes the compiler to abort compilation on the first error
3578 occurred rather than trying to keep going and printing further error
3583 You can request many specific warnings with options beginning with
3584 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3585 implicit declarations. Each of these specific warning options also
3586 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3587 example, @option{-Wno-implicit}. This manual lists only one of the
3588 two forms, whichever is not the default. For further
3589 language-specific options also refer to @ref{C++ Dialect Options} and
3590 @ref{Objective-C and Objective-C++ Dialect Options}.
3592 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3593 options, such as @option{-Wunused}, which may turn on further options,
3594 such as @option{-Wunused-value}. The combined effect of positive and
3595 negative forms is that more specific options have priority over less
3596 specific ones, independently of their position in the command-line. For
3597 options of the same specificity, the last one takes effect. Options
3598 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3599 as if they appeared at the end of the command-line.
3601 When an unrecognized warning option is requested (e.g.,
3602 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3603 that the option is not recognized. However, if the @option{-Wno-} form
3604 is used, the behavior is slightly different: no diagnostic is
3605 produced for @option{-Wno-unknown-warning} unless other diagnostics
3606 are being produced. This allows the use of new @option{-Wno-} options
3607 with old compilers, but if something goes wrong, the compiler
3608 warns that an unrecognized option is present.
3615 Issue all the warnings demanded by strict ISO C and ISO C++;
3616 reject all programs that use forbidden extensions, and some other
3617 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3618 version of the ISO C standard specified by any @option{-std} option used.
3620 Valid ISO C and ISO C++ programs should compile properly with or without
3621 this option (though a rare few require @option{-ansi} or a
3622 @option{-std} option specifying the required version of ISO C)@. However,
3623 without this option, certain GNU extensions and traditional C and C++
3624 features are supported as well. With this option, they are rejected.
3626 @option{-Wpedantic} does not cause warning messages for use of the
3627 alternate keywords whose names begin and end with @samp{__}. Pedantic
3628 warnings are also disabled in the expression that follows
3629 @code{__extension__}. However, only system header files should use
3630 these escape routes; application programs should avoid them.
3631 @xref{Alternate Keywords}.
3633 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3634 C conformance. They soon find that it does not do quite what they want:
3635 it finds some non-ISO practices, but not all---only those for which
3636 ISO C @emph{requires} a diagnostic, and some others for which
3637 diagnostics have been added.
3639 A feature to report any failure to conform to ISO C might be useful in
3640 some instances, but would require considerable additional work and would
3641 be quite different from @option{-Wpedantic}. We don't have plans to
3642 support such a feature in the near future.
3644 Where the standard specified with @option{-std} represents a GNU
3645 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3646 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3647 extended dialect is based. Warnings from @option{-Wpedantic} are given
3648 where they are required by the base standard. (It does not make sense
3649 for such warnings to be given only for features not in the specified GNU
3650 C dialect, since by definition the GNU dialects of C include all
3651 features the compiler supports with the given option, and there would be
3652 nothing to warn about.)
3654 @item -pedantic-errors
3655 @opindex pedantic-errors
3656 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3657 requires a diagnostic, in some cases where there is undefined behavior
3658 at compile-time and in some other cases that do not prevent compilation
3659 of programs that are valid according to the standard. This is not
3660 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3661 by this option and not enabled by the latter and vice versa.
3666 This enables all the warnings about constructions that some users
3667 consider questionable, and that are easy to avoid (or modify to
3668 prevent the warning), even in conjunction with macros. This also
3669 enables some language-specific warnings described in @ref{C++ Dialect
3670 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3672 @option{-Wall} turns on the following warning flags:
3674 @gccoptlist{-Waddress @gol
3675 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3677 -Wbool-operation @gol
3678 -Wc++11-compat -Wc++14-compat@gol
3679 -Wchar-subscripts @gol
3681 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3682 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3684 -Wint-in-bool-context @gol
3685 -Wimplicit @r{(C and Objective-C only)} @gol
3686 -Wimplicit-int @r{(C and Objective-C only)} @gol
3687 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3688 -Winit-self @r{(only for C++)} @gol
3689 -Wlogical-not-parentheses
3690 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3691 -Wmaybe-uninitialized @gol
3692 -Wmemset-elt-size @gol
3693 -Wmemset-transposed-args @gol
3694 -Wmisleading-indentation @r{(only for C/C++)} @gol
3695 -Wmissing-braces @r{(only for C/ObjC)} @gol
3696 -Wnarrowing @r{(only for C++)} @gol
3698 -Wnonnull-compare @gol
3704 -Wsequence-point @gol
3705 -Wsign-compare @r{(only in C++)} @gol
3706 -Wsizeof-pointer-memaccess @gol
3707 -Wstrict-aliasing @gol
3708 -Wstrict-overflow=1 @gol
3710 -Wtautological-compare @gol
3712 -Wuninitialized @gol
3713 -Wunknown-pragmas @gol
3714 -Wunused-function @gol
3717 -Wunused-variable @gol
3718 -Wvolatile-register-var @gol
3721 Note that some warning flags are not implied by @option{-Wall}. Some of
3722 them warn about constructions that users generally do not consider
3723 questionable, but which occasionally you might wish to check for;
3724 others warn about constructions that are necessary or hard to avoid in
3725 some cases, and there is no simple way to modify the code to suppress
3726 the warning. Some of them are enabled by @option{-Wextra} but many of
3727 them must be enabled individually.
3733 This enables some extra warning flags that are not enabled by
3734 @option{-Wall}. (This option used to be called @option{-W}. The older
3735 name is still supported, but the newer name is more descriptive.)
3737 @gccoptlist{-Wclobbered @gol
3739 -Wignored-qualifiers @gol
3740 -Wimplicit-fallthrough=3 @gol
3741 -Wmissing-field-initializers @gol
3742 -Wmissing-parameter-type @r{(C only)} @gol
3743 -Wold-style-declaration @r{(C only)} @gol
3744 -Woverride-init @gol
3745 -Wsign-compare @r{(C only)} @gol
3747 -Wuninitialized @gol
3748 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3749 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3750 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3753 The option @option{-Wextra} also prints warning messages for the
3759 A pointer is compared against integer zero with @code{<}, @code{<=},
3760 @code{>}, or @code{>=}.
3763 (C++ only) An enumerator and a non-enumerator both appear in a
3764 conditional expression.
3767 (C++ only) Ambiguous virtual bases.
3770 (C++ only) Subscripting an array that has been declared @code{register}.
3773 (C++ only) Taking the address of a variable that has been declared
3777 (C++ only) A base class is not initialized in the copy constructor
3782 @item -Wchar-subscripts
3783 @opindex Wchar-subscripts
3784 @opindex Wno-char-subscripts
3785 Warn if an array subscript has type @code{char}. This is a common cause
3786 of error, as programmers often forget that this type is signed on some
3788 This warning is enabled by @option{-Wall}.
3792 @opindex Wno-comment
3793 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3794 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3795 This warning is enabled by @option{-Wall}.
3797 @item -Wno-coverage-mismatch
3798 @opindex Wno-coverage-mismatch
3799 Warn if feedback profiles do not match when using the
3800 @option{-fprofile-use} option.
3801 If a source file is changed between compiling with @option{-fprofile-gen} and
3802 with @option{-fprofile-use}, the files with the profile feedback can fail
3803 to match the source file and GCC cannot use the profile feedback
3804 information. By default, this warning is enabled and is treated as an
3805 error. @option{-Wno-coverage-mismatch} can be used to disable the
3806 warning or @option{-Wno-error=coverage-mismatch} can be used to
3807 disable the error. Disabling the error for this warning can result in
3808 poorly optimized code and is useful only in the
3809 case of very minor changes such as bug fixes to an existing code-base.
3810 Completely disabling the warning is not recommended.
3813 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3815 Suppress warning messages emitted by @code{#warning} directives.
3817 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3818 @opindex Wdouble-promotion
3819 @opindex Wno-double-promotion
3820 Give a warning when a value of type @code{float} is implicitly
3821 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3822 floating-point unit implement @code{float} in hardware, but emulate
3823 @code{double} in software. On such a machine, doing computations
3824 using @code{double} values is much more expensive because of the
3825 overhead required for software emulation.
3827 It is easy to accidentally do computations with @code{double} because
3828 floating-point literals are implicitly of type @code{double}. For
3832 float area(float radius)
3834 return 3.14159 * radius * radius;
3838 the compiler performs the entire computation with @code{double}
3839 because the floating-point literal is a @code{double}.
3841 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3842 @opindex Wduplicate-decl-specifier
3843 @opindex Wno-duplicate-decl-specifier
3844 Warn if a declaration has duplicate @code{const}, @code{volatile},
3845 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3849 @itemx -Wformat=@var{n}
3852 @opindex ffreestanding
3853 @opindex fno-builtin
3855 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3856 the arguments supplied have types appropriate to the format string
3857 specified, and that the conversions specified in the format string make
3858 sense. This includes standard functions, and others specified by format
3859 attributes (@pxref{Function Attributes}), in the @code{printf},
3860 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3861 not in the C standard) families (or other target-specific families).
3862 Which functions are checked without format attributes having been
3863 specified depends on the standard version selected, and such checks of
3864 functions without the attribute specified are disabled by
3865 @option{-ffreestanding} or @option{-fno-builtin}.
3867 The formats are checked against the format features supported by GNU
3868 libc version 2.2. These include all ISO C90 and C99 features, as well
3869 as features from the Single Unix Specification and some BSD and GNU
3870 extensions. Other library implementations may not support all these
3871 features; GCC does not support warning about features that go beyond a
3872 particular library's limitations. However, if @option{-Wpedantic} is used
3873 with @option{-Wformat}, warnings are given about format features not
3874 in the selected standard version (but not for @code{strfmon} formats,
3875 since those are not in any version of the C standard). @xref{C Dialect
3876 Options,,Options Controlling C Dialect}.
3883 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3884 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3885 @option{-Wformat} also checks for null format arguments for several
3886 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3887 aspects of this level of format checking can be disabled by the
3888 options: @option{-Wno-format-contains-nul},
3889 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3890 @option{-Wformat} is enabled by @option{-Wall}.
3892 @item -Wno-format-contains-nul
3893 @opindex Wno-format-contains-nul
3894 @opindex Wformat-contains-nul
3895 If @option{-Wformat} is specified, do not warn about format strings that
3898 @item -Wno-format-extra-args
3899 @opindex Wno-format-extra-args
3900 @opindex Wformat-extra-args
3901 If @option{-Wformat} is specified, do not warn about excess arguments to a
3902 @code{printf} or @code{scanf} format function. The C standard specifies
3903 that such arguments are ignored.
3905 Where the unused arguments lie between used arguments that are
3906 specified with @samp{$} operand number specifications, normally
3907 warnings are still given, since the implementation could not know what
3908 type to pass to @code{va_arg} to skip the unused arguments. However,
3909 in the case of @code{scanf} formats, this option suppresses the
3910 warning if the unused arguments are all pointers, since the Single
3911 Unix Specification says that such unused arguments are allowed.
3913 @item -Wformat-length
3914 @itemx -Wformat-length=@var{level}
3915 @opindex Wformat-length
3916 @opindex Wno-format-length
3917 Warn about calls to formatted input/output functions such as @code{sprintf}
3918 that might overflow the destination buffer, or about bounded functions such
3919 as @code{snprintf} that might result in output truncation. When the exact
3920 number of bytes written by a format directive cannot be determined at
3921 compile-time it is estimated based on heuristics that depend on the
3922 @var{level} argument and on optimization. While enabling optimization
3923 will in most cases improve the accuracy of the warning, it may also
3924 result in false positives.
3927 @item -Wformat-length
3928 @item -Wformat-length=1
3929 @opindex Wformat-length
3930 @opindex Wno-format-length
3931 Level @var{1} of @option{-Wformat-length} enabled by @option{-Wformat}
3932 employs a conservative approach that warns only about calls that most
3933 likely overflow the buffer or result in output truncation. At this
3934 level, numeric arguments to format directives with unknown values are
3935 assumed to have the value of one, and strings of unknown length to be
3936 empty. Numeric arguments that are known to be bounded to a subrange
3937 of their type, or string arguments whose output is bounded either by
3938 their directive's precision or by a finite set of string literals, are
3939 assumed to take on the value within the range that results in the most
3940 bytes on output. For example, the call to @code{sprintf} below is
3941 diagnosed because even with both @var{a} and @var{b} equal to zero,
3942 the terminating NUL character (@code{'\0'}) appended by the function
3943 to the destination buffer will be written past its end. Increasing
3944 the size of the buffer by a single byte is sufficient to avoid the
3945 warning, though it may not be sufficient to avoid the overflow.
3948 void f (int a, int b)
3951 sprintf (buf, "a = %i, b = %i\n", a, b);
3955 @item -Wformat-length=2
3956 Level @var{2} warns also about calls that might overflow the destination
3957 buffer or result in truncation given an argument of sufficient length
3958 or magnitude. At level @var{2}, unknown numeric arguments are assumed
3959 to have the minimum representable value for signed types with a precision
3960 greater than 1, and the maximum representable value otherwise. Unknown
3961 string arguments whose length cannot be assumed to be bounded either by
3962 the directive's precision, or by a finite set of string literals they
3963 may evaluate to, or the character array they may point to, are assumed
3964 to be 1 character long.
3966 At level @var{2}, the call in the example above is again diagnosed, but
3967 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
3968 @code{%i} directive will write some of its digits beyond the end of
3969 the destination buffer. To make the call safe regardless of the values
3970 of the two variables, the size of the destination buffer must be increased
3971 to at least 34 bytes. GCC includes the minimum size of the buffer in
3972 an informational note following the warning.
3974 An alternative to increasing the size of the destination buffer is to
3975 constrain the range of formatted values. The maximum length of string
3976 arguments can be bounded by specifying the precision in the format
3977 directive. When numeric arguments of format directives can be assumed
3978 to be bounded by less than the precision of their type, choosing
3979 an appropriate length modifier to the format specifier will reduce
3980 the required buffer size. For example, if @var{a} and @var{b} in the
3981 example above can be assumed to be within the precision of
3982 the @code{short int} type then using either the @code{%hi} format
3983 directive or casting the argument to @code{short} reduces the maximum
3984 required size of the buffer to 24 bytes.
3987 void f (int a, int b)
3990 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
3995 @item -Wno-format-zero-length
3996 @opindex Wno-format-zero-length
3997 @opindex Wformat-zero-length
3998 If @option{-Wformat} is specified, do not warn about zero-length formats.
3999 The C standard specifies that zero-length formats are allowed.
4004 Enable @option{-Wformat} plus additional format checks. Currently
4005 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4008 @item -Wformat-nonliteral
4009 @opindex Wformat-nonliteral
4010 @opindex Wno-format-nonliteral
4011 If @option{-Wformat} is specified, also warn if the format string is not a
4012 string literal and so cannot be checked, unless the format function
4013 takes its format arguments as a @code{va_list}.
4015 @item -Wformat-security
4016 @opindex Wformat-security
4017 @opindex Wno-format-security
4018 If @option{-Wformat} is specified, also warn about uses of format
4019 functions that represent possible security problems. At present, this
4020 warns about calls to @code{printf} and @code{scanf} functions where the
4021 format string is not a string literal and there are no format arguments,
4022 as in @code{printf (foo);}. This may be a security hole if the format
4023 string came from untrusted input and contains @samp{%n}. (This is
4024 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4025 in future warnings may be added to @option{-Wformat-security} that are not
4026 included in @option{-Wformat-nonliteral}.)
4028 @item -Wformat-signedness
4029 @opindex Wformat-signedness
4030 @opindex Wno-format-signedness
4031 If @option{-Wformat} is specified, also warn if the format string
4032 requires an unsigned argument and the argument is signed and vice versa.
4035 @opindex Wformat-y2k
4036 @opindex Wno-format-y2k
4037 If @option{-Wformat} is specified, also warn about @code{strftime}
4038 formats that may yield only a two-digit year.
4043 @opindex Wno-nonnull
4044 Warn about passing a null pointer for arguments marked as
4045 requiring a non-null value by the @code{nonnull} function attribute.
4047 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4048 can be disabled with the @option{-Wno-nonnull} option.
4050 @item -Wnonnull-compare
4051 @opindex Wnonnull-compare
4052 @opindex Wno-nonnull-compare
4053 Warn when comparing an argument marked with the @code{nonnull}
4054 function attribute against null inside the function.
4056 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4057 can be disabled with the @option{-Wno-nonnull-compare} option.
4059 @item -Wnull-dereference
4060 @opindex Wnull-dereference
4061 @opindex Wno-null-dereference
4062 Warn if the compiler detects paths that trigger erroneous or
4063 undefined behavior due to dereferencing a null pointer. This option
4064 is only active when @option{-fdelete-null-pointer-checks} is active,
4065 which is enabled by optimizations in most targets. The precision of
4066 the warnings depends on the optimization options used.
4068 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4070 @opindex Wno-init-self
4071 Warn about uninitialized variables that are initialized with themselves.
4072 Note this option can only be used with the @option{-Wuninitialized} option.
4074 For example, GCC warns about @code{i} being uninitialized in the
4075 following snippet only when @option{-Winit-self} has been specified:
4086 This warning is enabled by @option{-Wall} in C++.
4088 @item -Wimplicit-int @r{(C and Objective-C only)}
4089 @opindex Wimplicit-int
4090 @opindex Wno-implicit-int
4091 Warn when a declaration does not specify a type.
4092 This warning is enabled by @option{-Wall}.
4094 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4095 @opindex Wimplicit-function-declaration
4096 @opindex Wno-implicit-function-declaration
4097 Give a warning whenever a function is used before being declared. In
4098 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4099 enabled by default and it is made into an error by
4100 @option{-pedantic-errors}. This warning is also enabled by
4103 @item -Wimplicit @r{(C and Objective-C only)}
4105 @opindex Wno-implicit
4106 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4107 This warning is enabled by @option{-Wall}.
4109 @item -Wimplicit-fallthrough
4110 @opindex Wimplicit-fallthrough
4111 @opindex Wno-implicit-fallthrough
4112 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4113 and @option{-Wno-implicit-fallthrough} is the same as
4114 @option{-Wimplicit-fallthrough=0}.
4116 @item -Wimplicit-fallthrough=@var{n}
4117 @opindex Wimplicit-fallthrough=
4118 Warn when a switch case falls through. For example:
4136 This warning does not warn when the last statement of a case cannot
4137 fall through, e.g. when there is a return statement or a call to function
4138 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4139 also takes into account control flow statements, such as ifs, and only
4140 warns when appropriate. E.g.@:
4150 @} else if (i < 1) @{
4160 Since there are occasions where a switch case fall through is desirable,
4161 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4162 to be used along with a null statement to suppress this warning that
4163 would normally occur:
4171 __attribute__ ((fallthrough));
4178 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4179 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4180 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4181 Instead of the these attributes, it is also possible to add a fallthrough
4182 comment to silence the warning. The whole body of the C or C++ style comment
4183 should match the given regular expressions listed below. The option argument
4184 @var{n} specifies what kind of comments are accepted:
4188 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4190 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4191 expression, any comment is used as fallthrough comment.
4193 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4194 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4196 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4197 following regular expressions:
4201 @item @code{-fallthrough}
4203 @item @code{@@fallthrough@@}
4205 @item @code{lint -fallthrough[ \t]*}
4207 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4209 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4211 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4215 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4216 following regular expressions:
4220 @item @code{-fallthrough}
4222 @item @code{@@fallthrough@@}
4224 @item @code{lint -fallthrough[ \t]*}
4226 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4230 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4231 fallthrough comments, only attributes disable the warning.
4235 The comment needs to be followed after optional whitespace and other comments
4236 by @code{case} or @code{default} keywords or by a user label that preceeds some
4237 @code{case} or @code{default} label.
4252 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4254 @item -Wignored-qualifiers @r{(C and C++ only)}
4255 @opindex Wignored-qualifiers
4256 @opindex Wno-ignored-qualifiers
4257 Warn if the return type of a function has a type qualifier
4258 such as @code{const}. For ISO C such a type qualifier has no effect,
4259 since the value returned by a function is not an lvalue.
4260 For C++, the warning is only emitted for scalar types or @code{void}.
4261 ISO C prohibits qualified @code{void} return types on function
4262 definitions, so such return types always receive a warning
4263 even without this option.
4265 This warning is also enabled by @option{-Wextra}.
4267 @item -Wignored-attributes @r{(C and C++ only)}
4268 @opindex Wignored-attributes
4269 @opindex Wno-ignored-attributes
4270 Warn when an attribute is ignored. This is different from the
4271 @option{-Wattributes} option in that it warns whenever the compiler decides
4272 to drop an attribute, not that the attribute is either unknown, used in a
4273 wrong place, etc. This warning is enabled by default.
4278 Warn if the type of @code{main} is suspicious. @code{main} should be
4279 a function with external linkage, returning int, taking either zero
4280 arguments, two, or three arguments of appropriate types. This warning
4281 is enabled by default in C++ and is enabled by either @option{-Wall}
4282 or @option{-Wpedantic}.
4284 @item -Wmisleading-indentation @r{(C and C++ only)}
4285 @opindex Wmisleading-indentation
4286 @opindex Wno-misleading-indentation
4287 Warn when the indentation of the code does not reflect the block structure.
4288 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4289 @code{for} clauses with a guarded statement that does not use braces,
4290 followed by an unguarded statement with the same indentation.
4292 In the following example, the call to ``bar'' is misleadingly indented as
4293 if it were guarded by the ``if'' conditional.
4296 if (some_condition ())
4298 bar (); /* Gotcha: this is not guarded by the "if". */
4301 In the case of mixed tabs and spaces, the warning uses the
4302 @option{-ftabstop=} option to determine if the statements line up
4305 The warning is not issued for code involving multiline preprocessor logic
4306 such as the following example.
4311 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4317 The warning is not issued after a @code{#line} directive, since this
4318 typically indicates autogenerated code, and no assumptions can be made
4319 about the layout of the file that the directive references.
4321 This warning is enabled by @option{-Wall} in C and C++.
4323 @item -Wmissing-braces
4324 @opindex Wmissing-braces
4325 @opindex Wno-missing-braces
4326 Warn if an aggregate or union initializer is not fully bracketed. In
4327 the following example, the initializer for @code{a} is not fully
4328 bracketed, but that for @code{b} is fully bracketed. This warning is
4329 enabled by @option{-Wall} in C.
4332 int a[2][2] = @{ 0, 1, 2, 3 @};
4333 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4336 This warning is enabled by @option{-Wall}.
4338 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4339 @opindex Wmissing-include-dirs
4340 @opindex Wno-missing-include-dirs
4341 Warn if a user-supplied include directory does not exist.
4344 @opindex Wparentheses
4345 @opindex Wno-parentheses
4346 Warn if parentheses are omitted in certain contexts, such
4347 as when there is an assignment in a context where a truth value
4348 is expected, or when operators are nested whose precedence people
4349 often get confused about.
4351 Also warn if a comparison like @code{x<=y<=z} appears; this is
4352 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4353 interpretation from that of ordinary mathematical notation.
4355 Also warn for dangerous uses of the GNU extension to
4356 @code{?:} with omitted middle operand. When the condition
4357 in the @code{?}: operator is a boolean expression, the omitted value is
4358 always 1. Often programmers expect it to be a value computed
4359 inside the conditional expression instead.
4361 This warning is enabled by @option{-Wall}.
4363 @item -Wsequence-point
4364 @opindex Wsequence-point
4365 @opindex Wno-sequence-point
4366 Warn about code that may have undefined semantics because of violations
4367 of sequence point rules in the C and C++ standards.
4369 The C and C++ standards define the order in which expressions in a C/C++
4370 program are evaluated in terms of @dfn{sequence points}, which represent
4371 a partial ordering between the execution of parts of the program: those
4372 executed before the sequence point, and those executed after it. These
4373 occur after the evaluation of a full expression (one which is not part
4374 of a larger expression), after the evaluation of the first operand of a
4375 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4376 function is called (but after the evaluation of its arguments and the
4377 expression denoting the called function), and in certain other places.
4378 Other than as expressed by the sequence point rules, the order of
4379 evaluation of subexpressions of an expression is not specified. All
4380 these rules describe only a partial order rather than a total order,
4381 since, for example, if two functions are called within one expression
4382 with no sequence point between them, the order in which the functions
4383 are called is not specified. However, the standards committee have
4384 ruled that function calls do not overlap.
4386 It is not specified when between sequence points modifications to the
4387 values of objects take effect. Programs whose behavior depends on this
4388 have undefined behavior; the C and C++ standards specify that ``Between
4389 the previous and next sequence point an object shall have its stored
4390 value modified at most once by the evaluation of an expression.
4391 Furthermore, the prior value shall be read only to determine the value
4392 to be stored.''. If a program breaks these rules, the results on any
4393 particular implementation are entirely unpredictable.
4395 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4396 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4397 diagnosed by this option, and it may give an occasional false positive
4398 result, but in general it has been found fairly effective at detecting
4399 this sort of problem in programs.
4401 The C++17 standard will define the order of evaluation of operands in
4402 more cases: in particular it requires that the right-hand side of an
4403 assignment be evaluated before the left-hand side, so the above
4404 examples are no longer undefined. But this warning will still warn
4405 about them, to help people avoid writing code that is undefined in C
4406 and earlier revisions of C++.
4408 The standard is worded confusingly, therefore there is some debate
4409 over the precise meaning of the sequence point rules in subtle cases.
4410 Links to discussions of the problem, including proposed formal
4411 definitions, may be found on the GCC readings page, at
4412 @uref{http://gcc.gnu.org/@/readings.html}.
4414 This warning is enabled by @option{-Wall} for C and C++.
4416 @item -Wno-return-local-addr
4417 @opindex Wno-return-local-addr
4418 @opindex Wreturn-local-addr
4419 Do not warn about returning a pointer (or in C++, a reference) to a
4420 variable that goes out of scope after the function returns.
4423 @opindex Wreturn-type
4424 @opindex Wno-return-type
4425 Warn whenever a function is defined with a return type that defaults
4426 to @code{int}. Also warn about any @code{return} statement with no
4427 return value in a function whose return type is not @code{void}
4428 (falling off the end of the function body is considered returning
4431 For C only, warn about a @code{return} statement with an expression in a
4432 function whose return type is @code{void}, unless the expression type is
4433 also @code{void}. As a GNU extension, the latter case is accepted
4434 without a warning unless @option{-Wpedantic} is used.
4436 For C++, a function without return type always produces a diagnostic
4437 message, even when @option{-Wno-return-type} is specified. The only
4438 exceptions are @code{main} and functions defined in system headers.
4440 This warning is enabled by @option{-Wall}.
4442 @item -Wshift-count-negative
4443 @opindex Wshift-count-negative
4444 @opindex Wno-shift-count-negative
4445 Warn if shift count is negative. This warning is enabled by default.
4447 @item -Wshift-count-overflow
4448 @opindex Wshift-count-overflow
4449 @opindex Wno-shift-count-overflow
4450 Warn if shift count >= width of type. This warning is enabled by default.
4452 @item -Wshift-negative-value
4453 @opindex Wshift-negative-value
4454 @opindex Wno-shift-negative-value
4455 Warn if left shifting a negative value. This warning is enabled by
4456 @option{-Wextra} in C99 and C++11 modes (and newer).
4458 @item -Wshift-overflow
4459 @itemx -Wshift-overflow=@var{n}
4460 @opindex Wshift-overflow
4461 @opindex Wno-shift-overflow
4462 Warn about left shift overflows. This warning is enabled by
4463 default in C99 and C++11 modes (and newer).
4466 @item -Wshift-overflow=1
4467 This is the warning level of @option{-Wshift-overflow} and is enabled
4468 by default in C99 and C++11 modes (and newer). This warning level does
4469 not warn about left-shifting 1 into the sign bit. (However, in C, such
4470 an overflow is still rejected in contexts where an integer constant expression
4473 @item -Wshift-overflow=2
4474 This warning level also warns about left-shifting 1 into the sign bit,
4475 unless C++14 mode is active.
4481 Warn whenever a @code{switch} statement has an index of enumerated type
4482 and lacks a @code{case} for one or more of the named codes of that
4483 enumeration. (The presence of a @code{default} label prevents this
4484 warning.) @code{case} labels outside the enumeration range also
4485 provoke warnings when this option is used (even if there is a
4486 @code{default} label).
4487 This warning is enabled by @option{-Wall}.
4489 @item -Wswitch-default
4490 @opindex Wswitch-default
4491 @opindex Wno-switch-default
4492 Warn whenever a @code{switch} statement does not have a @code{default}
4496 @opindex Wswitch-enum
4497 @opindex Wno-switch-enum
4498 Warn whenever a @code{switch} statement has an index of enumerated type
4499 and lacks a @code{case} for one or more of the named codes of that
4500 enumeration. @code{case} labels outside the enumeration range also
4501 provoke warnings when this option is used. The only difference
4502 between @option{-Wswitch} and this option is that this option gives a
4503 warning about an omitted enumeration code even if there is a
4504 @code{default} label.
4507 @opindex Wswitch-bool
4508 @opindex Wno-switch-bool
4509 Warn whenever a @code{switch} statement has an index of boolean type
4510 and the case values are outside the range of a boolean type.
4511 It is possible to suppress this warning by casting the controlling
4512 expression to a type other than @code{bool}. For example:
4515 switch ((int) (a == 4))
4521 This warning is enabled by default for C and C++ programs.
4523 @item -Wswitch-unreachable
4524 @opindex Wswitch-unreachable
4525 @opindex Wno-switch-unreachable
4526 Warn whenever a @code{switch} statement contains statements between the
4527 controlling expression and the first case label, which will never be
4528 executed. For example:
4540 @option{-Wswitch-unreachable} does not warn if the statement between the
4541 controlling expression and the first case label is just a declaration:
4554 This warning is enabled by default for C and C++ programs.
4556 @item -Wsync-nand @r{(C and C++ only)}
4558 @opindex Wno-sync-nand
4559 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4560 built-in functions are used. These functions changed semantics in GCC 4.4.
4564 @opindex Wno-trigraphs
4565 Warn if any trigraphs are encountered that might change the meaning of
4566 the program (trigraphs within comments are not warned about).
4567 This warning is enabled by @option{-Wall}.
4569 @item -Wunused-but-set-parameter
4570 @opindex Wunused-but-set-parameter
4571 @opindex Wno-unused-but-set-parameter
4572 Warn whenever a function parameter is assigned to, but otherwise unused
4573 (aside from its declaration).
4575 To suppress this warning use the @code{unused} attribute
4576 (@pxref{Variable Attributes}).
4578 This warning is also enabled by @option{-Wunused} together with
4581 @item -Wunused-but-set-variable
4582 @opindex Wunused-but-set-variable
4583 @opindex Wno-unused-but-set-variable
4584 Warn whenever a local variable is assigned to, but otherwise unused
4585 (aside from its declaration).
4586 This warning is enabled by @option{-Wall}.
4588 To suppress this warning use the @code{unused} attribute
4589 (@pxref{Variable Attributes}).
4591 This warning is also enabled by @option{-Wunused}, which is enabled
4594 @item -Wunused-function
4595 @opindex Wunused-function
4596 @opindex Wno-unused-function
4597 Warn whenever a static function is declared but not defined or a
4598 non-inline static function is unused.
4599 This warning is enabled by @option{-Wall}.
4601 @item -Wunused-label
4602 @opindex Wunused-label
4603 @opindex Wno-unused-label
4604 Warn whenever a label is declared but not used.
4605 This warning is enabled by @option{-Wall}.
4607 To suppress this warning use the @code{unused} attribute
4608 (@pxref{Variable Attributes}).
4610 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4611 @opindex Wunused-local-typedefs
4612 Warn when a typedef locally defined in a function is not used.
4613 This warning is enabled by @option{-Wall}.
4615 @item -Wunused-parameter
4616 @opindex Wunused-parameter
4617 @opindex Wno-unused-parameter
4618 Warn whenever a function parameter is unused aside from its declaration.
4620 To suppress this warning use the @code{unused} attribute
4621 (@pxref{Variable Attributes}).
4623 @item -Wno-unused-result
4624 @opindex Wunused-result
4625 @opindex Wno-unused-result
4626 Do not warn if a caller of a function marked with attribute
4627 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4628 its return value. The default is @option{-Wunused-result}.
4630 @item -Wunused-variable
4631 @opindex Wunused-variable
4632 @opindex Wno-unused-variable
4633 Warn whenever a local or static variable is unused aside from its
4634 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4635 but not for C++. This warning is enabled by @option{-Wall}.
4637 To suppress this warning use the @code{unused} attribute
4638 (@pxref{Variable Attributes}).
4640 @item -Wunused-const-variable
4641 @itemx -Wunused-const-variable=@var{n}
4642 @opindex Wunused-const-variable
4643 @opindex Wno-unused-const-variable
4644 Warn whenever a constant static variable is unused aside from its declaration.
4645 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4646 for C, but not for C++. In C this declares variable storage, but in C++ this
4647 is not an error since const variables take the place of @code{#define}s.
4649 To suppress this warning use the @code{unused} attribute
4650 (@pxref{Variable Attributes}).
4653 @item -Wunused-const-variable=1
4654 This is the warning level that is enabled by @option{-Wunused-variable} for
4655 C. It warns only about unused static const variables defined in the main
4656 compilation unit, but not about static const variables declared in any
4659 @item -Wunused-const-variable=2
4660 This warning level also warns for unused constant static variables in
4661 headers (excluding system headers). This is the warning level of
4662 @option{-Wunused-const-variable} and must be explicitly requested since
4663 in C++ this isn't an error and in C it might be harder to clean up all
4667 @item -Wunused-value
4668 @opindex Wunused-value
4669 @opindex Wno-unused-value
4670 Warn whenever a statement computes a result that is explicitly not
4671 used. To suppress this warning cast the unused expression to
4672 @code{void}. This includes an expression-statement or the left-hand
4673 side of a comma expression that contains no side effects. For example,
4674 an expression such as @code{x[i,j]} causes a warning, while
4675 @code{x[(void)i,j]} does not.
4677 This warning is enabled by @option{-Wall}.
4682 All the above @option{-Wunused} options combined.
4684 In order to get a warning about an unused function parameter, you must
4685 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4686 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4688 @item -Wuninitialized
4689 @opindex Wuninitialized
4690 @opindex Wno-uninitialized
4691 Warn if an automatic variable is used without first being initialized
4692 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4693 warn if a non-static reference or non-static @code{const} member
4694 appears in a class without constructors.
4696 If you want to warn about code that uses the uninitialized value of the
4697 variable in its own initializer, use the @option{-Winit-self} option.
4699 These warnings occur for individual uninitialized or clobbered
4700 elements of structure, union or array variables as well as for
4701 variables that are uninitialized or clobbered as a whole. They do
4702 not occur for variables or elements declared @code{volatile}. Because
4703 these warnings depend on optimization, the exact variables or elements
4704 for which there are warnings depends on the precise optimization
4705 options and version of GCC used.
4707 Note that there may be no warning about a variable that is used only
4708 to compute a value that itself is never used, because such
4709 computations may be deleted by data flow analysis before the warnings
4712 @item -Winvalid-memory-model
4713 @opindex Winvalid-memory-model
4714 @opindex Wno-invalid-memory-model
4715 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4716 and the C11 atomic generic functions with a memory consistency argument
4717 that is either invalid for the operation or outside the range of values
4718 of the @code{memory_order} enumeration. For example, since the
4719 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4720 defined for the relaxed, release, and sequentially consistent memory
4721 orders the following code is diagnosed:
4726 __atomic_store_n (i, 0, memory_order_consume);
4730 @option{-Winvalid-memory-model} is enabled by default.
4732 @item -Wmaybe-uninitialized
4733 @opindex Wmaybe-uninitialized
4734 @opindex Wno-maybe-uninitialized
4735 For an automatic variable, if there exists a path from the function
4736 entry to a use of the variable that is initialized, but there exist
4737 some other paths for which the variable is not initialized, the compiler
4738 emits a warning if it cannot prove the uninitialized paths are not
4739 executed at run time. These warnings are made optional because GCC is
4740 not smart enough to see all the reasons why the code might be correct
4741 in spite of appearing to have an error. Here is one example of how
4762 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4763 always initialized, but GCC doesn't know this. To suppress the
4764 warning, you need to provide a default case with assert(0) or
4767 @cindex @code{longjmp} warnings
4768 This option also warns when a non-volatile automatic variable might be
4769 changed by a call to @code{longjmp}. These warnings as well are possible
4770 only in optimizing compilation.
4772 The compiler sees only the calls to @code{setjmp}. It cannot know
4773 where @code{longjmp} will be called; in fact, a signal handler could
4774 call it at any point in the code. As a result, you may get a warning
4775 even when there is in fact no problem because @code{longjmp} cannot
4776 in fact be called at the place that would cause a problem.
4778 Some spurious warnings can be avoided if you declare all the functions
4779 you use that never return as @code{noreturn}. @xref{Function
4782 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4784 @item -Wunknown-pragmas
4785 @opindex Wunknown-pragmas
4786 @opindex Wno-unknown-pragmas
4787 @cindex warning for unknown pragmas
4788 @cindex unknown pragmas, warning
4789 @cindex pragmas, warning of unknown
4790 Warn when a @code{#pragma} directive is encountered that is not understood by
4791 GCC@. If this command-line option is used, warnings are even issued
4792 for unknown pragmas in system header files. This is not the case if
4793 the warnings are only enabled by the @option{-Wall} command-line option.
4796 @opindex Wno-pragmas
4798 Do not warn about misuses of pragmas, such as incorrect parameters,
4799 invalid syntax, or conflicts between pragmas. See also
4800 @option{-Wunknown-pragmas}.
4802 @item -Wstrict-aliasing
4803 @opindex Wstrict-aliasing
4804 @opindex Wno-strict-aliasing
4805 This option is only active when @option{-fstrict-aliasing} is active.
4806 It warns about code that might break the strict aliasing rules that the
4807 compiler is using for optimization. The warning does not catch all
4808 cases, but does attempt to catch the more common pitfalls. It is
4809 included in @option{-Wall}.
4810 It is equivalent to @option{-Wstrict-aliasing=3}
4812 @item -Wstrict-aliasing=n
4813 @opindex Wstrict-aliasing=n
4814 This option is only active when @option{-fstrict-aliasing} is active.
4815 It warns about code that might break the strict aliasing rules that the
4816 compiler is using for optimization.
4817 Higher levels correspond to higher accuracy (fewer false positives).
4818 Higher levels also correspond to more effort, similar to the way @option{-O}
4820 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4822 Level 1: Most aggressive, quick, least accurate.
4823 Possibly useful when higher levels
4824 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4825 false negatives. However, it has many false positives.
4826 Warns for all pointer conversions between possibly incompatible types,
4827 even if never dereferenced. Runs in the front end only.
4829 Level 2: Aggressive, quick, not too precise.
4830 May still have many false positives (not as many as level 1 though),
4831 and few false negatives (but possibly more than level 1).
4832 Unlike level 1, it only warns when an address is taken. Warns about
4833 incomplete types. Runs in the front end only.
4835 Level 3 (default for @option{-Wstrict-aliasing}):
4836 Should have very few false positives and few false
4837 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4838 Takes care of the common pun+dereference pattern in the front end:
4839 @code{*(int*)&some_float}.
4840 If optimization is enabled, it also runs in the back end, where it deals
4841 with multiple statement cases using flow-sensitive points-to information.
4842 Only warns when the converted pointer is dereferenced.
4843 Does not warn about incomplete types.
4845 @item -Wstrict-overflow
4846 @itemx -Wstrict-overflow=@var{n}
4847 @opindex Wstrict-overflow
4848 @opindex Wno-strict-overflow
4849 This option is only active when @option{-fstrict-overflow} is active.
4850 It warns about cases where the compiler optimizes based on the
4851 assumption that signed overflow does not occur. Note that it does not
4852 warn about all cases where the code might overflow: it only warns
4853 about cases where the compiler implements some optimization. Thus
4854 this warning depends on the optimization level.
4856 An optimization that assumes that signed overflow does not occur is
4857 perfectly safe if the values of the variables involved are such that
4858 overflow never does, in fact, occur. Therefore this warning can
4859 easily give a false positive: a warning about code that is not
4860 actually a problem. To help focus on important issues, several
4861 warning levels are defined. No warnings are issued for the use of
4862 undefined signed overflow when estimating how many iterations a loop
4863 requires, in particular when determining whether a loop will be
4867 @item -Wstrict-overflow=1
4868 Warn about cases that are both questionable and easy to avoid. For
4869 example, with @option{-fstrict-overflow}, the compiler simplifies
4870 @code{x + 1 > x} to @code{1}. This level of
4871 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4872 are not, and must be explicitly requested.
4874 @item -Wstrict-overflow=2
4875 Also warn about other cases where a comparison is simplified to a
4876 constant. For example: @code{abs (x) >= 0}. This can only be
4877 simplified when @option{-fstrict-overflow} is in effect, because
4878 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4879 zero. @option{-Wstrict-overflow} (with no level) is the same as
4880 @option{-Wstrict-overflow=2}.
4882 @item -Wstrict-overflow=3
4883 Also warn about other cases where a comparison is simplified. For
4884 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4886 @item -Wstrict-overflow=4
4887 Also warn about other simplifications not covered by the above cases.
4888 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4890 @item -Wstrict-overflow=5
4891 Also warn about cases where the compiler reduces the magnitude of a
4892 constant involved in a comparison. For example: @code{x + 2 > y} is
4893 simplified to @code{x + 1 >= y}. This is reported only at the
4894 highest warning level because this simplification applies to many
4895 comparisons, so this warning level gives a very large number of
4899 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4900 @opindex Wsuggest-attribute=
4901 @opindex Wno-suggest-attribute=
4902 Warn for cases where adding an attribute may be beneficial. The
4903 attributes currently supported are listed below.
4906 @item -Wsuggest-attribute=pure
4907 @itemx -Wsuggest-attribute=const
4908 @itemx -Wsuggest-attribute=noreturn
4909 @opindex Wsuggest-attribute=pure
4910 @opindex Wno-suggest-attribute=pure
4911 @opindex Wsuggest-attribute=const
4912 @opindex Wno-suggest-attribute=const
4913 @opindex Wsuggest-attribute=noreturn
4914 @opindex Wno-suggest-attribute=noreturn
4916 Warn about functions that might be candidates for attributes
4917 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4918 functions visible in other compilation units or (in the case of @code{pure} and
4919 @code{const}) if it cannot prove that the function returns normally. A function
4920 returns normally if it doesn't contain an infinite loop or return abnormally
4921 by throwing, calling @code{abort} or trapping. This analysis requires option
4922 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4923 higher. Higher optimization levels improve the accuracy of the analysis.
4925 @item -Wsuggest-attribute=format
4926 @itemx -Wmissing-format-attribute
4927 @opindex Wsuggest-attribute=format
4928 @opindex Wmissing-format-attribute
4929 @opindex Wno-suggest-attribute=format
4930 @opindex Wno-missing-format-attribute
4934 Warn about function pointers that might be candidates for @code{format}
4935 attributes. Note these are only possible candidates, not absolute ones.
4936 GCC guesses that function pointers with @code{format} attributes that
4937 are used in assignment, initialization, parameter passing or return
4938 statements should have a corresponding @code{format} attribute in the
4939 resulting type. I.e.@: the left-hand side of the assignment or
4940 initialization, the type of the parameter variable, or the return type
4941 of the containing function respectively should also have a @code{format}
4942 attribute to avoid the warning.
4944 GCC also warns about function definitions that might be
4945 candidates for @code{format} attributes. Again, these are only
4946 possible candidates. GCC guesses that @code{format} attributes
4947 might be appropriate for any function that calls a function like
4948 @code{vprintf} or @code{vscanf}, but this might not always be the
4949 case, and some functions for which @code{format} attributes are
4950 appropriate may not be detected.
4953 @item -Wsuggest-final-types
4954 @opindex Wno-suggest-final-types
4955 @opindex Wsuggest-final-types
4956 Warn about types with virtual methods where code quality would be improved
4957 if the type were declared with the C++11 @code{final} specifier,
4959 declared in an anonymous namespace. This allows GCC to more aggressively
4960 devirtualize the polymorphic calls. This warning is more effective with link
4961 time optimization, where the information about the class hierarchy graph is
4964 @item -Wsuggest-final-methods
4965 @opindex Wno-suggest-final-methods
4966 @opindex Wsuggest-final-methods
4967 Warn about virtual methods where code quality would be improved if the method
4968 were declared with the C++11 @code{final} specifier,
4969 or, if possible, its type were
4970 declared in an anonymous namespace or with the @code{final} specifier.
4972 more effective with link time optimization, where the information about the
4973 class hierarchy graph is more complete. It is recommended to first consider
4974 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4977 @item -Wsuggest-override
4978 Warn about overriding virtual functions that are not marked with the override
4984 This option warns on all uses of @code{alloca} in the source.
4986 @item -Walloca-larger-than=@var{n}
4987 This option warns on calls to @code{alloca} that are not bounded by a
4988 controlling predicate limiting its size to @var{n} bytes, or calls to
4989 @code{alloca} where the bound is unknown.
4991 For example, a bounded case of @code{alloca} could be:
4994 void func (size_t n)
5005 In the above example, passing @code{-Walloca=1000} would not issue a
5006 warning because the call to @code{alloca} is known to be at most 1000
5007 bytes. However, if @code{-Walloca=500} was passed, the compiler would
5008 have emitted a warning.
5010 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5011 controlling predicate verifying its size. For example:
5016 void *p = alloca (n);
5021 If @code{-Walloca=500} was passed, the above would trigger a warning,
5022 but this time because of the lack of bounds checking.
5024 Note, that even seemingly correct code involving signed integers could
5028 void func (signed int n)
5038 In the above example, @var{n} could be negative, causing a larger than
5039 expected argument to be implicitly casted into the @code{alloca} call.
5041 This option also warns when @code{alloca} is used in a loop.
5043 This warning is not enabled by @option{-Wall}, and is only active when
5044 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5046 See also @option{-Wvla-larger-than=@var{n}}.
5048 @item -Warray-bounds
5049 @itemx -Warray-bounds=@var{n}
5050 @opindex Wno-array-bounds
5051 @opindex Warray-bounds
5052 This option is only active when @option{-ftree-vrp} is active
5053 (default for @option{-O2} and above). It warns about subscripts to arrays
5054 that are always out of bounds. This warning is enabled by @option{-Wall}.
5057 @item -Warray-bounds=1
5058 This is the warning level of @option{-Warray-bounds} and is enabled
5059 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5061 @item -Warray-bounds=2
5062 This warning level also warns about out of bounds access for
5063 arrays at the end of a struct and for arrays accessed through
5064 pointers. This warning level may give a larger number of
5065 false positives and is deactivated by default.
5068 @item -Wbool-compare
5069 @opindex Wno-bool-compare
5070 @opindex Wbool-compare
5071 Warn about boolean expression compared with an integer value different from
5072 @code{true}/@code{false}. For instance, the following comparison is
5077 if ((n > 1) == 2) @{ @dots{} @}
5079 This warning is enabled by @option{-Wall}.
5081 @item -Wbool-operation
5082 @opindex Wno-bool-operation
5083 @opindex Wbool-operation
5084 Warn about suspicious operations on expressions of a boolean type. For
5085 instance, bitwise negation of a boolean is very likely a bug in the program.
5086 For C, this warning also warns about incrementing or decrementing a boolean,
5087 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5088 Incrementing a boolean is invalid in C++1z, and deprecated otherwise.)
5090 This warning is enabled by @option{-Wall}.
5092 @item -Wduplicated-cond
5093 @opindex Wno-duplicated-cond
5094 @opindex Wduplicated-cond
5095 Warn about duplicated conditions in an if-else-if chain. For instance,
5096 warn for the following code:
5098 if (p->q != NULL) @{ @dots{} @}
5099 else if (p->q != NULL) @{ @dots{} @}
5102 @item -Wframe-address
5103 @opindex Wno-frame-address
5104 @opindex Wframe-address
5105 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5106 is called with an argument greater than 0. Such calls may return indeterminate
5107 values or crash the program. The warning is included in @option{-Wall}.
5109 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5110 @opindex Wno-discarded-qualifiers
5111 @opindex Wdiscarded-qualifiers
5112 Do not warn if type qualifiers on pointers are being discarded.
5113 Typically, the compiler warns if a @code{const char *} variable is
5114 passed to a function that takes a @code{char *} parameter. This option
5115 can be used to suppress such a warning.
5117 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5118 @opindex Wno-discarded-array-qualifiers
5119 @opindex Wdiscarded-array-qualifiers
5120 Do not warn if type qualifiers on arrays which are pointer targets
5121 are being discarded. Typically, the compiler warns if a
5122 @code{const int (*)[]} variable is passed to a function that
5123 takes a @code{int (*)[]} parameter. This option can be used to
5124 suppress such a warning.
5126 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5127 @opindex Wno-incompatible-pointer-types
5128 @opindex Wincompatible-pointer-types
5129 Do not warn when there is a conversion between pointers that have incompatible
5130 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5131 which warns for pointer argument passing or assignment with different
5134 @item -Wno-int-conversion @r{(C and Objective-C only)}
5135 @opindex Wno-int-conversion
5136 @opindex Wint-conversion
5137 Do not warn about incompatible integer to pointer and pointer to integer
5138 conversions. This warning is about implicit conversions; for explicit
5139 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5140 @option{-Wno-pointer-to-int-cast} may be used.
5142 @item -Wno-div-by-zero
5143 @opindex Wno-div-by-zero
5144 @opindex Wdiv-by-zero
5145 Do not warn about compile-time integer division by zero. Floating-point
5146 division by zero is not warned about, as it can be a legitimate way of
5147 obtaining infinities and NaNs.
5149 @item -Wsystem-headers
5150 @opindex Wsystem-headers
5151 @opindex Wno-system-headers
5152 @cindex warnings from system headers
5153 @cindex system headers, warnings from
5154 Print warning messages for constructs found in system header files.
5155 Warnings from system headers are normally suppressed, on the assumption
5156 that they usually do not indicate real problems and would only make the
5157 compiler output harder to read. Using this command-line option tells
5158 GCC to emit warnings from system headers as if they occurred in user
5159 code. However, note that using @option{-Wall} in conjunction with this
5160 option does @emph{not} warn about unknown pragmas in system
5161 headers---for that, @option{-Wunknown-pragmas} must also be used.
5163 @item -Wtautological-compare
5164 @opindex Wtautological-compare
5165 @opindex Wno-tautological-compare
5166 Warn if a self-comparison always evaluates to true or false. This
5167 warning detects various mistakes such as:
5171 if (i > i) @{ @dots{} @}
5173 This warning is enabled by @option{-Wall}.
5176 @opindex Wtrampolines
5177 @opindex Wno-trampolines
5178 Warn about trampolines generated for pointers to nested functions.
5179 A trampoline is a small piece of data or code that is created at run
5180 time on the stack when the address of a nested function is taken, and is
5181 used to call the nested function indirectly. For some targets, it is
5182 made up of data only and thus requires no special treatment. But, for
5183 most targets, it is made up of code and thus requires the stack to be
5184 made executable in order for the program to work properly.
5187 @opindex Wfloat-equal
5188 @opindex Wno-float-equal
5189 Warn if floating-point values are used in equality comparisons.
5191 The idea behind this is that sometimes it is convenient (for the
5192 programmer) to consider floating-point values as approximations to
5193 infinitely precise real numbers. If you are doing this, then you need
5194 to compute (by analyzing the code, or in some other way) the maximum or
5195 likely maximum error that the computation introduces, and allow for it
5196 when performing comparisons (and when producing output, but that's a
5197 different problem). In particular, instead of testing for equality, you
5198 should check to see whether the two values have ranges that overlap; and
5199 this is done with the relational operators, so equality comparisons are
5202 @item -Wtraditional @r{(C and Objective-C only)}
5203 @opindex Wtraditional
5204 @opindex Wno-traditional
5205 Warn about certain constructs that behave differently in traditional and
5206 ISO C@. Also warn about ISO C constructs that have no traditional C
5207 equivalent, and/or problematic constructs that should be avoided.
5211 Macro parameters that appear within string literals in the macro body.
5212 In traditional C macro replacement takes place within string literals,
5213 but in ISO C it does not.
5216 In traditional C, some preprocessor directives did not exist.
5217 Traditional preprocessors only considered a line to be a directive
5218 if the @samp{#} appeared in column 1 on the line. Therefore
5219 @option{-Wtraditional} warns about directives that traditional C
5220 understands but ignores because the @samp{#} does not appear as the
5221 first character on the line. It also suggests you hide directives like
5222 @code{#pragma} not understood by traditional C by indenting them. Some
5223 traditional implementations do not recognize @code{#elif}, so this option
5224 suggests avoiding it altogether.
5227 A function-like macro that appears without arguments.
5230 The unary plus operator.
5233 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5234 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5235 constants.) Note, these suffixes appear in macros defined in the system
5236 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5237 Use of these macros in user code might normally lead to spurious
5238 warnings, however GCC's integrated preprocessor has enough context to
5239 avoid warning in these cases.
5242 A function declared external in one block and then used after the end of
5246 A @code{switch} statement has an operand of type @code{long}.
5249 A non-@code{static} function declaration follows a @code{static} one.
5250 This construct is not accepted by some traditional C compilers.
5253 The ISO type of an integer constant has a different width or
5254 signedness from its traditional type. This warning is only issued if
5255 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5256 typically represent bit patterns, are not warned about.
5259 Usage of ISO string concatenation is detected.
5262 Initialization of automatic aggregates.
5265 Identifier conflicts with labels. Traditional C lacks a separate
5266 namespace for labels.
5269 Initialization of unions. If the initializer is zero, the warning is
5270 omitted. This is done under the assumption that the zero initializer in
5271 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5272 initializer warnings and relies on default initialization to zero in the
5276 Conversions by prototypes between fixed/floating-point values and vice
5277 versa. The absence of these prototypes when compiling with traditional
5278 C causes serious problems. This is a subset of the possible
5279 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5282 Use of ISO C style function definitions. This warning intentionally is
5283 @emph{not} issued for prototype declarations or variadic functions
5284 because these ISO C features appear in your code when using
5285 libiberty's traditional C compatibility macros, @code{PARAMS} and
5286 @code{VPARAMS}. This warning is also bypassed for nested functions
5287 because that feature is already a GCC extension and thus not relevant to
5288 traditional C compatibility.
5291 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5292 @opindex Wtraditional-conversion
5293 @opindex Wno-traditional-conversion
5294 Warn if a prototype causes a type conversion that is different from what
5295 would happen to the same argument in the absence of a prototype. This
5296 includes conversions of fixed point to floating and vice versa, and
5297 conversions changing the width or signedness of a fixed-point argument
5298 except when the same as the default promotion.
5300 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5301 @opindex Wdeclaration-after-statement
5302 @opindex Wno-declaration-after-statement
5303 Warn when a declaration is found after a statement in a block. This
5304 construct, known from C++, was introduced with ISO C99 and is by default
5305 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5310 Warn if an undefined identifier is evaluated in an @code{#if} directive.
5312 @item -Wno-endif-labels
5313 @opindex Wno-endif-labels
5314 @opindex Wendif-labels
5315 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
5320 Warn whenever a local variable or type declaration shadows another
5321 variable, parameter, type, class member (in C++), or instance variable
5322 (in Objective-C) or whenever a built-in function is shadowed. Note
5323 that in C++, the compiler warns if a local variable shadows an
5324 explicit typedef, but not if it shadows a struct/class/enum.
5325 Same as @option{-Wshadow=global}.
5327 @item -Wno-shadow-ivar @r{(Objective-C only)}
5328 @opindex Wno-shadow-ivar
5329 @opindex Wshadow-ivar
5330 Do not warn whenever a local variable shadows an instance variable in an
5333 @item -Wshadow=global
5334 @opindex Wshadow=local
5335 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5337 @item -Wshadow=local
5338 @opindex Wshadow=local
5339 Warn when a local variable shadows another local variable or parameter.
5340 This warning is enabled by @option{-Wshadow=global}.
5342 @item -Wshadow=compatible-local
5343 @opindex Wshadow=compatible-local
5344 Warn when a local variable shadows another local variable or parameter
5345 whose type is compatible with that of the shadowing variable. In C++,
5346 type compatibility here means the type of the shadowing variable can be
5347 converted to that of the shadowed variable. The creation of this flag
5348 (in addition to @option{-Wshadow=local}) is based on the idea that when
5349 a local variable shadows another one of incompatible type, it is most
5350 likely intentional, not a bug or typo, as shown in the following example:
5354 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5356 for (int i = 0; i < N; ++i)
5365 Since the two variable @code{i} in the example above have incompatible types,
5366 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5367 Because their types are incompatible, if a programmer accidentally uses one
5368 in place of the other, type checking will catch that and emit an error or
5369 warning. So not warning (about shadowing) in this case will not lead to
5370 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5371 possibly reduce the number of warnings triggered by intentional shadowing.
5373 This warning is enabled by @option{-Wshadow=local}.
5375 @item -Wlarger-than=@var{len}
5376 @opindex Wlarger-than=@var{len}
5377 @opindex Wlarger-than-@var{len}
5378 Warn whenever an object of larger than @var{len} bytes is defined.
5380 @item -Wframe-larger-than=@var{len}
5381 @opindex Wframe-larger-than
5382 Warn if the size of a function frame is larger than @var{len} bytes.
5383 The computation done to determine the stack frame size is approximate
5384 and not conservative.
5385 The actual requirements may be somewhat greater than @var{len}
5386 even if you do not get a warning. In addition, any space allocated
5387 via @code{alloca}, variable-length arrays, or related constructs
5388 is not included by the compiler when determining
5389 whether or not to issue a warning.
5391 @item -Wno-free-nonheap-object
5392 @opindex Wno-free-nonheap-object
5393 @opindex Wfree-nonheap-object
5394 Do not warn when attempting to free an object that was not allocated
5397 @item -Wstack-usage=@var{len}
5398 @opindex Wstack-usage
5399 Warn if the stack usage of a function might be larger than @var{len} bytes.
5400 The computation done to determine the stack usage is conservative.
5401 Any space allocated via @code{alloca}, variable-length arrays, or related
5402 constructs is included by the compiler when determining whether or not to
5405 The message is in keeping with the output of @option{-fstack-usage}.
5409 If the stack usage is fully static but exceeds the specified amount, it's:
5412 warning: stack usage is 1120 bytes
5415 If the stack usage is (partly) dynamic but bounded, it's:
5418 warning: stack usage might be 1648 bytes
5421 If the stack usage is (partly) dynamic and not bounded, it's:
5424 warning: stack usage might be unbounded
5428 @item -Wunsafe-loop-optimizations
5429 @opindex Wunsafe-loop-optimizations
5430 @opindex Wno-unsafe-loop-optimizations
5431 Warn if the loop cannot be optimized because the compiler cannot
5432 assume anything on the bounds of the loop indices. With
5433 @option{-funsafe-loop-optimizations} warn if the compiler makes
5436 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5437 @opindex Wno-pedantic-ms-format
5438 @opindex Wpedantic-ms-format
5439 When used in combination with @option{-Wformat}
5440 and @option{-pedantic} without GNU extensions, this option
5441 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5442 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5443 which depend on the MS runtime.
5446 @opindex Waligned-new
5447 @opindex Wno-aligned-new
5448 Warn about a new-expression of a type that requires greater alignment
5449 than the @code{alignof(std::max_align_t)} but uses an allocation
5450 function without an explicit alignment parameter. This option is
5451 enabled by @option{-Wall}.
5453 Normally this only warns about global allocation functions, but
5454 @option{-Waligned-new=all} also warns about class member allocation
5457 @item -Wplacement-new
5458 @itemx -Wplacement-new=@var{n}
5459 @opindex Wplacement-new
5460 @opindex Wno-placement-new
5461 Warn about placement new expressions with undefined behavior, such as
5462 constructing an object in a buffer that is smaller than the type of
5463 the object. For example, the placement new expression below is diagnosed
5464 because it attempts to construct an array of 64 integers in a buffer only
5470 This warning is enabled by default.
5473 @item -Wplacement-new=1
5474 This is the default warning level of @option{-Wplacement-new}. At this
5475 level the warning is not issued for some strictly undefined constructs that
5476 GCC allows as extensions for compatibility with legacy code. For example,
5477 the following @code{new} expression is not diagnosed at this level even
5478 though it has undefined behavior according to the C++ standard because
5479 it writes past the end of the one-element array.
5481 struct S @{ int n, a[1]; @};
5482 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5483 new (s->a)int [32]();
5486 @item -Wplacement-new=2
5487 At this level, in addition to diagnosing all the same constructs as at level
5488 1, a diagnostic is also issued for placement new expressions that construct
5489 an object in the last member of structure whose type is an array of a single
5490 element and whose size is less than the size of the object being constructed.
5491 While the previous example would be diagnosed, the following construct makes
5492 use of the flexible member array extension to avoid the warning at level 2.
5494 struct S @{ int n, a[]; @};
5495 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5496 new (s->a)int [32]();
5501 @item -Wpointer-arith
5502 @opindex Wpointer-arith
5503 @opindex Wno-pointer-arith
5504 Warn about anything that depends on the ``size of'' a function type or
5505 of @code{void}. GNU C assigns these types a size of 1, for
5506 convenience in calculations with @code{void *} pointers and pointers
5507 to functions. In C++, warn also when an arithmetic operation involves
5508 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5511 @opindex Wtype-limits
5512 @opindex Wno-type-limits
5513 Warn if a comparison is always true or always false due to the limited
5514 range of the data type, but do not warn for constant expressions. For
5515 example, warn if an unsigned variable is compared against zero with
5516 @code{<} or @code{>=}. This warning is also enabled by
5519 @item -Wbad-function-cast @r{(C and Objective-C only)}
5520 @opindex Wbad-function-cast
5521 @opindex Wno-bad-function-cast
5522 Warn when a function call is cast to a non-matching type.
5523 For example, warn if a call to a function returning an integer type
5524 is cast to a pointer type.
5526 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5527 @opindex Wc90-c99-compat
5528 @opindex Wno-c90-c99-compat
5529 Warn about features not present in ISO C90, but present in ISO C99.
5530 For instance, warn about use of variable length arrays, @code{long long}
5531 type, @code{bool} type, compound literals, designated initializers, and so
5532 on. This option is independent of the standards mode. Warnings are disabled
5533 in the expression that follows @code{__extension__}.
5535 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5536 @opindex Wc99-c11-compat
5537 @opindex Wno-c99-c11-compat
5538 Warn about features not present in ISO C99, but present in ISO C11.
5539 For instance, warn about use of anonymous structures and unions,
5540 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5541 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5542 and so on. This option is independent of the standards mode. Warnings are
5543 disabled in the expression that follows @code{__extension__}.
5545 @item -Wc++-compat @r{(C and Objective-C only)}
5546 @opindex Wc++-compat
5547 Warn about ISO C constructs that are outside of the common subset of
5548 ISO C and ISO C++, e.g.@: request for implicit conversion from
5549 @code{void *} to a pointer to non-@code{void} type.
5551 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5552 @opindex Wc++11-compat
5553 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5554 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5555 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5556 enabled by @option{-Wall}.
5558 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5559 @opindex Wc++14-compat
5560 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5561 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5565 @opindex Wno-cast-qual
5566 Warn whenever a pointer is cast so as to remove a type qualifier from
5567 the target type. For example, warn if a @code{const char *} is cast
5568 to an ordinary @code{char *}.
5570 Also warn when making a cast that introduces a type qualifier in an
5571 unsafe way. For example, casting @code{char **} to @code{const char **}
5572 is unsafe, as in this example:
5575 /* p is char ** value. */
5576 const char **q = (const char **) p;
5577 /* Assignment of readonly string to const char * is OK. */
5579 /* Now char** pointer points to read-only memory. */
5584 @opindex Wcast-align
5585 @opindex Wno-cast-align
5586 Warn whenever a pointer is cast such that the required alignment of the
5587 target is increased. For example, warn if a @code{char *} is cast to
5588 an @code{int *} on machines where integers can only be accessed at
5589 two- or four-byte boundaries.
5591 @item -Wwrite-strings
5592 @opindex Wwrite-strings
5593 @opindex Wno-write-strings
5594 When compiling C, give string constants the type @code{const
5595 char[@var{length}]} so that copying the address of one into a
5596 non-@code{const} @code{char *} pointer produces a warning. These
5597 warnings help you find at compile time code that can try to write
5598 into a string constant, but only if you have been very careful about
5599 using @code{const} in declarations and prototypes. Otherwise, it is
5600 just a nuisance. This is why we did not make @option{-Wall} request
5603 When compiling C++, warn about the deprecated conversion from string
5604 literals to @code{char *}. This warning is enabled by default for C++
5609 @opindex Wno-clobbered
5610 Warn for variables that might be changed by @code{longjmp} or
5611 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5613 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5614 @opindex Wconditionally-supported
5615 @opindex Wno-conditionally-supported
5616 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5619 @opindex Wconversion
5620 @opindex Wno-conversion
5621 Warn for implicit conversions that may alter a value. This includes
5622 conversions between real and integer, like @code{abs (x)} when
5623 @code{x} is @code{double}; conversions between signed and unsigned,
5624 like @code{unsigned ui = -1}; and conversions to smaller types, like
5625 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5626 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5627 changed by the conversion like in @code{abs (2.0)}. Warnings about
5628 conversions between signed and unsigned integers can be disabled by
5629 using @option{-Wno-sign-conversion}.
5631 For C++, also warn for confusing overload resolution for user-defined
5632 conversions; and conversions that never use a type conversion
5633 operator: conversions to @code{void}, the same type, a base class or a
5634 reference to them. Warnings about conversions between signed and
5635 unsigned integers are disabled by default in C++ unless
5636 @option{-Wsign-conversion} is explicitly enabled.
5638 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5639 @opindex Wconversion-null
5640 @opindex Wno-conversion-null
5641 Do not warn for conversions between @code{NULL} and non-pointer
5642 types. @option{-Wconversion-null} is enabled by default.
5644 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5645 @opindex Wzero-as-null-pointer-constant
5646 @opindex Wno-zero-as-null-pointer-constant
5647 Warn when a literal @samp{0} is used as null pointer constant. This can
5648 be useful to facilitate the conversion to @code{nullptr} in C++11.
5650 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5651 @opindex Wsubobject-linkage
5652 @opindex Wno-subobject-linkage
5653 Warn if a class type has a base or a field whose type uses the anonymous
5654 namespace or depends on a type with no linkage. If a type A depends on
5655 a type B with no or internal linkage, defining it in multiple
5656 translation units would be an ODR violation because the meaning of B
5657 is different in each translation unit. If A only appears in a single
5658 translation unit, the best way to silence the warning is to give it
5659 internal linkage by putting it in an anonymous namespace as well. The
5660 compiler doesn't give this warning for types defined in the main .C
5661 file, as those are unlikely to have multiple definitions.
5662 @option{-Wsubobject-linkage} is enabled by default.
5664 @item -Wdangling-else
5665 @opindex Wdangling-else
5666 @opindex Wno-dangling-else
5667 Warn about constructions where there may be confusion to which
5668 @code{if} statement an @code{else} branch belongs. Here is an example of
5683 In C/C++, every @code{else} branch belongs to the innermost possible
5684 @code{if} statement, which in this example is @code{if (b)}. This is
5685 often not what the programmer expected, as illustrated in the above
5686 example by indentation the programmer chose. When there is the
5687 potential for this confusion, GCC issues a warning when this flag
5688 is specified. To eliminate the warning, add explicit braces around
5689 the innermost @code{if} statement so there is no way the @code{else}
5690 can belong to the enclosing @code{if}. The resulting code
5707 This warning is enabled by @option{-Wparentheses}.
5711 @opindex Wno-date-time
5712 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5713 are encountered as they might prevent bit-wise-identical reproducible
5716 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5717 @opindex Wdelete-incomplete
5718 @opindex Wno-delete-incomplete
5719 Warn when deleting a pointer to incomplete type, which may cause
5720 undefined behavior at runtime. This warning is enabled by default.
5722 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5723 @opindex Wuseless-cast
5724 @opindex Wno-useless-cast
5725 Warn when an expression is casted to its own type.
5728 @opindex Wempty-body
5729 @opindex Wno-empty-body
5730 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5731 while} statement. This warning is also enabled by @option{-Wextra}.
5733 @item -Wenum-compare
5734 @opindex Wenum-compare
5735 @opindex Wno-enum-compare
5736 Warn about a comparison between values of different enumerated types.
5737 In C++ enumeral mismatches in conditional expressions are also
5738 diagnosed and the warning is enabled by default. In C this warning is
5739 enabled by @option{-Wall}.
5741 @item -Wjump-misses-init @r{(C, Objective-C only)}
5742 @opindex Wjump-misses-init
5743 @opindex Wno-jump-misses-init
5744 Warn if a @code{goto} statement or a @code{switch} statement jumps
5745 forward across the initialization of a variable, or jumps backward to a
5746 label after the variable has been initialized. This only warns about
5747 variables that are initialized when they are declared. This warning is
5748 only supported for C and Objective-C; in C++ this sort of branch is an
5751 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5752 can be disabled with the @option{-Wno-jump-misses-init} option.
5754 @item -Wsign-compare
5755 @opindex Wsign-compare
5756 @opindex Wno-sign-compare
5757 @cindex warning for comparison of signed and unsigned values
5758 @cindex comparison of signed and unsigned values, warning
5759 @cindex signed and unsigned values, comparison warning
5760 Warn when a comparison between signed and unsigned values could produce
5761 an incorrect result when the signed value is converted to unsigned.
5762 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5763 also enabled by @option{-Wextra}.
5765 @item -Wsign-conversion
5766 @opindex Wsign-conversion
5767 @opindex Wno-sign-conversion
5768 Warn for implicit conversions that may change the sign of an integer
5769 value, like assigning a signed integer expression to an unsigned
5770 integer variable. An explicit cast silences the warning. In C, this
5771 option is enabled also by @option{-Wconversion}.
5773 @item -Wfloat-conversion
5774 @opindex Wfloat-conversion
5775 @opindex Wno-float-conversion
5776 Warn for implicit conversions that reduce the precision of a real value.
5777 This includes conversions from real to integer, and from higher precision
5778 real to lower precision real values. This option is also enabled by
5779 @option{-Wconversion}.
5781 @item -Wno-scalar-storage-order
5782 @opindex -Wno-scalar-storage-order
5783 @opindex -Wscalar-storage-order
5784 Do not warn on suspicious constructs involving reverse scalar storage order.
5786 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5787 @opindex Wsized-deallocation
5788 @opindex Wno-sized-deallocation
5789 Warn about a definition of an unsized deallocation function
5791 void operator delete (void *) noexcept;
5792 void operator delete[] (void *) noexcept;
5794 without a definition of the corresponding sized deallocation function
5796 void operator delete (void *, std::size_t) noexcept;
5797 void operator delete[] (void *, std::size_t) noexcept;
5799 or vice versa. Enabled by @option{-Wextra} along with
5800 @option{-fsized-deallocation}.
5802 @item -Wsizeof-pointer-memaccess
5803 @opindex Wsizeof-pointer-memaccess
5804 @opindex Wno-sizeof-pointer-memaccess
5805 Warn for suspicious length parameters to certain string and memory built-in
5806 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5807 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5808 but a pointer, and suggests a possible fix, or about
5809 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5812 @item -Wsizeof-array-argument
5813 @opindex Wsizeof-array-argument
5814 @opindex Wno-sizeof-array-argument
5815 Warn when the @code{sizeof} operator is applied to a parameter that is
5816 declared as an array in a function definition. This warning is enabled by
5817 default for C and C++ programs.
5819 @item -Wmemset-elt-size
5820 @opindex Wmemset-elt-size
5821 @opindex Wno-memset-elt-size
5822 Warn for suspicious calls to the @code{memset} built-in function, if the
5823 first argument references an array, and the third argument is a number
5824 equal to the number of elements, but not equal to the size of the array
5825 in memory. This indicates that the user has omitted a multiplication by
5826 the element size. This warning is enabled by @option{-Wall}.
5828 @item -Wmemset-transposed-args
5829 @opindex Wmemset-transposed-args
5830 @opindex Wno-memset-transposed-args
5831 Warn for suspicious calls to the @code{memset} built-in function, if the
5832 second argument is not zero and the third argument is zero. This warns e.g.@
5833 about @code{memset (buf, sizeof buf, 0)} where most probably
5834 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5835 is only emitted if the third argument is literal zero. If it is some
5836 expression that is folded to zero, a cast of zero to some type, etc.,
5837 it is far less likely that the user has mistakenly exchanged the arguments
5838 and no warning is emitted. This warning is enabled by @option{-Wall}.
5842 @opindex Wno-address
5843 Warn about suspicious uses of memory addresses. These include using
5844 the address of a function in a conditional expression, such as
5845 @code{void func(void); if (func)}, and comparisons against the memory
5846 address of a string literal, such as @code{if (x == "abc")}. Such
5847 uses typically indicate a programmer error: the address of a function
5848 always evaluates to true, so their use in a conditional usually
5849 indicate that the programmer forgot the parentheses in a function
5850 call; and comparisons against string literals result in unspecified
5851 behavior and are not portable in C, so they usually indicate that the
5852 programmer intended to use @code{strcmp}. This warning is enabled by
5856 @opindex Wlogical-op
5857 @opindex Wno-logical-op
5858 Warn about suspicious uses of logical operators in expressions.
5859 This includes using logical operators in contexts where a
5860 bit-wise operator is likely to be expected. Also warns when
5861 the operands of a logical operator are the same:
5864 if (a < 0 && a < 0) @{ @dots{} @}
5867 @item -Wlogical-not-parentheses
5868 @opindex Wlogical-not-parentheses
5869 @opindex Wno-logical-not-parentheses
5870 Warn about logical not used on the left hand side operand of a comparison.
5871 This option does not warn if the right operand is considered to be a boolean
5872 expression. Its purpose is to detect suspicious code like the following:
5876 if (!a > 1) @{ @dots{} @}
5879 It is possible to suppress the warning by wrapping the LHS into
5882 if ((!a) > 1) @{ @dots{} @}
5885 This warning is enabled by @option{-Wall}.
5887 @item -Waggregate-return
5888 @opindex Waggregate-return
5889 @opindex Wno-aggregate-return
5890 Warn if any functions that return structures or unions are defined or
5891 called. (In languages where you can return an array, this also elicits
5894 @item -Wno-aggressive-loop-optimizations
5895 @opindex Wno-aggressive-loop-optimizations
5896 @opindex Waggressive-loop-optimizations
5897 Warn if in a loop with constant number of iterations the compiler detects
5898 undefined behavior in some statement during one or more of the iterations.
5900 @item -Wno-attributes
5901 @opindex Wno-attributes
5902 @opindex Wattributes
5903 Do not warn if an unexpected @code{__attribute__} is used, such as
5904 unrecognized attributes, function attributes applied to variables,
5905 etc. This does not stop errors for incorrect use of supported
5908 @item -Wno-builtin-macro-redefined
5909 @opindex Wno-builtin-macro-redefined
5910 @opindex Wbuiltin-macro-redefined
5911 Do not warn if certain built-in macros are redefined. This suppresses
5912 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5913 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5915 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5916 @opindex Wstrict-prototypes
5917 @opindex Wno-strict-prototypes
5918 Warn if a function is declared or defined without specifying the
5919 argument types. (An old-style function definition is permitted without
5920 a warning if preceded by a declaration that specifies the argument
5923 @item -Wold-style-declaration @r{(C and Objective-C only)}
5924 @opindex Wold-style-declaration
5925 @opindex Wno-old-style-declaration
5926 Warn for obsolescent usages, according to the C Standard, in a
5927 declaration. For example, warn if storage-class specifiers like
5928 @code{static} are not the first things in a declaration. This warning
5929 is also enabled by @option{-Wextra}.
5931 @item -Wold-style-definition @r{(C and Objective-C only)}
5932 @opindex Wold-style-definition
5933 @opindex Wno-old-style-definition
5934 Warn if an old-style function definition is used. A warning is given
5935 even if there is a previous prototype.
5937 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5938 @opindex Wmissing-parameter-type
5939 @opindex Wno-missing-parameter-type
5940 A function parameter is declared without a type specifier in K&R-style
5947 This warning is also enabled by @option{-Wextra}.
5949 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5950 @opindex Wmissing-prototypes
5951 @opindex Wno-missing-prototypes
5952 Warn if a global function is defined without a previous prototype
5953 declaration. This warning is issued even if the definition itself
5954 provides a prototype. Use this option to detect global functions
5955 that do not have a matching prototype declaration in a header file.
5956 This option is not valid for C++ because all function declarations
5957 provide prototypes and a non-matching declaration declares an
5958 overload rather than conflict with an earlier declaration.
5959 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5961 @item -Wmissing-declarations
5962 @opindex Wmissing-declarations
5963 @opindex Wno-missing-declarations
5964 Warn if a global function is defined without a previous declaration.
5965 Do so even if the definition itself provides a prototype.
5966 Use this option to detect global functions that are not declared in
5967 header files. In C, no warnings are issued for functions with previous
5968 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5969 missing prototypes. In C++, no warnings are issued for function templates,
5970 or for inline functions, or for functions in anonymous namespaces.
5972 @item -Wmissing-field-initializers
5973 @opindex Wmissing-field-initializers
5974 @opindex Wno-missing-field-initializers
5978 Warn if a structure's initializer has some fields missing. For
5979 example, the following code causes such a warning, because
5980 @code{x.h} is implicitly zero:
5983 struct s @{ int f, g, h; @};
5984 struct s x = @{ 3, 4 @};
5987 This option does not warn about designated initializers, so the following
5988 modification does not trigger a warning:
5991 struct s @{ int f, g, h; @};
5992 struct s x = @{ .f = 3, .g = 4 @};
5995 In C++ this option does not warn either about the empty @{ @}
5996 initializer, for example:
5999 struct s @{ int f, g, h; @};
6003 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6004 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6006 @item -Wno-multichar
6007 @opindex Wno-multichar
6009 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6010 Usually they indicate a typo in the user's code, as they have
6011 implementation-defined values, and should not be used in portable code.
6013 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
6014 @opindex Wnormalized=
6015 @opindex Wnormalized
6016 @opindex Wno-normalized
6019 @cindex character set, input normalization
6020 In ISO C and ISO C++, two identifiers are different if they are
6021 different sequences of characters. However, sometimes when characters
6022 outside the basic ASCII character set are used, you can have two
6023 different character sequences that look the same. To avoid confusion,
6024 the ISO 10646 standard sets out some @dfn{normalization rules} which
6025 when applied ensure that two sequences that look the same are turned into
6026 the same sequence. GCC can warn you if you are using identifiers that
6027 have not been normalized; this option controls that warning.
6029 There are four levels of warning supported by GCC@. The default is
6030 @option{-Wnormalized=nfc}, which warns about any identifier that is
6031 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6032 recommended form for most uses. It is equivalent to
6033 @option{-Wnormalized}.
6035 Unfortunately, there are some characters allowed in identifiers by
6036 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6037 identifiers. That is, there's no way to use these symbols in portable
6038 ISO C or C++ and have all your identifiers in NFC@.
6039 @option{-Wnormalized=id} suppresses the warning for these characters.
6040 It is hoped that future versions of the standards involved will correct
6041 this, which is why this option is not the default.
6043 You can switch the warning off for all characters by writing
6044 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6045 only do this if you are using some other normalization scheme (like
6046 ``D''), because otherwise you can easily create bugs that are
6047 literally impossible to see.
6049 Some characters in ISO 10646 have distinct meanings but look identical
6050 in some fonts or display methodologies, especially once formatting has
6051 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6052 LETTER N'', displays just like a regular @code{n} that has been
6053 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6054 normalization scheme to convert all these into a standard form as
6055 well, and GCC warns if your code is not in NFKC if you use
6056 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6057 about every identifier that contains the letter O because it might be
6058 confused with the digit 0, and so is not the default, but may be
6059 useful as a local coding convention if the programming environment
6060 cannot be fixed to display these characters distinctly.
6062 @item -Wno-deprecated
6063 @opindex Wno-deprecated
6064 @opindex Wdeprecated
6065 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6067 @item -Wno-deprecated-declarations
6068 @opindex Wno-deprecated-declarations
6069 @opindex Wdeprecated-declarations
6070 Do not warn about uses of functions (@pxref{Function Attributes}),
6071 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6072 Attributes}) marked as deprecated by using the @code{deprecated}
6076 @opindex Wno-overflow
6078 Do not warn about compile-time overflow in constant expressions.
6083 Warn about One Definition Rule violations during link-time optimization.
6084 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6087 @opindex Wopenm-simd
6088 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6089 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6090 option can be used to relax the cost model.
6092 @item -Woverride-init @r{(C and Objective-C only)}
6093 @opindex Woverride-init
6094 @opindex Wno-override-init
6098 Warn if an initialized field without side effects is overridden when
6099 using designated initializers (@pxref{Designated Inits, , Designated
6102 This warning is included in @option{-Wextra}. To get other
6103 @option{-Wextra} warnings without this one, use @option{-Wextra
6104 -Wno-override-init}.
6106 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6107 @opindex Woverride-init-side-effects
6108 @opindex Wno-override-init-side-effects
6109 Warn if an initialized field with side effects is overridden when
6110 using designated initializers (@pxref{Designated Inits, , Designated
6111 Initializers}). This warning is enabled by default.
6116 Warn if a structure is given the packed attribute, but the packed
6117 attribute has no effect on the layout or size of the structure.
6118 Such structures may be mis-aligned for little benefit. For
6119 instance, in this code, the variable @code{f.x} in @code{struct bar}
6120 is misaligned even though @code{struct bar} does not itself
6121 have the packed attribute:
6128 @} __attribute__((packed));
6136 @item -Wpacked-bitfield-compat
6137 @opindex Wpacked-bitfield-compat
6138 @opindex Wno-packed-bitfield-compat
6139 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6140 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6141 the change can lead to differences in the structure layout. GCC
6142 informs you when the offset of such a field has changed in GCC 4.4.
6143 For example there is no longer a 4-bit padding between field @code{a}
6144 and @code{b} in this structure:
6151 @} __attribute__ ((packed));
6154 This warning is enabled by default. Use
6155 @option{-Wno-packed-bitfield-compat} to disable this warning.
6160 Warn if padding is included in a structure, either to align an element
6161 of the structure or to align the whole structure. Sometimes when this
6162 happens it is possible to rearrange the fields of the structure to
6163 reduce the padding and so make the structure smaller.
6165 @item -Wredundant-decls
6166 @opindex Wredundant-decls
6167 @opindex Wno-redundant-decls
6168 Warn if anything is declared more than once in the same scope, even in
6169 cases where multiple declaration is valid and changes nothing.
6171 @item -Wnested-externs @r{(C and Objective-C only)}
6172 @opindex Wnested-externs
6173 @opindex Wno-nested-externs
6174 Warn if an @code{extern} declaration is encountered within a function.
6176 @item -Wno-inherited-variadic-ctor
6177 @opindex Winherited-variadic-ctor
6178 @opindex Wno-inherited-variadic-ctor
6179 Suppress warnings about use of C++11 inheriting constructors when the
6180 base class inherited from has a C variadic constructor; the warning is
6181 on by default because the ellipsis is not inherited.
6186 Warn if a function that is declared as inline cannot be inlined.
6187 Even with this option, the compiler does not warn about failures to
6188 inline functions declared in system headers.
6190 The compiler uses a variety of heuristics to determine whether or not
6191 to inline a function. For example, the compiler takes into account
6192 the size of the function being inlined and the amount of inlining
6193 that has already been done in the current function. Therefore,
6194 seemingly insignificant changes in the source program can cause the
6195 warnings produced by @option{-Winline} to appear or disappear.
6197 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6198 @opindex Wno-invalid-offsetof
6199 @opindex Winvalid-offsetof
6200 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6201 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6202 to a non-standard-layout type is undefined. In existing C++ implementations,
6203 however, @code{offsetof} typically gives meaningful results.
6204 This flag is for users who are aware that they are
6205 writing nonportable code and who have deliberately chosen to ignore the
6208 The restrictions on @code{offsetof} may be relaxed in a future version
6209 of the C++ standard.
6211 @item -Wint-in-bool-context
6212 @opindex Wint-in-bool-context
6213 @opindex Wno-int-in-bool-context
6214 Warn for suspicious use of integer values where boolean values are expected,
6215 such as conditional expressions (?:) using non-boolean integer constants in
6216 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6217 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6218 for all kinds of multiplications regardless of the data type.
6219 This warning is enabled by @option{-Wall}.
6221 @item -Wno-int-to-pointer-cast
6222 @opindex Wno-int-to-pointer-cast
6223 @opindex Wint-to-pointer-cast
6224 Suppress warnings from casts to pointer type of an integer of a
6225 different size. In C++, casting to a pointer type of smaller size is
6226 an error. @option{Wint-to-pointer-cast} is enabled by default.
6229 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6230 @opindex Wno-pointer-to-int-cast
6231 @opindex Wpointer-to-int-cast
6232 Suppress warnings from casts from a pointer to an integer type of a
6236 @opindex Winvalid-pch
6237 @opindex Wno-invalid-pch
6238 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6239 the search path but can't be used.
6243 @opindex Wno-long-long
6244 Warn if @code{long long} type is used. This is enabled by either
6245 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6246 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6248 @item -Wvariadic-macros
6249 @opindex Wvariadic-macros
6250 @opindex Wno-variadic-macros
6251 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6252 alternate syntax is used in ISO C99 mode. This is enabled by either
6253 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6254 messages, use @option{-Wno-variadic-macros}.
6258 @opindex Wno-varargs
6259 Warn upon questionable usage of the macros used to handle variable
6260 arguments like @code{va_start}. This is default. To inhibit the
6261 warning messages, use @option{-Wno-varargs}.
6263 @item -Wvector-operation-performance
6264 @opindex Wvector-operation-performance
6265 @opindex Wno-vector-operation-performance
6266 Warn if vector operation is not implemented via SIMD capabilities of the
6267 architecture. Mainly useful for the performance tuning.
6268 Vector operation can be implemented @code{piecewise}, which means that the
6269 scalar operation is performed on every vector element;
6270 @code{in parallel}, which means that the vector operation is implemented
6271 using scalars of wider type, which normally is more performance efficient;
6272 and @code{as a single scalar}, which means that vector fits into a
6275 @item -Wno-virtual-move-assign
6276 @opindex Wvirtual-move-assign
6277 @opindex Wno-virtual-move-assign
6278 Suppress warnings about inheriting from a virtual base with a
6279 non-trivial C++11 move assignment operator. This is dangerous because
6280 if the virtual base is reachable along more than one path, it is
6281 moved multiple times, which can mean both objects end up in the
6282 moved-from state. If the move assignment operator is written to avoid
6283 moving from a moved-from object, this warning can be disabled.
6288 Warn if a variable-length array is used in the code.
6289 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6290 the variable-length array.
6292 @item -Wvla-larger-than=@var{n}
6293 If this option is used, the compiler will warn on uses of
6294 variable-length arrays where the size is either unbounded, or bounded
6295 by an argument that can be larger than @var{n} bytes. This is similar
6296 to how @option{-Walloca-larger-than=@var{n}} works, but with
6297 variable-length arrays.
6299 Note that GCC may optimize small variable-length arrays of a known
6300 value into plain arrays, so this warning may not get triggered for
6303 This warning is not enabled by @option{-Wall}, and is only active when
6304 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6306 See also @option{-Walloca-larger-than=@var{n}}.
6308 @item -Wvolatile-register-var
6309 @opindex Wvolatile-register-var
6310 @opindex Wno-volatile-register-var
6311 Warn if a register variable is declared volatile. The volatile
6312 modifier does not inhibit all optimizations that may eliminate reads
6313 and/or writes to register variables. This warning is enabled by
6316 @item -Wdisabled-optimization
6317 @opindex Wdisabled-optimization
6318 @opindex Wno-disabled-optimization
6319 Warn if a requested optimization pass is disabled. This warning does
6320 not generally indicate that there is anything wrong with your code; it
6321 merely indicates that GCC's optimizers are unable to handle the code
6322 effectively. Often, the problem is that your code is too big or too
6323 complex; GCC refuses to optimize programs when the optimization
6324 itself is likely to take inordinate amounts of time.
6326 @item -Wpointer-sign @r{(C and Objective-C only)}
6327 @opindex Wpointer-sign
6328 @opindex Wno-pointer-sign
6329 Warn for pointer argument passing or assignment with different signedness.
6330 This option is only supported for C and Objective-C@. It is implied by
6331 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6332 @option{-Wno-pointer-sign}.
6334 @item -Wstack-protector
6335 @opindex Wstack-protector
6336 @opindex Wno-stack-protector
6337 This option is only active when @option{-fstack-protector} is active. It
6338 warns about functions that are not protected against stack smashing.
6340 @item -Woverlength-strings
6341 @opindex Woverlength-strings
6342 @opindex Wno-overlength-strings
6343 Warn about string constants that are longer than the ``minimum
6344 maximum'' length specified in the C standard. Modern compilers
6345 generally allow string constants that are much longer than the
6346 standard's minimum limit, but very portable programs should avoid
6347 using longer strings.
6349 The limit applies @emph{after} string constant concatenation, and does
6350 not count the trailing NUL@. In C90, the limit was 509 characters; in
6351 C99, it was raised to 4095. C++98 does not specify a normative
6352 minimum maximum, so we do not diagnose overlength strings in C++@.
6354 This option is implied by @option{-Wpedantic}, and can be disabled with
6355 @option{-Wno-overlength-strings}.
6357 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6358 @opindex Wunsuffixed-float-constants
6360 Issue a warning for any floating constant that does not have
6361 a suffix. When used together with @option{-Wsystem-headers} it
6362 warns about such constants in system header files. This can be useful
6363 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6364 from the decimal floating-point extension to C99.
6366 @item -Wno-designated-init @r{(C and Objective-C only)}
6367 Suppress warnings when a positional initializer is used to initialize
6368 a structure that has been marked with the @code{designated_init}
6372 Issue a warning when HSAIL cannot be emitted for the compiled function or
6377 @node Debugging Options
6378 @section Options for Debugging Your Program
6379 @cindex options, debugging
6380 @cindex debugging information options
6382 To tell GCC to emit extra information for use by a debugger, in almost
6383 all cases you need only to add @option{-g} to your other options.
6385 GCC allows you to use @option{-g} with
6386 @option{-O}. The shortcuts taken by optimized code may occasionally
6387 be surprising: some variables you declared may not exist
6388 at all; flow of control may briefly move where you did not expect it;
6389 some statements may not be executed because they compute constant
6390 results or their values are already at hand; some statements may
6391 execute in different places because they have been moved out of loops.
6392 Nevertheless it is possible to debug optimized output. This makes
6393 it reasonable to use the optimizer for programs that might have bugs.
6395 If you are not using some other optimization option, consider
6396 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6397 With no @option{-O} option at all, some compiler passes that collect
6398 information useful for debugging do not run at all, so that
6399 @option{-Og} may result in a better debugging experience.
6404 Produce debugging information in the operating system's native format
6405 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6408 On most systems that use stabs format, @option{-g} enables use of extra
6409 debugging information that only GDB can use; this extra information
6410 makes debugging work better in GDB but probably makes other debuggers
6412 refuse to read the program. If you want to control for certain whether
6413 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6414 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6418 Produce debugging information for use by GDB@. This means to use the
6419 most expressive format available (DWARF, stabs, or the native format
6420 if neither of those are supported), including GDB extensions if at all
6424 @itemx -gdwarf-@var{version}
6426 Produce debugging information in DWARF format (if that is supported).
6427 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6428 for most targets is 4. DWARF Version 5 is only experimental.
6430 Note that with DWARF Version 2, some ports require and always
6431 use some non-conflicting DWARF 3 extensions in the unwind tables.
6433 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6434 for maximum benefit.
6436 GCC no longer supports DWARF Version 1, which is substantially
6437 different than Version 2 and later. For historical reasons, some
6438 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
6439 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6440 in their names, but apply to all currently-supported versions of DWARF.
6444 Produce debugging information in stabs format (if that is supported),
6445 without GDB extensions. This is the format used by DBX on most BSD
6446 systems. On MIPS, Alpha and System V Release 4 systems this option
6447 produces stabs debugging output that is not understood by DBX or SDB@.
6448 On System V Release 4 systems this option requires the GNU assembler.
6452 Produce debugging information in stabs format (if that is supported),
6453 using GNU extensions understood only by the GNU debugger (GDB)@. The
6454 use of these extensions is likely to make other debuggers crash or
6455 refuse to read the program.
6459 Produce debugging information in COFF format (if that is supported).
6460 This is the format used by SDB on most System V systems prior to
6465 Produce debugging information in XCOFF format (if that is supported).
6466 This is the format used by the DBX debugger on IBM RS/6000 systems.
6470 Produce debugging information in XCOFF format (if that is supported),
6471 using GNU extensions understood only by the GNU debugger (GDB)@. The
6472 use of these extensions is likely to make other debuggers crash or
6473 refuse to read the program, and may cause assemblers other than the GNU
6474 assembler (GAS) to fail with an error.
6478 Produce debugging information in Alpha/VMS debug format (if that is
6479 supported). This is the format used by DEBUG on Alpha/VMS systems.
6482 @itemx -ggdb@var{level}
6483 @itemx -gstabs@var{level}
6484 @itemx -gcoff@var{level}
6485 @itemx -gxcoff@var{level}
6486 @itemx -gvms@var{level}
6487 Request debugging information and also use @var{level} to specify how
6488 much information. The default level is 2.
6490 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6493 Level 1 produces minimal information, enough for making backtraces in
6494 parts of the program that you don't plan to debug. This includes
6495 descriptions of functions and external variables, and line number
6496 tables, but no information about local variables.
6498 Level 3 includes extra information, such as all the macro definitions
6499 present in the program. Some debuggers support macro expansion when
6500 you use @option{-g3}.
6502 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6503 confusion with @option{-gdwarf-@var{level}}.
6504 Instead use an additional @option{-g@var{level}} option to change the
6505 debug level for DWARF.
6507 @item -feliminate-unused-debug-symbols
6508 @opindex feliminate-unused-debug-symbols
6509 Produce debugging information in stabs format (if that is supported),
6510 for only symbols that are actually used.
6512 @item -femit-class-debug-always
6513 @opindex femit-class-debug-always
6514 Instead of emitting debugging information for a C++ class in only one
6515 object file, emit it in all object files using the class. This option
6516 should be used only with debuggers that are unable to handle the way GCC
6517 normally emits debugging information for classes because using this
6518 option increases the size of debugging information by as much as a
6521 @item -fno-merge-debug-strings
6522 @opindex fmerge-debug-strings
6523 @opindex fno-merge-debug-strings
6524 Direct the linker to not merge together strings in the debugging
6525 information that are identical in different object files. Merging is
6526 not supported by all assemblers or linkers. Merging decreases the size
6527 of the debug information in the output file at the cost of increasing
6528 link processing time. Merging is enabled by default.
6530 @item -fdebug-prefix-map=@var{old}=@var{new}
6531 @opindex fdebug-prefix-map
6532 When compiling files in directory @file{@var{old}}, record debugging
6533 information describing them as in @file{@var{new}} instead.
6535 @item -fvar-tracking
6536 @opindex fvar-tracking
6537 Run variable tracking pass. It computes where variables are stored at each
6538 position in code. Better debugging information is then generated
6539 (if the debugging information format supports this information).
6541 It is enabled by default when compiling with optimization (@option{-Os},
6542 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6543 the debug info format supports it.
6545 @item -fvar-tracking-assignments
6546 @opindex fvar-tracking-assignments
6547 @opindex fno-var-tracking-assignments
6548 Annotate assignments to user variables early in the compilation and
6549 attempt to carry the annotations over throughout the compilation all the
6550 way to the end, in an attempt to improve debug information while
6551 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6553 It can be enabled even if var-tracking is disabled, in which case
6554 annotations are created and maintained, but discarded at the end.
6555 By default, this flag is enabled together with @option{-fvar-tracking},
6556 except when selective scheduling is enabled.
6559 @opindex gsplit-dwarf
6560 Separate as much DWARF debugging information as possible into a
6561 separate output file with the extension @file{.dwo}. This option allows
6562 the build system to avoid linking files with debug information. To
6563 be useful, this option requires a debugger capable of reading @file{.dwo}
6568 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6570 @item -ggnu-pubnames
6571 @opindex ggnu-pubnames
6572 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6573 suitable for conversion into a GDB@ index. This option is only useful
6574 with a linker that can produce GDB@ index version 7.
6576 @item -fdebug-types-section
6577 @opindex fdebug-types-section
6578 @opindex fno-debug-types-section
6579 When using DWARF Version 4 or higher, type DIEs can be put into
6580 their own @code{.debug_types} section instead of making them part of the
6581 @code{.debug_info} section. It is more efficient to put them in a separate
6582 comdat sections since the linker can then remove duplicates.
6583 But not all DWARF consumers support @code{.debug_types} sections yet
6584 and on some objects @code{.debug_types} produces larger instead of smaller
6585 debugging information.
6587 @item -grecord-gcc-switches
6588 @item -gno-record-gcc-switches
6589 @opindex grecord-gcc-switches
6590 @opindex gno-record-gcc-switches
6591 This switch causes the command-line options used to invoke the
6592 compiler that may affect code generation to be appended to the
6593 DW_AT_producer attribute in DWARF debugging information. The options
6594 are concatenated with spaces separating them from each other and from
6595 the compiler version.
6596 It is enabled by default.
6597 See also @option{-frecord-gcc-switches} for another
6598 way of storing compiler options into the object file.
6600 @item -gstrict-dwarf
6601 @opindex gstrict-dwarf
6602 Disallow using extensions of later DWARF standard version than selected
6603 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6604 DWARF extensions from later standard versions is allowed.
6606 @item -gno-strict-dwarf
6607 @opindex gno-strict-dwarf
6608 Allow using extensions of later DWARF standard version than selected with
6609 @option{-gdwarf-@var{version}}.
6611 @item -gz@r{[}=@var{type}@r{]}
6613 Produce compressed debug sections in DWARF format, if that is supported.
6614 If @var{type} is not given, the default type depends on the capabilities
6615 of the assembler and linker used. @var{type} may be one of
6616 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6617 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6618 compression in traditional GNU format). If the linker doesn't support
6619 writing compressed debug sections, the option is rejected. Otherwise,
6620 if the assembler does not support them, @option{-gz} is silently ignored
6621 when producing object files.
6623 @item -feliminate-dwarf2-dups
6624 @opindex feliminate-dwarf2-dups
6625 Compress DWARF debugging information by eliminating duplicated
6626 information about each symbol. This option only makes sense when
6627 generating DWARF debugging information.
6629 @item -femit-struct-debug-baseonly
6630 @opindex femit-struct-debug-baseonly
6631 Emit debug information for struct-like types
6632 only when the base name of the compilation source file
6633 matches the base name of file in which the struct is defined.
6635 This option substantially reduces the size of debugging information,
6636 but at significant potential loss in type information to the debugger.
6637 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6638 See @option{-femit-struct-debug-detailed} for more detailed control.
6640 This option works only with DWARF debug output.
6642 @item -femit-struct-debug-reduced
6643 @opindex femit-struct-debug-reduced
6644 Emit debug information for struct-like types
6645 only when the base name of the compilation source file
6646 matches the base name of file in which the type is defined,
6647 unless the struct is a template or defined in a system header.
6649 This option significantly reduces the size of debugging information,
6650 with some potential loss in type information to the debugger.
6651 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6652 See @option{-femit-struct-debug-detailed} for more detailed control.
6654 This option works only with DWARF debug output.
6656 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6657 @opindex femit-struct-debug-detailed
6658 Specify the struct-like types
6659 for which the compiler generates debug information.
6660 The intent is to reduce duplicate struct debug information
6661 between different object files within the same program.
6663 This option is a detailed version of
6664 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6665 which serves for most needs.
6667 A specification has the syntax@*
6668 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6670 The optional first word limits the specification to
6671 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6672 A struct type is used directly when it is the type of a variable, member.
6673 Indirect uses arise through pointers to structs.
6674 That is, when use of an incomplete struct is valid, the use is indirect.
6676 @samp{struct one direct; struct two * indirect;}.
6678 The optional second word limits the specification to
6679 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6680 Generic structs are a bit complicated to explain.
6681 For C++, these are non-explicit specializations of template classes,
6682 or non-template classes within the above.
6683 Other programming languages have generics,
6684 but @option{-femit-struct-debug-detailed} does not yet implement them.
6686 The third word specifies the source files for those
6687 structs for which the compiler should emit debug information.
6688 The values @samp{none} and @samp{any} have the normal meaning.
6689 The value @samp{base} means that
6690 the base of name of the file in which the type declaration appears
6691 must match the base of the name of the main compilation file.
6692 In practice, this means that when compiling @file{foo.c}, debug information
6693 is generated for types declared in that file and @file{foo.h},
6694 but not other header files.
6695 The value @samp{sys} means those types satisfying @samp{base}
6696 or declared in system or compiler headers.
6698 You may need to experiment to determine the best settings for your application.
6700 The default is @option{-femit-struct-debug-detailed=all}.
6702 This option works only with DWARF debug output.
6704 @item -fno-dwarf2-cfi-asm
6705 @opindex fdwarf2-cfi-asm
6706 @opindex fno-dwarf2-cfi-asm
6707 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6708 instead of using GAS @code{.cfi_*} directives.
6710 @item -fno-eliminate-unused-debug-types
6711 @opindex feliminate-unused-debug-types
6712 @opindex fno-eliminate-unused-debug-types
6713 Normally, when producing DWARF output, GCC avoids producing debug symbol
6714 output for types that are nowhere used in the source file being compiled.
6715 Sometimes it is useful to have GCC emit debugging
6716 information for all types declared in a compilation
6717 unit, regardless of whether or not they are actually used
6718 in that compilation unit, for example
6719 if, in the debugger, you want to cast a value to a type that is
6720 not actually used in your program (but is declared). More often,
6721 however, this results in a significant amount of wasted space.
6724 @node Optimize Options
6725 @section Options That Control Optimization
6726 @cindex optimize options
6727 @cindex options, optimization
6729 These options control various sorts of optimizations.
6731 Without any optimization option, the compiler's goal is to reduce the
6732 cost of compilation and to make debugging produce the expected
6733 results. Statements are independent: if you stop the program with a
6734 breakpoint between statements, you can then assign a new value to any
6735 variable or change the program counter to any other statement in the
6736 function and get exactly the results you expect from the source
6739 Turning on optimization flags makes the compiler attempt to improve
6740 the performance and/or code size at the expense of compilation time
6741 and possibly the ability to debug the program.
6743 The compiler performs optimization based on the knowledge it has of the
6744 program. Compiling multiple files at once to a single output file mode allows
6745 the compiler to use information gained from all of the files when compiling
6748 Not all optimizations are controlled directly by a flag. Only
6749 optimizations that have a flag are listed in this section.
6751 Most optimizations are only enabled if an @option{-O} level is set on
6752 the command line. Otherwise they are disabled, even if individual
6753 optimization flags are specified.
6755 Depending on the target and how GCC was configured, a slightly different
6756 set of optimizations may be enabled at each @option{-O} level than
6757 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6758 to find out the exact set of optimizations that are enabled at each level.
6759 @xref{Overall Options}, for examples.
6766 Optimize. Optimizing compilation takes somewhat more time, and a lot
6767 more memory for a large function.
6769 With @option{-O}, the compiler tries to reduce code size and execution
6770 time, without performing any optimizations that take a great deal of
6773 @option{-O} turns on the following optimization flags:
6776 -fbranch-count-reg @gol
6777 -fcombine-stack-adjustments @gol
6779 -fcprop-registers @gol
6782 -fdelayed-branch @gol
6784 -fforward-propagate @gol
6785 -fguess-branch-probability @gol
6786 -fif-conversion2 @gol
6787 -fif-conversion @gol
6788 -finline-functions-called-once @gol
6789 -fipa-pure-const @gol
6791 -fipa-reference @gol
6792 -fmerge-constants @gol
6793 -fmove-loop-invariants @gol
6794 -freorder-blocks @gol
6796 -fshrink-wrap-separate @gol
6797 -fsplit-wide-types @gol
6800 -fstore-merging @gol
6804 -ftree-coalesce-vars @gol
6805 -ftree-copy-prop @gol
6807 -ftree-dominator-opts @gol
6809 -ftree-forwprop @gol
6819 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6820 where doing so does not interfere with debugging.
6824 Optimize even more. GCC performs nearly all supported optimizations
6825 that do not involve a space-speed tradeoff.
6826 As compared to @option{-O}, this option increases both compilation time
6827 and the performance of the generated code.
6829 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6830 also turns on the following optimization flags:
6831 @gccoptlist{-fthread-jumps @gol
6832 -falign-functions -falign-jumps @gol
6833 -falign-loops -falign-labels @gol
6836 -fcse-follow-jumps -fcse-skip-blocks @gol
6837 -fdelete-null-pointer-checks @gol
6838 -fdevirtualize -fdevirtualize-speculatively @gol
6839 -fexpensive-optimizations @gol
6840 -fgcse -fgcse-lm @gol
6841 -fhoist-adjacent-loads @gol
6842 -finline-small-functions @gol
6843 -findirect-inlining @gol
6845 -fipa-cp-alignment @gol
6849 -fisolate-erroneous-paths-dereference @gol
6851 -foptimize-sibling-calls @gol
6852 -foptimize-strlen @gol
6853 -fpartial-inlining @gol
6855 -freorder-blocks-algorithm=stc @gol
6856 -freorder-blocks-and-partition -freorder-functions @gol
6857 -frerun-cse-after-loop @gol
6858 -fsched-interblock -fsched-spec @gol
6859 -fschedule-insns -fschedule-insns2 @gol
6860 -fstrict-aliasing -fstrict-overflow @gol
6861 -ftree-builtin-call-dce @gol
6862 -ftree-switch-conversion -ftree-tail-merge @gol
6863 -fcode-hoisting @gol
6868 Please note the warning under @option{-fgcse} about
6869 invoking @option{-O2} on programs that use computed gotos.
6873 Optimize yet more. @option{-O3} turns on all optimizations specified
6874 by @option{-O2} and also turns on the @option{-finline-functions},
6875 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6876 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6877 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
6878 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6879 @option{-ftree-partial-pre}, @option{-fpeel-loops}
6880 and @option{-fipa-cp-clone} options.
6884 Reduce compilation time and make debugging produce the expected
6885 results. This is the default.
6889 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6890 do not typically increase code size. It also performs further
6891 optimizations designed to reduce code size.
6893 @option{-Os} disables the following optimization flags:
6894 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6895 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
6896 -freorder-blocks-and-partition -fprefetch-loop-arrays}
6900 Disregard strict standards compliance. @option{-Ofast} enables all
6901 @option{-O3} optimizations. It also enables optimizations that are not
6902 valid for all standard-compliant programs.
6903 It turns on @option{-ffast-math} and the Fortran-specific
6904 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6908 Optimize debugging experience. @option{-Og} enables optimizations
6909 that do not interfere with debugging. It should be the optimization
6910 level of choice for the standard edit-compile-debug cycle, offering
6911 a reasonable level of optimization while maintaining fast compilation
6912 and a good debugging experience.
6915 If you use multiple @option{-O} options, with or without level numbers,
6916 the last such option is the one that is effective.
6918 Options of the form @option{-f@var{flag}} specify machine-independent
6919 flags. Most flags have both positive and negative forms; the negative
6920 form of @option{-ffoo} is @option{-fno-foo}. In the table
6921 below, only one of the forms is listed---the one you typically
6922 use. You can figure out the other form by either removing @samp{no-}
6925 The following options control specific optimizations. They are either
6926 activated by @option{-O} options or are related to ones that are. You
6927 can use the following flags in the rare cases when ``fine-tuning'' of
6928 optimizations to be performed is desired.
6931 @item -fno-defer-pop
6932 @opindex fno-defer-pop
6933 Always pop the arguments to each function call as soon as that function
6934 returns. For machines that must pop arguments after a function call,
6935 the compiler normally lets arguments accumulate on the stack for several
6936 function calls and pops them all at once.
6938 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6940 @item -fforward-propagate
6941 @opindex fforward-propagate
6942 Perform a forward propagation pass on RTL@. The pass tries to combine two
6943 instructions and checks if the result can be simplified. If loop unrolling
6944 is active, two passes are performed and the second is scheduled after
6947 This option is enabled by default at optimization levels @option{-O},
6948 @option{-O2}, @option{-O3}, @option{-Os}.
6950 @item -ffp-contract=@var{style}
6951 @opindex ffp-contract
6952 @option{-ffp-contract=off} disables floating-point expression contraction.
6953 @option{-ffp-contract=fast} enables floating-point expression contraction
6954 such as forming of fused multiply-add operations if the target has
6955 native support for them.
6956 @option{-ffp-contract=on} enables floating-point expression contraction
6957 if allowed by the language standard. This is currently not implemented
6958 and treated equal to @option{-ffp-contract=off}.
6960 The default is @option{-ffp-contract=fast}.
6962 @item -fomit-frame-pointer
6963 @opindex fomit-frame-pointer
6964 Don't keep the frame pointer in a register for functions that
6965 don't need one. This avoids the instructions to save, set up and
6966 restore frame pointers; it also makes an extra register available
6967 in many functions. @strong{It also makes debugging impossible on
6970 On some machines, such as the VAX, this flag has no effect, because
6971 the standard calling sequence automatically handles the frame pointer
6972 and nothing is saved by pretending it doesn't exist. The
6973 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6974 whether a target machine supports this flag. @xref{Registers,,Register
6975 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6977 The default setting (when not optimizing for
6978 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
6979 @option{-fomit-frame-pointer}. You can configure GCC with the
6980 @option{--enable-frame-pointer} configure option to change the default.
6982 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6984 @item -foptimize-sibling-calls
6985 @opindex foptimize-sibling-calls
6986 Optimize sibling and tail recursive calls.
6988 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6990 @item -foptimize-strlen
6991 @opindex foptimize-strlen
6992 Optimize various standard C string functions (e.g. @code{strlen},
6993 @code{strchr} or @code{strcpy}) and
6994 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
6996 Enabled at levels @option{-O2}, @option{-O3}.
7000 Do not expand any functions inline apart from those marked with
7001 the @code{always_inline} attribute. This is the default when not
7004 Single functions can be exempted from inlining by marking them
7005 with the @code{noinline} attribute.
7007 @item -finline-small-functions
7008 @opindex finline-small-functions
7009 Integrate functions into their callers when their body is smaller than expected
7010 function call code (so overall size of program gets smaller). The compiler
7011 heuristically decides which functions are simple enough to be worth integrating
7012 in this way. This inlining applies to all functions, even those not declared
7015 Enabled at level @option{-O2}.
7017 @item -findirect-inlining
7018 @opindex findirect-inlining
7019 Inline also indirect calls that are discovered to be known at compile
7020 time thanks to previous inlining. This option has any effect only
7021 when inlining itself is turned on by the @option{-finline-functions}
7022 or @option{-finline-small-functions} options.
7024 Enabled at level @option{-O2}.
7026 @item -finline-functions
7027 @opindex finline-functions
7028 Consider all functions for inlining, even if they are not declared inline.
7029 The compiler heuristically decides which functions are worth integrating
7032 If all calls to a given function are integrated, and the function is
7033 declared @code{static}, then the function is normally not output as
7034 assembler code in its own right.
7036 Enabled at level @option{-O3}.
7038 @item -finline-functions-called-once
7039 @opindex finline-functions-called-once
7040 Consider all @code{static} functions called once for inlining into their
7041 caller even if they are not marked @code{inline}. If a call to a given
7042 function is integrated, then the function is not output as assembler code
7045 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7047 @item -fearly-inlining
7048 @opindex fearly-inlining
7049 Inline functions marked by @code{always_inline} and functions whose body seems
7050 smaller than the function call overhead early before doing
7051 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7052 makes profiling significantly cheaper and usually inlining faster on programs
7053 having large chains of nested wrapper functions.
7059 Perform interprocedural scalar replacement of aggregates, removal of
7060 unused parameters and replacement of parameters passed by reference
7061 by parameters passed by value.
7063 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7065 @item -finline-limit=@var{n}
7066 @opindex finline-limit
7067 By default, GCC limits the size of functions that can be inlined. This flag
7068 allows coarse control of this limit. @var{n} is the size of functions that
7069 can be inlined in number of pseudo instructions.
7071 Inlining is actually controlled by a number of parameters, which may be
7072 specified individually by using @option{--param @var{name}=@var{value}}.
7073 The @option{-finline-limit=@var{n}} option sets some of these parameters
7077 @item max-inline-insns-single
7078 is set to @var{n}/2.
7079 @item max-inline-insns-auto
7080 is set to @var{n}/2.
7083 See below for a documentation of the individual
7084 parameters controlling inlining and for the defaults of these parameters.
7086 @emph{Note:} there may be no value to @option{-finline-limit} that results
7087 in default behavior.
7089 @emph{Note:} pseudo instruction represents, in this particular context, an
7090 abstract measurement of function's size. In no way does it represent a count
7091 of assembly instructions and as such its exact meaning might change from one
7092 release to an another.
7094 @item -fno-keep-inline-dllexport
7095 @opindex fno-keep-inline-dllexport
7096 This is a more fine-grained version of @option{-fkeep-inline-functions},
7097 which applies only to functions that are declared using the @code{dllexport}
7098 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7101 @item -fkeep-inline-functions
7102 @opindex fkeep-inline-functions
7103 In C, emit @code{static} functions that are declared @code{inline}
7104 into the object file, even if the function has been inlined into all
7105 of its callers. This switch does not affect functions using the
7106 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7107 inline functions into the object file.
7109 @item -fkeep-static-functions
7110 @opindex fkeep-static-functions
7111 Emit @code{static} functions into the object file, even if the function
7114 @item -fkeep-static-consts
7115 @opindex fkeep-static-consts
7116 Emit variables declared @code{static const} when optimization isn't turned
7117 on, even if the variables aren't referenced.
7119 GCC enables this option by default. If you want to force the compiler to
7120 check if a variable is referenced, regardless of whether or not
7121 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7123 @item -fmerge-constants
7124 @opindex fmerge-constants
7125 Attempt to merge identical constants (string constants and floating-point
7126 constants) across compilation units.
7128 This option is the default for optimized compilation if the assembler and
7129 linker support it. Use @option{-fno-merge-constants} to inhibit this
7132 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7134 @item -fmerge-all-constants
7135 @opindex fmerge-all-constants
7136 Attempt to merge identical constants and identical variables.
7138 This option implies @option{-fmerge-constants}. In addition to
7139 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7140 arrays or initialized constant variables with integral or floating-point
7141 types. Languages like C or C++ require each variable, including multiple
7142 instances of the same variable in recursive calls, to have distinct locations,
7143 so using this option results in non-conforming
7146 @item -fmodulo-sched
7147 @opindex fmodulo-sched
7148 Perform swing modulo scheduling immediately before the first scheduling
7149 pass. This pass looks at innermost loops and reorders their
7150 instructions by overlapping different iterations.
7152 @item -fmodulo-sched-allow-regmoves
7153 @opindex fmodulo-sched-allow-regmoves
7154 Perform more aggressive SMS-based modulo scheduling with register moves
7155 allowed. By setting this flag certain anti-dependences edges are
7156 deleted, which triggers the generation of reg-moves based on the
7157 life-range analysis. This option is effective only with
7158 @option{-fmodulo-sched} enabled.
7160 @item -fno-branch-count-reg
7161 @opindex fno-branch-count-reg
7162 Avoid running a pass scanning for opportunities to use ``decrement and
7163 branch'' instructions on a count register instead of generating sequences
7164 of instructions that decrement a register, compare it against zero, and
7165 then branch based upon the result. This option is only meaningful on
7166 architectures that support such instructions, which include x86, PowerPC,
7167 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7168 doesn't remove the decrement and branch instructions from the generated
7169 instruction stream introduced by other optimization passes.
7171 Enabled by default at @option{-O1} and higher.
7173 The default is @option{-fbranch-count-reg}.
7175 @item -fno-function-cse
7176 @opindex fno-function-cse
7177 Do not put function addresses in registers; make each instruction that
7178 calls a constant function contain the function's address explicitly.
7180 This option results in less efficient code, but some strange hacks
7181 that alter the assembler output may be confused by the optimizations
7182 performed when this option is not used.
7184 The default is @option{-ffunction-cse}
7186 @item -fno-zero-initialized-in-bss
7187 @opindex fno-zero-initialized-in-bss
7188 If the target supports a BSS section, GCC by default puts variables that
7189 are initialized to zero into BSS@. This can save space in the resulting
7192 This option turns off this behavior because some programs explicitly
7193 rely on variables going to the data section---e.g., so that the
7194 resulting executable can find the beginning of that section and/or make
7195 assumptions based on that.
7197 The default is @option{-fzero-initialized-in-bss}.
7199 @item -fthread-jumps
7200 @opindex fthread-jumps
7201 Perform optimizations that check to see if a jump branches to a
7202 location where another comparison subsumed by the first is found. If
7203 so, the first branch is redirected to either the destination of the
7204 second branch or a point immediately following it, depending on whether
7205 the condition is known to be true or false.
7207 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7209 @item -fsplit-wide-types
7210 @opindex fsplit-wide-types
7211 When using a type that occupies multiple registers, such as @code{long
7212 long} on a 32-bit system, split the registers apart and allocate them
7213 independently. This normally generates better code for those types,
7214 but may make debugging more difficult.
7216 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7219 @item -fcse-follow-jumps
7220 @opindex fcse-follow-jumps
7221 In common subexpression elimination (CSE), scan through jump instructions
7222 when the target of the jump is not reached by any other path. For
7223 example, when CSE encounters an @code{if} statement with an
7224 @code{else} clause, CSE follows the jump when the condition
7227 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7229 @item -fcse-skip-blocks
7230 @opindex fcse-skip-blocks
7231 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7232 follow jumps that conditionally skip over blocks. When CSE
7233 encounters a simple @code{if} statement with no else clause,
7234 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7235 body of the @code{if}.
7237 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7239 @item -frerun-cse-after-loop
7240 @opindex frerun-cse-after-loop
7241 Re-run common subexpression elimination after loop optimizations are
7244 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7248 Perform a global common subexpression elimination pass.
7249 This pass also performs global constant and copy propagation.
7251 @emph{Note:} When compiling a program using computed gotos, a GCC
7252 extension, you may get better run-time performance if you disable
7253 the global common subexpression elimination pass by adding
7254 @option{-fno-gcse} to the command line.
7256 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7260 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7261 attempts to move loads that are only killed by stores into themselves. This
7262 allows a loop containing a load/store sequence to be changed to a load outside
7263 the loop, and a copy/store within the loop.
7265 Enabled by default when @option{-fgcse} is enabled.
7269 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7270 global common subexpression elimination. This pass attempts to move
7271 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7272 loops containing a load/store sequence can be changed to a load before
7273 the loop and a store after the loop.
7275 Not enabled at any optimization level.
7279 When @option{-fgcse-las} is enabled, the global common subexpression
7280 elimination pass eliminates redundant loads that come after stores to the
7281 same memory location (both partial and full redundancies).
7283 Not enabled at any optimization level.
7285 @item -fgcse-after-reload
7286 @opindex fgcse-after-reload
7287 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7288 pass is performed after reload. The purpose of this pass is to clean up
7291 @item -faggressive-loop-optimizations
7292 @opindex faggressive-loop-optimizations
7293 This option tells the loop optimizer to use language constraints to
7294 derive bounds for the number of iterations of a loop. This assumes that
7295 loop code does not invoke undefined behavior by for example causing signed
7296 integer overflows or out-of-bound array accesses. The bounds for the
7297 number of iterations of a loop are used to guide loop unrolling and peeling
7298 and loop exit test optimizations.
7299 This option is enabled by default.
7301 @item -funconstrained-commons
7302 @opindex funconstrained-commons
7303 This option tells the compiler that variables declared in common blocks
7304 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7305 prevents certain optimizations that depend on knowing the array bounds.
7307 @item -fcrossjumping
7308 @opindex fcrossjumping
7309 Perform cross-jumping transformation.
7310 This transformation unifies equivalent code and saves code size. The
7311 resulting code may or may not perform better than without cross-jumping.
7313 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7315 @item -fauto-inc-dec
7316 @opindex fauto-inc-dec
7317 Combine increments or decrements of addresses with memory accesses.
7318 This pass is always skipped on architectures that do not have
7319 instructions to support this. Enabled by default at @option{-O} and
7320 higher on architectures that support this.
7324 Perform dead code elimination (DCE) on RTL@.
7325 Enabled by default at @option{-O} and higher.
7329 Perform dead store elimination (DSE) on RTL@.
7330 Enabled by default at @option{-O} and higher.
7332 @item -fif-conversion
7333 @opindex fif-conversion
7334 Attempt to transform conditional jumps into branch-less equivalents. This
7335 includes use of conditional moves, min, max, set flags and abs instructions, and
7336 some tricks doable by standard arithmetics. The use of conditional execution
7337 on chips where it is available is controlled by @option{-fif-conversion2}.
7339 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7341 @item -fif-conversion2
7342 @opindex fif-conversion2
7343 Use conditional execution (where available) to transform conditional jumps into
7344 branch-less equivalents.
7346 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7348 @item -fdeclone-ctor-dtor
7349 @opindex fdeclone-ctor-dtor
7350 The C++ ABI requires multiple entry points for constructors and
7351 destructors: one for a base subobject, one for a complete object, and
7352 one for a virtual destructor that calls operator delete afterwards.
7353 For a hierarchy with virtual bases, the base and complete variants are
7354 clones, which means two copies of the function. With this option, the
7355 base and complete variants are changed to be thunks that call a common
7358 Enabled by @option{-Os}.
7360 @item -fdelete-null-pointer-checks
7361 @opindex fdelete-null-pointer-checks
7362 Assume that programs cannot safely dereference null pointers, and that
7363 no code or data element resides at address zero.
7364 This option enables simple constant
7365 folding optimizations at all optimization levels. In addition, other
7366 optimization passes in GCC use this flag to control global dataflow
7367 analyses that eliminate useless checks for null pointers; these assume
7368 that a memory access to address zero always results in a trap, so
7369 that if a pointer is checked after it has already been dereferenced,
7372 Note however that in some environments this assumption is not true.
7373 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7374 for programs that depend on that behavior.
7376 This option is enabled by default on most targets. On Nios II ELF, it
7377 defaults to off. On AVR and CR16, this option is completely disabled.
7379 Passes that use the dataflow information
7380 are enabled independently at different optimization levels.
7382 @item -fdevirtualize
7383 @opindex fdevirtualize
7384 Attempt to convert calls to virtual functions to direct calls. This
7385 is done both within a procedure and interprocedurally as part of
7386 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7387 propagation (@option{-fipa-cp}).
7388 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7390 @item -fdevirtualize-speculatively
7391 @opindex fdevirtualize-speculatively
7392 Attempt to convert calls to virtual functions to speculative direct calls.
7393 Based on the analysis of the type inheritance graph, determine for a given call
7394 the set of likely targets. If the set is small, preferably of size 1, change
7395 the call into a conditional deciding between direct and indirect calls. The
7396 speculative calls enable more optimizations, such as inlining. When they seem
7397 useless after further optimization, they are converted back into original form.
7399 @item -fdevirtualize-at-ltrans
7400 @opindex fdevirtualize-at-ltrans
7401 Stream extra information needed for aggressive devirtualization when running
7402 the link-time optimizer in local transformation mode.
7403 This option enables more devirtualization but
7404 significantly increases the size of streamed data. For this reason it is
7405 disabled by default.
7407 @item -fexpensive-optimizations
7408 @opindex fexpensive-optimizations
7409 Perform a number of minor optimizations that are relatively expensive.
7411 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7415 Attempt to remove redundant extension instructions. This is especially
7416 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7417 registers after writing to their lower 32-bit half.
7419 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7420 @option{-O3}, @option{-Os}.
7422 @item -fno-lifetime-dse
7423 @opindex fno-lifetime-dse
7424 In C++ the value of an object is only affected by changes within its
7425 lifetime: when the constructor begins, the object has an indeterminate
7426 value, and any changes during the lifetime of the object are dead when
7427 the object is destroyed. Normally dead store elimination will take
7428 advantage of this; if your code relies on the value of the object
7429 storage persisting beyond the lifetime of the object, you can use this
7430 flag to disable this optimization. To preserve stores before the
7431 constructor starts (e.g. because your operator new clears the object
7432 storage) but still treat the object as dead after the destructor you,
7433 can use @option{-flifetime-dse=1}. The default behavior can be
7434 explicitly selected with @option{-flifetime-dse=2}.
7435 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7437 @item -flive-range-shrinkage
7438 @opindex flive-range-shrinkage
7439 Attempt to decrease register pressure through register live range
7440 shrinkage. This is helpful for fast processors with small or moderate
7443 @item -fira-algorithm=@var{algorithm}
7444 @opindex fira-algorithm
7445 Use the specified coloring algorithm for the integrated register
7446 allocator. The @var{algorithm} argument can be @samp{priority}, which
7447 specifies Chow's priority coloring, or @samp{CB}, which specifies
7448 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7449 for all architectures, but for those targets that do support it, it is
7450 the default because it generates better code.
7452 @item -fira-region=@var{region}
7453 @opindex fira-region
7454 Use specified regions for the integrated register allocator. The
7455 @var{region} argument should be one of the following:
7460 Use all loops as register allocation regions.
7461 This can give the best results for machines with a small and/or
7462 irregular register set.
7465 Use all loops except for loops with small register pressure
7466 as the regions. This value usually gives
7467 the best results in most cases and for most architectures,
7468 and is enabled by default when compiling with optimization for speed
7469 (@option{-O}, @option{-O2}, @dots{}).
7472 Use all functions as a single region.
7473 This typically results in the smallest code size, and is enabled by default for
7474 @option{-Os} or @option{-O0}.
7478 @item -fira-hoist-pressure
7479 @opindex fira-hoist-pressure
7480 Use IRA to evaluate register pressure in the code hoisting pass for
7481 decisions to hoist expressions. This option usually results in smaller
7482 code, but it can slow the compiler down.
7484 This option is enabled at level @option{-Os} for all targets.
7486 @item -fira-loop-pressure
7487 @opindex fira-loop-pressure
7488 Use IRA to evaluate register pressure in loops for decisions to move
7489 loop invariants. This option usually results in generation
7490 of faster and smaller code on machines with large register files (>= 32
7491 registers), but it can slow the compiler down.
7493 This option is enabled at level @option{-O3} for some targets.
7495 @item -fno-ira-share-save-slots
7496 @opindex fno-ira-share-save-slots
7497 Disable sharing of stack slots used for saving call-used hard
7498 registers living through a call. Each hard register gets a
7499 separate stack slot, and as a result function stack frames are
7502 @item -fno-ira-share-spill-slots
7503 @opindex fno-ira-share-spill-slots
7504 Disable sharing of stack slots allocated for pseudo-registers. Each
7505 pseudo-register that does not get a hard register gets a separate
7506 stack slot, and as a result function stack frames are larger.
7510 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7511 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7512 values if it is profitable.
7514 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7516 @item -fdelayed-branch
7517 @opindex fdelayed-branch
7518 If supported for the target machine, attempt to reorder instructions
7519 to exploit instruction slots available after delayed branch
7522 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7524 @item -fschedule-insns
7525 @opindex fschedule-insns
7526 If supported for the target machine, attempt to reorder instructions to
7527 eliminate execution stalls due to required data being unavailable. This
7528 helps machines that have slow floating point or memory load instructions
7529 by allowing other instructions to be issued until the result of the load
7530 or floating-point instruction is required.
7532 Enabled at levels @option{-O2}, @option{-O3}.
7534 @item -fschedule-insns2
7535 @opindex fschedule-insns2
7536 Similar to @option{-fschedule-insns}, but requests an additional pass of
7537 instruction scheduling after register allocation has been done. This is
7538 especially useful on machines with a relatively small number of
7539 registers and where memory load instructions take more than one cycle.
7541 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7543 @item -fno-sched-interblock
7544 @opindex fno-sched-interblock
7545 Don't schedule instructions across basic blocks. This is normally
7546 enabled by default when scheduling before register allocation, i.e.@:
7547 with @option{-fschedule-insns} or at @option{-O2} or higher.
7549 @item -fno-sched-spec
7550 @opindex fno-sched-spec
7551 Don't allow speculative motion of non-load instructions. This is normally
7552 enabled by default when scheduling before register allocation, i.e.@:
7553 with @option{-fschedule-insns} or at @option{-O2} or higher.
7555 @item -fsched-pressure
7556 @opindex fsched-pressure
7557 Enable register pressure sensitive insn scheduling before register
7558 allocation. This only makes sense when scheduling before register
7559 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7560 @option{-O2} or higher. Usage of this option can improve the
7561 generated code and decrease its size by preventing register pressure
7562 increase above the number of available hard registers and subsequent
7563 spills in register allocation.
7565 @item -fsched-spec-load
7566 @opindex fsched-spec-load
7567 Allow speculative motion of some load instructions. This only makes
7568 sense when scheduling before register allocation, i.e.@: with
7569 @option{-fschedule-insns} or at @option{-O2} or higher.
7571 @item -fsched-spec-load-dangerous
7572 @opindex fsched-spec-load-dangerous
7573 Allow speculative motion of more load instructions. This only makes
7574 sense when scheduling before register allocation, i.e.@: with
7575 @option{-fschedule-insns} or at @option{-O2} or higher.
7577 @item -fsched-stalled-insns
7578 @itemx -fsched-stalled-insns=@var{n}
7579 @opindex fsched-stalled-insns
7580 Define how many insns (if any) can be moved prematurely from the queue
7581 of stalled insns into the ready list during the second scheduling pass.
7582 @option{-fno-sched-stalled-insns} means that no insns are moved
7583 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7584 on how many queued insns can be moved prematurely.
7585 @option{-fsched-stalled-insns} without a value is equivalent to
7586 @option{-fsched-stalled-insns=1}.
7588 @item -fsched-stalled-insns-dep
7589 @itemx -fsched-stalled-insns-dep=@var{n}
7590 @opindex fsched-stalled-insns-dep
7591 Define how many insn groups (cycles) are examined for a dependency
7592 on a stalled insn that is a candidate for premature removal from the queue
7593 of stalled insns. This has an effect only during the second scheduling pass,
7594 and only if @option{-fsched-stalled-insns} is used.
7595 @option{-fno-sched-stalled-insns-dep} is equivalent to
7596 @option{-fsched-stalled-insns-dep=0}.
7597 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7598 @option{-fsched-stalled-insns-dep=1}.
7600 @item -fsched2-use-superblocks
7601 @opindex fsched2-use-superblocks
7602 When scheduling after register allocation, use superblock scheduling.
7603 This allows motion across basic block boundaries,
7604 resulting in faster schedules. This option is experimental, as not all machine
7605 descriptions used by GCC model the CPU closely enough to avoid unreliable
7606 results from the algorithm.
7608 This only makes sense when scheduling after register allocation, i.e.@: with
7609 @option{-fschedule-insns2} or at @option{-O2} or higher.
7611 @item -fsched-group-heuristic
7612 @opindex fsched-group-heuristic
7613 Enable the group heuristic in the scheduler. This heuristic favors
7614 the instruction that belongs to a schedule group. This is enabled
7615 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7616 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7618 @item -fsched-critical-path-heuristic
7619 @opindex fsched-critical-path-heuristic
7620 Enable the critical-path heuristic in the scheduler. This heuristic favors
7621 instructions on the critical path. This is enabled by default when
7622 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7623 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7625 @item -fsched-spec-insn-heuristic
7626 @opindex fsched-spec-insn-heuristic
7627 Enable the speculative instruction heuristic in the scheduler. This
7628 heuristic favors speculative instructions with greater dependency weakness.
7629 This is enabled by default when scheduling is enabled, i.e.@:
7630 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7631 or at @option{-O2} or higher.
7633 @item -fsched-rank-heuristic
7634 @opindex fsched-rank-heuristic
7635 Enable the rank heuristic in the scheduler. This heuristic favors
7636 the instruction belonging to a basic block with greater size or frequency.
7637 This is enabled by default when scheduling is enabled, i.e.@:
7638 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7639 at @option{-O2} or higher.
7641 @item -fsched-last-insn-heuristic
7642 @opindex fsched-last-insn-heuristic
7643 Enable the last-instruction heuristic in the scheduler. This heuristic
7644 favors the instruction that is less dependent on the last instruction
7645 scheduled. This is enabled by default when scheduling is enabled,
7646 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7647 at @option{-O2} or higher.
7649 @item -fsched-dep-count-heuristic
7650 @opindex fsched-dep-count-heuristic
7651 Enable the dependent-count heuristic in the scheduler. This heuristic
7652 favors the instruction that has more instructions depending on it.
7653 This is enabled by default when scheduling is enabled, i.e.@:
7654 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7655 at @option{-O2} or higher.
7657 @item -freschedule-modulo-scheduled-loops
7658 @opindex freschedule-modulo-scheduled-loops
7659 Modulo scheduling is performed before traditional scheduling. If a loop
7660 is modulo scheduled, later scheduling passes may change its schedule.
7661 Use this option to control that behavior.
7663 @item -fselective-scheduling
7664 @opindex fselective-scheduling
7665 Schedule instructions using selective scheduling algorithm. Selective
7666 scheduling runs instead of the first scheduler pass.
7668 @item -fselective-scheduling2
7669 @opindex fselective-scheduling2
7670 Schedule instructions using selective scheduling algorithm. Selective
7671 scheduling runs instead of the second scheduler pass.
7673 @item -fsel-sched-pipelining
7674 @opindex fsel-sched-pipelining
7675 Enable software pipelining of innermost loops during selective scheduling.
7676 This option has no effect unless one of @option{-fselective-scheduling} or
7677 @option{-fselective-scheduling2} is turned on.
7679 @item -fsel-sched-pipelining-outer-loops
7680 @opindex fsel-sched-pipelining-outer-loops
7681 When pipelining loops during selective scheduling, also pipeline outer loops.
7682 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7684 @item -fsemantic-interposition
7685 @opindex fsemantic-interposition
7686 Some object formats, like ELF, allow interposing of symbols by the
7688 This means that for symbols exported from the DSO, the compiler cannot perform
7689 interprocedural propagation, inlining and other optimizations in anticipation
7690 that the function or variable in question may change. While this feature is
7691 useful, for example, to rewrite memory allocation functions by a debugging
7692 implementation, it is expensive in the terms of code quality.
7693 With @option{-fno-semantic-interposition} the compiler assumes that
7694 if interposition happens for functions the overwriting function will have
7695 precisely the same semantics (and side effects).
7696 Similarly if interposition happens
7697 for variables, the constructor of the variable will be the same. The flag
7698 has no effect for functions explicitly declared inline
7699 (where it is never allowed for interposition to change semantics)
7700 and for symbols explicitly declared weak.
7703 @opindex fshrink-wrap
7704 Emit function prologues only before parts of the function that need it,
7705 rather than at the top of the function. This flag is enabled by default at
7706 @option{-O} and higher.
7708 @item -fshrink-wrap-separate
7709 @opindex fshrink-wrap-separate
7710 Shrink-wrap separate parts of the prologue and epilogue separately, so that
7711 those parts are only executed when needed.
7712 This option is on by default, but has no effect unless @option{-fshrink-wrap}
7713 is also turned on and the target supports this.
7715 @item -fcaller-saves
7716 @opindex fcaller-saves
7717 Enable allocation of values to registers that are clobbered by
7718 function calls, by emitting extra instructions to save and restore the
7719 registers around such calls. Such allocation is done only when it
7720 seems to result in better code.
7722 This option is always enabled by default on certain machines, usually
7723 those which have no call-preserved registers to use instead.
7725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7727 @item -fcombine-stack-adjustments
7728 @opindex fcombine-stack-adjustments
7729 Tracks stack adjustments (pushes and pops) and stack memory references
7730 and then tries to find ways to combine them.
7732 Enabled by default at @option{-O1} and higher.
7736 Use caller save registers for allocation if those registers are not used by
7737 any called function. In that case it is not necessary to save and restore
7738 them around calls. This is only possible if called functions are part of
7739 same compilation unit as current function and they are compiled before it.
7741 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
7742 is disabled if generated code will be instrumented for profiling
7743 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
7744 exactly (this happens on targets that do not expose prologues
7745 and epilogues in RTL).
7747 @item -fconserve-stack
7748 @opindex fconserve-stack
7749 Attempt to minimize stack usage. The compiler attempts to use less
7750 stack space, even if that makes the program slower. This option
7751 implies setting the @option{large-stack-frame} parameter to 100
7752 and the @option{large-stack-frame-growth} parameter to 400.
7754 @item -ftree-reassoc
7755 @opindex ftree-reassoc
7756 Perform reassociation on trees. This flag is enabled by default
7757 at @option{-O} and higher.
7759 @item -fcode-hoisting
7760 @opindex fcode-hoisting
7761 Perform code hoisting. Code hoisting tries to move the
7762 evaluation of expressions executed on all paths to the function exit
7763 as early as possible. This is especially useful as a code size
7764 optimization, but it often helps for code speed as well.
7765 This flag is enabled by default at @option{-O2} and higher.
7769 Perform partial redundancy elimination (PRE) on trees. This flag is
7770 enabled by default at @option{-O2} and @option{-O3}.
7772 @item -ftree-partial-pre
7773 @opindex ftree-partial-pre
7774 Make partial redundancy elimination (PRE) more aggressive. This flag is
7775 enabled by default at @option{-O3}.
7777 @item -ftree-forwprop
7778 @opindex ftree-forwprop
7779 Perform forward propagation on trees. This flag is enabled by default
7780 at @option{-O} and higher.
7784 Perform full redundancy elimination (FRE) on trees. The difference
7785 between FRE and PRE is that FRE only considers expressions
7786 that are computed on all paths leading to the redundant computation.
7787 This analysis is faster than PRE, though it exposes fewer redundancies.
7788 This flag is enabled by default at @option{-O} and higher.
7790 @item -ftree-phiprop
7791 @opindex ftree-phiprop
7792 Perform hoisting of loads from conditional pointers on trees. This
7793 pass is enabled by default at @option{-O} and higher.
7795 @item -fhoist-adjacent-loads
7796 @opindex fhoist-adjacent-loads
7797 Speculatively hoist loads from both branches of an if-then-else if the
7798 loads are from adjacent locations in the same structure and the target
7799 architecture has a conditional move instruction. This flag is enabled
7800 by default at @option{-O2} and higher.
7802 @item -ftree-copy-prop
7803 @opindex ftree-copy-prop
7804 Perform copy propagation on trees. This pass eliminates unnecessary
7805 copy operations. This flag is enabled by default at @option{-O} and
7808 @item -fipa-pure-const
7809 @opindex fipa-pure-const
7810 Discover which functions are pure or constant.
7811 Enabled by default at @option{-O} and higher.
7813 @item -fipa-reference
7814 @opindex fipa-reference
7815 Discover which static variables do not escape the
7817 Enabled by default at @option{-O} and higher.
7821 Perform interprocedural pointer analysis and interprocedural modification
7822 and reference analysis. This option can cause excessive memory and
7823 compile-time usage on large compilation units. It is not enabled by
7824 default at any optimization level.
7827 @opindex fipa-profile
7828 Perform interprocedural profile propagation. The functions called only from
7829 cold functions are marked as cold. Also functions executed once (such as
7830 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7831 functions and loop less parts of functions executed once are then optimized for
7833 Enabled by default at @option{-O} and higher.
7837 Perform interprocedural constant propagation.
7838 This optimization analyzes the program to determine when values passed
7839 to functions are constants and then optimizes accordingly.
7840 This optimization can substantially increase performance
7841 if the application has constants passed to functions.
7842 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7844 @item -fipa-cp-clone
7845 @opindex fipa-cp-clone
7846 Perform function cloning to make interprocedural constant propagation stronger.
7847 When enabled, interprocedural constant propagation performs function cloning
7848 when externally visible function can be called with constant arguments.
7849 Because this optimization can create multiple copies of functions,
7850 it may significantly increase code size
7851 (see @option{--param ipcp-unit-growth=@var{value}}).
7852 This flag is enabled by default at @option{-O3}.
7854 @item -fipa-cp-alignment
7855 @opindex -fipa-cp-alignment
7856 When enabled, this optimization propagates alignment of function
7857 parameters to support better vectorization and string operations.
7859 This flag is enabled by default at @option{-O2} and @option{-Os}. It
7860 requires that @option{-fipa-cp} is enabled.
7861 @option{-fipa-cp-alignment} is obsolete, use @option{-fipa-bit-cp} instead.
7864 @opindex -fipa-bit-cp
7865 When enabled, perform ipa bitwise constant propagation. This flag is
7866 enabled by default at @option{-O2}. It requires that @option{-fipa-cp}
7871 Perform Identical Code Folding for functions and read-only variables.
7872 The optimization reduces code size and may disturb unwind stacks by replacing
7873 a function by equivalent one with a different name. The optimization works
7874 more effectively with link time optimization enabled.
7876 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
7877 works on different levels and thus the optimizations are not same - there are
7878 equivalences that are found only by GCC and equivalences found only by Gold.
7880 This flag is enabled by default at @option{-O2} and @option{-Os}.
7882 @item -fisolate-erroneous-paths-dereference
7883 @opindex fisolate-erroneous-paths-dereference
7884 Detect paths that trigger erroneous or undefined behavior due to
7885 dereferencing a null pointer. Isolate those paths from the main control
7886 flow and turn the statement with erroneous or undefined behavior into a trap.
7887 This flag is enabled by default at @option{-O2} and higher and depends on
7888 @option{-fdelete-null-pointer-checks} also being enabled.
7890 @item -fisolate-erroneous-paths-attribute
7891 @opindex fisolate-erroneous-paths-attribute
7892 Detect paths that trigger erroneous or undefined behavior due a null value
7893 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
7894 attribute. Isolate those paths from the main control flow and turn the
7895 statement with erroneous or undefined behavior into a trap. This is not
7896 currently enabled, but may be enabled by @option{-O2} in the future.
7900 Perform forward store motion on trees. This flag is
7901 enabled by default at @option{-O} and higher.
7903 @item -ftree-bit-ccp
7904 @opindex ftree-bit-ccp
7905 Perform sparse conditional bit constant propagation on trees and propagate
7906 pointer alignment information.
7907 This pass only operates on local scalar variables and is enabled by default
7908 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7912 Perform sparse conditional constant propagation (CCP) on trees. This
7913 pass only operates on local scalar variables and is enabled by default
7914 at @option{-O} and higher.
7916 @item -fssa-backprop
7917 @opindex fssa-backprop
7918 Propagate information about uses of a value up the definition chain
7919 in order to simplify the definitions. For example, this pass strips
7920 sign operations if the sign of a value never matters. The flag is
7921 enabled by default at @option{-O} and higher.
7924 @opindex fssa-phiopt
7925 Perform pattern matching on SSA PHI nodes to optimize conditional
7926 code. This pass is enabled by default at @option{-O} and higher.
7928 @item -ftree-switch-conversion
7929 @opindex ftree-switch-conversion
7930 Perform conversion of simple initializations in a switch to
7931 initializations from a scalar array. This flag is enabled by default
7932 at @option{-O2} and higher.
7934 @item -ftree-tail-merge
7935 @opindex ftree-tail-merge
7936 Look for identical code sequences. When found, replace one with a jump to the
7937 other. This optimization is known as tail merging or cross jumping. This flag
7938 is enabled by default at @option{-O2} and higher. The compilation time
7940 be limited using @option{max-tail-merge-comparisons} parameter and
7941 @option{max-tail-merge-iterations} parameter.
7945 Perform dead code elimination (DCE) on trees. This flag is enabled by
7946 default at @option{-O} and higher.
7948 @item -ftree-builtin-call-dce
7949 @opindex ftree-builtin-call-dce
7950 Perform conditional dead code elimination (DCE) for calls to built-in functions
7951 that may set @code{errno} but are otherwise side-effect free. This flag is
7952 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7955 @item -ftree-dominator-opts
7956 @opindex ftree-dominator-opts
7957 Perform a variety of simple scalar cleanups (constant/copy
7958 propagation, redundancy elimination, range propagation and expression
7959 simplification) based on a dominator tree traversal. This also
7960 performs jump threading (to reduce jumps to jumps). This flag is
7961 enabled by default at @option{-O} and higher.
7965 Perform dead store elimination (DSE) on trees. A dead store is a store into
7966 a memory location that is later overwritten by another store without
7967 any intervening loads. In this case the earlier store can be deleted. This
7968 flag is enabled by default at @option{-O} and higher.
7972 Perform loop header copying on trees. This is beneficial since it increases
7973 effectiveness of code motion optimizations. It also saves one jump. This flag
7974 is enabled by default at @option{-O} and higher. It is not enabled
7975 for @option{-Os}, since it usually increases code size.
7977 @item -ftree-loop-optimize
7978 @opindex ftree-loop-optimize
7979 Perform loop optimizations on trees. This flag is enabled by default
7980 at @option{-O} and higher.
7982 @item -ftree-loop-linear
7983 @itemx -floop-interchange
7984 @itemx -floop-strip-mine
7986 @itemx -floop-unroll-and-jam
7987 @opindex ftree-loop-linear
7988 @opindex floop-interchange
7989 @opindex floop-strip-mine
7990 @opindex floop-block
7991 @opindex floop-unroll-and-jam
7992 Perform loop nest optimizations. Same as
7993 @option{-floop-nest-optimize}. To use this code transformation, GCC has
7994 to be configured with @option{--with-isl} to enable the Graphite loop
7995 transformation infrastructure.
7997 @item -fgraphite-identity
7998 @opindex fgraphite-identity
7999 Enable the identity transformation for graphite. For every SCoP we generate
8000 the polyhedral representation and transform it back to gimple. Using
8001 @option{-fgraphite-identity} we can check the costs or benefits of the
8002 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8003 are also performed by the code generator isl, like index splitting and
8004 dead code elimination in loops.
8006 @item -floop-nest-optimize
8007 @opindex floop-nest-optimize
8008 Enable the isl based loop nest optimizer. This is a generic loop nest
8009 optimizer based on the Pluto optimization algorithms. It calculates a loop
8010 structure optimized for data-locality and parallelism. This option
8013 @item -floop-parallelize-all
8014 @opindex floop-parallelize-all
8015 Use the Graphite data dependence analysis to identify loops that can
8016 be parallelized. Parallelize all the loops that can be analyzed to
8017 not contain loop carried dependences without checking that it is
8018 profitable to parallelize the loops.
8020 @item -ftree-coalesce-vars
8021 @opindex ftree-coalesce-vars
8022 While transforming the program out of the SSA representation, attempt to
8023 reduce copying by coalescing versions of different user-defined
8024 variables, instead of just compiler temporaries. This may severely
8025 limit the ability to debug an optimized program compiled with
8026 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8027 prevents SSA coalescing of user variables. This option is enabled by
8028 default if optimization is enabled, and it does very little otherwise.
8030 @item -ftree-loop-if-convert
8031 @opindex ftree-loop-if-convert
8032 Attempt to transform conditional jumps in the innermost loops to
8033 branch-less equivalents. The intent is to remove control-flow from
8034 the innermost loops in order to improve the ability of the
8035 vectorization pass to handle these loops. This is enabled by default
8036 if vectorization is enabled.
8038 @item -ftree-loop-if-convert-stores
8039 @opindex ftree-loop-if-convert-stores
8040 Attempt to also if-convert conditional jumps containing memory writes.
8041 This transformation can be unsafe for multi-threaded programs as it
8042 transforms conditional memory writes into unconditional memory writes.
8045 for (i = 0; i < N; i++)
8051 for (i = 0; i < N; i++)
8052 A[i] = cond ? expr : A[i];
8054 potentially producing data races.
8056 @item -ftree-loop-distribution
8057 @opindex ftree-loop-distribution
8058 Perform loop distribution. This flag can improve cache performance on
8059 big loop bodies and allow further loop optimizations, like
8060 parallelization or vectorization, to take place. For example, the loop
8077 @item -ftree-loop-distribute-patterns
8078 @opindex ftree-loop-distribute-patterns
8079 Perform loop distribution of patterns that can be code generated with
8080 calls to a library. This flag is enabled by default at @option{-O3}.
8082 This pass distributes the initialization loops and generates a call to
8083 memset zero. For example, the loop
8099 and the initialization loop is transformed into a call to memset zero.
8101 @item -ftree-loop-im
8102 @opindex ftree-loop-im
8103 Perform loop invariant motion on trees. This pass moves only invariants that
8104 are hard to handle at RTL level (function calls, operations that expand to
8105 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8106 operands of conditions that are invariant out of the loop, so that we can use
8107 just trivial invariantness analysis in loop unswitching. The pass also includes
8110 @item -ftree-loop-ivcanon
8111 @opindex ftree-loop-ivcanon
8112 Create a canonical counter for number of iterations in loops for which
8113 determining number of iterations requires complicated analysis. Later
8114 optimizations then may determine the number easily. Useful especially
8115 in connection with unrolling.
8119 Perform induction variable optimizations (strength reduction, induction
8120 variable merging and induction variable elimination) on trees.
8122 @item -ftree-parallelize-loops=n
8123 @opindex ftree-parallelize-loops
8124 Parallelize loops, i.e., split their iteration space to run in n threads.
8125 This is only possible for loops whose iterations are independent
8126 and can be arbitrarily reordered. The optimization is only
8127 profitable on multiprocessor machines, for loops that are CPU-intensive,
8128 rather than constrained e.g.@: by memory bandwidth. This option
8129 implies @option{-pthread}, and thus is only supported on targets
8130 that have support for @option{-pthread}.
8134 Perform function-local points-to analysis on trees. This flag is
8135 enabled by default at @option{-O} and higher.
8139 Perform scalar replacement of aggregates. This pass replaces structure
8140 references with scalars to prevent committing structures to memory too
8141 early. This flag is enabled by default at @option{-O} and higher.
8143 @item -fstore-merging
8144 @opindex fstore-merging
8145 Perform merging of narrow stores to consecutive memory addresses. This pass
8146 merges contiguous stores of immediate values narrower than a word into fewer
8147 wider stores to reduce the number of instructions. This is enabled by default
8148 at @option{-O} and higher.
8152 Perform temporary expression replacement during the SSA->normal phase. Single
8153 use/single def temporaries are replaced at their use location with their
8154 defining expression. This results in non-GIMPLE code, but gives the expanders
8155 much more complex trees to work on resulting in better RTL generation. This is
8156 enabled by default at @option{-O} and higher.
8160 Perform straight-line strength reduction on trees. This recognizes related
8161 expressions involving multiplications and replaces them by less expensive
8162 calculations when possible. This is enabled by default at @option{-O} and
8165 @item -ftree-vectorize
8166 @opindex ftree-vectorize
8167 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8168 and @option{-ftree-slp-vectorize} if not explicitly specified.
8170 @item -ftree-loop-vectorize
8171 @opindex ftree-loop-vectorize
8172 Perform loop vectorization on trees. This flag is enabled by default at
8173 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8175 @item -ftree-slp-vectorize
8176 @opindex ftree-slp-vectorize
8177 Perform basic block vectorization on trees. This flag is enabled by default at
8178 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8180 @item -fvect-cost-model=@var{model}
8181 @opindex fvect-cost-model
8182 Alter the cost model used for vectorization. The @var{model} argument
8183 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8184 With the @samp{unlimited} model the vectorized code-path is assumed
8185 to be profitable while with the @samp{dynamic} model a runtime check
8186 guards the vectorized code-path to enable it only for iteration
8187 counts that will likely execute faster than when executing the original
8188 scalar loop. The @samp{cheap} model disables vectorization of
8189 loops where doing so would be cost prohibitive for example due to
8190 required runtime checks for data dependence or alignment but otherwise
8191 is equal to the @samp{dynamic} model.
8192 The default cost model depends on other optimization flags and is
8193 either @samp{dynamic} or @samp{cheap}.
8195 @item -fsimd-cost-model=@var{model}
8196 @opindex fsimd-cost-model
8197 Alter the cost model used for vectorization of loops marked with the OpenMP
8198 or Cilk Plus simd directive. The @var{model} argument should be one of
8199 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8200 have the same meaning as described in @option{-fvect-cost-model} and by
8201 default a cost model defined with @option{-fvect-cost-model} is used.
8205 Perform Value Range Propagation on trees. This is similar to the
8206 constant propagation pass, but instead of values, ranges of values are
8207 propagated. This allows the optimizers to remove unnecessary range
8208 checks like array bound checks and null pointer checks. This is
8209 enabled by default at @option{-O2} and higher. Null pointer check
8210 elimination is only done if @option{-fdelete-null-pointer-checks} is
8214 @opindex fsplit-paths
8215 Split paths leading to loop backedges. This can improve dead code
8216 elimination and common subexpression elimination. This is enabled by
8217 default at @option{-O2} and above.
8219 @item -fsplit-ivs-in-unroller
8220 @opindex fsplit-ivs-in-unroller
8221 Enables expression of values of induction variables in later iterations
8222 of the unrolled loop using the value in the first iteration. This breaks
8223 long dependency chains, thus improving efficiency of the scheduling passes.
8225 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8226 same effect. However, that is not reliable in cases where the loop body
8227 is more complicated than a single basic block. It also does not work at all
8228 on some architectures due to restrictions in the CSE pass.
8230 This optimization is enabled by default.
8232 @item -fvariable-expansion-in-unroller
8233 @opindex fvariable-expansion-in-unroller
8234 With this option, the compiler creates multiple copies of some
8235 local variables when unrolling a loop, which can result in superior code.
8237 @item -fpartial-inlining
8238 @opindex fpartial-inlining
8239 Inline parts of functions. This option has any effect only
8240 when inlining itself is turned on by the @option{-finline-functions}
8241 or @option{-finline-small-functions} options.
8243 Enabled at level @option{-O2}.
8245 @item -fpredictive-commoning
8246 @opindex fpredictive-commoning
8247 Perform predictive commoning optimization, i.e., reusing computations
8248 (especially memory loads and stores) performed in previous
8249 iterations of loops.
8251 This option is enabled at level @option{-O3}.
8253 @item -fprefetch-loop-arrays
8254 @opindex fprefetch-loop-arrays
8255 If supported by the target machine, generate instructions to prefetch
8256 memory to improve the performance of loops that access large arrays.
8258 This option may generate better or worse code; results are highly
8259 dependent on the structure of loops within the source code.
8261 Disabled at level @option{-Os}.
8263 @item -fprintf-return-value
8264 @opindex fprintf-return-value
8265 Substitute constants for known return value of formatted output functions
8266 such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and @code{vsnprintf}
8267 (but not @code{printf} of @code{fprintf}). This transformation allows GCC
8268 to optimize or even eliminate branches based on the known return value of
8269 these functions called with arguments that are either constant, or whose
8270 values are known to be in a range that makes determining the exact return
8271 value possible. For example, both the branch and the body of the @code{if}
8272 statement (but not the call to @code{snprint}) can be optimized away when
8273 @code{i} is a 32-bit or smaller integer because the return value is guaranteed
8278 if (snprintf (buf, "%08x", i) >= sizeof buf)
8282 The @option{-fprintf-return-value} option relies on other optimizations
8283 and yields best results with @option{-O2}. It works in tandem with the
8284 @option{-Wformat-length} option. The @option{-fprintf-return-value}
8285 option is disabled by default.
8288 @itemx -fno-peephole2
8289 @opindex fno-peephole
8290 @opindex fno-peephole2
8291 Disable any machine-specific peephole optimizations. The difference
8292 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8293 are implemented in the compiler; some targets use one, some use the
8294 other, a few use both.
8296 @option{-fpeephole} is enabled by default.
8297 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8299 @item -fno-guess-branch-probability
8300 @opindex fno-guess-branch-probability
8301 Do not guess branch probabilities using heuristics.
8303 GCC uses heuristics to guess branch probabilities if they are
8304 not provided by profiling feedback (@option{-fprofile-arcs}). These
8305 heuristics are based on the control flow graph. If some branch probabilities
8306 are specified by @code{__builtin_expect}, then the heuristics are
8307 used to guess branch probabilities for the rest of the control flow graph,
8308 taking the @code{__builtin_expect} info into account. The interactions
8309 between the heuristics and @code{__builtin_expect} can be complex, and in
8310 some cases, it may be useful to disable the heuristics so that the effects
8311 of @code{__builtin_expect} are easier to understand.
8313 The default is @option{-fguess-branch-probability} at levels
8314 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8316 @item -freorder-blocks
8317 @opindex freorder-blocks
8318 Reorder basic blocks in the compiled function in order to reduce number of
8319 taken branches and improve code locality.
8321 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8323 @item -freorder-blocks-algorithm=@var{algorithm}
8324 @opindex freorder-blocks-algorithm
8325 Use the specified algorithm for basic block reordering. The
8326 @var{algorithm} argument can be @samp{simple}, which does not increase
8327 code size (except sometimes due to secondary effects like alignment),
8328 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8329 put all often executed code together, minimizing the number of branches
8330 executed by making extra copies of code.
8332 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8333 @samp{stc} at levels @option{-O2}, @option{-O3}.
8335 @item -freorder-blocks-and-partition
8336 @opindex freorder-blocks-and-partition
8337 In addition to reordering basic blocks in the compiled function, in order
8338 to reduce number of taken branches, partitions hot and cold basic blocks
8339 into separate sections of the assembly and @file{.o} files, to improve
8340 paging and cache locality performance.
8342 This optimization is automatically turned off in the presence of
8343 exception handling, for linkonce sections, for functions with a user-defined
8344 section attribute and on any architecture that does not support named
8347 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8349 @item -freorder-functions
8350 @opindex freorder-functions
8351 Reorder functions in the object file in order to
8352 improve code locality. This is implemented by using special
8353 subsections @code{.text.hot} for most frequently executed functions and
8354 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8355 the linker so object file format must support named sections and linker must
8356 place them in a reasonable way.
8358 Also profile feedback must be available to make this option effective. See
8359 @option{-fprofile-arcs} for details.
8361 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8363 @item -fstrict-aliasing
8364 @opindex fstrict-aliasing
8365 Allow the compiler to assume the strictest aliasing rules applicable to
8366 the language being compiled. For C (and C++), this activates
8367 optimizations based on the type of expressions. In particular, an
8368 object of one type is assumed never to reside at the same address as an
8369 object of a different type, unless the types are almost the same. For
8370 example, an @code{unsigned int} can alias an @code{int}, but not a
8371 @code{void*} or a @code{double}. A character type may alias any other
8374 @anchor{Type-punning}Pay special attention to code like this:
8387 The practice of reading from a different union member than the one most
8388 recently written to (called ``type-punning'') is common. Even with
8389 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8390 is accessed through the union type. So, the code above works as
8391 expected. @xref{Structures unions enumerations and bit-fields
8392 implementation}. However, this code might not:
8403 Similarly, access by taking the address, casting the resulting pointer
8404 and dereferencing the result has undefined behavior, even if the cast
8405 uses a union type, e.g.:
8409 return ((union a_union *) &d)->i;
8413 The @option{-fstrict-aliasing} option is enabled at levels
8414 @option{-O2}, @option{-O3}, @option{-Os}.
8416 @item -fstrict-overflow
8417 @opindex fstrict-overflow
8418 Allow the compiler to assume strict signed overflow rules, depending
8419 on the language being compiled. For C (and C++) this means that
8420 overflow when doing arithmetic with signed numbers is undefined, which
8421 means that the compiler may assume that it does not happen. This
8422 permits various optimizations. For example, the compiler assumes
8423 that an expression like @code{i + 10 > i} is always true for
8424 signed @code{i}. This assumption is only valid if signed overflow is
8425 undefined, as the expression is false if @code{i + 10} overflows when
8426 using twos complement arithmetic. When this option is in effect any
8427 attempt to determine whether an operation on signed numbers
8428 overflows must be written carefully to not actually involve overflow.
8430 This option also allows the compiler to assume strict pointer
8431 semantics: given a pointer to an object, if adding an offset to that
8432 pointer does not produce a pointer to the same object, the addition is
8433 undefined. This permits the compiler to conclude that @code{p + u >
8434 p} is always true for a pointer @code{p} and unsigned integer
8435 @code{u}. This assumption is only valid because pointer wraparound is
8436 undefined, as the expression is false if @code{p + u} overflows using
8437 twos complement arithmetic.
8439 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8440 that integer signed overflow is fully defined: it wraps. When
8441 @option{-fwrapv} is used, there is no difference between
8442 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8443 integers. With @option{-fwrapv} certain types of overflow are
8444 permitted. For example, if the compiler gets an overflow when doing
8445 arithmetic on constants, the overflowed value can still be used with
8446 @option{-fwrapv}, but not otherwise.
8448 The @option{-fstrict-overflow} option is enabled at levels
8449 @option{-O2}, @option{-O3}, @option{-Os}.
8451 @item -falign-functions
8452 @itemx -falign-functions=@var{n}
8453 @opindex falign-functions
8454 Align the start of functions to the next power-of-two greater than
8455 @var{n}, skipping up to @var{n} bytes. For instance,
8456 @option{-falign-functions=32} aligns functions to the next 32-byte
8457 boundary, but @option{-falign-functions=24} aligns to the next
8458 32-byte boundary only if this can be done by skipping 23 bytes or less.
8460 @option{-fno-align-functions} and @option{-falign-functions=1} are
8461 equivalent and mean that functions are not aligned.
8463 Some assemblers only support this flag when @var{n} is a power of two;
8464 in that case, it is rounded up.
8466 If @var{n} is not specified or is zero, use a machine-dependent default.
8468 Enabled at levels @option{-O2}, @option{-O3}.
8470 @item -falign-labels
8471 @itemx -falign-labels=@var{n}
8472 @opindex falign-labels
8473 Align all branch targets to a power-of-two boundary, skipping up to
8474 @var{n} bytes like @option{-falign-functions}. This option can easily
8475 make code slower, because it must insert dummy operations for when the
8476 branch target is reached in the usual flow of the code.
8478 @option{-fno-align-labels} and @option{-falign-labels=1} are
8479 equivalent and mean that labels are not aligned.
8481 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8482 are greater than this value, then their values are used instead.
8484 If @var{n} is not specified or is zero, use a machine-dependent default
8485 which is very likely to be @samp{1}, meaning no alignment.
8487 Enabled at levels @option{-O2}, @option{-O3}.
8490 @itemx -falign-loops=@var{n}
8491 @opindex falign-loops
8492 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8493 like @option{-falign-functions}. If the loops are
8494 executed many times, this makes up for any execution of the dummy
8497 @option{-fno-align-loops} and @option{-falign-loops=1} are
8498 equivalent and mean that loops are not aligned.
8500 If @var{n} is not specified or is zero, use a machine-dependent default.
8502 Enabled at levels @option{-O2}, @option{-O3}.
8505 @itemx -falign-jumps=@var{n}
8506 @opindex falign-jumps
8507 Align branch targets to a power-of-two boundary, for branch targets
8508 where the targets can only be reached by jumping, skipping up to @var{n}
8509 bytes like @option{-falign-functions}. In this case, no dummy operations
8512 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8513 equivalent and mean that loops are not aligned.
8515 If @var{n} is not specified or is zero, use a machine-dependent default.
8517 Enabled at levels @option{-O2}, @option{-O3}.
8519 @item -funit-at-a-time
8520 @opindex funit-at-a-time
8521 This option is left for compatibility reasons. @option{-funit-at-a-time}
8522 has no effect, while @option{-fno-unit-at-a-time} implies
8523 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8527 @item -fno-toplevel-reorder
8528 @opindex fno-toplevel-reorder
8529 Do not reorder top-level functions, variables, and @code{asm}
8530 statements. Output them in the same order that they appear in the
8531 input file. When this option is used, unreferenced static variables
8532 are not removed. This option is intended to support existing code
8533 that relies on a particular ordering. For new code, it is better to
8534 use attributes when possible.
8536 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8537 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8542 Constructs webs as commonly used for register allocation purposes and assign
8543 each web individual pseudo register. This allows the register allocation pass
8544 to operate on pseudos directly, but also strengthens several other optimization
8545 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8546 however, make debugging impossible, since variables no longer stay in a
8549 Enabled by default with @option{-funroll-loops}.
8551 @item -fwhole-program
8552 @opindex fwhole-program
8553 Assume that the current compilation unit represents the whole program being
8554 compiled. All public functions and variables with the exception of @code{main}
8555 and those merged by attribute @code{externally_visible} become static functions
8556 and in effect are optimized more aggressively by interprocedural optimizers.
8558 This option should not be used in combination with @option{-flto}.
8559 Instead relying on a linker plugin should provide safer and more precise
8562 @item -flto[=@var{n}]
8564 This option runs the standard link-time optimizer. When invoked
8565 with source code, it generates GIMPLE (one of GCC's internal
8566 representations) and writes it to special ELF sections in the object
8567 file. When the object files are linked together, all the function
8568 bodies are read from these ELF sections and instantiated as if they
8569 had been part of the same translation unit.
8571 To use the link-time optimizer, @option{-flto} and optimization
8572 options should be specified at compile time and during the final link.
8573 It is recommended that you compile all the files participating in the
8574 same link with the same options and also specify those options at
8579 gcc -c -O2 -flto foo.c
8580 gcc -c -O2 -flto bar.c
8581 gcc -o myprog -flto -O2 foo.o bar.o
8584 The first two invocations to GCC save a bytecode representation
8585 of GIMPLE into special ELF sections inside @file{foo.o} and
8586 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8587 @file{foo.o} and @file{bar.o}, merges the two files into a single
8588 internal image, and compiles the result as usual. Since both
8589 @file{foo.o} and @file{bar.o} are merged into a single image, this
8590 causes all the interprocedural analyses and optimizations in GCC to
8591 work across the two files as if they were a single one. This means,
8592 for example, that the inliner is able to inline functions in
8593 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8595 Another (simpler) way to enable link-time optimization is:
8598 gcc -o myprog -flto -O2 foo.c bar.c
8601 The above generates bytecode for @file{foo.c} and @file{bar.c},
8602 merges them together into a single GIMPLE representation and optimizes
8603 them as usual to produce @file{myprog}.
8605 The only important thing to keep in mind is that to enable link-time
8606 optimizations you need to use the GCC driver to perform the link step.
8607 GCC then automatically performs link-time optimization if any of the
8608 objects involved were compiled with the @option{-flto} command-line option.
8610 should specify the optimization options to be used for link-time
8611 optimization though GCC tries to be clever at guessing an
8612 optimization level to use from the options used at compile time
8613 if you fail to specify one at link time. You can always override
8614 the automatic decision to do link-time optimization at link time
8615 by passing @option{-fno-lto} to the link command.
8617 To make whole program optimization effective, it is necessary to make
8618 certain whole program assumptions. The compiler needs to know
8619 what functions and variables can be accessed by libraries and runtime
8620 outside of the link-time optimized unit. When supported by the linker,
8621 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8622 to the compiler about used and externally visible symbols. When
8623 the linker plugin is not available, @option{-fwhole-program} should be
8624 used to allow the compiler to make these assumptions, which leads
8625 to more aggressive optimization decisions.
8627 When @option{-fuse-linker-plugin} is not enabled, when a file is
8628 compiled with @option{-flto}, the generated object file is larger than
8629 a regular object file because it contains GIMPLE bytecodes and the usual
8630 final code (see @option{-ffat-lto-objects}. This means that
8631 object files with LTO information can be linked as normal object
8632 files; if @option{-fno-lto} is passed to the linker, no
8633 interprocedural optimizations are applied. Note that when
8634 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8635 but you cannot perform a regular, non-LTO link on them.
8637 Additionally, the optimization flags used to compile individual files
8638 are not necessarily related to those used at link time. For instance,
8641 gcc -c -O0 -ffat-lto-objects -flto foo.c
8642 gcc -c -O0 -ffat-lto-objects -flto bar.c
8643 gcc -o myprog -O3 foo.o bar.o
8646 This produces individual object files with unoptimized assembler
8647 code, but the resulting binary @file{myprog} is optimized at
8648 @option{-O3}. If, instead, the final binary is generated with
8649 @option{-fno-lto}, then @file{myprog} is not optimized.
8651 When producing the final binary, GCC only
8652 applies link-time optimizations to those files that contain bytecode.
8653 Therefore, you can mix and match object files and libraries with
8654 GIMPLE bytecodes and final object code. GCC automatically selects
8655 which files to optimize in LTO mode and which files to link without
8658 There are some code generation flags preserved by GCC when
8659 generating bytecodes, as they need to be used during the final link
8660 stage. Generally options specified at link time override those
8661 specified at compile time.
8663 If you do not specify an optimization level option @option{-O} at
8664 link time, then GCC uses the highest optimization level
8665 used when compiling the object files.
8667 Currently, the following options and their settings are taken from
8668 the first object file that explicitly specifies them:
8669 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8670 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8671 and all the @option{-m} target flags.
8673 Certain ABI-changing flags are required to match in all compilation units,
8674 and trying to override this at link time with a conflicting value
8675 is ignored. This includes options such as @option{-freg-struct-return}
8676 and @option{-fpcc-struct-return}.
8678 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8679 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8680 are passed through to the link stage and merged conservatively for
8681 conflicting translation units. Specifically
8682 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8683 precedence; and for example @option{-ffp-contract=off} takes precedence
8684 over @option{-ffp-contract=fast}. You can override them at link time.
8686 If LTO encounters objects with C linkage declared with incompatible
8687 types in separate translation units to be linked together (undefined
8688 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8689 issued. The behavior is still undefined at run time. Similar
8690 diagnostics may be raised for other languages.
8692 Another feature of LTO is that it is possible to apply interprocedural
8693 optimizations on files written in different languages:
8698 gfortran -c -flto baz.f90
8699 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8702 Notice that the final link is done with @command{g++} to get the C++
8703 runtime libraries and @option{-lgfortran} is added to get the Fortran
8704 runtime libraries. In general, when mixing languages in LTO mode, you
8705 should use the same link command options as when mixing languages in a
8706 regular (non-LTO) compilation.
8708 If object files containing GIMPLE bytecode are stored in a library archive, say
8709 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8710 are using a linker with plugin support. To create static libraries suitable
8711 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8712 and @command{ranlib};
8713 to show the symbols of object files with GIMPLE bytecode, use
8714 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8715 and @command{nm} have been compiled with plugin support. At link time, use the the
8716 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8717 the LTO optimization process:
8720 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8723 With the linker plugin enabled, the linker extracts the needed
8724 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8725 to make them part of the aggregated GIMPLE image to be optimized.
8727 If you are not using a linker with plugin support and/or do not
8728 enable the linker plugin, then the objects inside @file{libfoo.a}
8729 are extracted and linked as usual, but they do not participate
8730 in the LTO optimization process. In order to make a static library suitable
8731 for both LTO optimization and usual linkage, compile its object files with
8732 @option{-flto} @option{-ffat-lto-objects}.
8734 Link-time optimizations do not require the presence of the whole program to
8735 operate. If the program does not require any symbols to be exported, it is
8736 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8737 the interprocedural optimizers to use more aggressive assumptions which may
8738 lead to improved optimization opportunities.
8739 Use of @option{-fwhole-program} is not needed when linker plugin is
8740 active (see @option{-fuse-linker-plugin}).
8742 The current implementation of LTO makes no
8743 attempt to generate bytecode that is portable between different
8744 types of hosts. The bytecode files are versioned and there is a
8745 strict version check, so bytecode files generated in one version of
8746 GCC do not work with an older or newer version of GCC.
8748 Link-time optimization does not work well with generation of debugging
8749 information. Combining @option{-flto} with
8750 @option{-g} is currently experimental and expected to produce unexpected
8753 If you specify the optional @var{n}, the optimization and code
8754 generation done at link time is executed in parallel using @var{n}
8755 parallel jobs by utilizing an installed @command{make} program. The
8756 environment variable @env{MAKE} may be used to override the program
8757 used. The default value for @var{n} is 1.
8759 You can also specify @option{-flto=jobserver} to use GNU make's
8760 job server mode to determine the number of parallel jobs. This
8761 is useful when the Makefile calling GCC is already executing in parallel.
8762 You must prepend a @samp{+} to the command recipe in the parent Makefile
8763 for this to work. This option likely only works if @env{MAKE} is
8766 @item -flto-partition=@var{alg}
8767 @opindex flto-partition
8768 Specify the partitioning algorithm used by the link-time optimizer.
8769 The value is either @samp{1to1} to specify a partitioning mirroring
8770 the original source files or @samp{balanced} to specify partitioning
8771 into equally sized chunks (whenever possible) or @samp{max} to create
8772 new partition for every symbol where possible. Specifying @samp{none}
8773 as an algorithm disables partitioning and streaming completely.
8774 The default value is @samp{balanced}. While @samp{1to1} can be used
8775 as an workaround for various code ordering issues, the @samp{max}
8776 partitioning is intended for internal testing only.
8777 The value @samp{one} specifies that exactly one partition should be
8778 used while the value @samp{none} bypasses partitioning and executes
8779 the link-time optimization step directly from the WPA phase.
8781 @item -flto-odr-type-merging
8782 @opindex flto-odr-type-merging
8783 Enable streaming of mangled types names of C++ types and their unification
8784 at link time. This increases size of LTO object files, but enables
8785 diagnostics about One Definition Rule violations.
8787 @item -flto-compression-level=@var{n}
8788 @opindex flto-compression-level
8789 This option specifies the level of compression used for intermediate
8790 language written to LTO object files, and is only meaningful in
8791 conjunction with LTO mode (@option{-flto}). Valid
8792 values are 0 (no compression) to 9 (maximum compression). Values
8793 outside this range are clamped to either 0 or 9. If the option is not
8794 given, a default balanced compression setting is used.
8796 @item -fuse-linker-plugin
8797 @opindex fuse-linker-plugin
8798 Enables the use of a linker plugin during link-time optimization. This
8799 option relies on plugin support in the linker, which is available in gold
8800 or in GNU ld 2.21 or newer.
8802 This option enables the extraction of object files with GIMPLE bytecode out
8803 of library archives. This improves the quality of optimization by exposing
8804 more code to the link-time optimizer. This information specifies what
8805 symbols can be accessed externally (by non-LTO object or during dynamic
8806 linking). Resulting code quality improvements on binaries (and shared
8807 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8808 See @option{-flto} for a description of the effect of this flag and how to
8811 This option is enabled by default when LTO support in GCC is enabled
8812 and GCC was configured for use with
8813 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8815 @item -ffat-lto-objects
8816 @opindex ffat-lto-objects
8817 Fat LTO objects are object files that contain both the intermediate language
8818 and the object code. This makes them usable for both LTO linking and normal
8819 linking. This option is effective only when compiling with @option{-flto}
8820 and is ignored at link time.
8822 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8823 requires the complete toolchain to be aware of LTO. It requires a linker with
8824 linker plugin support for basic functionality. Additionally,
8825 @command{nm}, @command{ar} and @command{ranlib}
8826 need to support linker plugins to allow a full-featured build environment
8827 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8828 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8829 to these tools. With non fat LTO makefiles need to be modified to use them.
8831 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8834 @item -fcompare-elim
8835 @opindex fcompare-elim
8836 After register allocation and post-register allocation instruction splitting,
8837 identify arithmetic instructions that compute processor flags similar to a
8838 comparison operation based on that arithmetic. If possible, eliminate the
8839 explicit comparison operation.
8841 This pass only applies to certain targets that cannot explicitly represent
8842 the comparison operation before register allocation is complete.
8844 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8846 @item -fcprop-registers
8847 @opindex fcprop-registers
8848 After register allocation and post-register allocation instruction splitting,
8849 perform a copy-propagation pass to try to reduce scheduling dependencies
8850 and occasionally eliminate the copy.
8852 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8854 @item -fprofile-correction
8855 @opindex fprofile-correction
8856 Profiles collected using an instrumented binary for multi-threaded programs may
8857 be inconsistent due to missed counter updates. When this option is specified,
8858 GCC uses heuristics to correct or smooth out such inconsistencies. By
8859 default, GCC emits an error message when an inconsistent profile is detected.
8862 @itemx -fprofile-use=@var{path}
8863 @opindex fprofile-use
8864 Enable profile feedback-directed optimizations,
8865 and the following optimizations
8866 which are generally profitable only with profile feedback available:
8867 @option{-fbranch-probabilities}, @option{-fvpt},
8868 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8869 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
8871 Before you can use this option, you must first generate profiling information.
8872 @xref{Optimize Options}, for information about the @option{-fprofile-generate}
8875 By default, GCC emits an error message if the feedback profiles do not
8876 match the source code. This error can be turned into a warning by using
8877 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8880 If @var{path} is specified, GCC looks at the @var{path} to find
8881 the profile feedback data files. See @option{-fprofile-dir}.
8883 @item -fauto-profile
8884 @itemx -fauto-profile=@var{path}
8885 @opindex fauto-profile
8886 Enable sampling-based feedback-directed optimizations,
8887 and the following optimizations
8888 which are generally profitable only with profile feedback available:
8889 @option{-fbranch-probabilities}, @option{-fvpt},
8890 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8891 @option{-ftree-vectorize},
8892 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
8893 @option{-fpredictive-commoning}, @option{-funswitch-loops},
8894 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
8896 @var{path} is the name of a file containing AutoFDO profile information.
8897 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
8899 Producing an AutoFDO profile data file requires running your program
8900 with the @command{perf} utility on a supported GNU/Linux target system.
8901 For more information, see @uref{https://perf.wiki.kernel.org/}.
8905 perf record -e br_inst_retired:near_taken -b -o perf.data \
8909 Then use the @command{create_gcov} tool to convert the raw profile data
8910 to a format that can be used by GCC.@ You must also supply the
8911 unstripped binary for your program to this tool.
8912 See @uref{https://github.com/google/autofdo}.
8916 create_gcov --binary=your_program.unstripped --profile=perf.data \
8921 The following options control compiler behavior regarding floating-point
8922 arithmetic. These options trade off between speed and
8923 correctness. All must be specifically enabled.
8927 @opindex ffloat-store
8928 Do not store floating-point variables in registers, and inhibit other
8929 options that might change whether a floating-point value is taken from a
8932 @cindex floating-point precision
8933 This option prevents undesirable excess precision on machines such as
8934 the 68000 where the floating registers (of the 68881) keep more
8935 precision than a @code{double} is supposed to have. Similarly for the
8936 x86 architecture. For most programs, the excess precision does only
8937 good, but a few programs rely on the precise definition of IEEE floating
8938 point. Use @option{-ffloat-store} for such programs, after modifying
8939 them to store all pertinent intermediate computations into variables.
8941 @item -fexcess-precision=@var{style}
8942 @opindex fexcess-precision
8943 This option allows further control over excess precision on machines
8944 where floating-point registers have more precision than the IEEE
8945 @code{float} and @code{double} types and the processor does not
8946 support operations rounding to those types. By default,
8947 @option{-fexcess-precision=fast} is in effect; this means that
8948 operations are carried out in the precision of the registers and that
8949 it is unpredictable when rounding to the types specified in the source
8950 code takes place. When compiling C, if
8951 @option{-fexcess-precision=standard} is specified then excess
8952 precision follows the rules specified in ISO C99; in particular,
8953 both casts and assignments cause values to be rounded to their
8954 semantic types (whereas @option{-ffloat-store} only affects
8955 assignments). This option is enabled by default for C if a strict
8956 conformance option such as @option{-std=c99} is used.
8959 @option{-fexcess-precision=standard} is not implemented for languages
8960 other than C, and has no effect if
8961 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8962 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8963 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8964 semantics apply without excess precision, and in the latter, rounding
8969 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8970 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8971 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8973 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8975 This option is not turned on by any @option{-O} option besides
8976 @option{-Ofast} since it can result in incorrect output for programs
8977 that depend on an exact implementation of IEEE or ISO rules/specifications
8978 for math functions. It may, however, yield faster code for programs
8979 that do not require the guarantees of these specifications.
8981 @item -fno-math-errno
8982 @opindex fno-math-errno
8983 Do not set @code{errno} after calling math functions that are executed
8984 with a single instruction, e.g., @code{sqrt}. A program that relies on
8985 IEEE exceptions for math error handling may want to use this flag
8986 for speed while maintaining IEEE arithmetic compatibility.
8988 This option is not turned on by any @option{-O} option since
8989 it can result in incorrect output for programs that depend on
8990 an exact implementation of IEEE or ISO rules/specifications for
8991 math functions. It may, however, yield faster code for programs
8992 that do not require the guarantees of these specifications.
8994 The default is @option{-fmath-errno}.
8996 On Darwin systems, the math library never sets @code{errno}. There is
8997 therefore no reason for the compiler to consider the possibility that
8998 it might, and @option{-fno-math-errno} is the default.
9000 @item -funsafe-math-optimizations
9001 @opindex funsafe-math-optimizations
9003 Allow optimizations for floating-point arithmetic that (a) assume
9004 that arguments and results are valid and (b) may violate IEEE or
9005 ANSI standards. When used at link time, it may include libraries
9006 or startup files that change the default FPU control word or other
9007 similar optimizations.
9009 This option is not turned on by any @option{-O} option since
9010 it can result in incorrect output for programs that depend on
9011 an exact implementation of IEEE or ISO rules/specifications for
9012 math functions. It may, however, yield faster code for programs
9013 that do not require the guarantees of these specifications.
9014 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9015 @option{-fassociative-math} and @option{-freciprocal-math}.
9017 The default is @option{-fno-unsafe-math-optimizations}.
9019 @item -fassociative-math
9020 @opindex fassociative-math
9022 Allow re-association of operands in series of floating-point operations.
9023 This violates the ISO C and C++ language standard by possibly changing
9024 computation result. NOTE: re-ordering may change the sign of zero as
9025 well as ignore NaNs and inhibit or create underflow or overflow (and
9026 thus cannot be used on code that relies on rounding behavior like
9027 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9028 and thus may not be used when ordered comparisons are required.
9029 This option requires that both @option{-fno-signed-zeros} and
9030 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9031 much sense with @option{-frounding-math}. For Fortran the option
9032 is automatically enabled when both @option{-fno-signed-zeros} and
9033 @option{-fno-trapping-math} are in effect.
9035 The default is @option{-fno-associative-math}.
9037 @item -freciprocal-math
9038 @opindex freciprocal-math
9040 Allow the reciprocal of a value to be used instead of dividing by
9041 the value if this enables optimizations. For example @code{x / y}
9042 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9043 is subject to common subexpression elimination. Note that this loses
9044 precision and increases the number of flops operating on the value.
9046 The default is @option{-fno-reciprocal-math}.
9048 @item -ffinite-math-only
9049 @opindex ffinite-math-only
9050 Allow optimizations for floating-point arithmetic that assume
9051 that arguments and results are not NaNs or +-Infs.
9053 This option is not turned on by any @option{-O} option since
9054 it can result in incorrect output for programs that depend on
9055 an exact implementation of IEEE or ISO rules/specifications for
9056 math functions. It may, however, yield faster code for programs
9057 that do not require the guarantees of these specifications.
9059 The default is @option{-fno-finite-math-only}.
9061 @item -fno-signed-zeros
9062 @opindex fno-signed-zeros
9063 Allow optimizations for floating-point arithmetic that ignore the
9064 signedness of zero. IEEE arithmetic specifies the behavior of
9065 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9066 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9067 This option implies that the sign of a zero result isn't significant.
9069 The default is @option{-fsigned-zeros}.
9071 @item -fno-trapping-math
9072 @opindex fno-trapping-math
9073 Compile code assuming that floating-point operations cannot generate
9074 user-visible traps. These traps include division by zero, overflow,
9075 underflow, inexact result and invalid operation. This option requires
9076 that @option{-fno-signaling-nans} be in effect. Setting this option may
9077 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9079 This option should never be turned on by any @option{-O} option since
9080 it can result in incorrect output for programs that depend on
9081 an exact implementation of IEEE or ISO rules/specifications for
9084 The default is @option{-ftrapping-math}.
9086 @item -frounding-math
9087 @opindex frounding-math
9088 Disable transformations and optimizations that assume default floating-point
9089 rounding behavior. This is round-to-zero for all floating point
9090 to integer conversions, and round-to-nearest for all other arithmetic
9091 truncations. This option should be specified for programs that change
9092 the FP rounding mode dynamically, or that may be executed with a
9093 non-default rounding mode. This option disables constant folding of
9094 floating-point expressions at compile time (which may be affected by
9095 rounding mode) and arithmetic transformations that are unsafe in the
9096 presence of sign-dependent rounding modes.
9098 The default is @option{-fno-rounding-math}.
9100 This option is experimental and does not currently guarantee to
9101 disable all GCC optimizations that are affected by rounding mode.
9102 Future versions of GCC may provide finer control of this setting
9103 using C99's @code{FENV_ACCESS} pragma. This command-line option
9104 will be used to specify the default state for @code{FENV_ACCESS}.
9106 @item -fsignaling-nans
9107 @opindex fsignaling-nans
9108 Compile code assuming that IEEE signaling NaNs may generate user-visible
9109 traps during floating-point operations. Setting this option disables
9110 optimizations that may change the number of exceptions visible with
9111 signaling NaNs. This option implies @option{-ftrapping-math}.
9113 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9116 The default is @option{-fno-signaling-nans}.
9118 This option is experimental and does not currently guarantee to
9119 disable all GCC optimizations that affect signaling NaN behavior.
9121 @item -fno-fp-int-builtin-inexact
9122 @opindex fno-fp-int-builtin-inexact
9123 Do not allow the built-in functions @code{ceil}, @code{floor},
9124 @code{round} and @code{trunc}, and their @code{float} and @code{long
9125 double} variants, to generate code that raises the ``inexact''
9126 floating-point exception for noninteger arguments. ISO C99 and C11
9127 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9128 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9131 The default is @option{-ffp-int-builtin-inexact}, allowing the
9132 exception to be raised. This option does nothing unless
9133 @option{-ftrapping-math} is in effect.
9135 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9136 generate a call to a library function then the ``inexact'' exception
9137 may be raised if the library implementation does not follow TS 18661.
9139 @item -fsingle-precision-constant
9140 @opindex fsingle-precision-constant
9141 Treat floating-point constants as single precision instead of
9142 implicitly converting them to double-precision constants.
9144 @item -fcx-limited-range
9145 @opindex fcx-limited-range
9146 When enabled, this option states that a range reduction step is not
9147 needed when performing complex division. Also, there is no checking
9148 whether the result of a complex multiplication or division is @code{NaN
9149 + I*NaN}, with an attempt to rescue the situation in that case. The
9150 default is @option{-fno-cx-limited-range}, but is enabled by
9151 @option{-ffast-math}.
9153 This option controls the default setting of the ISO C99
9154 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9157 @item -fcx-fortran-rules
9158 @opindex fcx-fortran-rules
9159 Complex multiplication and division follow Fortran rules. Range
9160 reduction is done as part of complex division, but there is no checking
9161 whether the result of a complex multiplication or division is @code{NaN
9162 + I*NaN}, with an attempt to rescue the situation in that case.
9164 The default is @option{-fno-cx-fortran-rules}.
9168 The following options control optimizations that may improve
9169 performance, but are not enabled by any @option{-O} options. This
9170 section includes experimental options that may produce broken code.
9173 @item -fbranch-probabilities
9174 @opindex fbranch-probabilities
9175 After running a program compiled with @option{-fprofile-arcs}
9176 (@pxref{Instrumentation Options}),
9177 you can compile it a second time using
9178 @option{-fbranch-probabilities}, to improve optimizations based on
9179 the number of times each branch was taken. When a program
9180 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9181 counts to a file called @file{@var{sourcename}.gcda} for each source
9182 file. The information in this data file is very dependent on the
9183 structure of the generated code, so you must use the same source code
9184 and the same optimization options for both compilations.
9186 With @option{-fbranch-probabilities}, GCC puts a
9187 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9188 These can be used to improve optimization. Currently, they are only
9189 used in one place: in @file{reorg.c}, instead of guessing which path a
9190 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9191 exactly determine which path is taken more often.
9193 @item -fprofile-values
9194 @opindex fprofile-values
9195 If combined with @option{-fprofile-arcs}, it adds code so that some
9196 data about values of expressions in the program is gathered.
9198 With @option{-fbranch-probabilities}, it reads back the data gathered
9199 from profiling values of expressions for usage in optimizations.
9201 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9203 @item -fprofile-reorder-functions
9204 @opindex fprofile-reorder-functions
9205 Function reordering based on profile instrumentation collects
9206 first time of execution of a function and orders these functions
9209 Enabled with @option{-fprofile-use}.
9213 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9214 to add code to gather information about values of expressions.
9216 With @option{-fbranch-probabilities}, it reads back the data gathered
9217 and actually performs the optimizations based on them.
9218 Currently the optimizations include specialization of division operations
9219 using the knowledge about the value of the denominator.
9221 @item -frename-registers
9222 @opindex frename-registers
9223 Attempt to avoid false dependencies in scheduled code by making use
9224 of registers left over after register allocation. This optimization
9225 most benefits processors with lots of registers. Depending on the
9226 debug information format adopted by the target, however, it can
9227 make debugging impossible, since variables no longer stay in
9228 a ``home register''.
9230 Enabled by default with @option{-funroll-loops}.
9232 @item -fschedule-fusion
9233 @opindex fschedule-fusion
9234 Performs a target dependent pass over the instruction stream to schedule
9235 instructions of same type together because target machine can execute them
9236 more efficiently if they are adjacent to each other in the instruction flow.
9238 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9242 Perform tail duplication to enlarge superblock size. This transformation
9243 simplifies the control flow of the function allowing other optimizations to do
9246 Enabled with @option{-fprofile-use}.
9248 @item -funroll-loops
9249 @opindex funroll-loops
9250 Unroll loops whose number of iterations can be determined at compile time or
9251 upon entry to the loop. @option{-funroll-loops} implies
9252 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9253 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9254 a small constant number of iterations). This option makes code larger, and may
9255 or may not make it run faster.
9257 Enabled with @option{-fprofile-use}.
9259 @item -funroll-all-loops
9260 @opindex funroll-all-loops
9261 Unroll all loops, even if their number of iterations is uncertain when
9262 the loop is entered. This usually makes programs run more slowly.
9263 @option{-funroll-all-loops} implies the same options as
9264 @option{-funroll-loops}.
9267 @opindex fpeel-loops
9268 Peels loops for which there is enough information that they do not
9269 roll much (from profile feedback or static analysis). It also turns on
9270 complete loop peeling (i.e.@: complete removal of loops with small constant
9271 number of iterations).
9273 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9275 @item -fmove-loop-invariants
9276 @opindex fmove-loop-invariants
9277 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9278 at level @option{-O1}
9281 @opindex fsplit-loops
9282 Split a loop into two if it contains a condition that's always true
9283 for one side of the iteration space and false for the other.
9285 @item -funswitch-loops
9286 @opindex funswitch-loops
9287 Move branches with loop invariant conditions out of the loop, with duplicates
9288 of the loop on both branches (modified according to result of the condition).
9290 @item -ffunction-sections
9291 @itemx -fdata-sections
9292 @opindex ffunction-sections
9293 @opindex fdata-sections
9294 Place each function or data item into its own section in the output
9295 file if the target supports arbitrary sections. The name of the
9296 function or the name of the data item determines the section's name
9299 Use these options on systems where the linker can perform optimizations
9300 to improve locality of reference in the instruction space. Most systems
9301 using the ELF object format and SPARC processors running Solaris 2 have
9302 linkers with such optimizations. AIX may have these optimizations in
9305 Only use these options when there are significant benefits from doing
9306 so. When you specify these options, the assembler and linker
9307 create larger object and executable files and are also slower.
9308 You cannot use @command{gprof} on all systems if you
9309 specify this option, and you may have problems with debugging if
9310 you specify both this option and @option{-g}.
9312 @item -fbranch-target-load-optimize
9313 @opindex fbranch-target-load-optimize
9314 Perform branch target register load optimization before prologue / epilogue
9316 The use of target registers can typically be exposed only during reload,
9317 thus hoisting loads out of loops and doing inter-block scheduling needs
9318 a separate optimization pass.
9320 @item -fbranch-target-load-optimize2
9321 @opindex fbranch-target-load-optimize2
9322 Perform branch target register load optimization after prologue / epilogue
9325 @item -fbtr-bb-exclusive
9326 @opindex fbtr-bb-exclusive
9327 When performing branch target register load optimization, don't reuse
9328 branch target registers within any basic block.
9331 @opindex fstdarg-opt
9332 Optimize the prologue of variadic argument functions with respect to usage of
9335 @item -fsection-anchors
9336 @opindex fsection-anchors
9337 Try to reduce the number of symbolic address calculations by using
9338 shared ``anchor'' symbols to address nearby objects. This transformation
9339 can help to reduce the number of GOT entries and GOT accesses on some
9342 For example, the implementation of the following function @code{foo}:
9346 int foo (void) @{ return a + b + c; @}
9350 usually calculates the addresses of all three variables, but if you
9351 compile it with @option{-fsection-anchors}, it accesses the variables
9352 from a common anchor point instead. The effect is similar to the
9353 following pseudocode (which isn't valid C):
9358 register int *xr = &x;
9359 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9363 Not all targets support this option.
9365 @item --param @var{name}=@var{value}
9367 In some places, GCC uses various constants to control the amount of
9368 optimization that is done. For example, GCC does not inline functions
9369 that contain more than a certain number of instructions. You can
9370 control some of these constants on the command line using the
9371 @option{--param} option.
9373 The names of specific parameters, and the meaning of the values, are
9374 tied to the internals of the compiler, and are subject to change
9375 without notice in future releases.
9377 In each case, the @var{value} is an integer. The allowable choices for
9381 @item predictable-branch-outcome
9382 When branch is predicted to be taken with probability lower than this threshold
9383 (in percent), then it is considered well predictable. The default is 10.
9385 @item max-rtl-if-conversion-insns
9386 RTL if-conversion tries to remove conditional branches around a block and
9387 replace them with conditionally executed instructions. This parameter
9388 gives the maximum number of instructions in a block which should be
9389 considered for if-conversion. The default is 10, though the compiler will
9390 also use other heuristics to decide whether if-conversion is likely to be
9393 @item max-rtl-if-conversion-predictable-cost
9394 @item max-rtl-if-conversion-unpredictable-cost
9395 RTL if-conversion will try to remove conditional branches around a block
9396 and replace them with conditionally executed instructions. These parameters
9397 give the maximum permissible cost for the sequence that would be generated
9398 by if-conversion depending on whether the branch is statically determined
9399 to be predictable or not. The units for this parameter are the same as
9400 those for the GCC internal seq_cost metric. The compiler will try to
9401 provide a reasonable default for this parameter using the BRANCH_COST
9404 @item max-crossjump-edges
9405 The maximum number of incoming edges to consider for cross-jumping.
9406 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9407 the number of edges incoming to each block. Increasing values mean
9408 more aggressive optimization, making the compilation time increase with
9409 probably small improvement in executable size.
9411 @item min-crossjump-insns
9412 The minimum number of instructions that must be matched at the end
9413 of two blocks before cross-jumping is performed on them. This
9414 value is ignored in the case where all instructions in the block being
9415 cross-jumped from are matched. The default value is 5.
9417 @item max-grow-copy-bb-insns
9418 The maximum code size expansion factor when copying basic blocks
9419 instead of jumping. The expansion is relative to a jump instruction.
9420 The default value is 8.
9422 @item max-goto-duplication-insns
9423 The maximum number of instructions to duplicate to a block that jumps
9424 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9425 passes, GCC factors computed gotos early in the compilation process,
9426 and unfactors them as late as possible. Only computed jumps at the
9427 end of a basic blocks with no more than max-goto-duplication-insns are
9428 unfactored. The default value is 8.
9430 @item max-delay-slot-insn-search
9431 The maximum number of instructions to consider when looking for an
9432 instruction to fill a delay slot. If more than this arbitrary number of
9433 instructions are searched, the time savings from filling the delay slot
9434 are minimal, so stop searching. Increasing values mean more
9435 aggressive optimization, making the compilation time increase with probably
9436 small improvement in execution time.
9438 @item max-delay-slot-live-search
9439 When trying to fill delay slots, the maximum number of instructions to
9440 consider when searching for a block with valid live register
9441 information. Increasing this arbitrarily chosen value means more
9442 aggressive optimization, increasing the compilation time. This parameter
9443 should be removed when the delay slot code is rewritten to maintain the
9446 @item max-gcse-memory
9447 The approximate maximum amount of memory that can be allocated in
9448 order to perform the global common subexpression elimination
9449 optimization. If more memory than specified is required, the
9450 optimization is not done.
9452 @item max-gcse-insertion-ratio
9453 If the ratio of expression insertions to deletions is larger than this value
9454 for any expression, then RTL PRE inserts or removes the expression and thus
9455 leaves partially redundant computations in the instruction stream. The default value is 20.
9457 @item max-pending-list-length
9458 The maximum number of pending dependencies scheduling allows
9459 before flushing the current state and starting over. Large functions
9460 with few branches or calls can create excessively large lists which
9461 needlessly consume memory and resources.
9463 @item max-modulo-backtrack-attempts
9464 The maximum number of backtrack attempts the scheduler should make
9465 when modulo scheduling a loop. Larger values can exponentially increase
9468 @item max-inline-insns-single
9469 Several parameters control the tree inliner used in GCC@.
9470 This number sets the maximum number of instructions (counted in GCC's
9471 internal representation) in a single function that the tree inliner
9472 considers for inlining. This only affects functions declared
9473 inline and methods implemented in a class declaration (C++).
9474 The default value is 400.
9476 @item max-inline-insns-auto
9477 When you use @option{-finline-functions} (included in @option{-O3}),
9478 a lot of functions that would otherwise not be considered for inlining
9479 by the compiler are investigated. To those functions, a different
9480 (more restrictive) limit compared to functions declared inline can
9482 The default value is 40.
9484 @item inline-min-speedup
9485 When estimated performance improvement of caller + callee runtime exceeds this
9486 threshold (in precent), the function can be inlined regardless the limit on
9487 @option{--param max-inline-insns-single} and @option{--param
9488 max-inline-insns-auto}.
9490 @item large-function-insns
9491 The limit specifying really large functions. For functions larger than this
9492 limit after inlining, inlining is constrained by
9493 @option{--param large-function-growth}. This parameter is useful primarily
9494 to avoid extreme compilation time caused by non-linear algorithms used by the
9496 The default value is 2700.
9498 @item large-function-growth
9499 Specifies maximal growth of large function caused by inlining in percents.
9500 The default value is 100 which limits large function growth to 2.0 times
9503 @item large-unit-insns
9504 The limit specifying large translation unit. Growth caused by inlining of
9505 units larger than this limit is limited by @option{--param inline-unit-growth}.
9506 For small units this might be too tight.
9507 For example, consider a unit consisting of function A
9508 that is inline and B that just calls A three times. If B is small relative to
9509 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9510 large units consisting of small inlineable functions, however, the overall unit
9511 growth limit is needed to avoid exponential explosion of code size. Thus for
9512 smaller units, the size is increased to @option{--param large-unit-insns}
9513 before applying @option{--param inline-unit-growth}. The default is 10000.
9515 @item inline-unit-growth
9516 Specifies maximal overall growth of the compilation unit caused by inlining.
9517 The default value is 20 which limits unit growth to 1.2 times the original
9518 size. Cold functions (either marked cold via an attribute or by profile
9519 feedback) are not accounted into the unit size.
9521 @item ipcp-unit-growth
9522 Specifies maximal overall growth of the compilation unit caused by
9523 interprocedural constant propagation. The default value is 10 which limits
9524 unit growth to 1.1 times the original size.
9526 @item large-stack-frame
9527 The limit specifying large stack frames. While inlining the algorithm is trying
9528 to not grow past this limit too much. The default value is 256 bytes.
9530 @item large-stack-frame-growth
9531 Specifies maximal growth of large stack frames caused by inlining in percents.
9532 The default value is 1000 which limits large stack frame growth to 11 times
9535 @item max-inline-insns-recursive
9536 @itemx max-inline-insns-recursive-auto
9537 Specifies the maximum number of instructions an out-of-line copy of a
9538 self-recursive inline
9539 function can grow into by performing recursive inlining.
9541 @option{--param max-inline-insns-recursive} applies to functions
9543 For functions not declared inline, recursive inlining
9544 happens only when @option{-finline-functions} (included in @option{-O3}) is
9545 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9546 default value is 450.
9548 @item max-inline-recursive-depth
9549 @itemx max-inline-recursive-depth-auto
9550 Specifies the maximum recursion depth used for recursive inlining.
9552 @option{--param max-inline-recursive-depth} applies to functions
9553 declared inline. For functions not declared inline, recursive inlining
9554 happens only when @option{-finline-functions} (included in @option{-O3}) is
9555 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9558 @item min-inline-recursive-probability
9559 Recursive inlining is profitable only for function having deep recursion
9560 in average and can hurt for function having little recursion depth by
9561 increasing the prologue size or complexity of function body to other
9564 When profile feedback is available (see @option{-fprofile-generate}) the actual
9565 recursion depth can be guessed from the probability that function recurses
9566 via a given call expression. This parameter limits inlining only to call
9567 expressions whose probability exceeds the given threshold (in percents).
9568 The default value is 10.
9570 @item early-inlining-insns
9571 Specify growth that the early inliner can make. In effect it increases
9572 the amount of inlining for code having a large abstraction penalty.
9573 The default value is 14.
9575 @item max-early-inliner-iterations
9576 Limit of iterations of the early inliner. This basically bounds
9577 the number of nested indirect calls the early inliner can resolve.
9578 Deeper chains are still handled by late inlining.
9580 @item comdat-sharing-probability
9581 Probability (in percent) that C++ inline function with comdat visibility
9582 are shared across multiple compilation units. The default value is 20.
9584 @item profile-func-internal-id
9585 A parameter to control whether to use function internal id in profile
9586 database lookup. If the value is 0, the compiler uses an id that
9587 is based on function assembler name and filename, which makes old profile
9588 data more tolerant to source changes such as function reordering etc.
9589 The default value is 0.
9591 @item min-vect-loop-bound
9592 The minimum number of iterations under which loops are not vectorized
9593 when @option{-ftree-vectorize} is used. The number of iterations after
9594 vectorization needs to be greater than the value specified by this option
9595 to allow vectorization. The default value is 0.
9597 @item gcse-cost-distance-ratio
9598 Scaling factor in calculation of maximum distance an expression
9599 can be moved by GCSE optimizations. This is currently supported only in the
9600 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9601 is with simple expressions, i.e., the expressions that have cost
9602 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9603 hoisting of simple expressions. The default value is 10.
9605 @item gcse-unrestricted-cost
9606 Cost, roughly measured as the cost of a single typical machine
9607 instruction, at which GCSE optimizations do not constrain
9608 the distance an expression can travel. This is currently
9609 supported only in the code hoisting pass. The lesser the cost,
9610 the more aggressive code hoisting is. Specifying 0
9611 allows all expressions to travel unrestricted distances.
9612 The default value is 3.
9614 @item max-hoist-depth
9615 The depth of search in the dominator tree for expressions to hoist.
9616 This is used to avoid quadratic behavior in hoisting algorithm.
9617 The value of 0 does not limit on the search, but may slow down compilation
9618 of huge functions. The default value is 30.
9620 @item max-tail-merge-comparisons
9621 The maximum amount of similar bbs to compare a bb with. This is used to
9622 avoid quadratic behavior in tree tail merging. The default value is 10.
9624 @item max-tail-merge-iterations
9625 The maximum amount of iterations of the pass over the function. This is used to
9626 limit compilation time in tree tail merging. The default value is 2.
9628 @item store-merging-allow-unaligned
9629 Allow the store merging pass to introduce unaligned stores if it is legal to
9630 do so. The default value is 1.
9632 @item max-stores-to-merge
9633 The maximum number of stores to attempt to merge into wider stores in the store
9634 merging pass. The minimum value is 2 and the default is 64.
9636 @item max-unrolled-insns
9637 The maximum number of instructions that a loop may have to be unrolled.
9638 If a loop is unrolled, this parameter also determines how many times
9639 the loop code is unrolled.
9641 @item max-average-unrolled-insns
9642 The maximum number of instructions biased by probabilities of their execution
9643 that a loop may have to be unrolled. If a loop is unrolled,
9644 this parameter also determines how many times the loop code is unrolled.
9646 @item max-unroll-times
9647 The maximum number of unrollings of a single loop.
9649 @item max-peeled-insns
9650 The maximum number of instructions that a loop may have to be peeled.
9651 If a loop is peeled, this parameter also determines how many times
9652 the loop code is peeled.
9654 @item max-peel-times
9655 The maximum number of peelings of a single loop.
9657 @item max-peel-branches
9658 The maximum number of branches on the hot path through the peeled sequence.
9660 @item max-completely-peeled-insns
9661 The maximum number of insns of a completely peeled loop.
9663 @item max-completely-peel-times
9664 The maximum number of iterations of a loop to be suitable for complete peeling.
9666 @item max-completely-peel-loop-nest-depth
9667 The maximum depth of a loop nest suitable for complete peeling.
9669 @item max-unswitch-insns
9670 The maximum number of insns of an unswitched loop.
9672 @item max-unswitch-level
9673 The maximum number of branches unswitched in a single loop.
9675 @item max-loop-headers-insns
9676 The maximum number of insns in loop header duplicated by he copy loop headers
9680 The minimum cost of an expensive expression in the loop invariant motion.
9682 @item iv-consider-all-candidates-bound
9683 Bound on number of candidates for induction variables, below which
9684 all candidates are considered for each use in induction variable
9685 optimizations. If there are more candidates than this,
9686 only the most relevant ones are considered to avoid quadratic time complexity.
9688 @item iv-max-considered-uses
9689 The induction variable optimizations give up on loops that contain more
9690 induction variable uses.
9692 @item iv-always-prune-cand-set-bound
9693 If the number of candidates in the set is smaller than this value,
9694 always try to remove unnecessary ivs from the set
9695 when adding a new one.
9697 @item avg-loop-niter
9698 Average number of iterations of a loop.
9700 @item scev-max-expr-size
9701 Bound on size of expressions used in the scalar evolutions analyzer.
9702 Large expressions slow the analyzer.
9704 @item scev-max-expr-complexity
9705 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9706 Complex expressions slow the analyzer.
9708 @item max-tree-if-conversion-phi-args
9709 Maximum number of arguments in a PHI supported by TREE if conversion
9710 unless the loop is marked with simd pragma.
9712 @item vect-max-version-for-alignment-checks
9713 The maximum number of run-time checks that can be performed when
9714 doing loop versioning for alignment in the vectorizer.
9716 @item vect-max-version-for-alias-checks
9717 The maximum number of run-time checks that can be performed when
9718 doing loop versioning for alias in the vectorizer.
9720 @item vect-max-peeling-for-alignment
9721 The maximum number of loop peels to enhance access alignment
9722 for vectorizer. Value -1 means no limit.
9724 @item max-iterations-to-track
9725 The maximum number of iterations of a loop the brute-force algorithm
9726 for analysis of the number of iterations of the loop tries to evaluate.
9728 @item hot-bb-count-ws-permille
9729 A basic block profile count is considered hot if it contributes to
9730 the given permillage (i.e. 0...1000) of the entire profiled execution.
9732 @item hot-bb-frequency-fraction
9733 Select fraction of the entry block frequency of executions of basic block in
9734 function given basic block needs to have to be considered hot.
9736 @item max-predicted-iterations
9737 The maximum number of loop iterations we predict statically. This is useful
9738 in cases where a function contains a single loop with known bound and
9739 another loop with unknown bound.
9740 The known number of iterations is predicted correctly, while
9741 the unknown number of iterations average to roughly 10. This means that the
9742 loop without bounds appears artificially cold relative to the other one.
9744 @item builtin-expect-probability
9745 Control the probability of the expression having the specified value. This
9746 parameter takes a percentage (i.e. 0 ... 100) as input.
9747 The default probability of 90 is obtained empirically.
9749 @item align-threshold
9751 Select fraction of the maximal frequency of executions of a basic block in
9752 a function to align the basic block.
9754 @item align-loop-iterations
9756 A loop expected to iterate at least the selected number of iterations is
9759 @item tracer-dynamic-coverage
9760 @itemx tracer-dynamic-coverage-feedback
9762 This value is used to limit superblock formation once the given percentage of
9763 executed instructions is covered. This limits unnecessary code size
9766 The @option{tracer-dynamic-coverage-feedback} parameter
9767 is used only when profile
9768 feedback is available. The real profiles (as opposed to statically estimated
9769 ones) are much less balanced allowing the threshold to be larger value.
9771 @item tracer-max-code-growth
9772 Stop tail duplication once code growth has reached given percentage. This is
9773 a rather artificial limit, as most of the duplicates are eliminated later in
9774 cross jumping, so it may be set to much higher values than is the desired code
9777 @item tracer-min-branch-ratio
9779 Stop reverse growth when the reverse probability of best edge is less than this
9780 threshold (in percent).
9782 @item tracer-min-branch-probability
9783 @itemx tracer-min-branch-probability-feedback
9785 Stop forward growth if the best edge has probability lower than this
9788 Similarly to @option{tracer-dynamic-coverage} two parameters are
9789 provided. @option{tracer-min-branch-probability-feedback} is used for
9790 compilation with profile feedback and @option{tracer-min-branch-probability}
9791 compilation without. The value for compilation with profile feedback
9792 needs to be more conservative (higher) in order to make tracer
9795 @item max-cse-path-length
9797 The maximum number of basic blocks on path that CSE considers.
9801 The maximum number of instructions CSE processes before flushing.
9802 The default is 1000.
9804 @item ggc-min-expand
9806 GCC uses a garbage collector to manage its own memory allocation. This
9807 parameter specifies the minimum percentage by which the garbage
9808 collector's heap should be allowed to expand between collections.
9809 Tuning this may improve compilation speed; it has no effect on code
9812 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9813 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9814 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9815 GCC is not able to calculate RAM on a particular platform, the lower
9816 bound of 30% is used. Setting this parameter and
9817 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9818 every opportunity. This is extremely slow, but can be useful for
9821 @item ggc-min-heapsize
9823 Minimum size of the garbage collector's heap before it begins bothering
9824 to collect garbage. The first collection occurs after the heap expands
9825 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9826 tuning this may improve compilation speed, and has no effect on code
9829 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9830 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9831 with a lower bound of 4096 (four megabytes) and an upper bound of
9832 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9833 particular platform, the lower bound is used. Setting this parameter
9834 very large effectively disables garbage collection. Setting this
9835 parameter and @option{ggc-min-expand} to zero causes a full collection
9836 to occur at every opportunity.
9838 @item max-reload-search-insns
9839 The maximum number of instruction reload should look backward for equivalent
9840 register. Increasing values mean more aggressive optimization, making the
9841 compilation time increase with probably slightly better performance.
9842 The default value is 100.
9844 @item max-cselib-memory-locations
9845 The maximum number of memory locations cselib should take into account.
9846 Increasing values mean more aggressive optimization, making the compilation time
9847 increase with probably slightly better performance. The default value is 500.
9849 @item max-sched-ready-insns
9850 The maximum number of instructions ready to be issued the scheduler should
9851 consider at any given time during the first scheduling pass. Increasing
9852 values mean more thorough searches, making the compilation time increase
9853 with probably little benefit. The default value is 100.
9855 @item max-sched-region-blocks
9856 The maximum number of blocks in a region to be considered for
9857 interblock scheduling. The default value is 10.
9859 @item max-pipeline-region-blocks
9860 The maximum number of blocks in a region to be considered for
9861 pipelining in the selective scheduler. The default value is 15.
9863 @item max-sched-region-insns
9864 The maximum number of insns in a region to be considered for
9865 interblock scheduling. The default value is 100.
9867 @item max-pipeline-region-insns
9868 The maximum number of insns in a region to be considered for
9869 pipelining in the selective scheduler. The default value is 200.
9872 The minimum probability (in percents) of reaching a source block
9873 for interblock speculative scheduling. The default value is 40.
9875 @item max-sched-extend-regions-iters
9876 The maximum number of iterations through CFG to extend regions.
9877 A value of 0 (the default) disables region extensions.
9879 @item max-sched-insn-conflict-delay
9880 The maximum conflict delay for an insn to be considered for speculative motion.
9881 The default value is 3.
9883 @item sched-spec-prob-cutoff
9884 The minimal probability of speculation success (in percents), so that
9885 speculative insns are scheduled.
9886 The default value is 40.
9888 @item sched-state-edge-prob-cutoff
9889 The minimum probability an edge must have for the scheduler to save its
9891 The default value is 10.
9893 @item sched-mem-true-dep-cost
9894 Minimal distance (in CPU cycles) between store and load targeting same
9895 memory locations. The default value is 1.
9897 @item selsched-max-lookahead
9898 The maximum size of the lookahead window of selective scheduling. It is a
9899 depth of search for available instructions.
9900 The default value is 50.
9902 @item selsched-max-sched-times
9903 The maximum number of times that an instruction is scheduled during
9904 selective scheduling. This is the limit on the number of iterations
9905 through which the instruction may be pipelined. The default value is 2.
9907 @item selsched-insns-to-rename
9908 The maximum number of best instructions in the ready list that are considered
9909 for renaming in the selective scheduler. The default value is 2.
9912 The minimum value of stage count that swing modulo scheduler
9913 generates. The default value is 2.
9915 @item max-last-value-rtl
9916 The maximum size measured as number of RTLs that can be recorded in an expression
9917 in combiner for a pseudo register as last known value of that register. The default
9920 @item max-combine-insns
9921 The maximum number of instructions the RTL combiner tries to combine.
9922 The default value is 2 at @option{-Og} and 4 otherwise.
9924 @item integer-share-limit
9925 Small integer constants can use a shared data structure, reducing the
9926 compiler's memory usage and increasing its speed. This sets the maximum
9927 value of a shared integer constant. The default value is 256.
9929 @item ssp-buffer-size
9930 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9931 protection when @option{-fstack-protection} is used.
9933 @item min-size-for-stack-sharing
9934 The minimum size of variables taking part in stack slot sharing when not
9935 optimizing. The default value is 32.
9937 @item max-jump-thread-duplication-stmts
9938 Maximum number of statements allowed in a block that needs to be
9939 duplicated when threading jumps.
9941 @item max-fields-for-field-sensitive
9942 Maximum number of fields in a structure treated in
9943 a field sensitive manner during pointer analysis. The default is zero
9944 for @option{-O0} and @option{-O1},
9945 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9947 @item prefetch-latency
9948 Estimate on average number of instructions that are executed before
9949 prefetch finishes. The distance prefetched ahead is proportional
9950 to this constant. Increasing this number may also lead to less
9951 streams being prefetched (see @option{simultaneous-prefetches}).
9953 @item simultaneous-prefetches
9954 Maximum number of prefetches that can run at the same time.
9956 @item l1-cache-line-size
9957 The size of cache line in L1 cache, in bytes.
9960 The size of L1 cache, in kilobytes.
9963 The size of L2 cache, in kilobytes.
9965 @item min-insn-to-prefetch-ratio
9966 The minimum ratio between the number of instructions and the
9967 number of prefetches to enable prefetching in a loop.
9969 @item prefetch-min-insn-to-mem-ratio
9970 The minimum ratio between the number of instructions and the
9971 number of memory references to enable prefetching in a loop.
9973 @item use-canonical-types
9974 Whether the compiler should use the ``canonical'' type system. By
9975 default, this should always be 1, which uses a more efficient internal
9976 mechanism for comparing types in C++ and Objective-C++. However, if
9977 bugs in the canonical type system are causing compilation failures,
9978 set this value to 0 to disable canonical types.
9980 @item switch-conversion-max-branch-ratio
9981 Switch initialization conversion refuses to create arrays that are
9982 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9983 branches in the switch.
9985 @item max-partial-antic-length
9986 Maximum length of the partial antic set computed during the tree
9987 partial redundancy elimination optimization (@option{-ftree-pre}) when
9988 optimizing at @option{-O3} and above. For some sorts of source code
9989 the enhanced partial redundancy elimination optimization can run away,
9990 consuming all of the memory available on the host machine. This
9991 parameter sets a limit on the length of the sets that are computed,
9992 which prevents the runaway behavior. Setting a value of 0 for
9993 this parameter allows an unlimited set length.
9995 @item sccvn-max-scc-size
9996 Maximum size of a strongly connected component (SCC) during SCCVN
9997 processing. If this limit is hit, SCCVN processing for the whole
9998 function is not done and optimizations depending on it are
9999 disabled. The default maximum SCC size is 10000.
10001 @item sccvn-max-alias-queries-per-access
10002 Maximum number of alias-oracle queries we perform when looking for
10003 redundancies for loads and stores. If this limit is hit the search
10004 is aborted and the load or store is not considered redundant. The
10005 number of queries is algorithmically limited to the number of
10006 stores on all paths from the load to the function entry.
10007 The default maximum number of queries is 1000.
10009 @item ira-max-loops-num
10010 IRA uses regional register allocation by default. If a function
10011 contains more loops than the number given by this parameter, only at most
10012 the given number of the most frequently-executed loops form regions
10013 for regional register allocation. The default value of the
10016 @item ira-max-conflict-table-size
10017 Although IRA uses a sophisticated algorithm to compress the conflict
10018 table, the table can still require excessive amounts of memory for
10019 huge functions. If the conflict table for a function could be more
10020 than the size in MB given by this parameter, the register allocator
10021 instead uses a faster, simpler, and lower-quality
10022 algorithm that does not require building a pseudo-register conflict table.
10023 The default value of the parameter is 2000.
10025 @item ira-loop-reserved-regs
10026 IRA can be used to evaluate more accurate register pressure in loops
10027 for decisions to move loop invariants (see @option{-O3}). The number
10028 of available registers reserved for some other purposes is given
10029 by this parameter. The default value of the parameter is 2, which is
10030 the minimal number of registers needed by typical instructions.
10031 This value is the best found from numerous experiments.
10033 @item lra-inheritance-ebb-probability-cutoff
10034 LRA tries to reuse values reloaded in registers in subsequent insns.
10035 This optimization is called inheritance. EBB is used as a region to
10036 do this optimization. The parameter defines a minimal fall-through
10037 edge probability in percentage used to add BB to inheritance EBB in
10038 LRA. The default value of the parameter is 40. The value was chosen
10039 from numerous runs of SPEC2000 on x86-64.
10041 @item loop-invariant-max-bbs-in-loop
10042 Loop invariant motion can be very expensive, both in compilation time and
10043 in amount of needed compile-time memory, with very large loops. Loops
10044 with more basic blocks than this parameter won't have loop invariant
10045 motion optimization performed on them. The default value of the
10046 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10048 @item loop-max-datarefs-for-datadeps
10049 Building data dependencies is expensive for very large loops. This
10050 parameter limits the number of data references in loops that are
10051 considered for data dependence analysis. These large loops are no
10052 handled by the optimizations using loop data dependencies.
10053 The default value is 1000.
10055 @item max-vartrack-size
10056 Sets a maximum number of hash table slots to use during variable
10057 tracking dataflow analysis of any function. If this limit is exceeded
10058 with variable tracking at assignments enabled, analysis for that
10059 function is retried without it, after removing all debug insns from
10060 the function. If the limit is exceeded even without debug insns, var
10061 tracking analysis is completely disabled for the function. Setting
10062 the parameter to zero makes it unlimited.
10064 @item max-vartrack-expr-depth
10065 Sets a maximum number of recursion levels when attempting to map
10066 variable names or debug temporaries to value expressions. This trades
10067 compilation time for more complete debug information. If this is set too
10068 low, value expressions that are available and could be represented in
10069 debug information may end up not being used; setting this higher may
10070 enable the compiler to find more complex debug expressions, but compile
10071 time and memory use may grow. The default is 12.
10073 @item min-nondebug-insn-uid
10074 Use uids starting at this parameter for nondebug insns. The range below
10075 the parameter is reserved exclusively for debug insns created by
10076 @option{-fvar-tracking-assignments}, but debug insns may get
10077 (non-overlapping) uids above it if the reserved range is exhausted.
10079 @item ipa-sra-ptr-growth-factor
10080 IPA-SRA replaces a pointer to an aggregate with one or more new
10081 parameters only when their cumulative size is less or equal to
10082 @option{ipa-sra-ptr-growth-factor} times the size of the original
10085 @item sra-max-scalarization-size-Ospeed
10086 @item sra-max-scalarization-size-Osize
10087 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10088 replace scalar parts of aggregates with uses of independent scalar
10089 variables. These parameters control the maximum size, in storage units,
10090 of aggregate which is considered for replacement when compiling for
10092 (@option{sra-max-scalarization-size-Ospeed}) or size
10093 (@option{sra-max-scalarization-size-Osize}) respectively.
10095 @item tm-max-aggregate-size
10096 When making copies of thread-local variables in a transaction, this
10097 parameter specifies the size in bytes after which variables are
10098 saved with the logging functions as opposed to save/restore code
10099 sequence pairs. This option only applies when using
10102 @item graphite-max-nb-scop-params
10103 To avoid exponential effects in the Graphite loop transforms, the
10104 number of parameters in a Static Control Part (SCoP) is bounded. The
10105 default value is 10 parameters. A variable whose value is unknown at
10106 compilation time and defined outside a SCoP is a parameter of the SCoP.
10108 @item graphite-max-bbs-per-function
10109 To avoid exponential effects in the detection of SCoPs, the size of
10110 the functions analyzed by Graphite is bounded. The default value is
10113 @item loop-block-tile-size
10114 Loop blocking or strip mining transforms, enabled with
10115 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10116 loop in the loop nest by a given number of iterations. The strip
10117 length can be changed using the @option{loop-block-tile-size}
10118 parameter. The default value is 51 iterations.
10120 @item loop-unroll-jam-size
10121 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10122 default value is 4.
10124 @item loop-unroll-jam-depth
10125 Specify the dimension to be unrolled (counting from the most inner loop)
10126 for the @option{-floop-unroll-and-jam}. The default value is 2.
10128 @item ipa-cp-value-list-size
10129 IPA-CP attempts to track all possible values and types passed to a function's
10130 parameter in order to propagate them and perform devirtualization.
10131 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10132 stores per one formal parameter of a function.
10134 @item ipa-cp-eval-threshold
10135 IPA-CP calculates its own score of cloning profitability heuristics
10136 and performs those cloning opportunities with scores that exceed
10137 @option{ipa-cp-eval-threshold}.
10139 @item ipa-cp-recursion-penalty
10140 Percentage penalty the recursive functions will receive when they
10141 are evaluated for cloning.
10143 @item ipa-cp-single-call-penalty
10144 Percentage penalty functions containg a single call to another
10145 function will receive when they are evaluated for cloning.
10148 @item ipa-max-agg-items
10149 IPA-CP is also capable to propagate a number of scalar values passed
10150 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10151 number of such values per one parameter.
10153 @item ipa-cp-loop-hint-bonus
10154 When IPA-CP determines that a cloning candidate would make the number
10155 of iterations of a loop known, it adds a bonus of
10156 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10159 @item ipa-cp-array-index-hint-bonus
10160 When IPA-CP determines that a cloning candidate would make the index of
10161 an array access known, it adds a bonus of
10162 @option{ipa-cp-array-index-hint-bonus} to the profitability
10163 score of the candidate.
10165 @item ipa-max-aa-steps
10166 During its analysis of function bodies, IPA-CP employs alias analysis
10167 in order to track values pointed to by function parameters. In order
10168 not spend too much time analyzing huge functions, it gives up and
10169 consider all memory clobbered after examining
10170 @option{ipa-max-aa-steps} statements modifying memory.
10172 @item lto-partitions
10173 Specify desired number of partitions produced during WHOPR compilation.
10174 The number of partitions should exceed the number of CPUs used for compilation.
10175 The default value is 32.
10177 @item lto-min-partition
10178 Size of minimal partition for WHOPR (in estimated instructions).
10179 This prevents expenses of splitting very small programs into too many
10182 @item lto-max-partition
10183 Size of max partition for WHOPR (in estimated instructions).
10184 to provide an upper bound for individual size of partition.
10185 Meant to be used only with balanced partitioning.
10187 @item cxx-max-namespaces-for-diagnostic-help
10188 The maximum number of namespaces to consult for suggestions when C++
10189 name lookup fails for an identifier. The default is 1000.
10191 @item sink-frequency-threshold
10192 The maximum relative execution frequency (in percents) of the target block
10193 relative to a statement's original block to allow statement sinking of a
10194 statement. Larger numbers result in more aggressive statement sinking.
10195 The default value is 75. A small positive adjustment is applied for
10196 statements with memory operands as those are even more profitable so sink.
10198 @item max-stores-to-sink
10199 The maximum number of conditional store pairs that can be sunk. Set to 0
10200 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10201 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10203 @item allow-store-data-races
10204 Allow optimizers to introduce new data races on stores.
10205 Set to 1 to allow, otherwise to 0. This option is enabled by default
10206 at optimization level @option{-Ofast}.
10208 @item case-values-threshold
10209 The smallest number of different values for which it is best to use a
10210 jump-table instead of a tree of conditional branches. If the value is
10211 0, use the default for the machine. The default is 0.
10213 @item tree-reassoc-width
10214 Set the maximum number of instructions executed in parallel in
10215 reassociated tree. This parameter overrides target dependent
10216 heuristics used by default if has non zero value.
10218 @item sched-pressure-algorithm
10219 Choose between the two available implementations of
10220 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10221 and is the more likely to prevent instructions from being reordered.
10222 Algorithm 2 was designed to be a compromise between the relatively
10223 conservative approach taken by algorithm 1 and the rather aggressive
10224 approach taken by the default scheduler. It relies more heavily on
10225 having a regular register file and accurate register pressure classes.
10226 See @file{haifa-sched.c} in the GCC sources for more details.
10228 The default choice depends on the target.
10230 @item max-slsr-cand-scan
10231 Set the maximum number of existing candidates that are considered when
10232 seeking a basis for a new straight-line strength reduction candidate.
10235 Enable buffer overflow detection for global objects. This kind
10236 of protection is enabled by default if you are using
10237 @option{-fsanitize=address} option.
10238 To disable global objects protection use @option{--param asan-globals=0}.
10241 Enable buffer overflow detection for stack objects. This kind of
10242 protection is enabled by default when using @option{-fsanitize=address}.
10243 To disable stack protection use @option{--param asan-stack=0} option.
10245 @item asan-instrument-reads
10246 Enable buffer overflow detection for memory reads. This kind of
10247 protection is enabled by default when using @option{-fsanitize=address}.
10248 To disable memory reads protection use
10249 @option{--param asan-instrument-reads=0}.
10251 @item asan-instrument-writes
10252 Enable buffer overflow detection for memory writes. This kind of
10253 protection is enabled by default when using @option{-fsanitize=address}.
10254 To disable memory writes protection use
10255 @option{--param asan-instrument-writes=0} option.
10257 @item asan-memintrin
10258 Enable detection for built-in functions. This kind of protection
10259 is enabled by default when using @option{-fsanitize=address}.
10260 To disable built-in functions protection use
10261 @option{--param asan-memintrin=0}.
10263 @item asan-use-after-return
10264 Enable detection of use-after-return. This kind of protection
10265 is enabled by default when using @option{-fsanitize=address} option.
10266 To disable use-after-return detection use
10267 @option{--param asan-use-after-return=0}.
10269 Note: The check is disabled by default at runtime. To enable the check,
10270 you should set environment variable @env{ASAN_OPTIONS} to
10271 @code{detect_stack_use_after_return=1}.
10273 @item asan-instrumentation-with-call-threshold
10274 If number of memory accesses in function being instrumented
10275 is greater or equal to this number, use callbacks instead of inline checks.
10276 E.g. to disable inline code use
10277 @option{--param asan-instrumentation-with-call-threshold=0}.
10279 @item chkp-max-ctor-size
10280 Static constructors generated by Pointer Bounds Checker may become very
10281 large and significantly increase compile time at optimization level
10282 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10283 in a single generated constructor. Default value is 5000.
10285 @item max-fsm-thread-path-insns
10286 Maximum number of instructions to copy when duplicating blocks on a
10287 finite state automaton jump thread path. The default is 100.
10289 @item max-fsm-thread-length
10290 Maximum number of basic blocks on a finite state automaton jump thread
10291 path. The default is 10.
10293 @item max-fsm-thread-paths
10294 Maximum number of new jump thread paths to create for a finite state
10295 automaton. The default is 50.
10297 @item parloops-chunk-size
10298 Chunk size of omp schedule for loops parallelized by parloops. The default
10301 @item parloops-schedule
10302 Schedule type of omp schedule for loops parallelized by parloops (static,
10303 dynamic, guided, auto, runtime). The default is static.
10305 @item max-ssa-name-query-depth
10306 Maximum depth of recursion when querying properties of SSA names in things
10307 like fold routines. One level of recursion corresponds to following a
10310 @item hsa-gen-debug-stores
10311 Enable emission of special debug stores within HSA kernels which are
10312 then read and reported by libgomp plugin. Generation of these stores
10313 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10316 @item max-speculative-devirt-maydefs
10317 The maximum number of may-defs we analyze when looking for a must-def
10318 specifying the dynamic type of an object that invokes a virtual call
10319 we may be able to devirtualize speculatively.
10321 @item max-vrp-switch-assertions
10322 The maximum number of assertions to add along the default edge of a switch
10323 statement during VRP. The default is 10.
10327 @node Instrumentation Options
10328 @section Program Instrumentation Options
10329 @cindex instrumentation options
10330 @cindex program instrumentation options
10331 @cindex run-time error checking options
10332 @cindex profiling options
10333 @cindex options, program instrumentation
10334 @cindex options, run-time error checking
10335 @cindex options, profiling
10337 GCC supports a number of command-line options that control adding
10338 run-time instrumentation to the code it normally generates.
10339 For example, one purpose of instrumentation is collect profiling
10340 statistics for use in finding program hot spots, code coverage
10341 analysis, or profile-guided optimizations.
10342 Another class of program instrumentation is adding run-time checking
10343 to detect programming errors like invalid pointer
10344 dereferences or out-of-bounds array accesses, as well as deliberately
10345 hostile attacks such as stack smashing or C++ vtable hijacking.
10346 There is also a general hook which can be used to implement other
10347 forms of tracing or function-level instrumentation for debug or
10348 program analysis purposes.
10351 @cindex @command{prof}
10354 Generate extra code to write profile information suitable for the
10355 analysis program @command{prof}. You must use this option when compiling
10356 the source files you want data about, and you must also use it when
10359 @cindex @command{gprof}
10362 Generate extra code to write profile information suitable for the
10363 analysis program @command{gprof}. You must use this option when compiling
10364 the source files you want data about, and you must also use it when
10367 @item -fprofile-arcs
10368 @opindex fprofile-arcs
10369 Add code so that program flow @dfn{arcs} are instrumented. During
10370 execution the program records how many times each branch and call is
10371 executed and how many times it is taken or returns. When the compiled
10372 program exits it saves this data to a file called
10373 @file{@var{auxname}.gcda} for each source file. The data may be used for
10374 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10375 test coverage analysis (@option{-ftest-coverage}). Each object file's
10376 @var{auxname} is generated from the name of the output file, if
10377 explicitly specified and it is not the final executable, otherwise it is
10378 the basename of the source file. In both cases any suffix is removed
10379 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10380 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10381 @xref{Cross-profiling}.
10383 @cindex @command{gcov}
10387 This option is used to compile and link code instrumented for coverage
10388 analysis. The option is a synonym for @option{-fprofile-arcs}
10389 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10390 linking). See the documentation for those options for more details.
10395 Compile the source files with @option{-fprofile-arcs} plus optimization
10396 and code generation options. For test coverage analysis, use the
10397 additional @option{-ftest-coverage} option. You do not need to profile
10398 every source file in a program.
10401 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10402 (the latter implies the former).
10405 Run the program on a representative workload to generate the arc profile
10406 information. This may be repeated any number of times. You can run
10407 concurrent instances of your program, and provided that the file system
10408 supports locking, the data files will be correctly updated. Also
10409 @code{fork} calls are detected and correctly handled (double counting
10413 For profile-directed optimizations, compile the source files again with
10414 the same optimization and code generation options plus
10415 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10416 Control Optimization}).
10419 For test coverage analysis, use @command{gcov} to produce human readable
10420 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10421 @command{gcov} documentation for further information.
10425 With @option{-fprofile-arcs}, for each function of your program GCC
10426 creates a program flow graph, then finds a spanning tree for the graph.
10427 Only arcs that are not on the spanning tree have to be instrumented: the
10428 compiler adds code to count the number of times that these arcs are
10429 executed. When an arc is the only exit or only entrance to a block, the
10430 instrumentation code can be added to the block; otherwise, a new basic
10431 block must be created to hold the instrumentation code.
10434 @item -ftest-coverage
10435 @opindex ftest-coverage
10436 Produce a notes file that the @command{gcov} code-coverage utility
10437 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10438 show program coverage. Each source file's note file is called
10439 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10440 above for a description of @var{auxname} and instructions on how to
10441 generate test coverage data. Coverage data matches the source files
10442 more closely if you do not optimize.
10444 @item -fprofile-dir=@var{path}
10445 @opindex fprofile-dir
10447 Set the directory to search for the profile data files in to @var{path}.
10448 This option affects only the profile data generated by
10449 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10450 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10451 and its related options. Both absolute and relative paths can be used.
10452 By default, GCC uses the current directory as @var{path}, thus the
10453 profile data file appears in the same directory as the object file.
10455 @item -fprofile-generate
10456 @itemx -fprofile-generate=@var{path}
10457 @opindex fprofile-generate
10459 Enable options usually used for instrumenting application to produce
10460 profile useful for later recompilation with profile feedback based
10461 optimization. You must use @option{-fprofile-generate} both when
10462 compiling and when linking your program.
10464 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10466 If @var{path} is specified, GCC looks at the @var{path} to find
10467 the profile feedback data files. See @option{-fprofile-dir}.
10469 To optimize the program based on the collected profile information, use
10470 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10472 @item -fprofile-update=@var{method}
10473 @opindex fprofile-update
10475 Alter the update method for an application instrumented for profile
10476 feedback based optimization. The @var{method} argument should be one of
10477 @samp{single} or @samp{atomic}. The first one is useful for single-threaded
10478 applications, while the second one prevents profile corruption by emitting
10481 @strong{Warning:} When an application does not properly join all threads
10482 (or creates an detached thread), a profile file can be still corrupted.
10484 @item -fsanitize=address
10485 @opindex fsanitize=address
10486 Enable AddressSanitizer, a fast memory error detector.
10487 Memory access instructions are instrumented to detect
10488 out-of-bounds and use-after-free bugs.
10489 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10490 more details. The run-time behavior can be influenced using the
10491 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10492 the available options are shown at startup of the instrumented program. See
10493 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10494 for a list of supported options.
10495 The option can't be combined with @option{-fsanitize=thread}.
10497 @item -fsanitize=kernel-address
10498 @opindex fsanitize=kernel-address
10499 Enable AddressSanitizer for Linux kernel.
10500 See @uref{https://github.com/google/kasan/wiki} for more details.
10502 @item -fsanitize=thread
10503 @opindex fsanitize=thread
10504 Enable ThreadSanitizer, a fast data race detector.
10505 Memory access instructions are instrumented to detect
10506 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10507 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10508 environment variable; see
10509 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10511 The option can't be combined with @option{-fsanitize=address}
10512 and/or @option{-fsanitize=leak}.
10514 @item -fsanitize=leak
10515 @opindex fsanitize=leak
10516 Enable LeakSanitizer, a memory leak detector.
10517 This option only matters for linking of executables and
10518 the executable is linked against a library that overrides @code{malloc}
10519 and other allocator functions. See
10520 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
10521 details. The run-time behavior can be influenced using the
10522 @env{LSAN_OPTIONS} environment variable.
10523 The option can't be combined with @option{-fsanitize=thread}.
10525 @item -fsanitize=undefined
10526 @opindex fsanitize=undefined
10527 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
10528 Various computations are instrumented to detect undefined behavior
10529 at runtime. Current suboptions are:
10533 @item -fsanitize=shift
10534 @opindex fsanitize=shift
10535 This option enables checking that the result of a shift operation is
10536 not undefined. Note that what exactly is considered undefined differs
10537 slightly between C and C++, as well as between ISO C90 and C99, etc.
10539 @item -fsanitize=integer-divide-by-zero
10540 @opindex fsanitize=integer-divide-by-zero
10541 Detect integer division by zero as well as @code{INT_MIN / -1} division.
10543 @item -fsanitize=unreachable
10544 @opindex fsanitize=unreachable
10545 With this option, the compiler turns the @code{__builtin_unreachable}
10546 call into a diagnostics message call instead. When reaching the
10547 @code{__builtin_unreachable} call, the behavior is undefined.
10549 @item -fsanitize=vla-bound
10550 @opindex fsanitize=vla-bound
10551 This option instructs the compiler to check that the size of a variable
10552 length array is positive.
10554 @item -fsanitize=null
10555 @opindex fsanitize=null
10556 This option enables pointer checking. Particularly, the application
10557 built with this option turned on will issue an error message when it
10558 tries to dereference a NULL pointer, or if a reference (possibly an
10559 rvalue reference) is bound to a NULL pointer, or if a method is invoked
10560 on an object pointed by a NULL pointer.
10562 @item -fsanitize=return
10563 @opindex fsanitize=return
10564 This option enables return statement checking. Programs
10565 built with this option turned on will issue an error message
10566 when the end of a non-void function is reached without actually
10567 returning a value. This option works in C++ only.
10569 @item -fsanitize=signed-integer-overflow
10570 @opindex fsanitize=signed-integer-overflow
10571 This option enables signed integer overflow checking. We check that
10572 the result of @code{+}, @code{*}, and both unary and binary @code{-}
10573 does not overflow in the signed arithmetics. Note, integer promotion
10574 rules must be taken into account. That is, the following is not an
10577 signed char a = SCHAR_MAX;
10581 @item -fsanitize=bounds
10582 @opindex fsanitize=bounds
10583 This option enables instrumentation of array bounds. Various out of bounds
10584 accesses are detected. Flexible array members, flexible array member-like
10585 arrays, and initializers of variables with static storage are not instrumented.
10587 @item -fsanitize=bounds-strict
10588 @opindex fsanitize=bounds-strict
10589 This option enables strict instrumentation of array bounds. Most out of bounds
10590 accesses are detected, including flexible array members and flexible array
10591 member-like arrays. Initializers of variables with static storage are not
10594 @item -fsanitize=alignment
10595 @opindex fsanitize=alignment
10597 This option enables checking of alignment of pointers when they are
10598 dereferenced, or when a reference is bound to insufficiently aligned target,
10599 or when a method or constructor is invoked on insufficiently aligned object.
10601 @item -fsanitize=object-size
10602 @opindex fsanitize=object-size
10603 This option enables instrumentation of memory references using the
10604 @code{__builtin_object_size} function. Various out of bounds pointer
10605 accesses are detected.
10607 @item -fsanitize=float-divide-by-zero
10608 @opindex fsanitize=float-divide-by-zero
10609 Detect floating-point division by zero. Unlike other similar options,
10610 @option{-fsanitize=float-divide-by-zero} is not enabled by
10611 @option{-fsanitize=undefined}, since floating-point division by zero can
10612 be a legitimate way of obtaining infinities and NaNs.
10614 @item -fsanitize=float-cast-overflow
10615 @opindex fsanitize=float-cast-overflow
10616 This option enables floating-point type to integer conversion checking.
10617 We check that the result of the conversion does not overflow.
10618 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10619 not enabled by @option{-fsanitize=undefined}.
10620 This option does not work well with @code{FE_INVALID} exceptions enabled.
10622 @item -fsanitize=nonnull-attribute
10623 @opindex fsanitize=nonnull-attribute
10625 This option enables instrumentation of calls, checking whether null values
10626 are not passed to arguments marked as requiring a non-null value by the
10627 @code{nonnull} function attribute.
10629 @item -fsanitize=returns-nonnull-attribute
10630 @opindex fsanitize=returns-nonnull-attribute
10632 This option enables instrumentation of return statements in functions
10633 marked with @code{returns_nonnull} function attribute, to detect returning
10634 of null values from such functions.
10636 @item -fsanitize=bool
10637 @opindex fsanitize=bool
10639 This option enables instrumentation of loads from bool. If a value other
10640 than 0/1 is loaded, a run-time error is issued.
10642 @item -fsanitize=enum
10643 @opindex fsanitize=enum
10645 This option enables instrumentation of loads from an enum type. If
10646 a value outside the range of values for the enum type is loaded,
10647 a run-time error is issued.
10649 @item -fsanitize=vptr
10650 @opindex fsanitize=vptr
10652 This option enables instrumentation of C++ member function calls, member
10653 accesses and some conversions between pointers to base and derived classes,
10654 to verify the referenced object has the correct dynamic type.
10658 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10659 @option{-fsanitize=undefined} gives a diagnostic message.
10660 This currently works only for the C family of languages.
10662 @item -fno-sanitize=all
10663 @opindex fno-sanitize=all
10665 This option disables all previously enabled sanitizers.
10666 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10669 @item -fasan-shadow-offset=@var{number}
10670 @opindex fasan-shadow-offset
10671 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10672 It is useful for experimenting with different shadow memory layouts in
10673 Kernel AddressSanitizer.
10675 @item -fsanitize-sections=@var{s1},@var{s2},...
10676 @opindex fsanitize-sections
10677 Sanitize global variables in selected user-defined sections. @var{si} may
10680 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10681 @opindex fsanitize-recover
10682 @opindex fno-sanitize-recover
10683 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10684 mentioned in comma-separated list of @var{opts}. Enabling this option
10685 for a sanitizer component causes it to attempt to continue
10686 running the program as if no error happened. This means multiple
10687 runtime errors can be reported in a single program run, and the exit
10688 code of the program may indicate success even when errors
10689 have been reported. The @option{-fno-sanitize-recover=} option
10690 can be used to alter
10691 this behavior: only the first detected error is reported
10692 and program then exits with a non-zero exit code.
10694 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10695 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10696 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10697 @option{-fsanitize=bounds-strict},
10698 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10699 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
10700 for which this feature is experimental.
10701 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10702 accepted, the former enables recovery for all sanitizers that support it,
10703 the latter disables recovery for all sanitizers that support it.
10705 Even if a recovery mode is turned on the compiler side, it needs to be also
10706 enabled on the runtime library side, otherwise the failures are still fatal.
10707 The runtime library defaults to @code{halt_on_error=0} for
10708 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10709 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10710 setting the @code{halt_on_error} flag in the corresponding environment variable.
10712 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
10714 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10717 Similarly @option{-fno-sanitize-recover} is equivalent to
10719 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10722 @item -fsanitize-undefined-trap-on-error
10723 @opindex fsanitize-undefined-trap-on-error
10724 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10725 report undefined behavior using @code{__builtin_trap} rather than
10726 a @code{libubsan} library routine. The advantage of this is that the
10727 @code{libubsan} library is not needed and is not linked in, so this
10728 is usable even in freestanding environments.
10730 @item -fsanitize-coverage=trace-pc
10731 @opindex fsanitize-coverage=trace-pc
10732 Enable coverage-guided fuzzing code instrumentation.
10733 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10735 @item -fbounds-check
10736 @opindex fbounds-check
10737 For front ends that support it, generate additional code to check that
10738 indices used to access arrays are within the declared range. This is
10739 currently only supported by the Java and Fortran front ends, where
10740 this option defaults to true and false respectively.
10742 @item -fcheck-pointer-bounds
10743 @opindex fcheck-pointer-bounds
10744 @opindex fno-check-pointer-bounds
10745 @cindex Pointer Bounds Checker options
10746 Enable Pointer Bounds Checker instrumentation. Each memory reference
10747 is instrumented with checks of the pointer used for memory access against
10748 bounds associated with that pointer.
10751 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10752 and @option{-mmpx} are required to enable this feature.
10753 MPX-based instrumentation requires
10754 a runtime library to enable MPX in hardware and handle bounds
10755 violation signals. By default when @option{-fcheck-pointer-bounds}
10756 and @option{-mmpx} options are used to link a program, the GCC driver
10757 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10758 Bounds checking on calls to dynamic libraries requires a linker
10759 with @option{-z bndplt} support; if GCC was configured with a linker
10760 without support for this option (including the Gold linker and older
10761 versions of ld), a warning is given if you link with @option{-mmpx}
10762 without also specifying @option{-static}, since the overall effectiveness
10763 of the bounds checking protection is reduced.
10764 See also @option{-static-libmpxwrappers}.
10766 MPX-based instrumentation
10767 may be used for debugging and also may be included in production code
10768 to increase program security. Depending on usage, you may
10769 have different requirements for the runtime library. The current version
10770 of the MPX runtime library is more oriented for use as a debugging
10771 tool. MPX runtime library usage implies @option{-lpthread}. See
10772 also @option{-static-libmpx}. The runtime library behavior can be
10773 influenced using various @env{CHKP_RT_*} environment variables. See
10774 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10777 Generated instrumentation may be controlled by various
10778 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10779 structure field attribute (@pxref{Type Attributes}) and
10780 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10781 (@pxref{Function Attributes}). GCC also provides a number of built-in
10782 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10783 Bounds Checker builtins}, for more information.
10785 @item -fchkp-check-incomplete-type
10786 @opindex fchkp-check-incomplete-type
10787 @opindex fno-chkp-check-incomplete-type
10788 Generate pointer bounds checks for variables with incomplete type.
10789 Enabled by default.
10791 @item -fchkp-narrow-bounds
10792 @opindex fchkp-narrow-bounds
10793 @opindex fno-chkp-narrow-bounds
10794 Controls bounds used by Pointer Bounds Checker for pointers to object
10795 fields. If narrowing is enabled then field bounds are used. Otherwise
10796 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10797 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10799 @item -fchkp-first-field-has-own-bounds
10800 @opindex fchkp-first-field-has-own-bounds
10801 @opindex fno-chkp-first-field-has-own-bounds
10802 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10803 first field in the structure. By default a pointer to the first field has
10804 the same bounds as a pointer to the whole structure.
10806 @item -fchkp-narrow-to-innermost-array
10807 @opindex fchkp-narrow-to-innermost-array
10808 @opindex fno-chkp-narrow-to-innermost-array
10809 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10810 case of nested static array access. By default this option is disabled and
10811 bounds of the outermost array are used.
10813 @item -fchkp-optimize
10814 @opindex fchkp-optimize
10815 @opindex fno-chkp-optimize
10816 Enables Pointer Bounds Checker optimizations. Enabled by default at
10817 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
10819 @item -fchkp-use-fast-string-functions
10820 @opindex fchkp-use-fast-string-functions
10821 @opindex fno-chkp-use-fast-string-functions
10822 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
10823 by Pointer Bounds Checker. Disabled by default.
10825 @item -fchkp-use-nochk-string-functions
10826 @opindex fchkp-use-nochk-string-functions
10827 @opindex fno-chkp-use-nochk-string-functions
10828 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
10829 by Pointer Bounds Checker. Disabled by default.
10831 @item -fchkp-use-static-bounds
10832 @opindex fchkp-use-static-bounds
10833 @opindex fno-chkp-use-static-bounds
10834 Allow Pointer Bounds Checker to generate static bounds holding
10835 bounds of static variables. Enabled by default.
10837 @item -fchkp-use-static-const-bounds
10838 @opindex fchkp-use-static-const-bounds
10839 @opindex fno-chkp-use-static-const-bounds
10840 Use statically-initialized bounds for constant bounds instead of
10841 generating them each time they are required. By default enabled when
10842 @option{-fchkp-use-static-bounds} is enabled.
10844 @item -fchkp-treat-zero-dynamic-size-as-infinite
10845 @opindex fchkp-treat-zero-dynamic-size-as-infinite
10846 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
10847 With this option, objects with incomplete type whose
10848 dynamically-obtained size is zero are treated as having infinite size
10849 instead by Pointer Bounds
10850 Checker. This option may be helpful if a program is linked with a library
10851 missing size information for some symbols. Disabled by default.
10853 @item -fchkp-check-read
10854 @opindex fchkp-check-read
10855 @opindex fno-chkp-check-read
10856 Instructs Pointer Bounds Checker to generate checks for all read
10857 accesses to memory. Enabled by default.
10859 @item -fchkp-check-write
10860 @opindex fchkp-check-write
10861 @opindex fno-chkp-check-write
10862 Instructs Pointer Bounds Checker to generate checks for all write
10863 accesses to memory. Enabled by default.
10865 @item -fchkp-store-bounds
10866 @opindex fchkp-store-bounds
10867 @opindex fno-chkp-store-bounds
10868 Instructs Pointer Bounds Checker to generate bounds stores for
10869 pointer writes. Enabled by default.
10871 @item -fchkp-instrument-calls
10872 @opindex fchkp-instrument-calls
10873 @opindex fno-chkp-instrument-calls
10874 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
10875 Enabled by default.
10877 @item -fchkp-instrument-marked-only
10878 @opindex fchkp-instrument-marked-only
10879 @opindex fno-chkp-instrument-marked-only
10880 Instructs Pointer Bounds Checker to instrument only functions
10881 marked with the @code{bnd_instrument} attribute
10882 (@pxref{Function Attributes}). Disabled by default.
10884 @item -fchkp-use-wrappers
10885 @opindex fchkp-use-wrappers
10886 @opindex fno-chkp-use-wrappers
10887 Allows Pointer Bounds Checker to replace calls to built-in functions
10888 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
10889 is used to link a program, the GCC driver automatically links
10890 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
10891 Enabled by default.
10893 @item -fstack-protector
10894 @opindex fstack-protector
10895 Emit extra code to check for buffer overflows, such as stack smashing
10896 attacks. This is done by adding a guard variable to functions with
10897 vulnerable objects. This includes functions that call @code{alloca}, and
10898 functions with buffers larger than 8 bytes. The guards are initialized
10899 when a function is entered and then checked when the function exits.
10900 If a guard check fails, an error message is printed and the program exits.
10902 @item -fstack-protector-all
10903 @opindex fstack-protector-all
10904 Like @option{-fstack-protector} except that all functions are protected.
10906 @item -fstack-protector-strong
10907 @opindex fstack-protector-strong
10908 Like @option{-fstack-protector} but includes additional functions to
10909 be protected --- those that have local array definitions, or have
10910 references to local frame addresses.
10912 @item -fstack-protector-explicit
10913 @opindex fstack-protector-explicit
10914 Like @option{-fstack-protector} but only protects those functions which
10915 have the @code{stack_protect} attribute.
10917 @item -fstack-check
10918 @opindex fstack-check
10919 Generate code to verify that you do not go beyond the boundary of the
10920 stack. You should specify this flag if you are running in an
10921 environment with multiple threads, but you only rarely need to specify it in
10922 a single-threaded environment since stack overflow is automatically
10923 detected on nearly all systems if there is only one stack.
10925 Note that this switch does not actually cause checking to be done; the
10926 operating system or the language runtime must do that. The switch causes
10927 generation of code to ensure that they see the stack being extended.
10929 You can additionally specify a string parameter: @samp{no} means no
10930 checking, @samp{generic} means force the use of old-style checking,
10931 @samp{specific} means use the best checking method and is equivalent
10932 to bare @option{-fstack-check}.
10934 Old-style checking is a generic mechanism that requires no specific
10935 target support in the compiler but comes with the following drawbacks:
10939 Modified allocation strategy for large objects: they are always
10940 allocated dynamically if their size exceeds a fixed threshold.
10943 Fixed limit on the size of the static frame of functions: when it is
10944 topped by a particular function, stack checking is not reliable and
10945 a warning is issued by the compiler.
10948 Inefficiency: because of both the modified allocation strategy and the
10949 generic implementation, code performance is hampered.
10952 Note that old-style stack checking is also the fallback method for
10953 @samp{specific} if no target support has been added in the compiler.
10955 @item -fstack-limit-register=@var{reg}
10956 @itemx -fstack-limit-symbol=@var{sym}
10957 @itemx -fno-stack-limit
10958 @opindex fstack-limit-register
10959 @opindex fstack-limit-symbol
10960 @opindex fno-stack-limit
10961 Generate code to ensure that the stack does not grow beyond a certain value,
10962 either the value of a register or the address of a symbol. If a larger
10963 stack is required, a signal is raised at run time. For most targets,
10964 the signal is raised before the stack overruns the boundary, so
10965 it is possible to catch the signal without taking special precautions.
10967 For instance, if the stack starts at absolute address @samp{0x80000000}
10968 and grows downwards, you can use the flags
10969 @option{-fstack-limit-symbol=__stack_limit} and
10970 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
10971 of 128KB@. Note that this may only work with the GNU linker.
10973 You can locally override stack limit checking by using the
10974 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
10976 @item -fsplit-stack
10977 @opindex fsplit-stack
10978 Generate code to automatically split the stack before it overflows.
10979 The resulting program has a discontiguous stack which can only
10980 overflow if the program is unable to allocate any more memory. This
10981 is most useful when running threaded programs, as it is no longer
10982 necessary to calculate a good stack size to use for each thread. This
10983 is currently only implemented for the x86 targets running
10986 When code compiled with @option{-fsplit-stack} calls code compiled
10987 without @option{-fsplit-stack}, there may not be much stack space
10988 available for the latter code to run. If compiling all code,
10989 including library code, with @option{-fsplit-stack} is not an option,
10990 then the linker can fix up these calls so that the code compiled
10991 without @option{-fsplit-stack} always has a large stack. Support for
10992 this is implemented in the gold linker in GNU binutils release 2.21
10995 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
10996 @opindex fvtable-verify
10997 This option is only available when compiling C++ code.
10998 It turns on (or off, if using @option{-fvtable-verify=none}) the security
10999 feature that verifies at run time, for every virtual call, that
11000 the vtable pointer through which the call is made is valid for the type of
11001 the object, and has not been corrupted or overwritten. If an invalid vtable
11002 pointer is detected at run time, an error is reported and execution of the
11003 program is immediately halted.
11005 This option causes run-time data structures to be built at program startup,
11006 which are used for verifying the vtable pointers.
11007 The options @samp{std} and @samp{preinit}
11008 control the timing of when these data structures are built. In both cases the
11009 data structures are built before execution reaches @code{main}. Using
11010 @option{-fvtable-verify=std} causes the data structures to be built after
11011 shared libraries have been loaded and initialized.
11012 @option{-fvtable-verify=preinit} causes them to be built before shared
11013 libraries have been loaded and initialized.
11015 If this option appears multiple times in the command line with different
11016 values specified, @samp{none} takes highest priority over both @samp{std} and
11017 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11020 @opindex fvtv-debug
11021 When used in conjunction with @option{-fvtable-verify=std} or
11022 @option{-fvtable-verify=preinit}, causes debug versions of the
11023 runtime functions for the vtable verification feature to be called.
11024 This flag also causes the compiler to log information about which
11025 vtable pointers it finds for each class.
11026 This information is written to a file named @file{vtv_set_ptr_data.log}
11027 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11028 if that is defined or the current working directory otherwise.
11030 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11031 file, be sure to delete any existing one.
11034 @opindex fvtv-counts
11035 This is a debugging flag. When used in conjunction with
11036 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11037 causes the compiler to keep track of the total number of virtual calls
11038 it encounters and the number of verifications it inserts. It also
11039 counts the number of calls to certain run-time library functions
11040 that it inserts and logs this information for each compilation unit.
11041 The compiler writes this information to a file named
11042 @file{vtv_count_data.log} in the directory named by the environment
11043 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11044 directory otherwise. It also counts the size of the vtable pointer sets
11045 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11046 in the same directory.
11048 Note: This feature @emph{appends} data to the log files. To get fresh log
11049 files, be sure to delete any existing ones.
11051 @item -finstrument-functions
11052 @opindex finstrument-functions
11053 Generate instrumentation calls for entry and exit to functions. Just
11054 after function entry and just before function exit, the following
11055 profiling functions are called with the address of the current
11056 function and its call site. (On some platforms,
11057 @code{__builtin_return_address} does not work beyond the current
11058 function, so the call site information may not be available to the
11059 profiling functions otherwise.)
11062 void __cyg_profile_func_enter (void *this_fn,
11064 void __cyg_profile_func_exit (void *this_fn,
11068 The first argument is the address of the start of the current function,
11069 which may be looked up exactly in the symbol table.
11071 This instrumentation is also done for functions expanded inline in other
11072 functions. The profiling calls indicate where, conceptually, the
11073 inline function is entered and exited. This means that addressable
11074 versions of such functions must be available. If all your uses of a
11075 function are expanded inline, this may mean an additional expansion of
11076 code size. If you use @code{extern inline} in your C code, an
11077 addressable version of such functions must be provided. (This is
11078 normally the case anyway, but if you get lucky and the optimizer always
11079 expands the functions inline, you might have gotten away without
11080 providing static copies.)
11082 A function may be given the attribute @code{no_instrument_function}, in
11083 which case this instrumentation is not done. This can be used, for
11084 example, for the profiling functions listed above, high-priority
11085 interrupt routines, and any functions from which the profiling functions
11086 cannot safely be called (perhaps signal handlers, if the profiling
11087 routines generate output or allocate memory).
11089 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11090 @opindex finstrument-functions-exclude-file-list
11092 Set the list of functions that are excluded from instrumentation (see
11093 the description of @option{-finstrument-functions}). If the file that
11094 contains a function definition matches with one of @var{file}, then
11095 that function is not instrumented. The match is done on substrings:
11096 if the @var{file} parameter is a substring of the file name, it is
11097 considered to be a match.
11102 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11106 excludes any inline function defined in files whose pathnames
11107 contain @file{/bits/stl} or @file{include/sys}.
11109 If, for some reason, you want to include letter @samp{,} in one of
11110 @var{sym}, write @samp{\,}. For example,
11111 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11112 (note the single quote surrounding the option).
11114 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11115 @opindex finstrument-functions-exclude-function-list
11117 This is similar to @option{-finstrument-functions-exclude-file-list},
11118 but this option sets the list of function names to be excluded from
11119 instrumentation. The function name to be matched is its user-visible
11120 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11121 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11122 match is done on substrings: if the @var{sym} parameter is a substring
11123 of the function name, it is considered to be a match. For C99 and C++
11124 extended identifiers, the function name must be given in UTF-8, not
11125 using universal character names.
11130 @node Preprocessor Options
11131 @section Options Controlling the Preprocessor
11132 @cindex preprocessor options
11133 @cindex options, preprocessor
11135 These options control the C preprocessor, which is run on each C source
11136 file before actual compilation.
11138 If you use the @option{-E} option, nothing is done except preprocessing.
11139 Some of these options make sense only together with @option{-E} because
11140 they cause the preprocessor output to be unsuitable for actual
11144 @item -Wp,@var{option}
11146 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11147 and pass @var{option} directly through to the preprocessor. If
11148 @var{option} contains commas, it is split into multiple options at the
11149 commas. However, many options are modified, translated or interpreted
11150 by the compiler driver before being passed to the preprocessor, and
11151 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11152 interface is undocumented and subject to change, so whenever possible
11153 you should avoid using @option{-Wp} and let the driver handle the
11156 @item -Xpreprocessor @var{option}
11157 @opindex Xpreprocessor
11158 Pass @var{option} as an option to the preprocessor. You can use this to
11159 supply system-specific preprocessor options that GCC does not
11162 If you want to pass an option that takes an argument, you must use
11163 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11165 @item -no-integrated-cpp
11166 @opindex no-integrated-cpp
11167 Perform preprocessing as a separate pass before compilation.
11168 By default, GCC performs preprocessing as an integrated part of
11169 input tokenization and parsing.
11170 If this option is provided, the appropriate language front end
11171 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11172 and Objective-C, respectively) is instead invoked twice,
11173 once for preprocessing only and once for actual compilation
11174 of the preprocessed input.
11175 This option may be useful in conjunction with the @option{-B} or
11176 @option{-wrapper} options to specify an alternate preprocessor or
11177 perform additional processing of the program source between
11178 normal preprocessing and compilation.
11181 @include cppopts.texi
11183 @node Assembler Options
11184 @section Passing Options to the Assembler
11186 @c prevent bad page break with this line
11187 You can pass options to the assembler.
11190 @item -Wa,@var{option}
11192 Pass @var{option} as an option to the assembler. If @var{option}
11193 contains commas, it is split into multiple options at the commas.
11195 @item -Xassembler @var{option}
11196 @opindex Xassembler
11197 Pass @var{option} as an option to the assembler. You can use this to
11198 supply system-specific assembler options that GCC does not
11201 If you want to pass an option that takes an argument, you must use
11202 @option{-Xassembler} twice, once for the option and once for the argument.
11207 @section Options for Linking
11208 @cindex link options
11209 @cindex options, linking
11211 These options come into play when the compiler links object files into
11212 an executable output file. They are meaningless if the compiler is
11213 not doing a link step.
11217 @item @var{object-file-name}
11218 A file name that does not end in a special recognized suffix is
11219 considered to name an object file or library. (Object files are
11220 distinguished from libraries by the linker according to the file
11221 contents.) If linking is done, these object files are used as input
11230 If any of these options is used, then the linker is not run, and
11231 object file names should not be used as arguments. @xref{Overall
11235 @opindex fuse-ld=bfd
11236 Use the @command{bfd} linker instead of the default linker.
11238 @item -fuse-ld=gold
11239 @opindex fuse-ld=gold
11240 Use the @command{gold} linker instead of the default linker.
11243 @item -l@var{library}
11244 @itemx -l @var{library}
11246 Search the library named @var{library} when linking. (The second
11247 alternative with the library as a separate argument is only for
11248 POSIX compliance and is not recommended.)
11250 It makes a difference where in the command you write this option; the
11251 linker searches and processes libraries and object files in the order they
11252 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11253 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11254 to functions in @samp{z}, those functions may not be loaded.
11256 The linker searches a standard list of directories for the library,
11257 which is actually a file named @file{lib@var{library}.a}. The linker
11258 then uses this file as if it had been specified precisely by name.
11260 The directories searched include several standard system directories
11261 plus any that you specify with @option{-L}.
11263 Normally the files found this way are library files---archive files
11264 whose members are object files. The linker handles an archive file by
11265 scanning through it for members which define symbols that have so far
11266 been referenced but not defined. But if the file that is found is an
11267 ordinary object file, it is linked in the usual fashion. The only
11268 difference between using an @option{-l} option and specifying a file name
11269 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11270 and searches several directories.
11274 You need this special case of the @option{-l} option in order to
11275 link an Objective-C or Objective-C++ program.
11277 @item -nostartfiles
11278 @opindex nostartfiles
11279 Do not use the standard system startup files when linking.
11280 The standard system libraries are used normally, unless @option{-nostdlib}
11281 or @option{-nodefaultlibs} is used.
11283 @item -nodefaultlibs
11284 @opindex nodefaultlibs
11285 Do not use the standard system libraries when linking.
11286 Only the libraries you specify are passed to the linker, and options
11287 specifying linkage of the system libraries, such as @option{-static-libgcc}
11288 or @option{-shared-libgcc}, are ignored.
11289 The standard startup files are used normally, unless @option{-nostartfiles}
11292 The compiler may generate calls to @code{memcmp},
11293 @code{memset}, @code{memcpy} and @code{memmove}.
11294 These entries are usually resolved by entries in
11295 libc. These entry points should be supplied through some other
11296 mechanism when this option is specified.
11300 Do not use the standard system startup files or libraries when linking.
11301 No startup files and only the libraries you specify are passed to
11302 the linker, and options specifying linkage of the system libraries, such as
11303 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11305 The compiler may generate calls to @code{memcmp}, @code{memset},
11306 @code{memcpy} and @code{memmove}.
11307 These entries are usually resolved by entries in
11308 libc. These entry points should be supplied through some other
11309 mechanism when this option is specified.
11311 @cindex @option{-lgcc}, use with @option{-nostdlib}
11312 @cindex @option{-nostdlib} and unresolved references
11313 @cindex unresolved references and @option{-nostdlib}
11314 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11315 @cindex @option{-nodefaultlibs} and unresolved references
11316 @cindex unresolved references and @option{-nodefaultlibs}
11317 One of the standard libraries bypassed by @option{-nostdlib} and
11318 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11319 which GCC uses to overcome shortcomings of particular machines, or special
11320 needs for some languages.
11321 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11322 Collection (GCC) Internals},
11323 for more discussion of @file{libgcc.a}.)
11324 In most cases, you need @file{libgcc.a} even when you want to avoid
11325 other standard libraries. In other words, when you specify @option{-nostdlib}
11326 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11327 This ensures that you have no unresolved references to internal GCC
11328 library subroutines.
11329 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11330 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11331 GNU Compiler Collection (GCC) Internals}.)
11335 Produce a position independent executable on targets that support it.
11336 For predictable results, you must also specify the same set of options
11337 used for compilation (@option{-fpie}, @option{-fPIE},
11338 or model suboptions) when you specify this linker option.
11342 Don't produce a position independent executable.
11346 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11347 that support it. This instructs the linker to add all symbols, not
11348 only used ones, to the dynamic symbol table. This option is needed
11349 for some uses of @code{dlopen} or to allow obtaining backtraces
11350 from within a program.
11354 Remove all symbol table and relocation information from the executable.
11358 On systems that support dynamic linking, this prevents linking with the shared
11359 libraries. On other systems, this option has no effect.
11363 Produce a shared object which can then be linked with other objects to
11364 form an executable. Not all systems support this option. For predictable
11365 results, you must also specify the same set of options used for compilation
11366 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11367 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11368 needs to build supplementary stub code for constructors to work. On
11369 multi-libbed systems, @samp{gcc -shared} must select the correct support
11370 libraries to link against. Failing to supply the correct flags may lead
11371 to subtle defects. Supplying them in cases where they are not necessary
11374 @item -shared-libgcc
11375 @itemx -static-libgcc
11376 @opindex shared-libgcc
11377 @opindex static-libgcc
11378 On systems that provide @file{libgcc} as a shared library, these options
11379 force the use of either the shared or static version, respectively.
11380 If no shared version of @file{libgcc} was built when the compiler was
11381 configured, these options have no effect.
11383 There are several situations in which an application should use the
11384 shared @file{libgcc} instead of the static version. The most common
11385 of these is when the application wishes to throw and catch exceptions
11386 across different shared libraries. In that case, each of the libraries
11387 as well as the application itself should use the shared @file{libgcc}.
11389 Therefore, the G++ and GCJ drivers automatically add
11390 @option{-shared-libgcc} whenever you build a shared library or a main
11391 executable, because C++ and Java programs typically use exceptions, so
11392 this is the right thing to do.
11394 If, instead, you use the GCC driver to create shared libraries, you may
11395 find that they are not always linked with the shared @file{libgcc}.
11396 If GCC finds, at its configuration time, that you have a non-GNU linker
11397 or a GNU linker that does not support option @option{--eh-frame-hdr},
11398 it links the shared version of @file{libgcc} into shared libraries
11399 by default. Otherwise, it takes advantage of the linker and optimizes
11400 away the linking with the shared version of @file{libgcc}, linking with
11401 the static version of libgcc by default. This allows exceptions to
11402 propagate through such shared libraries, without incurring relocation
11403 costs at library load time.
11405 However, if a library or main executable is supposed to throw or catch
11406 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11407 for the languages used in the program, or using the option
11408 @option{-shared-libgcc}, such that it is linked with the shared
11411 @item -static-libasan
11412 @opindex static-libasan
11413 When the @option{-fsanitize=address} option is used to link a program,
11414 the GCC driver automatically links against @option{libasan}. If
11415 @file{libasan} is available as a shared library, and the @option{-static}
11416 option is not used, then this links against the shared version of
11417 @file{libasan}. The @option{-static-libasan} option directs the GCC
11418 driver to link @file{libasan} statically, without necessarily linking
11419 other libraries statically.
11421 @item -static-libtsan
11422 @opindex static-libtsan
11423 When the @option{-fsanitize=thread} option is used to link a program,
11424 the GCC driver automatically links against @option{libtsan}. If
11425 @file{libtsan} is available as a shared library, and the @option{-static}
11426 option is not used, then this links against the shared version of
11427 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11428 driver to link @file{libtsan} statically, without necessarily linking
11429 other libraries statically.
11431 @item -static-liblsan
11432 @opindex static-liblsan
11433 When the @option{-fsanitize=leak} option is used to link a program,
11434 the GCC driver automatically links against @option{liblsan}. If
11435 @file{liblsan} is available as a shared library, and the @option{-static}
11436 option is not used, then this links against the shared version of
11437 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11438 driver to link @file{liblsan} statically, without necessarily linking
11439 other libraries statically.
11441 @item -static-libubsan
11442 @opindex static-libubsan
11443 When the @option{-fsanitize=undefined} option is used to link a program,
11444 the GCC driver automatically links against @option{libubsan}. If
11445 @file{libubsan} is available as a shared library, and the @option{-static}
11446 option is not used, then this links against the shared version of
11447 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11448 driver to link @file{libubsan} statically, without necessarily linking
11449 other libraries statically.
11451 @item -static-libmpx
11452 @opindex static-libmpx
11453 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11454 used to link a program, the GCC driver automatically links against
11455 @file{libmpx}. If @file{libmpx} is available as a shared library,
11456 and the @option{-static} option is not used, then this links against
11457 the shared version of @file{libmpx}. The @option{-static-libmpx}
11458 option directs the GCC driver to link @file{libmpx} statically,
11459 without necessarily linking other libraries statically.
11461 @item -static-libmpxwrappers
11462 @opindex static-libmpxwrappers
11463 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11464 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11465 GCC driver automatically links against @file{libmpxwrappers}. If
11466 @file{libmpxwrappers} is available as a shared library, and the
11467 @option{-static} option is not used, then this links against the shared
11468 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11469 option directs the GCC driver to link @file{libmpxwrappers} statically,
11470 without necessarily linking other libraries statically.
11472 @item -static-libstdc++
11473 @opindex static-libstdc++
11474 When the @command{g++} program is used to link a C++ program, it
11475 normally automatically links against @option{libstdc++}. If
11476 @file{libstdc++} is available as a shared library, and the
11477 @option{-static} option is not used, then this links against the
11478 shared version of @file{libstdc++}. That is normally fine. However, it
11479 is sometimes useful to freeze the version of @file{libstdc++} used by
11480 the program without going all the way to a fully static link. The
11481 @option{-static-libstdc++} option directs the @command{g++} driver to
11482 link @file{libstdc++} statically, without necessarily linking other
11483 libraries statically.
11487 Bind references to global symbols when building a shared object. Warn
11488 about any unresolved references (unless overridden by the link editor
11489 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11492 @item -T @var{script}
11494 @cindex linker script
11495 Use @var{script} as the linker script. This option is supported by most
11496 systems using the GNU linker. On some targets, such as bare-board
11497 targets without an operating system, the @option{-T} option may be required
11498 when linking to avoid references to undefined symbols.
11500 @item -Xlinker @var{option}
11502 Pass @var{option} as an option to the linker. You can use this to
11503 supply system-specific linker options that GCC does not recognize.
11505 If you want to pass an option that takes a separate argument, you must use
11506 @option{-Xlinker} twice, once for the option and once for the argument.
11507 For example, to pass @option{-assert definitions}, you must write
11508 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11509 @option{-Xlinker "-assert definitions"}, because this passes the entire
11510 string as a single argument, which is not what the linker expects.
11512 When using the GNU linker, it is usually more convenient to pass
11513 arguments to linker options using the @option{@var{option}=@var{value}}
11514 syntax than as separate arguments. For example, you can specify
11515 @option{-Xlinker -Map=output.map} rather than
11516 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11517 this syntax for command-line options.
11519 @item -Wl,@var{option}
11521 Pass @var{option} as an option to the linker. If @var{option} contains
11522 commas, it is split into multiple options at the commas. You can use this
11523 syntax to pass an argument to the option.
11524 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11525 linker. When using the GNU linker, you can also get the same effect with
11526 @option{-Wl,-Map=output.map}.
11528 @item -u @var{symbol}
11530 Pretend the symbol @var{symbol} is undefined, to force linking of
11531 library modules to define it. You can use @option{-u} multiple times with
11532 different symbols to force loading of additional library modules.
11534 @item -z @var{keyword}
11536 @option{-z} is passed directly on to the linker along with the keyword
11537 @var{keyword}. See the section in the documentation of your linker for
11538 permitted values and their meanings.
11541 @node Directory Options
11542 @section Options for Directory Search
11543 @cindex directory options
11544 @cindex options, directory search
11545 @cindex search path
11547 These options specify directories to search for header files, for
11548 libraries and for parts of the compiler:
11553 Add the directory @var{dir} to the head of the list of directories to be
11554 searched for header files. This can be used to override a system header
11555 file, substituting your own version, since these directories are
11556 searched before the system header file directories. However, you should
11557 not use this option to add directories that contain vendor-supplied
11558 system header files (use @option{-isystem} for that). If you use more than
11559 one @option{-I} option, the directories are scanned in left-to-right
11560 order; the standard system directories come after.
11562 If a standard system include directory, or a directory specified with
11563 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11564 option is ignored. The directory is still searched but as a
11565 system directory at its normal position in the system include chain.
11566 This is to ensure that GCC's procedure to fix buggy system headers and
11567 the ordering for the @code{include_next} directive are not inadvertently changed.
11568 If you really need to change the search order for system directories,
11569 use the @option{-nostdinc} and/or @option{-isystem} options.
11571 @item -iplugindir=@var{dir}
11572 @opindex iplugindir=
11573 Set the directory to search for plugins that are passed
11574 by @option{-fplugin=@var{name}} instead of
11575 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11576 to be used by the user, but only passed by the driver.
11578 @item -iquote@var{dir}
11580 Add the directory @var{dir} to the head of the list of directories to
11581 be searched for header files only for the case of @code{#include
11582 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11583 otherwise just like @option{-I}.
11587 Add directory @var{dir} to the list of directories to be searched
11590 @item -B@var{prefix}
11592 This option specifies where to find the executables, libraries,
11593 include files, and data files of the compiler itself.
11595 The compiler driver program runs one or more of the subprograms
11596 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11597 @var{prefix} as a prefix for each program it tries to run, both with and
11598 without @samp{@var{machine}/@var{version}/} for the corresponding target
11599 machine and compiler version.
11601 For each subprogram to be run, the compiler driver first tries the
11602 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11603 is not specified, the driver tries two standard prefixes,
11604 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11605 those results in a file name that is found, the unmodified program
11606 name is searched for using the directories specified in your
11607 @env{PATH} environment variable.
11609 The compiler checks to see if the path provided by @option{-B}
11610 refers to a directory, and if necessary it adds a directory
11611 separator character at the end of the path.
11613 @option{-B} prefixes that effectively specify directory names also apply
11614 to libraries in the linker, because the compiler translates these
11615 options into @option{-L} options for the linker. They also apply to
11616 include files in the preprocessor, because the compiler translates these
11617 options into @option{-isystem} options for the preprocessor. In this case,
11618 the compiler appends @samp{include} to the prefix.
11620 The runtime support file @file{libgcc.a} can also be searched for using
11621 the @option{-B} prefix, if needed. If it is not found there, the two
11622 standard prefixes above are tried, and that is all. The file is left
11623 out of the link if it is not found by those means.
11625 Another way to specify a prefix much like the @option{-B} prefix is to use
11626 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11629 As a special kludge, if the path provided by @option{-B} is
11630 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11631 9, then it is replaced by @file{[dir/]include}. This is to help
11632 with boot-strapping the compiler.
11634 @item -no-canonical-prefixes
11635 @opindex no-canonical-prefixes
11636 Do not expand any symbolic links, resolve references to @samp{/../}
11637 or @samp{/./}, or make the path absolute when generating a relative
11640 @item --sysroot=@var{dir}
11642 Use @var{dir} as the logical root directory for headers and libraries.
11643 For example, if the compiler normally searches for headers in
11644 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11645 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11647 If you use both this option and the @option{-isysroot} option, then
11648 the @option{--sysroot} option applies to libraries, but the
11649 @option{-isysroot} option applies to header files.
11651 The GNU linker (beginning with version 2.16) has the necessary support
11652 for this option. If your linker does not support this option, the
11653 header file aspect of @option{--sysroot} still works, but the
11654 library aspect does not.
11656 @item --no-sysroot-suffix
11657 @opindex no-sysroot-suffix
11658 For some targets, a suffix is added to the root directory specified
11659 with @option{--sysroot}, depending on the other options used, so that
11660 headers may for example be found in
11661 @file{@var{dir}/@var{suffix}/usr/include} instead of
11662 @file{@var{dir}/usr/include}. This option disables the addition of
11667 This option has been deprecated. Please use @option{-iquote} instead for
11668 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11670 Any directories you specify with @option{-I} options before the @option{-I-}
11671 option are searched only for the case of @code{#include "@var{file}"};
11672 they are not searched for @code{#include <@var{file}>}.
11674 If additional directories are specified with @option{-I} options after
11675 the @option{-I-} option, these directories are searched for all @code{#include}
11676 directives. (Ordinarily @emph{all} @option{-I} directories are used
11679 In addition, the @option{-I-} option inhibits the use of the current
11680 directory (where the current input file came from) as the first search
11681 directory for @code{#include "@var{file}"}. There is no way to
11682 override this effect of @option{-I-}. With @option{-I.} you can specify
11683 searching the directory that is current when the compiler is
11684 invoked. That is not exactly the same as what the preprocessor does
11685 by default, but it is often satisfactory.
11687 @option{-I-} does not inhibit the use of the standard system directories
11688 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11692 @node Code Gen Options
11693 @section Options for Code Generation Conventions
11694 @cindex code generation conventions
11695 @cindex options, code generation
11696 @cindex run-time options
11698 These machine-independent options control the interface conventions
11699 used in code generation.
11701 Most of them have both positive and negative forms; the negative form
11702 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11703 one of the forms is listed---the one that is not the default. You
11704 can figure out the other form by either removing @samp{no-} or adding
11708 @item -fstack-reuse=@var{reuse-level}
11709 @opindex fstack_reuse
11710 This option controls stack space reuse for user declared local/auto variables
11711 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11712 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11713 local variables and temporaries, @samp{named_vars} enables the reuse only for
11714 user defined local variables with names, and @samp{none} disables stack reuse
11715 completely. The default value is @samp{all}. The option is needed when the
11716 program extends the lifetime of a scoped local variable or a compiler generated
11717 temporary beyond the end point defined by the language. When a lifetime of
11718 a variable ends, and if the variable lives in memory, the optimizing compiler
11719 has the freedom to reuse its stack space with other temporaries or scoped
11720 local variables whose live range does not overlap with it. Legacy code extending
11721 local lifetime is likely to break with the stack reuse optimization.
11740 if (*p == 10) // out of scope use of local1
11751 A(int k) : i(k), j(k) @{ @}
11758 void foo(const A& ar)
11765 foo(A(10)); // temp object's lifetime ends when foo returns
11771 ap->i+= 10; // ap references out of scope temp whose space
11772 // is reused with a. What is the value of ap->i?
11777 The lifetime of a compiler generated temporary is well defined by the C++
11778 standard. When a lifetime of a temporary ends, and if the temporary lives
11779 in memory, the optimizing compiler has the freedom to reuse its stack
11780 space with other temporaries or scoped local variables whose live range
11781 does not overlap with it. However some of the legacy code relies on
11782 the behavior of older compilers in which temporaries' stack space is
11783 not reused, the aggressive stack reuse can lead to runtime errors. This
11784 option is used to control the temporary stack reuse optimization.
11788 This option generates traps for signed overflow on addition, subtraction,
11789 multiplication operations.
11790 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11791 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11792 @option{-fwrapv} being effective. Note that only active options override, so
11793 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11794 results in @option{-ftrapv} being effective.
11798 This option instructs the compiler to assume that signed arithmetic
11799 overflow of addition, subtraction and multiplication wraps around
11800 using twos-complement representation. This flag enables some optimizations
11801 and disables others. This option is enabled by default for the Java
11802 front end, as required by the Java language specification.
11803 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11804 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11805 @option{-fwrapv} being effective. Note that only active options override, so
11806 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11807 results in @option{-ftrapv} being effective.
11810 @opindex fexceptions
11811 Enable exception handling. Generates extra code needed to propagate
11812 exceptions. For some targets, this implies GCC generates frame
11813 unwind information for all functions, which can produce significant data
11814 size overhead, although it does not affect execution. If you do not
11815 specify this option, GCC enables it by default for languages like
11816 C++ that normally require exception handling, and disables it for
11817 languages like C that do not normally require it. However, you may need
11818 to enable this option when compiling C code that needs to interoperate
11819 properly with exception handlers written in C++. You may also wish to
11820 disable this option if you are compiling older C++ programs that don't
11821 use exception handling.
11823 @item -fnon-call-exceptions
11824 @opindex fnon-call-exceptions
11825 Generate code that allows trapping instructions to throw exceptions.
11826 Note that this requires platform-specific runtime support that does
11827 not exist everywhere. Moreover, it only allows @emph{trapping}
11828 instructions to throw exceptions, i.e.@: memory references or floating-point
11829 instructions. It does not allow exceptions to be thrown from
11830 arbitrary signal handlers such as @code{SIGALRM}.
11832 @item -fdelete-dead-exceptions
11833 @opindex fdelete-dead-exceptions
11834 Consider that instructions that may throw exceptions but don't otherwise
11835 contribute to the execution of the program can be optimized away.
11836 This option is enabled by default for the Ada front end, as permitted by
11837 the Ada language specification.
11838 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
11840 @item -funwind-tables
11841 @opindex funwind-tables
11842 Similar to @option{-fexceptions}, except that it just generates any needed
11843 static data, but does not affect the generated code in any other way.
11844 You normally do not need to enable this option; instead, a language processor
11845 that needs this handling enables it on your behalf.
11847 @item -fasynchronous-unwind-tables
11848 @opindex fasynchronous-unwind-tables
11849 Generate unwind table in DWARF format, if supported by target machine. The
11850 table is exact at each instruction boundary, so it can be used for stack
11851 unwinding from asynchronous events (such as debugger or garbage collector).
11853 @item -fno-gnu-unique
11854 @opindex fno-gnu-unique
11855 On systems with recent GNU assembler and C library, the C++ compiler
11856 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
11857 of template static data members and static local variables in inline
11858 functions are unique even in the presence of @code{RTLD_LOCAL}; this
11859 is necessary to avoid problems with a library used by two different
11860 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
11861 therefore disagreeing with the other one about the binding of the
11862 symbol. But this causes @code{dlclose} to be ignored for affected
11863 DSOs; if your program relies on reinitialization of a DSO via
11864 @code{dlclose} and @code{dlopen}, you can use
11865 @option{-fno-gnu-unique}.
11867 @item -fpcc-struct-return
11868 @opindex fpcc-struct-return
11869 Return ``short'' @code{struct} and @code{union} values in memory like
11870 longer ones, rather than in registers. This convention is less
11871 efficient, but it has the advantage of allowing intercallability between
11872 GCC-compiled files and files compiled with other compilers, particularly
11873 the Portable C Compiler (pcc).
11875 The precise convention for returning structures in memory depends
11876 on the target configuration macros.
11878 Short structures and unions are those whose size and alignment match
11879 that of some integer type.
11881 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11882 switch is not binary compatible with code compiled with the
11883 @option{-freg-struct-return} switch.
11884 Use it to conform to a non-default application binary interface.
11886 @item -freg-struct-return
11887 @opindex freg-struct-return
11888 Return @code{struct} and @code{union} values in registers when possible.
11889 This is more efficient for small structures than
11890 @option{-fpcc-struct-return}.
11892 If you specify neither @option{-fpcc-struct-return} nor
11893 @option{-freg-struct-return}, GCC defaults to whichever convention is
11894 standard for the target. If there is no standard convention, GCC
11895 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11896 the principal compiler. In those cases, we can choose the standard, and
11897 we chose the more efficient register return alternative.
11899 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11900 switch is not binary compatible with code compiled with the
11901 @option{-fpcc-struct-return} switch.
11902 Use it to conform to a non-default application binary interface.
11904 @item -fshort-enums
11905 @opindex fshort-enums
11906 Allocate to an @code{enum} type only as many bytes as it needs for the
11907 declared range of possible values. Specifically, the @code{enum} type
11908 is equivalent to the smallest integer type that has enough room.
11910 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11911 code that is not binary compatible with code generated without that switch.
11912 Use it to conform to a non-default application binary interface.
11914 @item -fshort-wchar
11915 @opindex fshort-wchar
11916 Override the underlying type for @code{wchar_t} to be @code{short
11917 unsigned int} instead of the default for the target. This option is
11918 useful for building programs to run under WINE@.
11920 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11921 code that is not binary compatible with code generated without that switch.
11922 Use it to conform to a non-default application binary interface.
11925 @opindex fno-common
11926 In C code, controls the placement of uninitialized global variables.
11927 Unix C compilers have traditionally permitted multiple definitions of
11928 such variables in different compilation units by placing the variables
11930 This is the behavior specified by @option{-fcommon}, and is the default
11931 for GCC on most targets.
11932 On the other hand, this behavior is not required by ISO C, and on some
11933 targets may carry a speed or code size penalty on variable references.
11934 The @option{-fno-common} option specifies that the compiler should place
11935 uninitialized global variables in the data section of the object file,
11936 rather than generating them as common blocks.
11937 This has the effect that if the same variable is declared
11938 (without @code{extern}) in two different compilations,
11939 you get a multiple-definition error when you link them.
11940 In this case, you must compile with @option{-fcommon} instead.
11941 Compiling with @option{-fno-common} is useful on targets for which
11942 it provides better performance, or if you wish to verify that the
11943 program will work on other systems that always treat uninitialized
11944 variable declarations this way.
11948 Ignore the @code{#ident} directive.
11950 @item -finhibit-size-directive
11951 @opindex finhibit-size-directive
11952 Don't output a @code{.size} assembler directive, or anything else that
11953 would cause trouble if the function is split in the middle, and the
11954 two halves are placed at locations far apart in memory. This option is
11955 used when compiling @file{crtstuff.c}; you should not need to use it
11958 @item -fverbose-asm
11959 @opindex fverbose-asm
11960 Put extra commentary information in the generated assembly code to
11961 make it more readable. This option is generally only of use to those
11962 who actually need to read the generated assembly code (perhaps while
11963 debugging the compiler itself).
11965 @option{-fno-verbose-asm}, the default, causes the
11966 extra information to be omitted and is useful when comparing two assembler
11969 The added comments include:
11974 information on the compiler version and command-line options,
11977 the source code lines associated with the assembly instructions,
11978 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
11981 hints on which high-level expressions correspond to
11982 the various assembly instruction operands.
11986 For example, given this C source file:
11994 for (i = 0; i < n; i++)
12001 compiling to (x86_64) assembly via @option{-S} and emitting the result
12002 direct to stdout via @option{-o} @option{-}
12005 gcc -S test.c -fverbose-asm -Os -o -
12008 gives output similar to this:
12012 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12019 .type test, @@function
12023 # test.c:4: int total = 0;
12024 xorl %eax, %eax # <retval>
12025 # test.c:6: for (i = 0; i < n; i++)
12026 xorl %edx, %edx # i
12028 # test.c:6: for (i = 0; i < n; i++)
12029 cmpl %edi, %edx # n, i
12031 # test.c:7: total += i * i;
12032 movl %edx, %ecx # i, tmp92
12033 imull %edx, %ecx # i, tmp92
12034 # test.c:6: for (i = 0; i < n; i++)
12036 # test.c:7: total += i * i;
12037 addl %ecx, %eax # tmp92, <retval>
12045 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12046 .section .note.GNU-stack,"",@@progbits
12049 The comments are intended for humans rather than machines and hence the
12050 precise format of the comments is subject to change.
12052 @item -frecord-gcc-switches
12053 @opindex frecord-gcc-switches
12054 This switch causes the command line used to invoke the
12055 compiler to be recorded into the object file that is being created.
12056 This switch is only implemented on some targets and the exact format
12057 of the recording is target and binary file format dependent, but it
12058 usually takes the form of a section containing ASCII text. This
12059 switch is related to the @option{-fverbose-asm} switch, but that
12060 switch only records information in the assembler output file as
12061 comments, so it never reaches the object file.
12062 See also @option{-grecord-gcc-switches} for another
12063 way of storing compiler options into the object file.
12067 @cindex global offset table
12069 Generate position-independent code (PIC) suitable for use in a shared
12070 library, if supported for the target machine. Such code accesses all
12071 constant addresses through a global offset table (GOT)@. The dynamic
12072 loader resolves the GOT entries when the program starts (the dynamic
12073 loader is not part of GCC; it is part of the operating system). If
12074 the GOT size for the linked executable exceeds a machine-specific
12075 maximum size, you get an error message from the linker indicating that
12076 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12077 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12078 on the m68k and RS/6000. The x86 has no such limit.)
12080 Position-independent code requires special support, and therefore works
12081 only on certain machines. For the x86, GCC supports PIC for System V
12082 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12083 position-independent.
12085 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12090 If supported for the target machine, emit position-independent code,
12091 suitable for dynamic linking and avoiding any limit on the size of the
12092 global offset table. This option makes a difference on AArch64, m68k,
12093 PowerPC and SPARC@.
12095 Position-independent code requires special support, and therefore works
12096 only on certain machines.
12098 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12105 These options are similar to @option{-fpic} and @option{-fPIC}, but
12106 generated position independent code can be only linked into executables.
12107 Usually these options are used when @option{-pie} GCC option is
12108 used during linking.
12110 @option{-fpie} and @option{-fPIE} both define the macros
12111 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12112 for @option{-fpie} and 2 for @option{-fPIE}.
12116 Do not use the PLT for external function calls in position-independent code.
12117 Instead, load the callee address at call sites from the GOT and branch to it.
12118 This leads to more efficient code by eliminating PLT stubs and exposing
12119 GOT loads to optimizations. On architectures such as 32-bit x86 where
12120 PLT stubs expect the GOT pointer in a specific register, this gives more
12121 register allocation freedom to the compiler.
12122 Lazy binding requires use of the PLT;
12123 with @option{-fno-plt} all external symbols are resolved at load time.
12125 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12126 through the PLT for specific external functions.
12128 In position-dependent code, a few targets also convert calls to
12129 functions that are marked to not use the PLT to use the GOT instead.
12131 @item -fno-jump-tables
12132 @opindex fno-jump-tables
12133 Do not use jump tables for switch statements even where it would be
12134 more efficient than other code generation strategies. This option is
12135 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12136 building code that forms part of a dynamic linker and cannot
12137 reference the address of a jump table. On some targets, jump tables
12138 do not require a GOT and this option is not needed.
12140 @item -ffixed-@var{reg}
12142 Treat the register named @var{reg} as a fixed register; generated code
12143 should never refer to it (except perhaps as a stack pointer, frame
12144 pointer or in some other fixed role).
12146 @var{reg} must be the name of a register. The register names accepted
12147 are machine-specific and are defined in the @code{REGISTER_NAMES}
12148 macro in the machine description macro file.
12150 This flag does not have a negative form, because it specifies a
12153 @item -fcall-used-@var{reg}
12154 @opindex fcall-used
12155 Treat the register named @var{reg} as an allocable register that is
12156 clobbered by function calls. It may be allocated for temporaries or
12157 variables that do not live across a call. Functions compiled this way
12158 do not save and restore the register @var{reg}.
12160 It is an error to use this flag with the frame pointer or stack pointer.
12161 Use of this flag for other registers that have fixed pervasive roles in
12162 the machine's execution model produces disastrous results.
12164 This flag does not have a negative form, because it specifies a
12167 @item -fcall-saved-@var{reg}
12168 @opindex fcall-saved
12169 Treat the register named @var{reg} as an allocable register saved by
12170 functions. It may be allocated even for temporaries or variables that
12171 live across a call. Functions compiled this way save and restore
12172 the register @var{reg} if they use it.
12174 It is an error to use this flag with the frame pointer or stack pointer.
12175 Use of this flag for other registers that have fixed pervasive roles in
12176 the machine's execution model produces disastrous results.
12178 A different sort of disaster results from the use of this flag for
12179 a register in which function values may be returned.
12181 This flag does not have a negative form, because it specifies a
12184 @item -fpack-struct[=@var{n}]
12185 @opindex fpack-struct
12186 Without a value specified, pack all structure members together without
12187 holes. When a value is specified (which must be a small power of two), pack
12188 structure members according to this value, representing the maximum
12189 alignment (that is, objects with default alignment requirements larger than
12190 this are output potentially unaligned at the next fitting location.
12192 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12193 code that is not binary compatible with code generated without that switch.
12194 Additionally, it makes the code suboptimal.
12195 Use it to conform to a non-default application binary interface.
12197 @item -fleading-underscore
12198 @opindex fleading-underscore
12199 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12200 change the way C symbols are represented in the object file. One use
12201 is to help link with legacy assembly code.
12203 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12204 generate code that is not binary compatible with code generated without that
12205 switch. Use it to conform to a non-default application binary interface.
12206 Not all targets provide complete support for this switch.
12208 @item -ftls-model=@var{model}
12209 @opindex ftls-model
12210 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12211 The @var{model} argument should be one of @samp{global-dynamic},
12212 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12213 Note that the choice is subject to optimization: the compiler may use
12214 a more efficient model for symbols not visible outside of the translation
12215 unit, or if @option{-fpic} is not given on the command line.
12217 The default without @option{-fpic} is @samp{initial-exec}; with
12218 @option{-fpic} the default is @samp{global-dynamic}.
12220 @item -ftrampolines
12221 @opindex ftrampolines
12222 For targets that normally need trampolines for nested functions, always
12223 generate them instead of using descriptors. Otherwise, for targets that
12224 do not need them, like for example HP-PA or IA-64, do nothing.
12226 A trampoline is a small piece of code that is created at run time on the
12227 stack when the address of a nested function is taken, and is used to call
12228 the nested function indirectly. Therefore, it requires the stack to be
12229 made executable in order for the program to work properly.
12231 @option{-fno-trampolines} is enabled by default on a language by language
12232 basis to let the compiler avoid generating them, if it computes that this
12233 is safe, and replace them with descriptors. Descriptors are made up of data
12234 only, but the generated code must be prepared to deal with them. As of this
12235 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12237 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12238 @option{-fno-trampolines} are not binary compatible if nested functions are
12239 present. This option must therefore be used on a program-wide basis and be
12240 manipulated with extreme care.
12242 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12243 @opindex fvisibility
12244 Set the default ELF image symbol visibility to the specified option---all
12245 symbols are marked with this unless overridden within the code.
12246 Using this feature can very substantially improve linking and
12247 load times of shared object libraries, produce more optimized
12248 code, provide near-perfect API export and prevent symbol clashes.
12249 It is @strong{strongly} recommended that you use this in any shared objects
12252 Despite the nomenclature, @samp{default} always means public; i.e.,
12253 available to be linked against from outside the shared object.
12254 @samp{protected} and @samp{internal} are pretty useless in real-world
12255 usage so the only other commonly used option is @samp{hidden}.
12256 The default if @option{-fvisibility} isn't specified is
12257 @samp{default}, i.e., make every symbol public.
12259 A good explanation of the benefits offered by ensuring ELF
12260 symbols have the correct visibility is given by ``How To Write
12261 Shared Libraries'' by Ulrich Drepper (which can be found at
12262 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
12263 solution made possible by this option to marking things hidden when
12264 the default is public is to make the default hidden and mark things
12265 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12266 and @code{__attribute__ ((visibility("default")))} instead of
12267 @code{__declspec(dllexport)} you get almost identical semantics with
12268 identical syntax. This is a great boon to those working with
12269 cross-platform projects.
12271 For those adding visibility support to existing code, you may find
12272 @code{#pragma GCC visibility} of use. This works by you enclosing
12273 the declarations you wish to set visibility for with (for example)
12274 @code{#pragma GCC visibility push(hidden)} and
12275 @code{#pragma GCC visibility pop}.
12276 Bear in mind that symbol visibility should be viewed @strong{as
12277 part of the API interface contract} and thus all new code should
12278 always specify visibility when it is not the default; i.e., declarations
12279 only for use within the local DSO should @strong{always} be marked explicitly
12280 as hidden as so to avoid PLT indirection overheads---making this
12281 abundantly clear also aids readability and self-documentation of the code.
12282 Note that due to ISO C++ specification requirements, @code{operator new} and
12283 @code{operator delete} must always be of default visibility.
12285 Be aware that headers from outside your project, in particular system
12286 headers and headers from any other library you use, may not be
12287 expecting to be compiled with visibility other than the default. You
12288 may need to explicitly say @code{#pragma GCC visibility push(default)}
12289 before including any such headers.
12291 @code{extern} declarations are not affected by @option{-fvisibility}, so
12292 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12293 no modifications. However, this means that calls to @code{extern}
12294 functions with no explicit visibility use the PLT, so it is more
12295 effective to use @code{__attribute ((visibility))} and/or
12296 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12297 declarations should be treated as hidden.
12299 Note that @option{-fvisibility} does affect C++ vague linkage
12300 entities. This means that, for instance, an exception class that is
12301 be thrown between DSOs must be explicitly marked with default
12302 visibility so that the @samp{type_info} nodes are unified between
12305 An overview of these techniques, their benefits and how to use them
12306 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12308 @item -fstrict-volatile-bitfields
12309 @opindex fstrict-volatile-bitfields
12310 This option should be used if accesses to volatile bit-fields (or other
12311 structure fields, although the compiler usually honors those types
12312 anyway) should use a single access of the width of the
12313 field's type, aligned to a natural alignment if possible. For
12314 example, targets with memory-mapped peripheral registers might require
12315 all such accesses to be 16 bits wide; with this flag you can
12316 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12317 is 16 bits on these targets) to force GCC to use 16-bit accesses
12318 instead of, perhaps, a more efficient 32-bit access.
12320 If this option is disabled, the compiler uses the most efficient
12321 instruction. In the previous example, that might be a 32-bit load
12322 instruction, even though that accesses bytes that do not contain
12323 any portion of the bit-field, or memory-mapped registers unrelated to
12324 the one being updated.
12326 In some cases, such as when the @code{packed} attribute is applied to a
12327 structure field, it may not be possible to access the field with a single
12328 read or write that is correctly aligned for the target machine. In this
12329 case GCC falls back to generating multiple accesses rather than code that
12330 will fault or truncate the result at run time.
12332 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12333 not allowed to touch non bit-field members. It is therefore recommended
12334 to define all bits of the field's type as bit-field members.
12336 The default value of this option is determined by the application binary
12337 interface for the target processor.
12339 @item -fsync-libcalls
12340 @opindex fsync-libcalls
12341 This option controls whether any out-of-line instance of the @code{__sync}
12342 family of functions may be used to implement the C++11 @code{__atomic}
12343 family of functions.
12345 The default value of this option is enabled, thus the only useful form
12346 of the option is @option{-fno-sync-libcalls}. This option is used in
12347 the implementation of the @file{libatomic} runtime library.
12351 @node Developer Options
12352 @section GCC Developer Options
12353 @cindex developer options
12354 @cindex debugging GCC
12355 @cindex debug dump options
12356 @cindex dump options
12357 @cindex compilation statistics
12359 This section describes command-line options that are primarily of
12360 interest to GCC developers, including options to support compiler
12361 testing and investigation of compiler bugs and compile-time
12362 performance problems. This includes options that produce debug dumps
12363 at various points in the compilation; that print statistics such as
12364 memory use and execution time; and that print information about GCC's
12365 configuration, such as where it searches for libraries. You should
12366 rarely need to use any of these options for ordinary compilation and
12371 @item -d@var{letters}
12372 @itemx -fdump-rtl-@var{pass}
12373 @itemx -fdump-rtl-@var{pass}=@var{filename}
12375 @opindex fdump-rtl-@var{pass}
12376 Says to make debugging dumps during compilation at times specified by
12377 @var{letters}. This is used for debugging the RTL-based passes of the
12378 compiler. The file names for most of the dumps are made by appending
12379 a pass number and a word to the @var{dumpname}, and the files are
12380 created in the directory of the output file. In case of
12381 @option{=@var{filename}} option, the dump is output on the given file
12382 instead of the pass numbered dump files. Note that the pass number is
12383 assigned as passes are registered into the pass manager. Most passes
12384 are registered in the order that they will execute and for these passes
12385 the number corresponds to the pass execution order. However, passes
12386 registered by plugins, passes specific to compilation targets, or
12387 passes that are otherwise registered after all the other passes are
12388 numbered higher than a pass named "final", even if they are executed
12389 earlier. @var{dumpname} is generated from the name of the output
12390 file if explicitly specified and not an executable, otherwise it is
12391 the basename of the source file. These switches may have different
12392 effects when @option{-E} is used for preprocessing.
12394 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12395 @option{-d} option @var{letters}. Here are the possible
12396 letters for use in @var{pass} and @var{letters}, and their meanings:
12400 @item -fdump-rtl-alignments
12401 @opindex fdump-rtl-alignments
12402 Dump after branch alignments have been computed.
12404 @item -fdump-rtl-asmcons
12405 @opindex fdump-rtl-asmcons
12406 Dump after fixing rtl statements that have unsatisfied in/out constraints.
12408 @item -fdump-rtl-auto_inc_dec
12409 @opindex fdump-rtl-auto_inc_dec
12410 Dump after auto-inc-dec discovery. This pass is only run on
12411 architectures that have auto inc or auto dec instructions.
12413 @item -fdump-rtl-barriers
12414 @opindex fdump-rtl-barriers
12415 Dump after cleaning up the barrier instructions.
12417 @item -fdump-rtl-bbpart
12418 @opindex fdump-rtl-bbpart
12419 Dump after partitioning hot and cold basic blocks.
12421 @item -fdump-rtl-bbro
12422 @opindex fdump-rtl-bbro
12423 Dump after block reordering.
12425 @item -fdump-rtl-btl1
12426 @itemx -fdump-rtl-btl2
12427 @opindex fdump-rtl-btl2
12428 @opindex fdump-rtl-btl2
12429 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
12430 after the two branch
12431 target load optimization passes.
12433 @item -fdump-rtl-bypass
12434 @opindex fdump-rtl-bypass
12435 Dump after jump bypassing and control flow optimizations.
12437 @item -fdump-rtl-combine
12438 @opindex fdump-rtl-combine
12439 Dump after the RTL instruction combination pass.
12441 @item -fdump-rtl-compgotos
12442 @opindex fdump-rtl-compgotos
12443 Dump after duplicating the computed gotos.
12445 @item -fdump-rtl-ce1
12446 @itemx -fdump-rtl-ce2
12447 @itemx -fdump-rtl-ce3
12448 @opindex fdump-rtl-ce1
12449 @opindex fdump-rtl-ce2
12450 @opindex fdump-rtl-ce3
12451 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
12452 @option{-fdump-rtl-ce3} enable dumping after the three
12453 if conversion passes.
12455 @item -fdump-rtl-cprop_hardreg
12456 @opindex fdump-rtl-cprop_hardreg
12457 Dump after hard register copy propagation.
12459 @item -fdump-rtl-csa
12460 @opindex fdump-rtl-csa
12461 Dump after combining stack adjustments.
12463 @item -fdump-rtl-cse1
12464 @itemx -fdump-rtl-cse2
12465 @opindex fdump-rtl-cse1
12466 @opindex fdump-rtl-cse2
12467 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
12468 the two common subexpression elimination passes.
12470 @item -fdump-rtl-dce
12471 @opindex fdump-rtl-dce
12472 Dump after the standalone dead code elimination passes.
12474 @item -fdump-rtl-dbr
12475 @opindex fdump-rtl-dbr
12476 Dump after delayed branch scheduling.
12478 @item -fdump-rtl-dce1
12479 @itemx -fdump-rtl-dce2
12480 @opindex fdump-rtl-dce1
12481 @opindex fdump-rtl-dce2
12482 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
12483 the two dead store elimination passes.
12485 @item -fdump-rtl-eh
12486 @opindex fdump-rtl-eh
12487 Dump after finalization of EH handling code.
12489 @item -fdump-rtl-eh_ranges
12490 @opindex fdump-rtl-eh_ranges
12491 Dump after conversion of EH handling range regions.
12493 @item -fdump-rtl-expand
12494 @opindex fdump-rtl-expand
12495 Dump after RTL generation.
12497 @item -fdump-rtl-fwprop1
12498 @itemx -fdump-rtl-fwprop2
12499 @opindex fdump-rtl-fwprop1
12500 @opindex fdump-rtl-fwprop2
12501 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
12502 dumping after the two forward propagation passes.
12504 @item -fdump-rtl-gcse1
12505 @itemx -fdump-rtl-gcse2
12506 @opindex fdump-rtl-gcse1
12507 @opindex fdump-rtl-gcse2
12508 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
12509 after global common subexpression elimination.
12511 @item -fdump-rtl-init-regs
12512 @opindex fdump-rtl-init-regs
12513 Dump after the initialization of the registers.
12515 @item -fdump-rtl-initvals
12516 @opindex fdump-rtl-initvals
12517 Dump after the computation of the initial value sets.
12519 @item -fdump-rtl-into_cfglayout
12520 @opindex fdump-rtl-into_cfglayout
12521 Dump after converting to cfglayout mode.
12523 @item -fdump-rtl-ira
12524 @opindex fdump-rtl-ira
12525 Dump after iterated register allocation.
12527 @item -fdump-rtl-jump
12528 @opindex fdump-rtl-jump
12529 Dump after the second jump optimization.
12531 @item -fdump-rtl-loop2
12532 @opindex fdump-rtl-loop2
12533 @option{-fdump-rtl-loop2} enables dumping after the rtl
12534 loop optimization passes.
12536 @item -fdump-rtl-mach
12537 @opindex fdump-rtl-mach
12538 Dump after performing the machine dependent reorganization pass, if that
12541 @item -fdump-rtl-mode_sw
12542 @opindex fdump-rtl-mode_sw
12543 Dump after removing redundant mode switches.
12545 @item -fdump-rtl-rnreg
12546 @opindex fdump-rtl-rnreg
12547 Dump after register renumbering.
12549 @item -fdump-rtl-outof_cfglayout
12550 @opindex fdump-rtl-outof_cfglayout
12551 Dump after converting from cfglayout mode.
12553 @item -fdump-rtl-peephole2
12554 @opindex fdump-rtl-peephole2
12555 Dump after the peephole pass.
12557 @item -fdump-rtl-postreload
12558 @opindex fdump-rtl-postreload
12559 Dump after post-reload optimizations.
12561 @item -fdump-rtl-pro_and_epilogue
12562 @opindex fdump-rtl-pro_and_epilogue
12563 Dump after generating the function prologues and epilogues.
12565 @item -fdump-rtl-sched1
12566 @itemx -fdump-rtl-sched2
12567 @opindex fdump-rtl-sched1
12568 @opindex fdump-rtl-sched2
12569 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
12570 after the basic block scheduling passes.
12572 @item -fdump-rtl-ree
12573 @opindex fdump-rtl-ree
12574 Dump after sign/zero extension elimination.
12576 @item -fdump-rtl-seqabstr
12577 @opindex fdump-rtl-seqabstr
12578 Dump after common sequence discovery.
12580 @item -fdump-rtl-shorten
12581 @opindex fdump-rtl-shorten
12582 Dump after shortening branches.
12584 @item -fdump-rtl-sibling
12585 @opindex fdump-rtl-sibling
12586 Dump after sibling call optimizations.
12588 @item -fdump-rtl-split1
12589 @itemx -fdump-rtl-split2
12590 @itemx -fdump-rtl-split3
12591 @itemx -fdump-rtl-split4
12592 @itemx -fdump-rtl-split5
12593 @opindex fdump-rtl-split1
12594 @opindex fdump-rtl-split2
12595 @opindex fdump-rtl-split3
12596 @opindex fdump-rtl-split4
12597 @opindex fdump-rtl-split5
12598 These options enable dumping after five rounds of
12599 instruction splitting.
12601 @item -fdump-rtl-sms
12602 @opindex fdump-rtl-sms
12603 Dump after modulo scheduling. This pass is only run on some
12606 @item -fdump-rtl-stack
12607 @opindex fdump-rtl-stack
12608 Dump after conversion from GCC's ``flat register file'' registers to the
12609 x87's stack-like registers. This pass is only run on x86 variants.
12611 @item -fdump-rtl-subreg1
12612 @itemx -fdump-rtl-subreg2
12613 @opindex fdump-rtl-subreg1
12614 @opindex fdump-rtl-subreg2
12615 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
12616 the two subreg expansion passes.
12618 @item -fdump-rtl-unshare
12619 @opindex fdump-rtl-unshare
12620 Dump after all rtl has been unshared.
12622 @item -fdump-rtl-vartrack
12623 @opindex fdump-rtl-vartrack
12624 Dump after variable tracking.
12626 @item -fdump-rtl-vregs
12627 @opindex fdump-rtl-vregs
12628 Dump after converting virtual registers to hard registers.
12630 @item -fdump-rtl-web
12631 @opindex fdump-rtl-web
12632 Dump after live range splitting.
12634 @item -fdump-rtl-regclass
12635 @itemx -fdump-rtl-subregs_of_mode_init
12636 @itemx -fdump-rtl-subregs_of_mode_finish
12637 @itemx -fdump-rtl-dfinit
12638 @itemx -fdump-rtl-dfinish
12639 @opindex fdump-rtl-regclass
12640 @opindex fdump-rtl-subregs_of_mode_init
12641 @opindex fdump-rtl-subregs_of_mode_finish
12642 @opindex fdump-rtl-dfinit
12643 @opindex fdump-rtl-dfinish
12644 These dumps are defined but always produce empty files.
12647 @itemx -fdump-rtl-all
12649 @opindex fdump-rtl-all
12650 Produce all the dumps listed above.
12654 Annotate the assembler output with miscellaneous debugging information.
12658 Dump all macro definitions, at the end of preprocessing, in addition to
12663 Produce a core dump whenever an error occurs.
12667 Annotate the assembler output with a comment indicating which
12668 pattern and alternative is used. The length of each instruction is
12673 Dump the RTL in the assembler output as a comment before each instruction.
12674 Also turns on @option{-dp} annotation.
12678 Just generate RTL for a function instead of compiling it. Usually used
12679 with @option{-fdump-rtl-expand}.
12682 @item -fdump-noaddr
12683 @opindex fdump-noaddr
12684 When doing debugging dumps, suppress address output. This makes it more
12685 feasible to use diff on debugging dumps for compiler invocations with
12686 different compiler binaries and/or different
12687 text / bss / data / heap / stack / dso start locations.
12690 @opindex freport-bug
12691 Collect and dump debug information into a temporary file if an
12692 internal compiler error (ICE) occurs.
12694 @item -fdump-unnumbered
12695 @opindex fdump-unnumbered
12696 When doing debugging dumps, suppress instruction numbers and address output.
12697 This makes it more feasible to use diff on debugging dumps for compiler
12698 invocations with different options, in particular with and without
12701 @item -fdump-unnumbered-links
12702 @opindex fdump-unnumbered-links
12703 When doing debugging dumps (see @option{-d} option above), suppress
12704 instruction numbers for the links to the previous and next instructions
12707 @item -fdump-translation-unit @r{(C++ only)}
12708 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12709 @opindex fdump-translation-unit
12710 Dump a representation of the tree structure for the entire translation
12711 unit to a file. The file name is made by appending @file{.tu} to the
12712 source file name, and the file is created in the same directory as the
12713 output file. If the @samp{-@var{options}} form is used, @var{options}
12714 controls the details of the dump as described for the
12715 @option{-fdump-tree} options.
12717 @item -fdump-class-hierarchy @r{(C++ only)}
12718 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12719 @opindex fdump-class-hierarchy
12720 Dump a representation of each class's hierarchy and virtual function
12721 table layout to a file. The file name is made by appending
12722 @file{.class} to the source file name, and the file is created in the
12723 same directory as the output file. If the @samp{-@var{options}} form
12724 is used, @var{options} controls the details of the dump as described
12725 for the @option{-fdump-tree} options.
12727 @item -fdump-ipa-@var{switch}
12729 Control the dumping at various stages of inter-procedural analysis
12730 language tree to a file. The file name is generated by appending a
12731 switch specific suffix to the source file name, and the file is created
12732 in the same directory as the output file. The following dumps are
12737 Enables all inter-procedural analysis dumps.
12740 Dumps information about call-graph optimization, unused function removal,
12741 and inlining decisions.
12744 Dump after function inlining.
12748 @item -fdump-passes
12749 @opindex fdump-passes
12750 Dump the list of optimization passes that are turned on and off by
12751 the current command-line options.
12753 @item -fdump-statistics-@var{option}
12754 @opindex fdump-statistics
12755 Enable and control dumping of pass statistics in a separate file. The
12756 file name is generated by appending a suffix ending in
12757 @samp{.statistics} to the source file name, and the file is created in
12758 the same directory as the output file. If the @samp{-@var{option}}
12759 form is used, @samp{-stats} causes counters to be summed over the
12760 whole compilation unit while @samp{-details} dumps every event as
12761 the passes generate them. The default with no option is to sum
12762 counters for each function compiled.
12764 @item -fdump-tree-@var{switch}
12765 @itemx -fdump-tree-@var{switch}-@var{options}
12766 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
12767 @opindex fdump-tree
12768 Control the dumping at various stages of processing the intermediate
12769 language tree to a file. The file name is generated by appending a
12770 switch-specific suffix to the source file name, and the file is
12771 created in the same directory as the output file. In case of
12772 @option{=@var{filename}} option, the dump is output on the given file
12773 instead of the auto named dump files. If the @samp{-@var{options}}
12774 form is used, @var{options} is a list of @samp{-} separated options
12775 which control the details of the dump. Not all options are applicable
12776 to all dumps; those that are not meaningful are ignored. The
12777 following options are available
12781 Print the address of each node. Usually this is not meaningful as it
12782 changes according to the environment and source file. Its primary use
12783 is for tying up a dump file with a debug environment.
12785 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
12786 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
12787 use working backward from mangled names in the assembly file.
12789 When dumping front-end intermediate representations, inhibit dumping
12790 of members of a scope or body of a function merely because that scope
12791 has been reached. Only dump such items when they are directly reachable
12792 by some other path.
12794 When dumping pretty-printed trees, this option inhibits dumping the
12795 bodies of control structures.
12797 When dumping RTL, print the RTL in slim (condensed) form instead of
12798 the default LISP-like representation.
12800 Print a raw representation of the tree. By default, trees are
12801 pretty-printed into a C-like representation.
12803 Enable more detailed dumps (not honored by every dump option). Also
12804 include information from the optimization passes.
12806 Enable dumping various statistics about the pass (not honored by every dump
12809 Enable showing basic block boundaries (disabled in raw dumps).
12811 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
12812 dump a representation of the control flow graph suitable for viewing with
12813 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
12814 the file is pretty-printed as a subgraph, so that GraphViz can render them
12815 all in a single plot.
12817 This option currently only works for RTL dumps, and the RTL is always
12818 dumped in slim form.
12820 Enable showing virtual operands for every statement.
12822 Enable showing line numbers for statements.
12824 Enable showing the unique ID (@code{DECL_UID}) for each variable.
12826 Enable showing the tree dump for each statement.
12828 Enable showing the EH region number holding each statement.
12830 Enable showing scalar evolution analysis details.
12832 Enable showing optimization information (only available in certain
12835 Enable showing missed optimization information (only available in certain
12838 Enable other detailed optimization information (only available in
12840 @item =@var{filename}
12841 Instead of an auto named dump file, output into the given file
12842 name. The file names @file{stdout} and @file{stderr} are treated
12843 specially and are considered already open standard streams. For
12847 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
12848 -fdump-tree-pre=stderr file.c
12851 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
12852 output on to @file{stderr}. If two conflicting dump filenames are
12853 given for the same pass, then the latter option overrides the earlier
12857 @opindex fdump-tree-split-paths
12858 Dump each function after splitting paths to loop backedges. The file
12859 name is made by appending @file{.split-paths} to the source file name.
12862 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
12863 and @option{lineno}.
12866 Turn on all optimization options, i.e., @option{optimized},
12867 @option{missed}, and @option{note}.
12870 The following tree dumps are possible:
12874 @opindex fdump-tree-original
12875 Dump before any tree based optimization, to @file{@var{file}.original}.
12878 @opindex fdump-tree-optimized
12879 Dump after all tree based optimization, to @file{@var{file}.optimized}.
12882 @opindex fdump-tree-gimple
12883 Dump each function before and after the gimplification pass to a file. The
12884 file name is made by appending @file{.gimple} to the source file name.
12887 @opindex fdump-tree-cfg
12888 Dump the control flow graph of each function to a file. The file name is
12889 made by appending @file{.cfg} to the source file name.
12892 @opindex fdump-tree-ch
12893 Dump each function after copying loop headers. The file name is made by
12894 appending @file{.ch} to the source file name.
12897 @opindex fdump-tree-ssa
12898 Dump SSA related information to a file. The file name is made by appending
12899 @file{.ssa} to the source file name.
12902 @opindex fdump-tree-alias
12903 Dump aliasing information for each function. The file name is made by
12904 appending @file{.alias} to the source file name.
12907 @opindex fdump-tree-ccp
12908 Dump each function after CCP@. The file name is made by appending
12909 @file{.ccp} to the source file name.
12912 @opindex fdump-tree-storeccp
12913 Dump each function after STORE-CCP@. The file name is made by appending
12914 @file{.storeccp} to the source file name.
12917 @opindex fdump-tree-pre
12918 Dump trees after partial redundancy elimination and/or code hoisting.
12919 The file name is made by appending @file{.pre} to the source file name.
12922 @opindex fdump-tree-fre
12923 Dump trees after full redundancy elimination. The file name is made
12924 by appending @file{.fre} to the source file name.
12927 @opindex fdump-tree-copyprop
12928 Dump trees after copy propagation. The file name is made
12929 by appending @file{.copyprop} to the source file name.
12931 @item store_copyprop
12932 @opindex fdump-tree-store_copyprop
12933 Dump trees after store copy-propagation. The file name is made
12934 by appending @file{.store_copyprop} to the source file name.
12937 @opindex fdump-tree-dce
12938 Dump each function after dead code elimination. The file name is made by
12939 appending @file{.dce} to the source file name.
12942 @opindex fdump-tree-sra
12943 Dump each function after performing scalar replacement of aggregates. The
12944 file name is made by appending @file{.sra} to the source file name.
12947 @opindex fdump-tree-sink
12948 Dump each function after performing code sinking. The file name is made
12949 by appending @file{.sink} to the source file name.
12952 @opindex fdump-tree-dom
12953 Dump each function after applying dominator tree optimizations. The file
12954 name is made by appending @file{.dom} to the source file name.
12957 @opindex fdump-tree-dse
12958 Dump each function after applying dead store elimination. The file
12959 name is made by appending @file{.dse} to the source file name.
12962 @opindex fdump-tree-phiopt
12963 Dump each function after optimizing PHI nodes into straightline code. The file
12964 name is made by appending @file{.phiopt} to the source file name.
12967 @opindex fdump-tree-backprop
12968 Dump each function after back-propagating use information up the definition
12969 chain. The file name is made by appending @file{.backprop} to the
12973 @opindex fdump-tree-forwprop
12974 Dump each function after forward propagating single use variables. The file
12975 name is made by appending @file{.forwprop} to the source file name.
12978 @opindex fdump-tree-nrv
12979 Dump each function after applying the named return value optimization on
12980 generic trees. The file name is made by appending @file{.nrv} to the source
12984 @opindex fdump-tree-vect
12985 Dump each function after applying vectorization of loops. The file name is
12986 made by appending @file{.vect} to the source file name.
12989 @opindex fdump-tree-slp
12990 Dump each function after applying vectorization of basic blocks. The file name
12991 is made by appending @file{.slp} to the source file name.
12994 @opindex fdump-tree-vrp
12995 Dump each function after Value Range Propagation (VRP). The file name
12996 is made by appending @file{.vrp} to the source file name.
12999 @opindex fdump-tree-evrp
13000 Dump each function after Early Value Range Propagation (EVRP). The file name
13001 is made by appending @file{.evrp} to the source file name.
13004 @opindex fdump-tree-oaccdevlow
13005 Dump each function after applying device-specific OpenACC transformations.
13006 The file name is made by appending @file{.oaccdevlow} to the source file name.
13009 @opindex fdump-tree-all
13010 Enable all the available tree dumps with the flags provided in this option.
13014 @itemx -fopt-info-@var{options}
13015 @itemx -fopt-info-@var{options}=@var{filename}
13017 Controls optimization dumps from various optimization passes. If the
13018 @samp{-@var{options}} form is used, @var{options} is a list of
13019 @samp{-} separated option keywords to select the dump details and
13022 The @var{options} can be divided into two groups: options describing the
13023 verbosity of the dump, and options describing which optimizations
13024 should be included. The options from both the groups can be freely
13025 mixed as they are non-overlapping. However, in case of any conflicts,
13026 the later options override the earlier options on the command
13029 The following options control the dump verbosity:
13033 Print information when an optimization is successfully applied. It is
13034 up to a pass to decide which information is relevant. For example, the
13035 vectorizer passes print the source location of loops which are
13036 successfully vectorized.
13038 Print information about missed optimizations. Individual passes
13039 control which information to include in the output.
13041 Print verbose information about optimizations, such as certain
13042 transformations, more detailed messages about decisions etc.
13044 Print detailed optimization information. This includes
13045 @samp{optimized}, @samp{missed}, and @samp{note}.
13048 One or more of the following option keywords can be used to describe a
13049 group of optimizations:
13053 Enable dumps from all interprocedural optimizations.
13055 Enable dumps from all loop optimizations.
13057 Enable dumps from all inlining optimizations.
13059 Enable dumps from all vectorization optimizations.
13061 Enable dumps from all optimizations. This is a superset of
13062 the optimization groups listed above.
13065 If @var{options} is
13066 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13067 info about successful optimizations from all the passes.
13069 If the @var{filename} is provided, then the dumps from all the
13070 applicable optimizations are concatenated into the @var{filename}.
13071 Otherwise the dump is output onto @file{stderr}. Though multiple
13072 @option{-fopt-info} options are accepted, only one of them can include
13073 a @var{filename}. If other filenames are provided then all but the
13074 first such option are ignored.
13076 Note that the output @var{filename} is overwritten
13077 in case of multiple translation units. If a combined output from
13078 multiple translation units is desired, @file{stderr} should be used
13081 In the following example, the optimization info is output to
13090 gcc -O3 -fopt-info-missed=missed.all
13094 outputs missed optimization report from all the passes into
13095 @file{missed.all}, and this one:
13098 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13102 prints information about missed optimization opportunities from
13103 vectorization passes on @file{stderr}.
13104 Note that @option{-fopt-info-vec-missed} is equivalent to
13105 @option{-fopt-info-missed-vec}.
13107 As another example,
13109 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13113 outputs information about missed optimizations as well as
13114 optimized locations from all the inlining passes into
13120 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13124 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13125 in conflict since only one output file is allowed. In this case, only
13126 the first option takes effect and the subsequent options are
13127 ignored. Thus only @file{vec.miss} is produced which contains
13128 dumps from the vectorizer about missed opportunities.
13130 @item -fsched-verbose=@var{n}
13131 @opindex fsched-verbose
13132 On targets that use instruction scheduling, this option controls the
13133 amount of debugging output the scheduler prints to the dump files.
13135 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13136 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13137 For @var{n} greater than one, it also output basic block probabilities,
13138 detailed ready list information and unit/insn info. For @var{n} greater
13139 than two, it includes RTL at abort point, control-flow and regions info.
13140 And for @var{n} over four, @option{-fsched-verbose} also includes
13145 @item -fenable-@var{kind}-@var{pass}
13146 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13150 This is a set of options that are used to explicitly disable/enable
13151 optimization passes. These options are intended for use for debugging GCC.
13152 Compiler users should use regular options for enabling/disabling
13157 @item -fdisable-ipa-@var{pass}
13158 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13159 statically invoked in the compiler multiple times, the pass name should be
13160 appended with a sequential number starting from 1.
13162 @item -fdisable-rtl-@var{pass}
13163 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13164 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13165 statically invoked in the compiler multiple times, the pass name should be
13166 appended with a sequential number starting from 1. @var{range-list} is a
13167 comma-separated list of function ranges or assembler names. Each range is a number
13168 pair separated by a colon. The range is inclusive in both ends. If the range
13169 is trivial, the number pair can be simplified as a single number. If the
13170 function's call graph node's @var{uid} falls within one of the specified ranges,
13171 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13172 function header of a dump file, and the pass names can be dumped by using
13173 option @option{-fdump-passes}.
13175 @item -fdisable-tree-@var{pass}
13176 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13177 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13180 @item -fenable-ipa-@var{pass}
13181 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13182 statically invoked in the compiler multiple times, the pass name should be
13183 appended with a sequential number starting from 1.
13185 @item -fenable-rtl-@var{pass}
13186 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13187 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13188 description and examples.
13190 @item -fenable-tree-@var{pass}
13191 @itemx -fenable-tree-@var{pass}=@var{range-list}
13192 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13193 of option arguments.
13197 Here are some examples showing uses of these options.
13201 # disable ccp1 for all functions
13202 -fdisable-tree-ccp1
13203 # disable complete unroll for function whose cgraph node uid is 1
13204 -fenable-tree-cunroll=1
13205 # disable gcse2 for functions at the following ranges [1,1],
13206 # [300,400], and [400,1000]
13207 # disable gcse2 for functions foo and foo2
13208 -fdisable-rtl-gcse2=foo,foo2
13209 # disable early inlining
13210 -fdisable-tree-einline
13211 # disable ipa inlining
13212 -fdisable-ipa-inline
13213 # enable tree full unroll
13214 -fenable-tree-unroll
13219 @itemx -fchecking=@var{n}
13221 @opindex fno-checking
13222 Enable internal consistency checking. The default depends on
13223 the compiler configuration. @option{-fchecking=2} enables further
13224 internal consistency checking that might affect code generation.
13226 @item -frandom-seed=@var{string}
13227 @opindex frandom-seed
13228 This option provides a seed that GCC uses in place of
13229 random numbers in generating certain symbol names
13230 that have to be different in every compiled file. It is also used to
13231 place unique stamps in coverage data files and the object files that
13232 produce them. You can use the @option{-frandom-seed} option to produce
13233 reproducibly identical object files.
13235 The @var{string} can either be a number (decimal, octal or hex) or an
13236 arbitrary string (in which case it's converted to a number by
13239 The @var{string} should be different for every file you compile.
13242 @itemx -save-temps=cwd
13243 @opindex save-temps
13244 Store the usual ``temporary'' intermediate files permanently; place them
13245 in the current directory and name them based on the source file. Thus,
13246 compiling @file{foo.c} with @option{-c -save-temps} produces files
13247 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13248 preprocessed @file{foo.i} output file even though the compiler now
13249 normally uses an integrated preprocessor.
13251 When used in combination with the @option{-x} command-line option,
13252 @option{-save-temps} is sensible enough to avoid over writing an
13253 input source file with the same extension as an intermediate file.
13254 The corresponding intermediate file may be obtained by renaming the
13255 source file before using @option{-save-temps}.
13257 If you invoke GCC in parallel, compiling several different source
13258 files that share a common base name in different subdirectories or the
13259 same source file compiled for multiple output destinations, it is
13260 likely that the different parallel compilers will interfere with each
13261 other, and overwrite the temporary files. For instance:
13264 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13265 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13268 may result in @file{foo.i} and @file{foo.o} being written to
13269 simultaneously by both compilers.
13271 @item -save-temps=obj
13272 @opindex save-temps=obj
13273 Store the usual ``temporary'' intermediate files permanently. If the
13274 @option{-o} option is used, the temporary files are based on the
13275 object file. If the @option{-o} option is not used, the
13276 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13281 gcc -save-temps=obj -c foo.c
13282 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13283 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13287 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13288 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13289 @file{dir2/yfoobar.o}.
13291 @item -time@r{[}=@var{file}@r{]}
13293 Report the CPU time taken by each subprocess in the compilation
13294 sequence. For C source files, this is the compiler proper and assembler
13295 (plus the linker if linking is done).
13297 Without the specification of an output file, the output looks like this:
13304 The first number on each line is the ``user time'', that is time spent
13305 executing the program itself. The second number is ``system time'',
13306 time spent executing operating system routines on behalf of the program.
13307 Both numbers are in seconds.
13309 With the specification of an output file, the output is appended to the
13310 named file, and it looks like this:
13313 0.12 0.01 cc1 @var{options}
13314 0.00 0.01 as @var{options}
13317 The ``user time'' and the ``system time'' are moved before the program
13318 name, and the options passed to the program are displayed, so that one
13319 can later tell what file was being compiled, and with which options.
13321 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13322 @opindex fdump-final-insns
13323 Dump the final internal representation (RTL) to @var{file}. If the
13324 optional argument is omitted (or if @var{file} is @code{.}), the name
13325 of the dump file is determined by appending @code{.gkd} to the
13326 compilation output file name.
13328 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13329 @opindex fcompare-debug
13330 @opindex fno-compare-debug
13331 If no error occurs during compilation, run the compiler a second time,
13332 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13333 passed to the second compilation. Dump the final internal
13334 representation in both compilations, and print an error if they differ.
13336 If the equal sign is omitted, the default @option{-gtoggle} is used.
13338 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13339 and nonzero, implicitly enables @option{-fcompare-debug}. If
13340 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13341 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13344 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13345 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13346 of the final representation and the second compilation, preventing even
13347 @env{GCC_COMPARE_DEBUG} from taking effect.
13349 To verify full coverage during @option{-fcompare-debug} testing, set
13350 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13351 which GCC rejects as an invalid option in any actual compilation
13352 (rather than preprocessing, assembly or linking). To get just a
13353 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13354 not overridden} will do.
13356 @item -fcompare-debug-second
13357 @opindex fcompare-debug-second
13358 This option is implicitly passed to the compiler for the second
13359 compilation requested by @option{-fcompare-debug}, along with options to
13360 silence warnings, and omitting other options that would cause
13361 side-effect compiler outputs to files or to the standard output. Dump
13362 files and preserved temporary files are renamed so as to contain the
13363 @code{.gk} additional extension during the second compilation, to avoid
13364 overwriting those generated by the first.
13366 When this option is passed to the compiler driver, it causes the
13367 @emph{first} compilation to be skipped, which makes it useful for little
13368 other than debugging the compiler proper.
13372 Turn off generation of debug info, if leaving out this option
13373 generates it, or turn it on at level 2 otherwise. The position of this
13374 argument in the command line does not matter; it takes effect after all
13375 other options are processed, and it does so only once, no matter how
13376 many times it is given. This is mainly intended to be used with
13377 @option{-fcompare-debug}.
13379 @item -fvar-tracking-assignments-toggle
13380 @opindex fvar-tracking-assignments-toggle
13381 @opindex fno-var-tracking-assignments-toggle
13382 Toggle @option{-fvar-tracking-assignments}, in the same way that
13383 @option{-gtoggle} toggles @option{-g}.
13387 Makes the compiler print out each function name as it is compiled, and
13388 print some statistics about each pass when it finishes.
13390 @item -ftime-report
13391 @opindex ftime-report
13392 Makes the compiler print some statistics about the time consumed by each
13393 pass when it finishes.
13395 @item -ftime-report-details
13396 @opindex ftime-report-details
13397 Record the time consumed by infrastructure parts separately for each pass.
13399 @item -fira-verbose=@var{n}
13400 @opindex fira-verbose
13401 Control the verbosity of the dump file for the integrated register allocator.
13402 The default value is 5. If the value @var{n} is greater or equal to 10,
13403 the dump output is sent to stderr using the same format as @var{n} minus 10.
13406 @opindex flto-report
13407 Prints a report with internal details on the workings of the link-time
13408 optimizer. The contents of this report vary from version to version.
13409 It is meant to be useful to GCC developers when processing object
13410 files in LTO mode (via @option{-flto}).
13412 Disabled by default.
13414 @item -flto-report-wpa
13415 @opindex flto-report-wpa
13416 Like @option{-flto-report}, but only print for the WPA phase of Link
13420 @opindex fmem-report
13421 Makes the compiler print some statistics about permanent memory
13422 allocation when it finishes.
13424 @item -fmem-report-wpa
13425 @opindex fmem-report-wpa
13426 Makes the compiler print some statistics about permanent memory
13427 allocation for the WPA phase only.
13429 @item -fpre-ipa-mem-report
13430 @opindex fpre-ipa-mem-report
13431 @item -fpost-ipa-mem-report
13432 @opindex fpost-ipa-mem-report
13433 Makes the compiler print some statistics about permanent memory
13434 allocation before or after interprocedural optimization.
13436 @item -fprofile-report
13437 @opindex fprofile-report
13438 Makes the compiler print some statistics about consistency of the
13439 (estimated) profile and effect of individual passes.
13441 @item -fstack-usage
13442 @opindex fstack-usage
13443 Makes the compiler output stack usage information for the program, on a
13444 per-function basis. The filename for the dump is made by appending
13445 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13446 the output file, if explicitly specified and it is not an executable,
13447 otherwise it is the basename of the source file. An entry is made up
13452 The name of the function.
13456 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13459 The qualifier @code{static} means that the function manipulates the stack
13460 statically: a fixed number of bytes are allocated for the frame on function
13461 entry and released on function exit; no stack adjustments are otherwise made
13462 in the function. The second field is this fixed number of bytes.
13464 The qualifier @code{dynamic} means that the function manipulates the stack
13465 dynamically: in addition to the static allocation described above, stack
13466 adjustments are made in the body of the function, for example to push/pop
13467 arguments around function calls. If the qualifier @code{bounded} is also
13468 present, the amount of these adjustments is bounded at compile time and
13469 the second field is an upper bound of the total amount of stack used by
13470 the function. If it is not present, the amount of these adjustments is
13471 not bounded at compile time and the second field only represents the
13476 Emit statistics about front-end processing at the end of the compilation.
13477 This option is supported only by the C++ front end, and
13478 the information is generally only useful to the G++ development team.
13480 @item -fdbg-cnt-list
13481 @opindex fdbg-cnt-list
13482 Print the name and the counter upper bound for all debug counters.
13485 @item -fdbg-cnt=@var{counter-value-list}
13487 Set the internal debug counter upper bound. @var{counter-value-list}
13488 is a comma-separated list of @var{name}:@var{value} pairs
13489 which sets the upper bound of each debug counter @var{name} to @var{value}.
13490 All debug counters have the initial upper bound of @code{UINT_MAX};
13491 thus @code{dbg_cnt} returns true always unless the upper bound
13492 is set by this option.
13493 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13494 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13496 @item -print-file-name=@var{library}
13497 @opindex print-file-name
13498 Print the full absolute name of the library file @var{library} that
13499 would be used when linking---and don't do anything else. With this
13500 option, GCC does not compile or link anything; it just prints the
13503 @item -print-multi-directory
13504 @opindex print-multi-directory
13505 Print the directory name corresponding to the multilib selected by any
13506 other switches present in the command line. This directory is supposed
13507 to exist in @env{GCC_EXEC_PREFIX}.
13509 @item -print-multi-lib
13510 @opindex print-multi-lib
13511 Print the mapping from multilib directory names to compiler switches
13512 that enable them. The directory name is separated from the switches by
13513 @samp{;}, and each switch starts with an @samp{@@} instead of the
13514 @samp{-}, without spaces between multiple switches. This is supposed to
13515 ease shell processing.
13517 @item -print-multi-os-directory
13518 @opindex print-multi-os-directory
13519 Print the path to OS libraries for the selected
13520 multilib, relative to some @file{lib} subdirectory. If OS libraries are
13521 present in the @file{lib} subdirectory and no multilibs are used, this is
13522 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
13523 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
13524 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
13525 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
13527 @item -print-multiarch
13528 @opindex print-multiarch
13529 Print the path to OS libraries for the selected multiarch,
13530 relative to some @file{lib} subdirectory.
13532 @item -print-prog-name=@var{program}
13533 @opindex print-prog-name
13534 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
13536 @item -print-libgcc-file-name
13537 @opindex print-libgcc-file-name
13538 Same as @option{-print-file-name=libgcc.a}.
13540 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
13541 but you do want to link with @file{libgcc.a}. You can do:
13544 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
13547 @item -print-search-dirs
13548 @opindex print-search-dirs
13549 Print the name of the configured installation directory and a list of
13550 program and library directories @command{gcc} searches---and don't do anything else.
13552 This is useful when @command{gcc} prints the error message
13553 @samp{installation problem, cannot exec cpp0: No such file or directory}.
13554 To resolve this you either need to put @file{cpp0} and the other compiler
13555 components where @command{gcc} expects to find them, or you can set the environment
13556 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
13557 Don't forget the trailing @samp{/}.
13558 @xref{Environment Variables}.
13560 @item -print-sysroot
13561 @opindex print-sysroot
13562 Print the target sysroot directory that is used during
13563 compilation. This is the target sysroot specified either at configure
13564 time or using the @option{--sysroot} option, possibly with an extra
13565 suffix that depends on compilation options. If no target sysroot is
13566 specified, the option prints nothing.
13568 @item -print-sysroot-headers-suffix
13569 @opindex print-sysroot-headers-suffix
13570 Print the suffix added to the target sysroot when searching for
13571 headers, or give an error if the compiler is not configured with such
13572 a suffix---and don't do anything else.
13575 @opindex dumpmachine
13576 Print the compiler's target machine (for example,
13577 @samp{i686-pc-linux-gnu})---and don't do anything else.
13580 @opindex dumpversion
13581 Print the compiler version (for example, @code{3.0})---and don't do
13586 Print the compiler's built-in specs---and don't do anything else. (This
13587 is used when GCC itself is being built.) @xref{Spec Files}.
13590 @node Submodel Options
13591 @section Machine-Dependent Options
13592 @cindex submodel options
13593 @cindex specifying hardware config
13594 @cindex hardware models and configurations, specifying
13595 @cindex target-dependent options
13596 @cindex machine-dependent options
13598 Each target machine supported by GCC can have its own options---for
13599 example, to allow you to compile for a particular processor variant or
13600 ABI, or to control optimizations specific to that machine. By
13601 convention, the names of machine-specific options start with
13604 Some configurations of the compiler also support additional target-specific
13605 options, usually for compatibility with other compilers on the same
13608 @c This list is ordered alphanumerically by subsection name.
13609 @c It should be the same order and spelling as these options are listed
13610 @c in Machine Dependent Options
13613 * AArch64 Options::
13614 * Adapteva Epiphany Options::
13618 * Blackfin Options::
13623 * DEC Alpha Options::
13627 * GNU/Linux Options::
13637 * MicroBlaze Options::
13640 * MN10300 Options::
13644 * Nios II Options::
13645 * Nvidia PTX Options::
13647 * picoChip Options::
13648 * PowerPC Options::
13650 * RS/6000 and PowerPC Options::
13652 * S/390 and zSeries Options::
13655 * Solaris 2 Options::
13658 * System V Options::
13659 * TILE-Gx Options::
13660 * TILEPro Options::
13665 * VxWorks Options::
13667 * x86 Windows Options::
13668 * Xstormy16 Options::
13670 * zSeries Options::
13673 @node AArch64 Options
13674 @subsection AArch64 Options
13675 @cindex AArch64 Options
13677 These options are defined for AArch64 implementations:
13681 @item -mabi=@var{name}
13683 Generate code for the specified data model. Permissible values
13684 are @samp{ilp32} for SysV-like data model where int, long int and pointers
13685 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
13686 but long int and pointers are 64 bits.
13688 The default depends on the specific target configuration. Note that
13689 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
13690 entire program with the same ABI, and link with a compatible set of libraries.
13693 @opindex mbig-endian
13694 Generate big-endian code. This is the default when GCC is configured for an
13695 @samp{aarch64_be-*-*} target.
13697 @item -mgeneral-regs-only
13698 @opindex mgeneral-regs-only
13699 Generate code which uses only the general-purpose registers. This will prevent
13700 the compiler from using floating-point and Advanced SIMD registers but will not
13701 impose any restrictions on the assembler.
13703 @item -mlittle-endian
13704 @opindex mlittle-endian
13705 Generate little-endian code. This is the default when GCC is configured for an
13706 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13708 @item -mcmodel=tiny
13709 @opindex mcmodel=tiny
13710 Generate code for the tiny code model. The program and its statically defined
13711 symbols must be within 1MB of each other. Programs can be statically or
13712 dynamically linked.
13714 @item -mcmodel=small
13715 @opindex mcmodel=small
13716 Generate code for the small code model. The program and its statically defined
13717 symbols must be within 4GB of each other. Programs can be statically or
13718 dynamically linked. This is the default code model.
13720 @item -mcmodel=large
13721 @opindex mcmodel=large
13722 Generate code for the large code model. This makes no assumptions about
13723 addresses and sizes of sections. Programs can be statically linked only.
13725 @item -mstrict-align
13726 @opindex mstrict-align
13727 Avoid generating memory accesses that may not be aligned on a natural object
13728 boundary as described in the architecture specification.
13730 @item -momit-leaf-frame-pointer
13731 @itemx -mno-omit-leaf-frame-pointer
13732 @opindex momit-leaf-frame-pointer
13733 @opindex mno-omit-leaf-frame-pointer
13734 Omit or keep the frame pointer in leaf functions. The former behavior is the
13737 @item -mtls-dialect=desc
13738 @opindex mtls-dialect=desc
13739 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13740 of TLS variables. This is the default.
13742 @item -mtls-dialect=traditional
13743 @opindex mtls-dialect=traditional
13744 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13747 @item -mtls-size=@var{size}
13749 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13750 This option requires binutils 2.26 or newer.
13752 @item -mfix-cortex-a53-835769
13753 @itemx -mno-fix-cortex-a53-835769
13754 @opindex mfix-cortex-a53-835769
13755 @opindex mno-fix-cortex-a53-835769
13756 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13757 This involves inserting a NOP instruction between memory instructions and
13758 64-bit integer multiply-accumulate instructions.
13760 @item -mfix-cortex-a53-843419
13761 @itemx -mno-fix-cortex-a53-843419
13762 @opindex mfix-cortex-a53-843419
13763 @opindex mno-fix-cortex-a53-843419
13764 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13765 This erratum workaround is made at link time and this will only pass the
13766 corresponding flag to the linker.
13768 @item -mlow-precision-recip-sqrt
13769 @item -mno-low-precision-recip-sqrt
13770 @opindex mlow-precision-recip-sqrt
13771 @opindex mno-low-precision-recip-sqrt
13772 Enable or disable the reciprocal square root approximation.
13773 This option only has an effect if @option{-ffast-math} or
13774 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13775 precision of reciprocal square root results to about 16 bits for
13776 single precision and to 32 bits for double precision.
13778 @item -mlow-precision-sqrt
13779 @item -mno-low-precision-sqrt
13780 @opindex -mlow-precision-sqrt
13781 @opindex -mno-low-precision-sqrt
13782 Enable or disable the square root approximation.
13783 This option only has an effect if @option{-ffast-math} or
13784 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13785 precision of square root results to about 16 bits for
13786 single precision and to 32 bits for double precision.
13787 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13789 @item -mlow-precision-div
13790 @item -mno-low-precision-div
13791 @opindex -mlow-precision-div
13792 @opindex -mno-low-precision-div
13793 Enable or disable the division approximation.
13794 This option only has an effect if @option{-ffast-math} or
13795 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13796 precision of division results to about 16 bits for
13797 single precision and to 32 bits for double precision.
13799 @item -march=@var{name}
13801 Specify the name of the target architecture and, optionally, one or
13802 more feature modifiers. This option has the form
13803 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13805 The permissible values for @var{arch} are @samp{armv8-a},
13806 @samp{armv8.1-a}, @samp{armv8.2-a} or @var{native}.
13808 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
13809 support for the ARMv8.2-A architecture extensions.
13811 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
13812 support for the ARMv8.1-A architecture extension. In particular, it
13813 enables the @samp{+crc} and @samp{+lse} features.
13815 The value @samp{native} is available on native AArch64 GNU/Linux and
13816 causes the compiler to pick the architecture of the host system. This
13817 option has no effect if the compiler is unable to recognize the
13818 architecture of the host system,
13820 The permissible values for @var{feature} are listed in the sub-section
13821 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13822 Feature Modifiers}. Where conflicting feature modifiers are
13823 specified, the right-most feature is used.
13825 GCC uses @var{name} to determine what kind of instructions it can emit
13826 when generating assembly code. If @option{-march} is specified
13827 without either of @option{-mtune} or @option{-mcpu} also being
13828 specified, the code is tuned to perform well across a range of target
13829 processors implementing the target architecture.
13831 @item -mtune=@var{name}
13833 Specify the name of the target processor for which GCC should tune the
13834 performance of the code. Permissible values for this option are:
13835 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
13836 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{qdf24xx},
13837 @samp{thunderx}, @samp{xgene1}, @samp{vulcan}, @samp{cortex-a57.cortex-a53},
13838 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
13839 @samp{cortex-a73.cortex-a53}, @samp{native}.
13841 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
13842 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
13843 specify that GCC should tune for a big.LITTLE system.
13845 Additionally on native AArch64 GNU/Linux systems the value
13846 @samp{native} tunes performance to the host system. This option has no effect
13847 if the compiler is unable to recognize the processor of the host system.
13849 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13850 are specified, the code is tuned to perform well across a range
13851 of target processors.
13853 This option cannot be suffixed by feature modifiers.
13855 @item -mcpu=@var{name}
13857 Specify the name of the target processor, optionally suffixed by one
13858 or more feature modifiers. This option has the form
13859 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13860 the permissible values for @var{cpu} are the same as those available
13861 for @option{-mtune}. The permissible values for @var{feature} are
13862 documented in the sub-section on
13863 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13864 Feature Modifiers}. Where conflicting feature modifiers are
13865 specified, the right-most feature is used.
13867 GCC uses @var{name} to determine what kind of instructions it can emit when
13868 generating assembly code (as if by @option{-march}) and to determine
13869 the target processor for which to tune for performance (as if
13870 by @option{-mtune}). Where this option is used in conjunction
13871 with @option{-march} or @option{-mtune}, those options take precedence
13872 over the appropriate part of this option.
13874 @item -moverride=@var{string}
13876 Override tuning decisions made by the back-end in response to a
13877 @option{-mtune=} switch. The syntax, semantics, and accepted values
13878 for @var{string} in this option are not guaranteed to be consistent
13881 This option is only intended to be useful when developing GCC.
13883 @item -mpc-relative-literal-loads
13884 @opindex mpc-relative-literal-loads
13885 Enable PC-relative literal loads. With this option literal pools are
13886 accessed using a single instruction and emitted after each function. This
13887 limits the maximum size of functions to 1MB. This is enabled by default for
13888 @option{-mcmodel=tiny}.
13892 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
13893 @anchor{aarch64-feature-modifiers}
13894 @cindex @option{-march} feature modifiers
13895 @cindex @option{-mcpu} feature modifiers
13896 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
13897 the following and their inverses @option{no@var{feature}}:
13901 Enable CRC extension. This is on by default for
13902 @option{-march=armv8.1-a}.
13904 Enable Crypto extension. This also enables Advanced SIMD and floating-point
13907 Enable floating-point instructions. This is on by default for all possible
13908 values for options @option{-march} and @option{-mcpu}.
13910 Enable Advanced SIMD instructions. This also enables floating-point
13911 instructions. This is on by default for all possible values for options
13912 @option{-march} and @option{-mcpu}.
13914 Enable Large System Extension instructions. This is on by default for
13915 @option{-march=armv8.1-a}.
13917 Enable FP16 extension. This also enables floating-point instructions.
13921 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
13922 Conversely, @option{nofp} implies @option{nosimd}, which implies
13925 @node Adapteva Epiphany Options
13926 @subsection Adapteva Epiphany Options
13928 These @samp{-m} options are defined for Adapteva Epiphany:
13931 @item -mhalf-reg-file
13932 @opindex mhalf-reg-file
13933 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
13934 That allows code to run on hardware variants that lack these registers.
13936 @item -mprefer-short-insn-regs
13937 @opindex mprefer-short-insn-regs
13938 Preferentially allocate registers that allow short instruction generation.
13939 This can result in increased instruction count, so this may either reduce or
13940 increase overall code size.
13942 @item -mbranch-cost=@var{num}
13943 @opindex mbranch-cost
13944 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13945 This cost is only a heuristic and is not guaranteed to produce
13946 consistent results across releases.
13950 Enable the generation of conditional moves.
13952 @item -mnops=@var{num}
13954 Emit @var{num} NOPs before every other generated instruction.
13956 @item -mno-soft-cmpsf
13957 @opindex mno-soft-cmpsf
13958 For single-precision floating-point comparisons, emit an @code{fsub} instruction
13959 and test the flags. This is faster than a software comparison, but can
13960 get incorrect results in the presence of NaNs, or when two different small
13961 numbers are compared such that their difference is calculated as zero.
13962 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
13963 software comparisons.
13965 @item -mstack-offset=@var{num}
13966 @opindex mstack-offset
13967 Set the offset between the top of the stack and the stack pointer.
13968 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
13969 can be used by leaf functions without stack allocation.
13970 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
13971 Note also that this option changes the ABI; compiling a program with a
13972 different stack offset than the libraries have been compiled with
13973 generally does not work.
13974 This option can be useful if you want to evaluate if a different stack
13975 offset would give you better code, but to actually use a different stack
13976 offset to build working programs, it is recommended to configure the
13977 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
13979 @item -mno-round-nearest
13980 @opindex mno-round-nearest
13981 Make the scheduler assume that the rounding mode has been set to
13982 truncating. The default is @option{-mround-nearest}.
13985 @opindex mlong-calls
13986 If not otherwise specified by an attribute, assume all calls might be beyond
13987 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
13988 function address into a register before performing a (otherwise direct) call.
13989 This is the default.
13991 @item -mshort-calls
13992 @opindex short-calls
13993 If not otherwise specified by an attribute, assume all direct calls are
13994 in the range of the @code{b} / @code{bl} instructions, so use these instructions
13995 for direct calls. The default is @option{-mlong-calls}.
13999 Assume addresses can be loaded as 16-bit unsigned values. This does not
14000 apply to function addresses for which @option{-mlong-calls} semantics
14003 @item -mfp-mode=@var{mode}
14005 Set the prevailing mode of the floating-point unit.
14006 This determines the floating-point mode that is provided and expected
14007 at function call and return time. Making this mode match the mode you
14008 predominantly need at function start can make your programs smaller and
14009 faster by avoiding unnecessary mode switches.
14011 @var{mode} can be set to one the following values:
14015 Any mode at function entry is valid, and retained or restored when
14016 the function returns, and when it calls other functions.
14017 This mode is useful for compiling libraries or other compilation units
14018 you might want to incorporate into different programs with different
14019 prevailing FPU modes, and the convenience of being able to use a single
14020 object file outweighs the size and speed overhead for any extra
14021 mode switching that might be needed, compared with what would be needed
14022 with a more specific choice of prevailing FPU mode.
14025 This is the mode used for floating-point calculations with
14026 truncating (i.e.@: round towards zero) rounding mode. That includes
14027 conversion from floating point to integer.
14029 @item round-nearest
14030 This is the mode used for floating-point calculations with
14031 round-to-nearest-or-even rounding mode.
14034 This is the mode used to perform integer calculations in the FPU, e.g.@:
14035 integer multiply, or integer multiply-and-accumulate.
14038 The default is @option{-mfp-mode=caller}
14040 @item -mnosplit-lohi
14041 @itemx -mno-postinc
14042 @itemx -mno-postmodify
14043 @opindex mnosplit-lohi
14044 @opindex mno-postinc
14045 @opindex mno-postmodify
14046 Code generation tweaks that disable, respectively, splitting of 32-bit
14047 loads, generation of post-increment addresses, and generation of
14048 post-modify addresses. The defaults are @option{msplit-lohi},
14049 @option{-mpost-inc}, and @option{-mpost-modify}.
14051 @item -mnovect-double
14052 @opindex mno-vect-double
14053 Change the preferred SIMD mode to SImode. The default is
14054 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14056 @item -max-vect-align=@var{num}
14057 @opindex max-vect-align
14058 The maximum alignment for SIMD vector mode types.
14059 @var{num} may be 4 or 8. The default is 8.
14060 Note that this is an ABI change, even though many library function
14061 interfaces are unaffected if they don't use SIMD vector modes
14062 in places that affect size and/or alignment of relevant types.
14064 @item -msplit-vecmove-early
14065 @opindex msplit-vecmove-early
14066 Split vector moves into single word moves before reload. In theory this
14067 can give better register allocation, but so far the reverse seems to be
14068 generally the case.
14070 @item -m1reg-@var{reg}
14072 Specify a register to hold the constant @minus{}1, which makes loading small negative
14073 constants and certain bitmasks faster.
14074 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14075 which specify use of that register as a fixed register,
14076 and @samp{none}, which means that no register is used for this
14077 purpose. The default is @option{-m1reg-none}.
14082 @subsection ARC Options
14083 @cindex ARC options
14085 The following options control the architecture variant for which code
14088 @c architecture variants
14091 @item -mbarrel-shifter
14092 @opindex mbarrel-shifter
14093 Generate instructions supported by barrel shifter. This is the default
14094 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14096 @item -mcpu=@var{cpu}
14098 Set architecture type, register usage, and instruction scheduling
14099 parameters for @var{cpu}. There are also shortcut alias options
14100 available for backward compatibility and convenience. Supported
14101 values for @var{cpu} are
14108 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14113 Compile for ARC601. Alias: @option{-mARC601}.
14119 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14120 This is the default when configured with @option{--with-cpu=arc700}@.
14124 Compile for ARC EM.
14128 Compile for ARC HS.
14133 @itemx -mdpfp-compact
14134 @opindex mdpfp-compact
14135 FPX: Generate Double Precision FPX instructions, tuned for the compact
14139 @opindex mdpfp-fast
14140 FPX: Generate Double Precision FPX instructions, tuned for the fast
14143 @item -mno-dpfp-lrsr
14144 @opindex mno-dpfp-lrsr
14145 Disable LR and SR instructions from using FPX extension aux registers.
14149 Generate Extended arithmetic instructions. Currently only
14150 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14151 supported. This is always enabled for @option{-mcpu=ARC700}.
14155 Do not generate mpy instructions for ARC700.
14159 Generate 32x16 bit multiply and mac instructions.
14163 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
14167 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
14172 @itemx -mspfp-compact
14173 @opindex mspfp-compact
14174 FPX: Generate Single Precision FPX instructions, tuned for the compact
14178 @opindex mspfp-fast
14179 FPX: Generate Single Precision FPX instructions, tuned for the fast
14184 Enable generation of ARC SIMD instructions via target-specific
14185 builtins. Only valid for @option{-mcpu=ARC700}.
14188 @opindex msoft-float
14189 This option ignored; it is provided for compatibility purposes only.
14190 Software floating point code is emitted by default, and this default
14191 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
14192 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
14193 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
14197 Generate swap instructions.
14201 This enables Locked Load/Store Conditional extension to implement
14202 atomic memopry built-in functions. Not available for ARC 6xx or ARC
14207 Enable DIV/REM instructions for ARCv2 cores.
14209 @item -mcode-density
14210 @opindex mcode-density
14211 Enable code density instructions for ARC EM, default on for ARC HS.
14215 Enable double load/store operations for ARC HS cores.
14217 @item -mtp-regno=@var{regno}
14219 Specify thread pointer register number.
14221 @item -mmpy-option=@var{multo}
14222 @opindex mmpy-option
14223 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
14224 the default value. The recognized values for @var{multo} are:
14228 No multiplier available.
14232 The multiply option is set to w: 16x16 multiplier, fully pipelined.
14233 The following instructions are enabled: MPYW, and MPYUW.
14237 The multiply option is set to wlh1: 32x32 multiplier, fully
14238 pipelined (1 stage). The following instructions are additionally
14239 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
14243 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
14244 (2 stages). The following instructions are additionally enabled: MPY,
14245 MPYU, MPYM, MPYMU, and MPY_S.
14249 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
14250 sequential. The following instructions are additionally enabled: MPY,
14251 MPYU, MPYM, MPYMU, and MPY_S.
14255 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
14256 sequential. The following instructions are additionally enabled: MPY,
14257 MPYU, MPYM, MPYMU, and MPY_S.
14261 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
14262 sequential. The following instructions are additionally enabled: MPY,
14263 MPYU, MPYM, MPYMU, and MPY_S.
14267 This option is only available for ARCv2 cores@.
14269 @item -mfpu=@var{fpu}
14271 Enables specific floating-point hardware extension for ARCv2
14272 core. Supported values for @var{fpu} are:
14278 Enables support for single precision floating point hardware
14283 Enables support for double precision floating point hardware
14284 extensions. The single precision floating point extension is also
14285 enabled. Not available for ARC EM@.
14289 Enables support for double precision floating point hardware
14290 extensions using double precision assist instructions. The single
14291 precision floating point extension is also enabled. This option is
14292 only available for ARC EM@.
14296 Enables support for double precision floating point hardware
14297 extensions using double precision assist instructions, and simple
14298 precision square-root and divide hardware extensions. The single
14299 precision floating point extension is also enabled. This option is
14300 only available for ARC EM@.
14304 Enables support for double precision floating point hardware
14305 extensions using double precision assist instructions, and simple
14306 precision fused multiple and add hardware extension. The single
14307 precision floating point extension is also enabled. This option is
14308 only available for ARC EM@.
14312 Enables support for double precision floating point hardware
14313 extensions using double precision assist instructions, and all simple
14314 precision hardware extensions. The single precision floating point
14315 extension is also enabled. This option is only available for ARC EM@.
14319 Enables support for single precision floating point, and single
14320 precision square-root and divide hardware extensions@.
14324 Enables support for double precision floating point, and double
14325 precision square-root and divide hardware extensions. This option
14326 includes option @samp{fpus_div}. Not available for ARC EM@.
14330 Enables support for single precision floating point, and single
14331 precision fused multiple and add hardware extensions@.
14335 Enables support for double precision floating point, and double
14336 precision fused multiple and add hardware extensions. This option
14337 includes option @samp{fpus_fma}. Not available for ARC EM@.
14341 Enables support for all single precision floating point hardware
14346 Enables support for all single and double precision floating point
14347 hardware extensions. Not available for ARC EM@.
14353 The following options are passed through to the assembler, and also
14354 define preprocessor macro symbols.
14356 @c Flags used by the assembler, but for which we define preprocessor
14357 @c macro symbols as well.
14360 @opindex mdsp-packa
14361 Passed down to the assembler to enable the DSP Pack A extensions.
14362 Also sets the preprocessor symbol @code{__Xdsp_packa}.
14366 Passed down to the assembler to enable the dual viterbi butterfly
14367 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
14369 @c ARC700 4.10 extension instruction
14372 Passed down to the assembler to enable the Locked Load/Store
14373 Conditional extension. Also sets the preprocessor symbol
14378 Passed down to the assembler. Also sets the preprocessor symbol
14379 @code{__Xxmac_d16}.
14383 Passed down to the assembler. Also sets the preprocessor symbol
14386 @c ARC700 4.10 extension instruction
14389 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
14390 extension instruction. Also sets the preprocessor symbol
14393 @c ARC700 4.10 extension instruction
14396 Passed down to the assembler to enable the swap byte ordering
14397 extension instruction. Also sets the preprocessor symbol
14401 @opindex mtelephony
14402 Passed down to the assembler to enable dual and single operand
14403 instructions for telephony. Also sets the preprocessor symbol
14404 @code{__Xtelephony}.
14408 Passed down to the assembler to enable the XY Memory extension. Also
14409 sets the preprocessor symbol @code{__Xxy}.
14413 The following options control how the assembly code is annotated:
14415 @c Assembly annotation options
14419 Annotate assembler instructions with estimated addresses.
14421 @item -mannotate-align
14422 @opindex mannotate-align
14423 Explain what alignment considerations lead to the decision to make an
14424 instruction short or long.
14428 The following options are passed through to the linker:
14430 @c options passed through to the linker
14434 Passed through to the linker, to specify use of the @code{arclinux} emulation.
14435 This option is enabled by default in tool chains built for
14436 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
14437 when profiling is not requested.
14439 @item -marclinux_prof
14440 @opindex marclinux_prof
14441 Passed through to the linker, to specify use of the
14442 @code{arclinux_prof} emulation. This option is enabled by default in
14443 tool chains built for @w{@code{arc-linux-uclibc}} and
14444 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
14448 The following options control the semantics of generated code:
14450 @c semantically relevant code generation options
14453 @opindex mlong-calls
14454 Generate call insns as register indirect calls, thus providing access
14455 to the full 32-bit address range.
14457 @item -mmedium-calls
14458 @opindex mmedium-calls
14459 Don't use less than 25 bit addressing range for calls, which is the
14460 offset available for an unconditional branch-and-link
14461 instruction. Conditional execution of function calls is suppressed, to
14462 allow use of the 25-bit range, rather than the 21-bit range with
14463 conditional branch-and-link. This is the default for tool chains built
14464 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
14468 Do not generate sdata references. This is the default for tool chains
14469 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
14473 @opindex mucb-mcount
14474 Instrument with mcount calls as used in UCB code. I.e. do the
14475 counting in the callee, not the caller. By default ARC instrumentation
14476 counts in the caller.
14478 @item -mvolatile-cache
14479 @opindex mvolatile-cache
14480 Use ordinarily cached memory accesses for volatile references. This is the
14483 @item -mno-volatile-cache
14484 @opindex mno-volatile-cache
14485 Enable cache bypass for volatile references.
14489 The following options fine tune code generation:
14490 @c code generation tuning options
14493 @opindex malign-call
14494 Do alignment optimizations for call instructions.
14496 @item -mauto-modify-reg
14497 @opindex mauto-modify-reg
14498 Enable the use of pre/post modify with register displacement.
14500 @item -mbbit-peephole
14501 @opindex mbbit-peephole
14502 Enable bbit peephole2.
14506 This option disables a target-specific pass in @file{arc_reorg} to
14507 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
14508 generation driven by the combiner pass.
14510 @item -mcase-vector-pcrel
14511 @opindex mcase-vector-pcrel
14512 Use pc-relative switch case tables - this enables case table shortening.
14513 This is the default for @option{-Os}.
14515 @item -mcompact-casesi
14516 @opindex mcompact-casesi
14517 Enable compact casesi pattern. This is the default for @option{-Os},
14518 and only available for ARCv1 cores.
14520 @item -mno-cond-exec
14521 @opindex mno-cond-exec
14522 Disable ARCompact specific pass to generate conditional execution instructions.
14523 Due to delay slot scheduling and interactions between operand numbers,
14524 literal sizes, instruction lengths, and the support for conditional execution,
14525 the target-independent pass to generate conditional execution is often lacking,
14526 so the ARC port has kept a special pass around that tries to find more
14527 conditional execution generating opportunities after register allocation,
14528 branch shortening, and delay slot scheduling have been done. This pass
14529 generally, but not always, improves performance and code size, at the cost of
14530 extra compilation time, which is why there is an option to switch it off.
14531 If you have a problem with call instructions exceeding their allowable
14532 offset range because they are conditionalized, you should consider using
14533 @option{-mmedium-calls} instead.
14535 @item -mearly-cbranchsi
14536 @opindex mearly-cbranchsi
14537 Enable pre-reload use of the cbranchsi pattern.
14539 @item -mexpand-adddi
14540 @opindex mexpand-adddi
14541 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
14542 @code{add.f}, @code{adc} etc.
14544 @item -mindexed-loads
14545 @opindex mindexed-loads
14546 Enable the use of indexed loads. This can be problematic because some
14547 optimizers then assume that indexed stores exist, which is not
14551 Enable Local Register Allocation. This is still experimental for ARC,
14552 so by default the compiler uses standard reload
14553 (i.e. @option{-mno-lra}).
14555 @item -mlra-priority-none
14556 @opindex mlra-priority-none
14557 Don't indicate any priority for target registers.
14559 @item -mlra-priority-compact
14560 @opindex mlra-priority-compact
14561 Indicate target register priority for r0..r3 / r12..r15.
14563 @item -mlra-priority-noncompact
14564 @opindex mlra-priority-noncompact
14565 Reduce target register priority for r0..r3 / r12..r15.
14567 @item -mno-millicode
14568 @opindex mno-millicode
14569 When optimizing for size (using @option{-Os}), prologues and epilogues
14570 that have to save or restore a large number of registers are often
14571 shortened by using call to a special function in libgcc; this is
14572 referred to as a @emph{millicode} call. As these calls can pose
14573 performance issues, and/or cause linking issues when linking in a
14574 nonstandard way, this option is provided to turn off millicode call
14578 @opindex mmixed-code
14579 Tweak register allocation to help 16-bit instruction generation.
14580 This generally has the effect of decreasing the average instruction size
14581 while increasing the instruction count.
14585 Enable 'q' instruction alternatives.
14586 This is the default for @option{-Os}.
14590 Enable Rcq constraint handling - most short code generation depends on this.
14591 This is the default.
14595 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
14596 This is the default.
14598 @item -msize-level=@var{level}
14599 @opindex msize-level
14600 Fine-tune size optimization with regards to instruction lengths and alignment.
14601 The recognized values for @var{level} are:
14604 No size optimization. This level is deprecated and treated like @samp{1}.
14607 Short instructions are used opportunistically.
14610 In addition, alignment of loops and of code after barriers are dropped.
14613 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
14617 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
14618 the behavior when this is not set is equivalent to level @samp{1}.
14620 @item -mtune=@var{cpu}
14622 Set instruction scheduling parameters for @var{cpu}, overriding any implied
14623 by @option{-mcpu=}.
14625 Supported values for @var{cpu} are
14629 Tune for ARC600 cpu.
14632 Tune for ARC601 cpu.
14635 Tune for ARC700 cpu with standard multiplier block.
14638 Tune for ARC700 cpu with XMAC block.
14641 Tune for ARC725D cpu.
14644 Tune for ARC750D cpu.
14648 @item -mmultcost=@var{num}
14650 Cost to assume for a multiply instruction, with @samp{4} being equal to a
14651 normal instruction.
14653 @item -munalign-prob-threshold=@var{probability}
14654 @opindex munalign-prob-threshold
14655 Set probability threshold for unaligning branches.
14656 When tuning for @samp{ARC700} and optimizing for speed, branches without
14657 filled delay slot are preferably emitted unaligned and long, unless
14658 profiling indicates that the probability for the branch to be taken
14659 is below @var{probability}. @xref{Cross-profiling}.
14660 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
14664 The following options are maintained for backward compatibility, but
14665 are now deprecated and will be removed in a future release:
14667 @c Deprecated options
14675 @opindex mbig-endian
14678 Compile code for big endian targets. Use of these options is now
14679 deprecated. Users wanting big-endian code, should use the
14680 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
14681 building the tool chain, for which big-endian is the default.
14683 @item -mlittle-endian
14684 @opindex mlittle-endian
14687 Compile code for little endian targets. Use of these options is now
14688 deprecated. Users wanting little-endian code should use the
14689 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
14690 building the tool chain, for which little-endian is the default.
14692 @item -mbarrel_shifter
14693 @opindex mbarrel_shifter
14694 Replaced by @option{-mbarrel-shifter}.
14696 @item -mdpfp_compact
14697 @opindex mdpfp_compact
14698 Replaced by @option{-mdpfp-compact}.
14701 @opindex mdpfp_fast
14702 Replaced by @option{-mdpfp-fast}.
14705 @opindex mdsp_packa
14706 Replaced by @option{-mdsp-packa}.
14710 Replaced by @option{-mea}.
14714 Replaced by @option{-mmac-24}.
14718 Replaced by @option{-mmac-d16}.
14720 @item -mspfp_compact
14721 @opindex mspfp_compact
14722 Replaced by @option{-mspfp-compact}.
14725 @opindex mspfp_fast
14726 Replaced by @option{-mspfp-fast}.
14728 @item -mtune=@var{cpu}
14730 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14731 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14732 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
14734 @item -multcost=@var{num}
14736 Replaced by @option{-mmultcost}.
14741 @subsection ARM Options
14742 @cindex ARM options
14744 These @samp{-m} options are defined for the ARM port:
14747 @item -mabi=@var{name}
14749 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
14750 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
14753 @opindex mapcs-frame
14754 Generate a stack frame that is compliant with the ARM Procedure Call
14755 Standard for all functions, even if this is not strictly necessary for
14756 correct execution of the code. Specifying @option{-fomit-frame-pointer}
14757 with this option causes the stack frames not to be generated for
14758 leaf functions. The default is @option{-mno-apcs-frame}.
14759 This option is deprecated.
14763 This is a synonym for @option{-mapcs-frame} and is deprecated.
14766 @c not currently implemented
14767 @item -mapcs-stack-check
14768 @opindex mapcs-stack-check
14769 Generate code to check the amount of stack space available upon entry to
14770 every function (that actually uses some stack space). If there is
14771 insufficient space available then either the function
14772 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
14773 called, depending upon the amount of stack space required. The runtime
14774 system is required to provide these functions. The default is
14775 @option{-mno-apcs-stack-check}, since this produces smaller code.
14777 @c not currently implemented
14779 @opindex mapcs-float
14780 Pass floating-point arguments using the floating-point registers. This is
14781 one of the variants of the APCS@. This option is recommended if the
14782 target hardware has a floating-point unit or if a lot of floating-point
14783 arithmetic is going to be performed by the code. The default is
14784 @option{-mno-apcs-float}, since the size of integer-only code is
14785 slightly increased if @option{-mapcs-float} is used.
14787 @c not currently implemented
14788 @item -mapcs-reentrant
14789 @opindex mapcs-reentrant
14790 Generate reentrant, position-independent code. The default is
14791 @option{-mno-apcs-reentrant}.
14794 @item -mthumb-interwork
14795 @opindex mthumb-interwork
14796 Generate code that supports calling between the ARM and Thumb
14797 instruction sets. Without this option, on pre-v5 architectures, the
14798 two instruction sets cannot be reliably used inside one program. The
14799 default is @option{-mno-thumb-interwork}, since slightly larger code
14800 is generated when @option{-mthumb-interwork} is specified. In AAPCS
14801 configurations this option is meaningless.
14803 @item -mno-sched-prolog
14804 @opindex mno-sched-prolog
14805 Prevent the reordering of instructions in the function prologue, or the
14806 merging of those instruction with the instructions in the function's
14807 body. This means that all functions start with a recognizable set
14808 of instructions (or in fact one of a choice from a small set of
14809 different function prologues), and this information can be used to
14810 locate the start of functions inside an executable piece of code. The
14811 default is @option{-msched-prolog}.
14813 @item -mfloat-abi=@var{name}
14814 @opindex mfloat-abi
14815 Specifies which floating-point ABI to use. Permissible values
14816 are: @samp{soft}, @samp{softfp} and @samp{hard}.
14818 Specifying @samp{soft} causes GCC to generate output containing
14819 library calls for floating-point operations.
14820 @samp{softfp} allows the generation of code using hardware floating-point
14821 instructions, but still uses the soft-float calling conventions.
14822 @samp{hard} allows generation of floating-point instructions
14823 and uses FPU-specific calling conventions.
14825 The default depends on the specific target configuration. Note that
14826 the hard-float and soft-float ABIs are not link-compatible; you must
14827 compile your entire program with the same ABI, and link with a
14828 compatible set of libraries.
14830 @item -mlittle-endian
14831 @opindex mlittle-endian
14832 Generate code for a processor running in little-endian mode. This is
14833 the default for all standard configurations.
14836 @opindex mbig-endian
14837 Generate code for a processor running in big-endian mode; the default is
14838 to compile code for a little-endian processor.
14840 @item -march=@var{name}
14842 This specifies the name of the target ARM architecture. GCC uses this
14843 name to determine what kind of instructions it can emit when generating
14844 assembly code. This option can be used in conjunction with or instead
14845 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
14846 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
14847 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
14848 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
14849 @samp{armv6kz}, @samp{armv6s-m},
14850 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
14851 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
14852 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
14853 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
14854 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
14856 Architecture revisions older than @samp{armv4t} are deprecated.
14858 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
14859 the (now mandatory) SVC instruction.
14861 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
14864 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
14867 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
14868 architecture together with the optional CRC32 extensions.
14870 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
14871 architecture. This also enables the features provided by
14872 @option{-march=armv8-a+crc}.
14874 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
14875 architecture. This also enables the features provided by
14876 @option{-march=armv8.1-a}.
14878 @option{-march=armv8.2-a+fp16} enables compiler support for the
14879 ARMv8.2-A architecture with the optional FP16 instructions extension.
14880 This also enables the features provided by @option{-march=armv8.1-a}
14881 and implies @option{-mfp16-format=ieee}.
14883 @option{-march=native} causes the compiler to auto-detect the architecture
14884 of the build computer. At present, this feature is only supported on
14885 GNU/Linux, and not all architectures are recognized. If the auto-detect
14886 is unsuccessful the option has no effect.
14888 @item -mtune=@var{name}
14890 This option specifies the name of the target ARM processor for
14891 which GCC should tune the performance of the code.
14892 For some ARM implementations better performance can be obtained by using
14894 Permissible names are: @samp{arm2}, @samp{arm250},
14895 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
14896 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
14897 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
14898 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
14900 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
14901 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
14902 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
14903 @samp{strongarm1110},
14904 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
14905 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
14906 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
14907 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
14908 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
14909 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
14910 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
14911 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
14912 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
14913 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14914 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
14915 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
14921 @samp{cortex-m0plus},
14922 @samp{cortex-m1.small-multiply},
14923 @samp{cortex-m0.small-multiply},
14924 @samp{cortex-m0plus.small-multiply},
14927 @samp{marvell-pj4},
14928 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
14929 @samp{fa526}, @samp{fa626},
14930 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
14933 Additionally, this option can specify that GCC should tune the performance
14934 of the code for a big.LITTLE system. Permissible names are:
14935 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
14936 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14937 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
14939 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
14940 performance for a blend of processors within architecture @var{arch}.
14941 The aim is to generate code that run well on the current most popular
14942 processors, balancing between optimizations that benefit some CPUs in the
14943 range, and avoiding performance pitfalls of other CPUs. The effects of
14944 this option may change in future GCC versions as CPU models come and go.
14946 @option{-mtune=native} causes the compiler to auto-detect the CPU
14947 of the build computer. At present, this feature is only supported on
14948 GNU/Linux, and not all architectures are recognized. If the auto-detect is
14949 unsuccessful the option has no effect.
14951 @item -mcpu=@var{name}
14953 This specifies the name of the target ARM processor. GCC uses this name
14954 to derive the name of the target ARM architecture (as if specified
14955 by @option{-march}) and the ARM processor type for which to tune for
14956 performance (as if specified by @option{-mtune}). Where this option
14957 is used in conjunction with @option{-march} or @option{-mtune},
14958 those options take precedence over the appropriate part of this option.
14960 Permissible names for this option are the same as those for
14963 @option{-mcpu=generic-@var{arch}} is also permissible, and is
14964 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
14965 See @option{-mtune} for more information.
14967 @option{-mcpu=native} causes the compiler to auto-detect the CPU
14968 of the build computer. At present, this feature is only supported on
14969 GNU/Linux, and not all architectures are recognized. If the auto-detect
14970 is unsuccessful the option has no effect.
14972 @item -mfpu=@var{name}
14974 This specifies what floating-point hardware (or hardware emulation) is
14975 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
14976 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
14977 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
14978 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
14979 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
14980 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
14982 If @option{-msoft-float} is specified this specifies the format of
14983 floating-point values.
14985 If the selected floating-point hardware includes the NEON extension
14986 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
14987 operations are not generated by GCC's auto-vectorization pass unless
14988 @option{-funsafe-math-optimizations} is also specified. This is
14989 because NEON hardware does not fully implement the IEEE 754 standard for
14990 floating-point arithmetic (in particular denormal values are treated as
14991 zero), so the use of NEON instructions may lead to a loss of precision.
14993 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}).
14995 @item -mfp16-format=@var{name}
14996 @opindex mfp16-format
14997 Specify the format of the @code{__fp16} half-precision floating-point type.
14998 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
14999 the default is @samp{none}, in which case the @code{__fp16} type is not
15000 defined. @xref{Half-Precision}, for more information.
15002 @item -mstructure-size-boundary=@var{n}
15003 @opindex mstructure-size-boundary
15004 The sizes of all structures and unions are rounded up to a multiple
15005 of the number of bits set by this option. Permissible values are 8, 32
15006 and 64. The default value varies for different toolchains. For the COFF
15007 targeted toolchain the default value is 8. A value of 64 is only allowed
15008 if the underlying ABI supports it.
15010 Specifying a larger number can produce faster, more efficient code, but
15011 can also increase the size of the program. Different values are potentially
15012 incompatible. Code compiled with one value cannot necessarily expect to
15013 work with code or libraries compiled with another value, if they exchange
15014 information using structures or unions.
15016 @item -mabort-on-noreturn
15017 @opindex mabort-on-noreturn
15018 Generate a call to the function @code{abort} at the end of a
15019 @code{noreturn} function. It is executed if the function tries to
15023 @itemx -mno-long-calls
15024 @opindex mlong-calls
15025 @opindex mno-long-calls
15026 Tells the compiler to perform function calls by first loading the
15027 address of the function into a register and then performing a subroutine
15028 call on this register. This switch is needed if the target function
15029 lies outside of the 64-megabyte addressing range of the offset-based
15030 version of subroutine call instruction.
15032 Even if this switch is enabled, not all function calls are turned
15033 into long calls. The heuristic is that static functions, functions
15034 that have the @code{short_call} attribute, functions that are inside
15035 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15036 definitions have already been compiled within the current compilation
15037 unit are not turned into long calls. The exceptions to this rule are
15038 that weak function definitions, functions with the @code{long_call}
15039 attribute or the @code{section} attribute, and functions that are within
15040 the scope of a @code{#pragma long_calls} directive are always
15041 turned into long calls.
15043 This feature is not enabled by default. Specifying
15044 @option{-mno-long-calls} restores the default behavior, as does
15045 placing the function calls within the scope of a @code{#pragma
15046 long_calls_off} directive. Note these switches have no effect on how
15047 the compiler generates code to handle function calls via function
15050 @item -msingle-pic-base
15051 @opindex msingle-pic-base
15052 Treat the register used for PIC addressing as read-only, rather than
15053 loading it in the prologue for each function. The runtime system is
15054 responsible for initializing this register with an appropriate value
15055 before execution begins.
15057 @item -mpic-register=@var{reg}
15058 @opindex mpic-register
15059 Specify the register to be used for PIC addressing.
15060 For standard PIC base case, the default is any suitable register
15061 determined by compiler. For single PIC base case, the default is
15062 @samp{R9} if target is EABI based or stack-checking is enabled,
15063 otherwise the default is @samp{R10}.
15065 @item -mpic-data-is-text-relative
15066 @opindex mpic-data-is-text-relative
15067 Assume that the displacement between the text and data segments is fixed
15068 at static link time. This permits using PC-relative addressing
15069 operations to access data known to be in the data segment. For
15070 non-VxWorks RTP targets, this option is enabled by default. When
15071 disabled on such targets, it will enable @option{-msingle-pic-base} by
15074 @item -mpoke-function-name
15075 @opindex mpoke-function-name
15076 Write the name of each function into the text section, directly
15077 preceding the function prologue. The generated code is similar to this:
15081 .ascii "arm_poke_function_name", 0
15084 .word 0xff000000 + (t1 - t0)
15085 arm_poke_function_name
15087 stmfd sp!, @{fp, ip, lr, pc@}
15091 When performing a stack backtrace, code can inspect the value of
15092 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
15093 location @code{pc - 12} and the top 8 bits are set, then we know that
15094 there is a function name embedded immediately preceding this location
15095 and has length @code{((pc[-3]) & 0xff000000)}.
15102 Select between generating code that executes in ARM and Thumb
15103 states. The default for most configurations is to generate code
15104 that executes in ARM state, but the default can be changed by
15105 configuring GCC with the @option{--with-mode=}@var{state}
15108 You can also override the ARM and Thumb mode for each function
15109 by using the @code{target("thumb")} and @code{target("arm")} function attributes
15110 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
15113 @opindex mtpcs-frame
15114 Generate a stack frame that is compliant with the Thumb Procedure Call
15115 Standard for all non-leaf functions. (A leaf function is one that does
15116 not call any other functions.) The default is @option{-mno-tpcs-frame}.
15118 @item -mtpcs-leaf-frame
15119 @opindex mtpcs-leaf-frame
15120 Generate a stack frame that is compliant with the Thumb Procedure Call
15121 Standard for all leaf functions. (A leaf function is one that does
15122 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
15124 @item -mcallee-super-interworking
15125 @opindex mcallee-super-interworking
15126 Gives all externally visible functions in the file being compiled an ARM
15127 instruction set header which switches to Thumb mode before executing the
15128 rest of the function. This allows these functions to be called from
15129 non-interworking code. This option is not valid in AAPCS configurations
15130 because interworking is enabled by default.
15132 @item -mcaller-super-interworking
15133 @opindex mcaller-super-interworking
15134 Allows calls via function pointers (including virtual functions) to
15135 execute correctly regardless of whether the target code has been
15136 compiled for interworking or not. There is a small overhead in the cost
15137 of executing a function pointer if this option is enabled. This option
15138 is not valid in AAPCS configurations because interworking is enabled
15141 @item -mtp=@var{name}
15143 Specify the access model for the thread local storage pointer. The valid
15144 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
15145 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
15146 (supported in the arm6k architecture), and @samp{auto}, which uses the
15147 best available method for the selected processor. The default setting is
15150 @item -mtls-dialect=@var{dialect}
15151 @opindex mtls-dialect
15152 Specify the dialect to use for accessing thread local storage. Two
15153 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
15154 @samp{gnu} dialect selects the original GNU scheme for supporting
15155 local and global dynamic TLS models. The @samp{gnu2} dialect
15156 selects the GNU descriptor scheme, which provides better performance
15157 for shared libraries. The GNU descriptor scheme is compatible with
15158 the original scheme, but does require new assembler, linker and
15159 library support. Initial and local exec TLS models are unaffected by
15160 this option and always use the original scheme.
15162 @item -mword-relocations
15163 @opindex mword-relocations
15164 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
15165 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
15166 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
15169 @item -mfix-cortex-m3-ldrd
15170 @opindex mfix-cortex-m3-ldrd
15171 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
15172 with overlapping destination and base registers are used. This option avoids
15173 generating these instructions. This option is enabled by default when
15174 @option{-mcpu=cortex-m3} is specified.
15176 @item -munaligned-access
15177 @itemx -mno-unaligned-access
15178 @opindex munaligned-access
15179 @opindex mno-unaligned-access
15180 Enables (or disables) reading and writing of 16- and 32- bit values
15181 from addresses that are not 16- or 32- bit aligned. By default
15182 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
15183 ARMv8-M Baseline architectures, and enabled for all other
15184 architectures. If unaligned access is not enabled then words in packed
15185 data structures are accessed a byte at a time.
15187 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
15188 generated object file to either true or false, depending upon the
15189 setting of this option. If unaligned access is enabled then the
15190 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
15193 @item -mneon-for-64bits
15194 @opindex mneon-for-64bits
15195 Enables using Neon to handle scalar 64-bits operations. This is
15196 disabled by default since the cost of moving data from core registers
15199 @item -mslow-flash-data
15200 @opindex mslow-flash-data
15201 Assume loading data from flash is slower than fetching instruction.
15202 Therefore literal load is minimized for better performance.
15203 This option is only supported when compiling for ARMv7 M-profile and
15206 @item -masm-syntax-unified
15207 @opindex masm-syntax-unified
15208 Assume inline assembler is using unified asm syntax. The default is
15209 currently off which implies divided syntax. This option has no impact
15210 on Thumb2. However, this may change in future releases of GCC.
15211 Divided syntax should be considered deprecated.
15213 @item -mrestrict-it
15214 @opindex mrestrict-it
15215 Restricts generation of IT blocks to conform to the rules of ARMv8.
15216 IT blocks can only contain a single 16-bit instruction from a select
15217 set of instructions. This option is on by default for ARMv8 Thumb mode.
15219 @item -mprint-tune-info
15220 @opindex mprint-tune-info
15221 Print CPU tuning information as comment in assembler file. This is
15222 an option used only for regression testing of the compiler and not
15223 intended for ordinary use in compiling code. This option is disabled
15227 @opindex mpure-code
15228 Do not allow constant data to be placed in code sections.
15229 Additionally, when compiling for ELF object format give all text sections the
15230 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
15231 is only available when generating non-pic code for ARMv7-M targets.
15236 @subsection AVR Options
15237 @cindex AVR Options
15239 These options are defined for AVR implementations:
15242 @item -mmcu=@var{mcu}
15244 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
15246 The default for this option is@tie{}@samp{avr2}.
15248 GCC supports the following AVR devices and ISAs:
15250 @include avr-mmcu.texi
15252 @item -maccumulate-args
15253 @opindex maccumulate-args
15254 Accumulate outgoing function arguments and acquire/release the needed
15255 stack space for outgoing function arguments once in function
15256 prologue/epilogue. Without this option, outgoing arguments are pushed
15257 before calling a function and popped afterwards.
15259 Popping the arguments after the function call can be expensive on
15260 AVR so that accumulating the stack space might lead to smaller
15261 executables because arguments need not to be removed from the
15262 stack after such a function call.
15264 This option can lead to reduced code size for functions that perform
15265 several calls to functions that get their arguments on the stack like
15266 calls to printf-like functions.
15268 @item -mbranch-cost=@var{cost}
15269 @opindex mbranch-cost
15270 Set the branch costs for conditional branch instructions to
15271 @var{cost}. Reasonable values for @var{cost} are small, non-negative
15272 integers. The default branch cost is 0.
15274 @item -mcall-prologues
15275 @opindex mcall-prologues
15276 Functions prologues/epilogues are expanded as calls to appropriate
15277 subroutines. Code size is smaller.
15281 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
15282 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
15283 and @code{long long} is 4 bytes. Please note that this option does not
15284 conform to the C standards, but it results in smaller code
15287 @item -mn-flash=@var{num}
15289 Assume that the flash memory has a size of
15290 @var{num} times 64@tie{}KiB.
15292 @item -mno-interrupts
15293 @opindex mno-interrupts
15294 Generated code is not compatible with hardware interrupts.
15295 Code size is smaller.
15299 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
15300 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
15301 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
15302 the assembler's command line and the @option{--relax} option to the
15303 linker's command line.
15305 Jump relaxing is performed by the linker because jump offsets are not
15306 known before code is located. Therefore, the assembler code generated by the
15307 compiler is the same, but the instructions in the executable may
15308 differ from instructions in the assembler code.
15310 Relaxing must be turned on if linker stubs are needed, see the
15311 section on @code{EIND} and linker stubs below.
15315 Assume that the device supports the Read-Modify-Write
15316 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
15320 Treat the stack pointer register as an 8-bit register,
15321 i.e.@: assume the high byte of the stack pointer is zero.
15322 In general, you don't need to set this option by hand.
15324 This option is used internally by the compiler to select and
15325 build multilibs for architectures @code{avr2} and @code{avr25}.
15326 These architectures mix devices with and without @code{SPH}.
15327 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
15328 the compiler driver adds or removes this option from the compiler
15329 proper's command line, because the compiler then knows if the device
15330 or architecture has an 8-bit stack pointer and thus no @code{SPH}
15335 Use address register @code{X} in a way proposed by the hardware. This means
15336 that @code{X} is only used in indirect, post-increment or
15337 pre-decrement addressing.
15339 Without this option, the @code{X} register may be used in the same way
15340 as @code{Y} or @code{Z} which then is emulated by additional
15342 For example, loading a value with @code{X+const} addressing with a
15343 small non-negative @code{const < 64} to a register @var{Rn} is
15347 adiw r26, const ; X += const
15348 ld @var{Rn}, X ; @var{Rn} = *X
15349 sbiw r26, const ; X -= const
15353 @opindex mtiny-stack
15354 Only change the lower 8@tie{}bits of the stack pointer.
15356 @item -mfract-convert-truncate
15357 @opindex mfract-convert-truncate
15358 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
15361 @opindex nodevicelib
15362 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
15364 @item -Waddr-space-convert
15365 @opindex Waddr-space-convert
15366 Warn about conversions between address spaces in the case where the
15367 resulting address space is not contained in the incoming address space.
15369 @item -Wmisspelled-isr
15370 @opindex Wmisspelled-isr
15371 Warn if the ISR is misspelled, i.e. without __vector prefix.
15372 Enabled by default.
15375 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
15376 @cindex @code{EIND}
15377 Pointers in the implementation are 16@tie{}bits wide.
15378 The address of a function or label is represented as word address so
15379 that indirect jumps and calls can target any code address in the
15380 range of 64@tie{}Ki words.
15382 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
15383 bytes of program memory space, there is a special function register called
15384 @code{EIND} that serves as most significant part of the target address
15385 when @code{EICALL} or @code{EIJMP} instructions are used.
15387 Indirect jumps and calls on these devices are handled as follows by
15388 the compiler and are subject to some limitations:
15393 The compiler never sets @code{EIND}.
15396 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
15397 instructions or might read @code{EIND} directly in order to emulate an
15398 indirect call/jump by means of a @code{RET} instruction.
15401 The compiler assumes that @code{EIND} never changes during the startup
15402 code or during the application. In particular, @code{EIND} is not
15403 saved/restored in function or interrupt service routine
15407 For indirect calls to functions and computed goto, the linker
15408 generates @emph{stubs}. Stubs are jump pads sometimes also called
15409 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
15410 The stub contains a direct jump to the desired address.
15413 Linker relaxation must be turned on so that the linker generates
15414 the stubs correctly in all situations. See the compiler option
15415 @option{-mrelax} and the linker option @option{--relax}.
15416 There are corner cases where the linker is supposed to generate stubs
15417 but aborts without relaxation and without a helpful error message.
15420 The default linker script is arranged for code with @code{EIND = 0}.
15421 If code is supposed to work for a setup with @code{EIND != 0}, a custom
15422 linker script has to be used in order to place the sections whose
15423 name start with @code{.trampolines} into the segment where @code{EIND}
15427 The startup code from libgcc never sets @code{EIND}.
15428 Notice that startup code is a blend of code from libgcc and AVR-LibC.
15429 For the impact of AVR-LibC on @code{EIND}, see the
15430 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
15433 It is legitimate for user-specific startup code to set up @code{EIND}
15434 early, for example by means of initialization code located in
15435 section @code{.init3}. Such code runs prior to general startup code
15436 that initializes RAM and calls constructors, but after the bit
15437 of startup code from AVR-LibC that sets @code{EIND} to the segment
15438 where the vector table is located.
15440 #include <avr/io.h>
15443 __attribute__((section(".init3"),naked,used,no_instrument_function))
15444 init3_set_eind (void)
15446 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
15447 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
15452 The @code{__trampolines_start} symbol is defined in the linker script.
15455 Stubs are generated automatically by the linker if
15456 the following two conditions are met:
15459 @item The address of a label is taken by means of the @code{gs} modifier
15460 (short for @emph{generate stubs}) like so:
15462 LDI r24, lo8(gs(@var{func}))
15463 LDI r25, hi8(gs(@var{func}))
15465 @item The final location of that label is in a code segment
15466 @emph{outside} the segment where the stubs are located.
15470 The compiler emits such @code{gs} modifiers for code labels in the
15471 following situations:
15473 @item Taking address of a function or code label.
15474 @item Computed goto.
15475 @item If prologue-save function is used, see @option{-mcall-prologues}
15476 command-line option.
15477 @item Switch/case dispatch tables. If you do not want such dispatch
15478 tables you can specify the @option{-fno-jump-tables} command-line option.
15479 @item C and C++ constructors/destructors called during startup/shutdown.
15480 @item If the tools hit a @code{gs()} modifier explained above.
15484 Jumping to non-symbolic addresses like so is @emph{not} supported:
15489 /* Call function at word address 0x2 */
15490 return ((int(*)(void)) 0x2)();
15494 Instead, a stub has to be set up, i.e.@: the function has to be called
15495 through a symbol (@code{func_4} in the example):
15500 extern int func_4 (void);
15502 /* Call function at byte address 0x4 */
15507 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
15508 Alternatively, @code{func_4} can be defined in the linker script.
15511 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
15512 @cindex @code{RAMPD}
15513 @cindex @code{RAMPX}
15514 @cindex @code{RAMPY}
15515 @cindex @code{RAMPZ}
15516 Some AVR devices support memories larger than the 64@tie{}KiB range
15517 that can be accessed with 16-bit pointers. To access memory locations
15518 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
15519 register is used as high part of the address:
15520 The @code{X}, @code{Y}, @code{Z} address register is concatenated
15521 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
15522 register, respectively, to get a wide address. Similarly,
15523 @code{RAMPD} is used together with direct addressing.
15527 The startup code initializes the @code{RAMP} special function
15528 registers with zero.
15531 If a @ref{AVR Named Address Spaces,named address space} other than
15532 generic or @code{__flash} is used, then @code{RAMPZ} is set
15533 as needed before the operation.
15536 If the device supports RAM larger than 64@tie{}KiB and the compiler
15537 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
15538 is reset to zero after the operation.
15541 If the device comes with a specific @code{RAMP} register, the ISR
15542 prologue/epilogue saves/restores that SFR and initializes it with
15543 zero in case the ISR code might (implicitly) use it.
15546 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
15547 If you use inline assembler to read from locations outside the
15548 16-bit address range and change one of the @code{RAMP} registers,
15549 you must reset it to zero after the access.
15553 @subsubsection AVR Built-in Macros
15555 GCC defines several built-in macros so that the user code can test
15556 for the presence or absence of features. Almost any of the following
15557 built-in macros are deduced from device capabilities and thus
15558 triggered by the @option{-mmcu=} command-line option.
15560 For even more AVR-specific built-in macros see
15561 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
15566 Build-in macro that resolves to a decimal number that identifies the
15567 architecture and depends on the @option{-mmcu=@var{mcu}} option.
15568 Possible values are:
15570 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
15571 @code{4}, @code{5}, @code{51}, @code{6}
15573 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
15574 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
15578 @code{100}, @code{102}, @code{104},
15579 @code{105}, @code{106}, @code{107}
15581 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
15582 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
15583 If @var{mcu} specifies a device, this built-in macro is set
15584 accordingly. For example, with @option{-mmcu=atmega8} the macro is
15585 defined to @code{4}.
15587 @item __AVR_@var{Device}__
15588 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
15589 the device's name. For example, @option{-mmcu=atmega8} defines the
15590 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
15591 @code{__AVR_ATtiny261A__}, etc.
15593 The built-in macros' names follow
15594 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
15595 the device name as from the AVR user manual. The difference between
15596 @var{Device} in the built-in macro and @var{device} in
15597 @option{-mmcu=@var{device}} is that the latter is always lowercase.
15599 If @var{device} is not a device but only a core architecture like
15600 @samp{avr51}, this macro is not defined.
15602 @item __AVR_DEVICE_NAME__
15603 Setting @option{-mmcu=@var{device}} defines this built-in macro to
15604 the device's name. For example, with @option{-mmcu=atmega8} the macro
15605 is defined to @code{atmega8}.
15607 If @var{device} is not a device but only a core architecture like
15608 @samp{avr51}, this macro is not defined.
15610 @item __AVR_XMEGA__
15611 The device / architecture belongs to the XMEGA family of devices.
15613 @item __AVR_HAVE_ELPM__
15614 The device has the @code{ELPM} instruction.
15616 @item __AVR_HAVE_ELPMX__
15617 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
15618 R@var{n},Z+} instructions.
15620 @item __AVR_HAVE_MOVW__
15621 The device has the @code{MOVW} instruction to perform 16-bit
15622 register-register moves.
15624 @item __AVR_HAVE_LPMX__
15625 The device has the @code{LPM R@var{n},Z} and
15626 @code{LPM R@var{n},Z+} instructions.
15628 @item __AVR_HAVE_MUL__
15629 The device has a hardware multiplier.
15631 @item __AVR_HAVE_JMP_CALL__
15632 The device has the @code{JMP} and @code{CALL} instructions.
15633 This is the case for devices with at least 16@tie{}KiB of program
15636 @item __AVR_HAVE_EIJMP_EICALL__
15637 @itemx __AVR_3_BYTE_PC__
15638 The device has the @code{EIJMP} and @code{EICALL} instructions.
15639 This is the case for devices with more than 128@tie{}KiB of program memory.
15640 This also means that the program counter
15641 (PC) is 3@tie{}bytes wide.
15643 @item __AVR_2_BYTE_PC__
15644 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
15645 with up to 128@tie{}KiB of program memory.
15647 @item __AVR_HAVE_8BIT_SP__
15648 @itemx __AVR_HAVE_16BIT_SP__
15649 The stack pointer (SP) register is treated as 8-bit respectively
15650 16-bit register by the compiler.
15651 The definition of these macros is affected by @option{-mtiny-stack}.
15653 @item __AVR_HAVE_SPH__
15655 The device has the SPH (high part of stack pointer) special function
15656 register or has an 8-bit stack pointer, respectively.
15657 The definition of these macros is affected by @option{-mmcu=} and
15658 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
15661 @item __AVR_HAVE_RAMPD__
15662 @itemx __AVR_HAVE_RAMPX__
15663 @itemx __AVR_HAVE_RAMPY__
15664 @itemx __AVR_HAVE_RAMPZ__
15665 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
15666 @code{RAMPZ} special function register, respectively.
15668 @item __NO_INTERRUPTS__
15669 This macro reflects the @option{-mno-interrupts} command-line option.
15671 @item __AVR_ERRATA_SKIP__
15672 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
15673 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
15674 instructions because of a hardware erratum. Skip instructions are
15675 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
15676 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
15679 @item __AVR_ISA_RMW__
15680 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
15682 @item __AVR_SFR_OFFSET__=@var{offset}
15683 Instructions that can address I/O special function registers directly
15684 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
15685 address as if addressed by an instruction to access RAM like @code{LD}
15686 or @code{STS}. This offset depends on the device architecture and has
15687 to be subtracted from the RAM address in order to get the
15688 respective I/O@tie{}address.
15690 @item __WITH_AVRLIBC__
15691 The compiler is configured to be used together with AVR-Libc.
15692 See the @option{--with-avrlibc} configure option.
15696 @node Blackfin Options
15697 @subsection Blackfin Options
15698 @cindex Blackfin Options
15701 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
15703 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
15704 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
15705 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
15706 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
15707 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
15708 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
15709 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
15710 @samp{bf561}, @samp{bf592}.
15712 The optional @var{sirevision} specifies the silicon revision of the target
15713 Blackfin processor. Any workarounds available for the targeted silicon revision
15714 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
15715 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
15716 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
15717 hexadecimal digits representing the major and minor numbers in the silicon
15718 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
15719 is not defined. If @var{sirevision} is @samp{any}, the
15720 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
15721 If this optional @var{sirevision} is not used, GCC assumes the latest known
15722 silicon revision of the targeted Blackfin processor.
15724 GCC defines a preprocessor macro for the specified @var{cpu}.
15725 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
15726 provided by libgloss to be linked in if @option{-msim} is not given.
15728 Without this option, @samp{bf532} is used as the processor by default.
15730 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
15731 only the preprocessor macro is defined.
15735 Specifies that the program will be run on the simulator. This causes
15736 the simulator BSP provided by libgloss to be linked in. This option
15737 has effect only for @samp{bfin-elf} toolchain.
15738 Certain other options, such as @option{-mid-shared-library} and
15739 @option{-mfdpic}, imply @option{-msim}.
15741 @item -momit-leaf-frame-pointer
15742 @opindex momit-leaf-frame-pointer
15743 Don't keep the frame pointer in a register for leaf functions. This
15744 avoids the instructions to save, set up and restore frame pointers and
15745 makes an extra register available in leaf functions. The option
15746 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
15747 which might make debugging harder.
15749 @item -mspecld-anomaly
15750 @opindex mspecld-anomaly
15751 When enabled, the compiler ensures that the generated code does not
15752 contain speculative loads after jump instructions. If this option is used,
15753 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
15755 @item -mno-specld-anomaly
15756 @opindex mno-specld-anomaly
15757 Don't generate extra code to prevent speculative loads from occurring.
15759 @item -mcsync-anomaly
15760 @opindex mcsync-anomaly
15761 When enabled, the compiler ensures that the generated code does not
15762 contain CSYNC or SSYNC instructions too soon after conditional branches.
15763 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
15765 @item -mno-csync-anomaly
15766 @opindex mno-csync-anomaly
15767 Don't generate extra code to prevent CSYNC or SSYNC instructions from
15768 occurring too soon after a conditional branch.
15772 When enabled, the compiler is free to take advantage of the knowledge that
15773 the entire program fits into the low 64k of memory.
15776 @opindex mno-low-64k
15777 Assume that the program is arbitrarily large. This is the default.
15779 @item -mstack-check-l1
15780 @opindex mstack-check-l1
15781 Do stack checking using information placed into L1 scratchpad memory by the
15784 @item -mid-shared-library
15785 @opindex mid-shared-library
15786 Generate code that supports shared libraries via the library ID method.
15787 This allows for execute in place and shared libraries in an environment
15788 without virtual memory management. This option implies @option{-fPIC}.
15789 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15791 @item -mno-id-shared-library
15792 @opindex mno-id-shared-library
15793 Generate code that doesn't assume ID-based shared libraries are being used.
15794 This is the default.
15796 @item -mleaf-id-shared-library
15797 @opindex mleaf-id-shared-library
15798 Generate code that supports shared libraries via the library ID method,
15799 but assumes that this library or executable won't link against any other
15800 ID shared libraries. That allows the compiler to use faster code for jumps
15803 @item -mno-leaf-id-shared-library
15804 @opindex mno-leaf-id-shared-library
15805 Do not assume that the code being compiled won't link against any ID shared
15806 libraries. Slower code is generated for jump and call insns.
15808 @item -mshared-library-id=n
15809 @opindex mshared-library-id
15810 Specifies the identification number of the ID-based shared library being
15811 compiled. Specifying a value of 0 generates more compact code; specifying
15812 other values forces the allocation of that number to the current
15813 library but is no more space- or time-efficient than omitting this option.
15817 Generate code that allows the data segment to be located in a different
15818 area of memory from the text segment. This allows for execute in place in
15819 an environment without virtual memory management by eliminating relocations
15820 against the text section.
15822 @item -mno-sep-data
15823 @opindex mno-sep-data
15824 Generate code that assumes that the data segment follows the text segment.
15825 This is the default.
15828 @itemx -mno-long-calls
15829 @opindex mlong-calls
15830 @opindex mno-long-calls
15831 Tells the compiler to perform function calls by first loading the
15832 address of the function into a register and then performing a subroutine
15833 call on this register. This switch is needed if the target function
15834 lies outside of the 24-bit addressing range of the offset-based
15835 version of subroutine call instruction.
15837 This feature is not enabled by default. Specifying
15838 @option{-mno-long-calls} restores the default behavior. Note these
15839 switches have no effect on how the compiler generates code to handle
15840 function calls via function pointers.
15844 Link with the fast floating-point library. This library relaxes some of
15845 the IEEE floating-point standard's rules for checking inputs against
15846 Not-a-Number (NAN), in the interest of performance.
15849 @opindex minline-plt
15850 Enable inlining of PLT entries in function calls to functions that are
15851 not known to bind locally. It has no effect without @option{-mfdpic}.
15854 @opindex mmulticore
15855 Build a standalone application for multicore Blackfin processors.
15856 This option causes proper start files and link scripts supporting
15857 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
15858 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
15860 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
15861 selects the one-application-per-core programming model. Without
15862 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
15863 programming model is used. In this model, the main function of Core B
15864 should be named as @code{coreb_main}.
15866 If this option is not used, the single-core application programming
15871 Build a standalone application for Core A of BF561 when using
15872 the one-application-per-core programming model. Proper start files
15873 and link scripts are used to support Core A, and the macro
15874 @code{__BFIN_COREA} is defined.
15875 This option can only be used in conjunction with @option{-mmulticore}.
15879 Build a standalone application for Core B of BF561 when using
15880 the one-application-per-core programming model. Proper start files
15881 and link scripts are used to support Core B, and the macro
15882 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
15883 should be used instead of @code{main}.
15884 This option can only be used in conjunction with @option{-mmulticore}.
15888 Build a standalone application for SDRAM. Proper start files and
15889 link scripts are used to put the application into SDRAM, and the macro
15890 @code{__BFIN_SDRAM} is defined.
15891 The loader should initialize SDRAM before loading the application.
15895 Assume that ICPLBs are enabled at run time. This has an effect on certain
15896 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
15897 are enabled; for standalone applications the default is off.
15901 @subsection C6X Options
15902 @cindex C6X Options
15905 @item -march=@var{name}
15907 This specifies the name of the target architecture. GCC uses this
15908 name to determine what kind of instructions it can emit when generating
15909 assembly code. Permissible names are: @samp{c62x},
15910 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
15913 @opindex mbig-endian
15914 Generate code for a big-endian target.
15916 @item -mlittle-endian
15917 @opindex mlittle-endian
15918 Generate code for a little-endian target. This is the default.
15922 Choose startup files and linker script suitable for the simulator.
15924 @item -msdata=default
15925 @opindex msdata=default
15926 Put small global and static data in the @code{.neardata} section,
15927 which is pointed to by register @code{B14}. Put small uninitialized
15928 global and static data in the @code{.bss} section, which is adjacent
15929 to the @code{.neardata} section. Put small read-only data into the
15930 @code{.rodata} section. The corresponding sections used for large
15931 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
15934 @opindex msdata=all
15935 Put all data, not just small objects, into the sections reserved for
15936 small data, and use addressing relative to the @code{B14} register to
15940 @opindex msdata=none
15941 Make no use of the sections reserved for small data, and use absolute
15942 addresses to access all data. Put all initialized global and static
15943 data in the @code{.fardata} section, and all uninitialized data in the
15944 @code{.far} section. Put all constant data into the @code{.const}
15949 @subsection CRIS Options
15950 @cindex CRIS Options
15952 These options are defined specifically for the CRIS ports.
15955 @item -march=@var{architecture-type}
15956 @itemx -mcpu=@var{architecture-type}
15959 Generate code for the specified architecture. The choices for
15960 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
15961 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
15962 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
15965 @item -mtune=@var{architecture-type}
15967 Tune to @var{architecture-type} everything applicable about the generated
15968 code, except for the ABI and the set of available instructions. The
15969 choices for @var{architecture-type} are the same as for
15970 @option{-march=@var{architecture-type}}.
15972 @item -mmax-stack-frame=@var{n}
15973 @opindex mmax-stack-frame
15974 Warn when the stack frame of a function exceeds @var{n} bytes.
15980 The options @option{-metrax4} and @option{-metrax100} are synonyms for
15981 @option{-march=v3} and @option{-march=v8} respectively.
15983 @item -mmul-bug-workaround
15984 @itemx -mno-mul-bug-workaround
15985 @opindex mmul-bug-workaround
15986 @opindex mno-mul-bug-workaround
15987 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
15988 models where it applies. This option is active by default.
15992 Enable CRIS-specific verbose debug-related information in the assembly
15993 code. This option also has the effect of turning off the @samp{#NO_APP}
15994 formatted-code indicator to the assembler at the beginning of the
15999 Do not use condition-code results from previous instruction; always emit
16000 compare and test instructions before use of condition codes.
16002 @item -mno-side-effects
16003 @opindex mno-side-effects
16004 Do not emit instructions with side effects in addressing modes other than
16007 @item -mstack-align
16008 @itemx -mno-stack-align
16009 @itemx -mdata-align
16010 @itemx -mno-data-align
16011 @itemx -mconst-align
16012 @itemx -mno-const-align
16013 @opindex mstack-align
16014 @opindex mno-stack-align
16015 @opindex mdata-align
16016 @opindex mno-data-align
16017 @opindex mconst-align
16018 @opindex mno-const-align
16019 These options (@samp{no-} options) arrange (eliminate arrangements) for the
16020 stack frame, individual data and constants to be aligned for the maximum
16021 single data access size for the chosen CPU model. The default is to
16022 arrange for 32-bit alignment. ABI details such as structure layout are
16023 not affected by these options.
16031 Similar to the stack- data- and const-align options above, these options
16032 arrange for stack frame, writable data and constants to all be 32-bit,
16033 16-bit or 8-bit aligned. The default is 32-bit alignment.
16035 @item -mno-prologue-epilogue
16036 @itemx -mprologue-epilogue
16037 @opindex mno-prologue-epilogue
16038 @opindex mprologue-epilogue
16039 With @option{-mno-prologue-epilogue}, the normal function prologue and
16040 epilogue which set up the stack frame are omitted and no return
16041 instructions or return sequences are generated in the code. Use this
16042 option only together with visual inspection of the compiled code: no
16043 warnings or errors are generated when call-saved registers must be saved,
16044 or storage for local variables needs to be allocated.
16048 @opindex mno-gotplt
16050 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
16051 instruction sequences that load addresses for functions from the PLT part
16052 of the GOT rather than (traditional on other architectures) calls to the
16053 PLT@. The default is @option{-mgotplt}.
16057 Legacy no-op option only recognized with the cris-axis-elf and
16058 cris-axis-linux-gnu targets.
16062 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
16066 This option, recognized for the cris-axis-elf, arranges
16067 to link with input-output functions from a simulator library. Code,
16068 initialized data and zero-initialized data are allocated consecutively.
16072 Like @option{-sim}, but pass linker options to locate initialized data at
16073 0x40000000 and zero-initialized data at 0x80000000.
16077 @subsection CR16 Options
16078 @cindex CR16 Options
16080 These options are defined specifically for the CR16 ports.
16086 Enable the use of multiply-accumulate instructions. Disabled by default.
16090 @opindex mcr16cplus
16092 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
16097 Links the library libsim.a which is in compatible with simulator. Applicable
16098 to ELF compiler only.
16102 Choose integer type as 32-bit wide.
16106 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
16108 @item -mdata-model=@var{model}
16109 @opindex mdata-model
16110 Choose a data model. The choices for @var{model} are @samp{near},
16111 @samp{far} or @samp{medium}. @samp{medium} is default.
16112 However, @samp{far} is not valid with @option{-mcr16c}, as the
16113 CR16C architecture does not support the far data model.
16116 @node Darwin Options
16117 @subsection Darwin Options
16118 @cindex Darwin options
16120 These options are defined for all architectures running the Darwin operating
16123 FSF GCC on Darwin does not create ``fat'' object files; it creates
16124 an object file for the single architecture that GCC was built to
16125 target. Apple's GCC on Darwin does create ``fat'' files if multiple
16126 @option{-arch} options are used; it does so by running the compiler or
16127 linker multiple times and joining the results together with
16130 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
16131 @samp{i686}) is determined by the flags that specify the ISA
16132 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
16133 @option{-force_cpusubtype_ALL} option can be used to override this.
16135 The Darwin tools vary in their behavior when presented with an ISA
16136 mismatch. The assembler, @file{as}, only permits instructions to
16137 be used that are valid for the subtype of the file it is generating,
16138 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
16139 The linker for shared libraries, @file{/usr/bin/libtool}, fails
16140 and prints an error if asked to create a shared library with a less
16141 restrictive subtype than its input files (for instance, trying to put
16142 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
16143 for executables, @command{ld}, quietly gives the executable the most
16144 restrictive subtype of any of its input files.
16149 Add the framework directory @var{dir} to the head of the list of
16150 directories to be searched for header files. These directories are
16151 interleaved with those specified by @option{-I} options and are
16152 scanned in a left-to-right order.
16154 A framework directory is a directory with frameworks in it. A
16155 framework is a directory with a @file{Headers} and/or
16156 @file{PrivateHeaders} directory contained directly in it that ends
16157 in @file{.framework}. The name of a framework is the name of this
16158 directory excluding the @file{.framework}. Headers associated with
16159 the framework are found in one of those two directories, with
16160 @file{Headers} being searched first. A subframework is a framework
16161 directory that is in a framework's @file{Frameworks} directory.
16162 Includes of subframework headers can only appear in a header of a
16163 framework that contains the subframework, or in a sibling subframework
16164 header. Two subframeworks are siblings if they occur in the same
16165 framework. A subframework should not have the same name as a
16166 framework; a warning is issued if this is violated. Currently a
16167 subframework cannot have subframeworks; in the future, the mechanism
16168 may be extended to support this. The standard frameworks can be found
16169 in @file{/System/Library/Frameworks} and
16170 @file{/Library/Frameworks}. An example include looks like
16171 @code{#include <Framework/header.h>}, where @file{Framework} denotes
16172 the name of the framework and @file{header.h} is found in the
16173 @file{PrivateHeaders} or @file{Headers} directory.
16175 @item -iframework@var{dir}
16176 @opindex iframework
16177 Like @option{-F} except the directory is a treated as a system
16178 directory. The main difference between this @option{-iframework} and
16179 @option{-F} is that with @option{-iframework} the compiler does not
16180 warn about constructs contained within header files found via
16181 @var{dir}. This option is valid only for the C family of languages.
16185 Emit debugging information for symbols that are used. For stabs
16186 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
16187 This is by default ON@.
16191 Emit debugging information for all symbols and types.
16193 @item -mmacosx-version-min=@var{version}
16194 The earliest version of MacOS X that this executable will run on
16195 is @var{version}. Typical values of @var{version} include @code{10.1},
16196 @code{10.2}, and @code{10.3.9}.
16198 If the compiler was built to use the system's headers by default,
16199 then the default for this option is the system version on which the
16200 compiler is running, otherwise the default is to make choices that
16201 are compatible with as many systems and code bases as possible.
16205 Enable kernel development mode. The @option{-mkernel} option sets
16206 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
16207 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
16208 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
16209 applicable. This mode also sets @option{-mno-altivec},
16210 @option{-msoft-float}, @option{-fno-builtin} and
16211 @option{-mlong-branch} for PowerPC targets.
16213 @item -mone-byte-bool
16214 @opindex mone-byte-bool
16215 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
16216 By default @code{sizeof(bool)} is @code{4} when compiling for
16217 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
16218 option has no effect on x86.
16220 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
16221 to generate code that is not binary compatible with code generated
16222 without that switch. Using this switch may require recompiling all
16223 other modules in a program, including system libraries. Use this
16224 switch to conform to a non-default data model.
16226 @item -mfix-and-continue
16227 @itemx -ffix-and-continue
16228 @itemx -findirect-data
16229 @opindex mfix-and-continue
16230 @opindex ffix-and-continue
16231 @opindex findirect-data
16232 Generate code suitable for fast turnaround development, such as to
16233 allow GDB to dynamically load @file{.o} files into already-running
16234 programs. @option{-findirect-data} and @option{-ffix-and-continue}
16235 are provided for backwards compatibility.
16239 Loads all members of static archive libraries.
16240 See man ld(1) for more information.
16242 @item -arch_errors_fatal
16243 @opindex arch_errors_fatal
16244 Cause the errors having to do with files that have the wrong architecture
16247 @item -bind_at_load
16248 @opindex bind_at_load
16249 Causes the output file to be marked such that the dynamic linker will
16250 bind all undefined references when the file is loaded or launched.
16254 Produce a Mach-o bundle format file.
16255 See man ld(1) for more information.
16257 @item -bundle_loader @var{executable}
16258 @opindex bundle_loader
16259 This option specifies the @var{executable} that will load the build
16260 output file being linked. See man ld(1) for more information.
16263 @opindex dynamiclib
16264 When passed this option, GCC produces a dynamic library instead of
16265 an executable when linking, using the Darwin @file{libtool} command.
16267 @item -force_cpusubtype_ALL
16268 @opindex force_cpusubtype_ALL
16269 This causes GCC's output file to have the @samp{ALL} subtype, instead of
16270 one controlled by the @option{-mcpu} or @option{-march} option.
16272 @item -allowable_client @var{client_name}
16273 @itemx -client_name
16274 @itemx -compatibility_version
16275 @itemx -current_version
16277 @itemx -dependency-file
16279 @itemx -dylinker_install_name
16281 @itemx -exported_symbols_list
16284 @itemx -flat_namespace
16285 @itemx -force_flat_namespace
16286 @itemx -headerpad_max_install_names
16289 @itemx -install_name
16290 @itemx -keep_private_externs
16291 @itemx -multi_module
16292 @itemx -multiply_defined
16293 @itemx -multiply_defined_unused
16296 @itemx -no_dead_strip_inits_and_terms
16297 @itemx -nofixprebinding
16298 @itemx -nomultidefs
16300 @itemx -noseglinkedit
16301 @itemx -pagezero_size
16303 @itemx -prebind_all_twolevel_modules
16304 @itemx -private_bundle
16306 @itemx -read_only_relocs
16308 @itemx -sectobjectsymbols
16312 @itemx -sectobjectsymbols
16315 @itemx -segs_read_only_addr
16317 @itemx -segs_read_write_addr
16318 @itemx -seg_addr_table
16319 @itemx -seg_addr_table_filename
16320 @itemx -seglinkedit
16322 @itemx -segs_read_only_addr
16323 @itemx -segs_read_write_addr
16324 @itemx -single_module
16326 @itemx -sub_library
16328 @itemx -sub_umbrella
16329 @itemx -twolevel_namespace
16332 @itemx -unexported_symbols_list
16333 @itemx -weak_reference_mismatches
16334 @itemx -whatsloaded
16335 @opindex allowable_client
16336 @opindex client_name
16337 @opindex compatibility_version
16338 @opindex current_version
16339 @opindex dead_strip
16340 @opindex dependency-file
16341 @opindex dylib_file
16342 @opindex dylinker_install_name
16344 @opindex exported_symbols_list
16346 @opindex flat_namespace
16347 @opindex force_flat_namespace
16348 @opindex headerpad_max_install_names
16349 @opindex image_base
16351 @opindex install_name
16352 @opindex keep_private_externs
16353 @opindex multi_module
16354 @opindex multiply_defined
16355 @opindex multiply_defined_unused
16356 @opindex noall_load
16357 @opindex no_dead_strip_inits_and_terms
16358 @opindex nofixprebinding
16359 @opindex nomultidefs
16361 @opindex noseglinkedit
16362 @opindex pagezero_size
16364 @opindex prebind_all_twolevel_modules
16365 @opindex private_bundle
16366 @opindex read_only_relocs
16368 @opindex sectobjectsymbols
16371 @opindex sectcreate
16372 @opindex sectobjectsymbols
16375 @opindex segs_read_only_addr
16376 @opindex segs_read_write_addr
16377 @opindex seg_addr_table
16378 @opindex seg_addr_table_filename
16379 @opindex seglinkedit
16381 @opindex segs_read_only_addr
16382 @opindex segs_read_write_addr
16383 @opindex single_module
16385 @opindex sub_library
16386 @opindex sub_umbrella
16387 @opindex twolevel_namespace
16390 @opindex unexported_symbols_list
16391 @opindex weak_reference_mismatches
16392 @opindex whatsloaded
16393 These options are passed to the Darwin linker. The Darwin linker man page
16394 describes them in detail.
16397 @node DEC Alpha Options
16398 @subsection DEC Alpha Options
16400 These @samp{-m} options are defined for the DEC Alpha implementations:
16403 @item -mno-soft-float
16404 @itemx -msoft-float
16405 @opindex mno-soft-float
16406 @opindex msoft-float
16407 Use (do not use) the hardware floating-point instructions for
16408 floating-point operations. When @option{-msoft-float} is specified,
16409 functions in @file{libgcc.a} are used to perform floating-point
16410 operations. Unless they are replaced by routines that emulate the
16411 floating-point operations, or compiled in such a way as to call such
16412 emulations routines, these routines issue floating-point
16413 operations. If you are compiling for an Alpha without floating-point
16414 operations, you must ensure that the library is built so as not to call
16417 Note that Alpha implementations without floating-point operations are
16418 required to have floating-point registers.
16421 @itemx -mno-fp-regs
16423 @opindex mno-fp-regs
16424 Generate code that uses (does not use) the floating-point register set.
16425 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
16426 register set is not used, floating-point operands are passed in integer
16427 registers as if they were integers and floating-point results are passed
16428 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
16429 so any function with a floating-point argument or return value called by code
16430 compiled with @option{-mno-fp-regs} must also be compiled with that
16433 A typical use of this option is building a kernel that does not use,
16434 and hence need not save and restore, any floating-point registers.
16438 The Alpha architecture implements floating-point hardware optimized for
16439 maximum performance. It is mostly compliant with the IEEE floating-point
16440 standard. However, for full compliance, software assistance is
16441 required. This option generates code fully IEEE-compliant code
16442 @emph{except} that the @var{inexact-flag} is not maintained (see below).
16443 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
16444 defined during compilation. The resulting code is less efficient but is
16445 able to correctly support denormalized numbers and exceptional IEEE
16446 values such as not-a-number and plus/minus infinity. Other Alpha
16447 compilers call this option @option{-ieee_with_no_inexact}.
16449 @item -mieee-with-inexact
16450 @opindex mieee-with-inexact
16451 This is like @option{-mieee} except the generated code also maintains
16452 the IEEE @var{inexact-flag}. Turning on this option causes the
16453 generated code to implement fully-compliant IEEE math. In addition to
16454 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
16455 macro. On some Alpha implementations the resulting code may execute
16456 significantly slower than the code generated by default. Since there is
16457 very little code that depends on the @var{inexact-flag}, you should
16458 normally not specify this option. Other Alpha compilers call this
16459 option @option{-ieee_with_inexact}.
16461 @item -mfp-trap-mode=@var{trap-mode}
16462 @opindex mfp-trap-mode
16463 This option controls what floating-point related traps are enabled.
16464 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
16465 The trap mode can be set to one of four values:
16469 This is the default (normal) setting. The only traps that are enabled
16470 are the ones that cannot be disabled in software (e.g., division by zero
16474 In addition to the traps enabled by @samp{n}, underflow traps are enabled
16478 Like @samp{u}, but the instructions are marked to be safe for software
16479 completion (see Alpha architecture manual for details).
16482 Like @samp{su}, but inexact traps are enabled as well.
16485 @item -mfp-rounding-mode=@var{rounding-mode}
16486 @opindex mfp-rounding-mode
16487 Selects the IEEE rounding mode. Other Alpha compilers call this option
16488 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
16493 Normal IEEE rounding mode. Floating-point numbers are rounded towards
16494 the nearest machine number or towards the even machine number in case
16498 Round towards minus infinity.
16501 Chopped rounding mode. Floating-point numbers are rounded towards zero.
16504 Dynamic rounding mode. A field in the floating-point control register
16505 (@var{fpcr}, see Alpha architecture reference manual) controls the
16506 rounding mode in effect. The C library initializes this register for
16507 rounding towards plus infinity. Thus, unless your program modifies the
16508 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
16511 @item -mtrap-precision=@var{trap-precision}
16512 @opindex mtrap-precision
16513 In the Alpha architecture, floating-point traps are imprecise. This
16514 means without software assistance it is impossible to recover from a
16515 floating trap and program execution normally needs to be terminated.
16516 GCC can generate code that can assist operating system trap handlers
16517 in determining the exact location that caused a floating-point trap.
16518 Depending on the requirements of an application, different levels of
16519 precisions can be selected:
16523 Program precision. This option is the default and means a trap handler
16524 can only identify which program caused a floating-point exception.
16527 Function precision. The trap handler can determine the function that
16528 caused a floating-point exception.
16531 Instruction precision. The trap handler can determine the exact
16532 instruction that caused a floating-point exception.
16535 Other Alpha compilers provide the equivalent options called
16536 @option{-scope_safe} and @option{-resumption_safe}.
16538 @item -mieee-conformant
16539 @opindex mieee-conformant
16540 This option marks the generated code as IEEE conformant. You must not
16541 use this option unless you also specify @option{-mtrap-precision=i} and either
16542 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
16543 is to emit the line @samp{.eflag 48} in the function prologue of the
16544 generated assembly file.
16546 @item -mbuild-constants
16547 @opindex mbuild-constants
16548 Normally GCC examines a 32- or 64-bit integer constant to
16549 see if it can construct it from smaller constants in two or three
16550 instructions. If it cannot, it outputs the constant as a literal and
16551 generates code to load it from the data segment at run time.
16553 Use this option to require GCC to construct @emph{all} integer constants
16554 using code, even if it takes more instructions (the maximum is six).
16556 You typically use this option to build a shared library dynamic
16557 loader. Itself a shared library, it must relocate itself in memory
16558 before it can find the variables and constants in its own data segment.
16576 Indicate whether GCC should generate code to use the optional BWX,
16577 CIX, FIX and MAX instruction sets. The default is to use the instruction
16578 sets supported by the CPU type specified via @option{-mcpu=} option or that
16579 of the CPU on which GCC was built if none is specified.
16582 @itemx -mfloat-ieee
16583 @opindex mfloat-vax
16584 @opindex mfloat-ieee
16585 Generate code that uses (does not use) VAX F and G floating-point
16586 arithmetic instead of IEEE single and double precision.
16588 @item -mexplicit-relocs
16589 @itemx -mno-explicit-relocs
16590 @opindex mexplicit-relocs
16591 @opindex mno-explicit-relocs
16592 Older Alpha assemblers provided no way to generate symbol relocations
16593 except via assembler macros. Use of these macros does not allow
16594 optimal instruction scheduling. GNU binutils as of version 2.12
16595 supports a new syntax that allows the compiler to explicitly mark
16596 which relocations should apply to which instructions. This option
16597 is mostly useful for debugging, as GCC detects the capabilities of
16598 the assembler when it is built and sets the default accordingly.
16601 @itemx -mlarge-data
16602 @opindex msmall-data
16603 @opindex mlarge-data
16604 When @option{-mexplicit-relocs} is in effect, static data is
16605 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
16606 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
16607 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
16608 16-bit relocations off of the @code{$gp} register. This limits the
16609 size of the small data area to 64KB, but allows the variables to be
16610 directly accessed via a single instruction.
16612 The default is @option{-mlarge-data}. With this option the data area
16613 is limited to just below 2GB@. Programs that require more than 2GB of
16614 data must use @code{malloc} or @code{mmap} to allocate the data in the
16615 heap instead of in the program's data segment.
16617 When generating code for shared libraries, @option{-fpic} implies
16618 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
16621 @itemx -mlarge-text
16622 @opindex msmall-text
16623 @opindex mlarge-text
16624 When @option{-msmall-text} is used, the compiler assumes that the
16625 code of the entire program (or shared library) fits in 4MB, and is
16626 thus reachable with a branch instruction. When @option{-msmall-data}
16627 is used, the compiler can assume that all local symbols share the
16628 same @code{$gp} value, and thus reduce the number of instructions
16629 required for a function call from 4 to 1.
16631 The default is @option{-mlarge-text}.
16633 @item -mcpu=@var{cpu_type}
16635 Set the instruction set and instruction scheduling parameters for
16636 machine type @var{cpu_type}. You can specify either the @samp{EV}
16637 style name or the corresponding chip number. GCC supports scheduling
16638 parameters for the EV4, EV5 and EV6 family of processors and
16639 chooses the default values for the instruction set from the processor
16640 you specify. If you do not specify a processor type, GCC defaults
16641 to the processor on which the compiler was built.
16643 Supported values for @var{cpu_type} are
16649 Schedules as an EV4 and has no instruction set extensions.
16653 Schedules as an EV5 and has no instruction set extensions.
16657 Schedules as an EV5 and supports the BWX extension.
16662 Schedules as an EV5 and supports the BWX and MAX extensions.
16666 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
16670 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
16673 Native toolchains also support the value @samp{native},
16674 which selects the best architecture option for the host processor.
16675 @option{-mcpu=native} has no effect if GCC does not recognize
16678 @item -mtune=@var{cpu_type}
16680 Set only the instruction scheduling parameters for machine type
16681 @var{cpu_type}. The instruction set is not changed.
16683 Native toolchains also support the value @samp{native},
16684 which selects the best architecture option for the host processor.
16685 @option{-mtune=native} has no effect if GCC does not recognize
16688 @item -mmemory-latency=@var{time}
16689 @opindex mmemory-latency
16690 Sets the latency the scheduler should assume for typical memory
16691 references as seen by the application. This number is highly
16692 dependent on the memory access patterns used by the application
16693 and the size of the external cache on the machine.
16695 Valid options for @var{time} are
16699 A decimal number representing clock cycles.
16705 The compiler contains estimates of the number of clock cycles for
16706 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
16707 (also called Dcache, Scache, and Bcache), as well as to main memory.
16708 Note that L3 is only valid for EV5.
16714 @subsection FR30 Options
16715 @cindex FR30 Options
16717 These options are defined specifically for the FR30 port.
16721 @item -msmall-model
16722 @opindex msmall-model
16723 Use the small address space model. This can produce smaller code, but
16724 it does assume that all symbolic values and addresses fit into a
16729 Assume that runtime support has been provided and so there is no need
16730 to include the simulator library (@file{libsim.a}) on the linker
16736 @subsection FT32 Options
16737 @cindex FT32 Options
16739 These options are defined specifically for the FT32 port.
16745 Specifies that the program will be run on the simulator. This causes
16746 an alternate runtime startup and library to be linked.
16747 You must not use this option when generating programs that will run on
16748 real hardware; you must provide your own runtime library for whatever
16749 I/O functions are needed.
16753 Enable Local Register Allocation. This is still experimental for FT32,
16754 so by default the compiler uses standard reload.
16758 Do not use div and mod instructions.
16763 @subsection FRV Options
16764 @cindex FRV Options
16770 Only use the first 32 general-purpose registers.
16775 Use all 64 general-purpose registers.
16780 Use only the first 32 floating-point registers.
16785 Use all 64 floating-point registers.
16788 @opindex mhard-float
16790 Use hardware instructions for floating-point operations.
16793 @opindex msoft-float
16795 Use library routines for floating-point operations.
16800 Dynamically allocate condition code registers.
16805 Do not try to dynamically allocate condition code registers, only
16806 use @code{icc0} and @code{fcc0}.
16811 Change ABI to use double word insns.
16816 Do not use double word instructions.
16821 Use floating-point double instructions.
16824 @opindex mno-double
16826 Do not use floating-point double instructions.
16831 Use media instructions.
16836 Do not use media instructions.
16841 Use multiply and add/subtract instructions.
16844 @opindex mno-muladd
16846 Do not use multiply and add/subtract instructions.
16851 Select the FDPIC ABI, which uses function descriptors to represent
16852 pointers to functions. Without any PIC/PIE-related options, it
16853 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
16854 assumes GOT entries and small data are within a 12-bit range from the
16855 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
16856 are computed with 32 bits.
16857 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16860 @opindex minline-plt
16862 Enable inlining of PLT entries in function calls to functions that are
16863 not known to bind locally. It has no effect without @option{-mfdpic}.
16864 It's enabled by default if optimizing for speed and compiling for
16865 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
16866 optimization option such as @option{-O3} or above is present in the
16872 Assume a large TLS segment when generating thread-local code.
16877 Do not assume a large TLS segment when generating thread-local code.
16882 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
16883 that is known to be in read-only sections. It's enabled by default,
16884 except for @option{-fpic} or @option{-fpie}: even though it may help
16885 make the global offset table smaller, it trades 1 instruction for 4.
16886 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
16887 one of which may be shared by multiple symbols, and it avoids the need
16888 for a GOT entry for the referenced symbol, so it's more likely to be a
16889 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
16891 @item -multilib-library-pic
16892 @opindex multilib-library-pic
16894 Link with the (library, not FD) pic libraries. It's implied by
16895 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
16896 @option{-fpic} without @option{-mfdpic}. You should never have to use
16900 @opindex mlinked-fp
16902 Follow the EABI requirement of always creating a frame pointer whenever
16903 a stack frame is allocated. This option is enabled by default and can
16904 be disabled with @option{-mno-linked-fp}.
16907 @opindex mlong-calls
16909 Use indirect addressing to call functions outside the current
16910 compilation unit. This allows the functions to be placed anywhere
16911 within the 32-bit address space.
16913 @item -malign-labels
16914 @opindex malign-labels
16916 Try to align labels to an 8-byte boundary by inserting NOPs into the
16917 previous packet. This option only has an effect when VLIW packing
16918 is enabled. It doesn't create new packets; it merely adds NOPs to
16921 @item -mlibrary-pic
16922 @opindex mlibrary-pic
16924 Generate position-independent EABI code.
16929 Use only the first four media accumulator registers.
16934 Use all eight media accumulator registers.
16939 Pack VLIW instructions.
16944 Do not pack VLIW instructions.
16947 @opindex mno-eflags
16949 Do not mark ABI switches in e_flags.
16952 @opindex mcond-move
16954 Enable the use of conditional-move instructions (default).
16956 This switch is mainly for debugging the compiler and will likely be removed
16957 in a future version.
16959 @item -mno-cond-move
16960 @opindex mno-cond-move
16962 Disable the use of conditional-move instructions.
16964 This switch is mainly for debugging the compiler and will likely be removed
16965 in a future version.
16970 Enable the use of conditional set instructions (default).
16972 This switch is mainly for debugging the compiler and will likely be removed
16973 in a future version.
16978 Disable the use of conditional set instructions.
16980 This switch is mainly for debugging the compiler and will likely be removed
16981 in a future version.
16984 @opindex mcond-exec
16986 Enable the use of conditional execution (default).
16988 This switch is mainly for debugging the compiler and will likely be removed
16989 in a future version.
16991 @item -mno-cond-exec
16992 @opindex mno-cond-exec
16994 Disable the use of conditional execution.
16996 This switch is mainly for debugging the compiler and will likely be removed
16997 in a future version.
16999 @item -mvliw-branch
17000 @opindex mvliw-branch
17002 Run a pass to pack branches into VLIW instructions (default).
17004 This switch is mainly for debugging the compiler and will likely be removed
17005 in a future version.
17007 @item -mno-vliw-branch
17008 @opindex mno-vliw-branch
17010 Do not run a pass to pack branches into VLIW instructions.
17012 This switch is mainly for debugging the compiler and will likely be removed
17013 in a future version.
17015 @item -mmulti-cond-exec
17016 @opindex mmulti-cond-exec
17018 Enable optimization of @code{&&} and @code{||} in conditional execution
17021 This switch is mainly for debugging the compiler and will likely be removed
17022 in a future version.
17024 @item -mno-multi-cond-exec
17025 @opindex mno-multi-cond-exec
17027 Disable optimization of @code{&&} and @code{||} in conditional execution.
17029 This switch is mainly for debugging the compiler and will likely be removed
17030 in a future version.
17032 @item -mnested-cond-exec
17033 @opindex mnested-cond-exec
17035 Enable nested conditional execution optimizations (default).
17037 This switch is mainly for debugging the compiler and will likely be removed
17038 in a future version.
17040 @item -mno-nested-cond-exec
17041 @opindex mno-nested-cond-exec
17043 Disable nested conditional execution optimizations.
17045 This switch is mainly for debugging the compiler and will likely be removed
17046 in a future version.
17048 @item -moptimize-membar
17049 @opindex moptimize-membar
17051 This switch removes redundant @code{membar} instructions from the
17052 compiler-generated code. It is enabled by default.
17054 @item -mno-optimize-membar
17055 @opindex mno-optimize-membar
17057 This switch disables the automatic removal of redundant @code{membar}
17058 instructions from the generated code.
17060 @item -mtomcat-stats
17061 @opindex mtomcat-stats
17063 Cause gas to print out tomcat statistics.
17065 @item -mcpu=@var{cpu}
17068 Select the processor type for which to generate code. Possible values are
17069 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
17070 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
17074 @node GNU/Linux Options
17075 @subsection GNU/Linux Options
17077 These @samp{-m} options are defined for GNU/Linux targets:
17082 Use the GNU C library. This is the default except
17083 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
17084 @samp{*-*-linux-*android*} targets.
17088 Use uClibc C library. This is the default on
17089 @samp{*-*-linux-*uclibc*} targets.
17093 Use the musl C library. This is the default on
17094 @samp{*-*-linux-*musl*} targets.
17098 Use Bionic C library. This is the default on
17099 @samp{*-*-linux-*android*} targets.
17103 Compile code compatible with Android platform. This is the default on
17104 @samp{*-*-linux-*android*} targets.
17106 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
17107 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
17108 this option makes the GCC driver pass Android-specific options to the linker.
17109 Finally, this option causes the preprocessor macro @code{__ANDROID__}
17112 @item -tno-android-cc
17113 @opindex tno-android-cc
17114 Disable compilation effects of @option{-mandroid}, i.e., do not enable
17115 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
17116 @option{-fno-rtti} by default.
17118 @item -tno-android-ld
17119 @opindex tno-android-ld
17120 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
17121 linking options to the linker.
17125 @node H8/300 Options
17126 @subsection H8/300 Options
17128 These @samp{-m} options are defined for the H8/300 implementations:
17133 Shorten some address references at link time, when possible; uses the
17134 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
17135 ld, Using ld}, for a fuller description.
17139 Generate code for the H8/300H@.
17143 Generate code for the H8S@.
17147 Generate code for the H8S and H8/300H in the normal mode. This switch
17148 must be used either with @option{-mh} or @option{-ms}.
17152 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
17156 Extended registers are stored on stack before execution of function
17157 with monitor attribute. Default option is @option{-mexr}.
17158 This option is valid only for H8S targets.
17162 Extended registers are not stored on stack before execution of function
17163 with monitor attribute. Default option is @option{-mno-exr}.
17164 This option is valid only for H8S targets.
17168 Make @code{int} data 32 bits by default.
17171 @opindex malign-300
17172 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
17173 The default for the H8/300H and H8S is to align longs and floats on
17175 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
17176 This option has no effect on the H8/300.
17180 @subsection HPPA Options
17181 @cindex HPPA Options
17183 These @samp{-m} options are defined for the HPPA family of computers:
17186 @item -march=@var{architecture-type}
17188 Generate code for the specified architecture. The choices for
17189 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
17190 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
17191 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
17192 architecture option for your machine. Code compiled for lower numbered
17193 architectures runs on higher numbered architectures, but not the
17196 @item -mpa-risc-1-0
17197 @itemx -mpa-risc-1-1
17198 @itemx -mpa-risc-2-0
17199 @opindex mpa-risc-1-0
17200 @opindex mpa-risc-1-1
17201 @opindex mpa-risc-2-0
17202 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
17204 @item -mjump-in-delay
17205 @opindex mjump-in-delay
17206 This option is ignored and provided for compatibility purposes only.
17208 @item -mdisable-fpregs
17209 @opindex mdisable-fpregs
17210 Prevent floating-point registers from being used in any manner. This is
17211 necessary for compiling kernels that perform lazy context switching of
17212 floating-point registers. If you use this option and attempt to perform
17213 floating-point operations, the compiler aborts.
17215 @item -mdisable-indexing
17216 @opindex mdisable-indexing
17217 Prevent the compiler from using indexing address modes. This avoids some
17218 rather obscure problems when compiling MIG generated code under MACH@.
17220 @item -mno-space-regs
17221 @opindex mno-space-regs
17222 Generate code that assumes the target has no space registers. This allows
17223 GCC to generate faster indirect calls and use unscaled index address modes.
17225 Such code is suitable for level 0 PA systems and kernels.
17227 @item -mfast-indirect-calls
17228 @opindex mfast-indirect-calls
17229 Generate code that assumes calls never cross space boundaries. This
17230 allows GCC to emit code that performs faster indirect calls.
17232 This option does not work in the presence of shared libraries or nested
17235 @item -mfixed-range=@var{register-range}
17236 @opindex mfixed-range
17237 Generate code treating the given register range as fixed registers.
17238 A fixed register is one that the register allocator cannot use. This is
17239 useful when compiling kernel code. A register range is specified as
17240 two registers separated by a dash. Multiple register ranges can be
17241 specified separated by a comma.
17243 @item -mlong-load-store
17244 @opindex mlong-load-store
17245 Generate 3-instruction load and store sequences as sometimes required by
17246 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
17249 @item -mportable-runtime
17250 @opindex mportable-runtime
17251 Use the portable calling conventions proposed by HP for ELF systems.
17255 Enable the use of assembler directives only GAS understands.
17257 @item -mschedule=@var{cpu-type}
17259 Schedule code according to the constraints for the machine type
17260 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
17261 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
17262 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
17263 proper scheduling option for your machine. The default scheduling is
17267 @opindex mlinker-opt
17268 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
17269 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
17270 linkers in which they give bogus error messages when linking some programs.
17273 @opindex msoft-float
17274 Generate output containing library calls for floating point.
17275 @strong{Warning:} the requisite libraries are not available for all HPPA
17276 targets. Normally the facilities of the machine's usual C compiler are
17277 used, but this cannot be done directly in cross-compilation. You must make
17278 your own arrangements to provide suitable library functions for
17281 @option{-msoft-float} changes the calling convention in the output file;
17282 therefore, it is only useful if you compile @emph{all} of a program with
17283 this option. In particular, you need to compile @file{libgcc.a}, the
17284 library that comes with GCC, with @option{-msoft-float} in order for
17289 Generate the predefine, @code{_SIO}, for server IO@. The default is
17290 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
17291 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
17292 options are available under HP-UX and HI-UX@.
17296 Use options specific to GNU @command{ld}.
17297 This passes @option{-shared} to @command{ld} when
17298 building a shared library. It is the default when GCC is configured,
17299 explicitly or implicitly, with the GNU linker. This option does not
17300 affect which @command{ld} is called; it only changes what parameters
17301 are passed to that @command{ld}.
17302 The @command{ld} that is called is determined by the
17303 @option{--with-ld} configure option, GCC's program search path, and
17304 finally by the user's @env{PATH}. The linker used by GCC can be printed
17305 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
17306 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17310 Use options specific to HP @command{ld}.
17311 This passes @option{-b} to @command{ld} when building
17312 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
17313 links. It is the default when GCC is configured, explicitly or
17314 implicitly, with the HP linker. This option does not affect
17315 which @command{ld} is called; it only changes what parameters are passed to that
17317 The @command{ld} that is called is determined by the @option{--with-ld}
17318 configure option, GCC's program search path, and finally by the user's
17319 @env{PATH}. The linker used by GCC can be printed using @samp{which
17320 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
17321 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17324 @opindex mno-long-calls
17325 Generate code that uses long call sequences. This ensures that a call
17326 is always able to reach linker generated stubs. The default is to generate
17327 long calls only when the distance from the call site to the beginning
17328 of the function or translation unit, as the case may be, exceeds a
17329 predefined limit set by the branch type being used. The limits for
17330 normal calls are 7,600,000 and 240,000 bytes, respectively for the
17331 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
17334 Distances are measured from the beginning of functions when using the
17335 @option{-ffunction-sections} option, or when using the @option{-mgas}
17336 and @option{-mno-portable-runtime} options together under HP-UX with
17339 It is normally not desirable to use this option as it degrades
17340 performance. However, it may be useful in large applications,
17341 particularly when partial linking is used to build the application.
17343 The types of long calls used depends on the capabilities of the
17344 assembler and linker, and the type of code being generated. The
17345 impact on systems that support long absolute calls, and long pic
17346 symbol-difference or pc-relative calls should be relatively small.
17347 However, an indirect call is used on 32-bit ELF systems in pic code
17348 and it is quite long.
17350 @item -munix=@var{unix-std}
17352 Generate compiler predefines and select a startfile for the specified
17353 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
17354 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
17355 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
17356 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
17357 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
17360 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
17361 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
17362 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
17363 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
17364 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
17365 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
17367 It is @emph{important} to note that this option changes the interfaces
17368 for various library routines. It also affects the operational behavior
17369 of the C library. Thus, @emph{extreme} care is needed in using this
17372 Library code that is intended to operate with more than one UNIX
17373 standard must test, set and restore the variable @code{__xpg4_extended_mask}
17374 as appropriate. Most GNU software doesn't provide this capability.
17378 Suppress the generation of link options to search libdld.sl when the
17379 @option{-static} option is specified on HP-UX 10 and later.
17383 The HP-UX implementation of setlocale in libc has a dependency on
17384 libdld.sl. There isn't an archive version of libdld.sl. Thus,
17385 when the @option{-static} option is specified, special link options
17386 are needed to resolve this dependency.
17388 On HP-UX 10 and later, the GCC driver adds the necessary options to
17389 link with libdld.sl when the @option{-static} option is specified.
17390 This causes the resulting binary to be dynamic. On the 64-bit port,
17391 the linkers generate dynamic binaries by default in any case. The
17392 @option{-nolibdld} option can be used to prevent the GCC driver from
17393 adding these link options.
17397 Add support for multithreading with the @dfn{dce thread} library
17398 under HP-UX@. This option sets flags for both the preprocessor and
17402 @node IA-64 Options
17403 @subsection IA-64 Options
17404 @cindex IA-64 Options
17406 These are the @samp{-m} options defined for the Intel IA-64 architecture.
17410 @opindex mbig-endian
17411 Generate code for a big-endian target. This is the default for HP-UX@.
17413 @item -mlittle-endian
17414 @opindex mlittle-endian
17415 Generate code for a little-endian target. This is the default for AIX5
17421 @opindex mno-gnu-as
17422 Generate (or don't) code for the GNU assembler. This is the default.
17423 @c Also, this is the default if the configure option @option{--with-gnu-as}
17429 @opindex mno-gnu-ld
17430 Generate (or don't) code for the GNU linker. This is the default.
17431 @c Also, this is the default if the configure option @option{--with-gnu-ld}
17436 Generate code that does not use a global pointer register. The result
17437 is not position independent code, and violates the IA-64 ABI@.
17439 @item -mvolatile-asm-stop
17440 @itemx -mno-volatile-asm-stop
17441 @opindex mvolatile-asm-stop
17442 @opindex mno-volatile-asm-stop
17443 Generate (or don't) a stop bit immediately before and after volatile asm
17446 @item -mregister-names
17447 @itemx -mno-register-names
17448 @opindex mregister-names
17449 @opindex mno-register-names
17450 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
17451 the stacked registers. This may make assembler output more readable.
17457 Disable (or enable) optimizations that use the small data section. This may
17458 be useful for working around optimizer bugs.
17460 @item -mconstant-gp
17461 @opindex mconstant-gp
17462 Generate code that uses a single constant global pointer value. This is
17463 useful when compiling kernel code.
17467 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
17468 This is useful when compiling firmware code.
17470 @item -minline-float-divide-min-latency
17471 @opindex minline-float-divide-min-latency
17472 Generate code for inline divides of floating-point values
17473 using the minimum latency algorithm.
17475 @item -minline-float-divide-max-throughput
17476 @opindex minline-float-divide-max-throughput
17477 Generate code for inline divides of floating-point values
17478 using the maximum throughput algorithm.
17480 @item -mno-inline-float-divide
17481 @opindex mno-inline-float-divide
17482 Do not generate inline code for divides of floating-point values.
17484 @item -minline-int-divide-min-latency
17485 @opindex minline-int-divide-min-latency
17486 Generate code for inline divides of integer values
17487 using the minimum latency algorithm.
17489 @item -minline-int-divide-max-throughput
17490 @opindex minline-int-divide-max-throughput
17491 Generate code for inline divides of integer values
17492 using the maximum throughput algorithm.
17494 @item -mno-inline-int-divide
17495 @opindex mno-inline-int-divide
17496 Do not generate inline code for divides of integer values.
17498 @item -minline-sqrt-min-latency
17499 @opindex minline-sqrt-min-latency
17500 Generate code for inline square roots
17501 using the minimum latency algorithm.
17503 @item -minline-sqrt-max-throughput
17504 @opindex minline-sqrt-max-throughput
17505 Generate code for inline square roots
17506 using the maximum throughput algorithm.
17508 @item -mno-inline-sqrt
17509 @opindex mno-inline-sqrt
17510 Do not generate inline code for @code{sqrt}.
17513 @itemx -mno-fused-madd
17514 @opindex mfused-madd
17515 @opindex mno-fused-madd
17516 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
17517 instructions. The default is to use these instructions.
17519 @item -mno-dwarf2-asm
17520 @itemx -mdwarf2-asm
17521 @opindex mno-dwarf2-asm
17522 @opindex mdwarf2-asm
17523 Don't (or do) generate assembler code for the DWARF line number debugging
17524 info. This may be useful when not using the GNU assembler.
17526 @item -mearly-stop-bits
17527 @itemx -mno-early-stop-bits
17528 @opindex mearly-stop-bits
17529 @opindex mno-early-stop-bits
17530 Allow stop bits to be placed earlier than immediately preceding the
17531 instruction that triggered the stop bit. This can improve instruction
17532 scheduling, but does not always do so.
17534 @item -mfixed-range=@var{register-range}
17535 @opindex mfixed-range
17536 Generate code treating the given register range as fixed registers.
17537 A fixed register is one that the register allocator cannot use. This is
17538 useful when compiling kernel code. A register range is specified as
17539 two registers separated by a dash. Multiple register ranges can be
17540 specified separated by a comma.
17542 @item -mtls-size=@var{tls-size}
17544 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
17547 @item -mtune=@var{cpu-type}
17549 Tune the instruction scheduling for a particular CPU, Valid values are
17550 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
17551 and @samp{mckinley}.
17557 Generate code for a 32-bit or 64-bit environment.
17558 The 32-bit environment sets int, long and pointer to 32 bits.
17559 The 64-bit environment sets int to 32 bits and long and pointer
17560 to 64 bits. These are HP-UX specific flags.
17562 @item -mno-sched-br-data-spec
17563 @itemx -msched-br-data-spec
17564 @opindex mno-sched-br-data-spec
17565 @opindex msched-br-data-spec
17566 (Dis/En)able data speculative scheduling before reload.
17567 This results in generation of @code{ld.a} instructions and
17568 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17569 The default setting is disabled.
17571 @item -msched-ar-data-spec
17572 @itemx -mno-sched-ar-data-spec
17573 @opindex msched-ar-data-spec
17574 @opindex mno-sched-ar-data-spec
17575 (En/Dis)able data speculative scheduling after reload.
17576 This results in generation of @code{ld.a} instructions and
17577 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17578 The default setting is enabled.
17580 @item -mno-sched-control-spec
17581 @itemx -msched-control-spec
17582 @opindex mno-sched-control-spec
17583 @opindex msched-control-spec
17584 (Dis/En)able control speculative scheduling. This feature is
17585 available only during region scheduling (i.e.@: before reload).
17586 This results in generation of the @code{ld.s} instructions and
17587 the corresponding check instructions @code{chk.s}.
17588 The default setting is disabled.
17590 @item -msched-br-in-data-spec
17591 @itemx -mno-sched-br-in-data-spec
17592 @opindex msched-br-in-data-spec
17593 @opindex mno-sched-br-in-data-spec
17594 (En/Dis)able speculative scheduling of the instructions that
17595 are dependent on the data speculative loads before reload.
17596 This is effective only with @option{-msched-br-data-spec} enabled.
17597 The default setting is enabled.
17599 @item -msched-ar-in-data-spec
17600 @itemx -mno-sched-ar-in-data-spec
17601 @opindex msched-ar-in-data-spec
17602 @opindex mno-sched-ar-in-data-spec
17603 (En/Dis)able speculative scheduling of the instructions that
17604 are dependent on the data speculative loads after reload.
17605 This is effective only with @option{-msched-ar-data-spec} enabled.
17606 The default setting is enabled.
17608 @item -msched-in-control-spec
17609 @itemx -mno-sched-in-control-spec
17610 @opindex msched-in-control-spec
17611 @opindex mno-sched-in-control-spec
17612 (En/Dis)able speculative scheduling of the instructions that
17613 are dependent on the control speculative loads.
17614 This is effective only with @option{-msched-control-spec} enabled.
17615 The default setting is enabled.
17617 @item -mno-sched-prefer-non-data-spec-insns
17618 @itemx -msched-prefer-non-data-spec-insns
17619 @opindex mno-sched-prefer-non-data-spec-insns
17620 @opindex msched-prefer-non-data-spec-insns
17621 If enabled, data-speculative instructions are chosen for schedule
17622 only if there are no other choices at the moment. This makes
17623 the use of the data speculation much more conservative.
17624 The default setting is disabled.
17626 @item -mno-sched-prefer-non-control-spec-insns
17627 @itemx -msched-prefer-non-control-spec-insns
17628 @opindex mno-sched-prefer-non-control-spec-insns
17629 @opindex msched-prefer-non-control-spec-insns
17630 If enabled, control-speculative instructions are chosen for schedule
17631 only if there are no other choices at the moment. This makes
17632 the use of the control speculation much more conservative.
17633 The default setting is disabled.
17635 @item -mno-sched-count-spec-in-critical-path
17636 @itemx -msched-count-spec-in-critical-path
17637 @opindex mno-sched-count-spec-in-critical-path
17638 @opindex msched-count-spec-in-critical-path
17639 If enabled, speculative dependencies are considered during
17640 computation of the instructions priorities. This makes the use of the
17641 speculation a bit more conservative.
17642 The default setting is disabled.
17644 @item -msched-spec-ldc
17645 @opindex msched-spec-ldc
17646 Use a simple data speculation check. This option is on by default.
17648 @item -msched-control-spec-ldc
17649 @opindex msched-spec-ldc
17650 Use a simple check for control speculation. This option is on by default.
17652 @item -msched-stop-bits-after-every-cycle
17653 @opindex msched-stop-bits-after-every-cycle
17654 Place a stop bit after every cycle when scheduling. This option is on
17657 @item -msched-fp-mem-deps-zero-cost
17658 @opindex msched-fp-mem-deps-zero-cost
17659 Assume that floating-point stores and loads are not likely to cause a conflict
17660 when placed into the same instruction group. This option is disabled by
17663 @item -msel-sched-dont-check-control-spec
17664 @opindex msel-sched-dont-check-control-spec
17665 Generate checks for control speculation in selective scheduling.
17666 This flag is disabled by default.
17668 @item -msched-max-memory-insns=@var{max-insns}
17669 @opindex msched-max-memory-insns
17670 Limit on the number of memory insns per instruction group, giving lower
17671 priority to subsequent memory insns attempting to schedule in the same
17672 instruction group. Frequently useful to prevent cache bank conflicts.
17673 The default value is 1.
17675 @item -msched-max-memory-insns-hard-limit
17676 @opindex msched-max-memory-insns-hard-limit
17677 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
17678 disallowing more than that number in an instruction group.
17679 Otherwise, the limit is ``soft'', meaning that non-memory operations
17680 are preferred when the limit is reached, but memory operations may still
17686 @subsection LM32 Options
17687 @cindex LM32 options
17689 These @option{-m} options are defined for the LatticeMico32 architecture:
17692 @item -mbarrel-shift-enabled
17693 @opindex mbarrel-shift-enabled
17694 Enable barrel-shift instructions.
17696 @item -mdivide-enabled
17697 @opindex mdivide-enabled
17698 Enable divide and modulus instructions.
17700 @item -mmultiply-enabled
17701 @opindex multiply-enabled
17702 Enable multiply instructions.
17704 @item -msign-extend-enabled
17705 @opindex msign-extend-enabled
17706 Enable sign extend instructions.
17708 @item -muser-enabled
17709 @opindex muser-enabled
17710 Enable user-defined instructions.
17715 @subsection M32C Options
17716 @cindex M32C options
17719 @item -mcpu=@var{name}
17721 Select the CPU for which code is generated. @var{name} may be one of
17722 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
17723 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
17724 the M32C/80 series.
17728 Specifies that the program will be run on the simulator. This causes
17729 an alternate runtime library to be linked in which supports, for
17730 example, file I/O@. You must not use this option when generating
17731 programs that will run on real hardware; you must provide your own
17732 runtime library for whatever I/O functions are needed.
17734 @item -memregs=@var{number}
17736 Specifies the number of memory-based pseudo-registers GCC uses
17737 during code generation. These pseudo-registers are used like real
17738 registers, so there is a tradeoff between GCC's ability to fit the
17739 code into available registers, and the performance penalty of using
17740 memory instead of registers. Note that all modules in a program must
17741 be compiled with the same value for this option. Because of that, you
17742 must not use this option with GCC's default runtime libraries.
17746 @node M32R/D Options
17747 @subsection M32R/D Options
17748 @cindex M32R/D options
17750 These @option{-m} options are defined for Renesas M32R/D architectures:
17755 Generate code for the M32R/2@.
17759 Generate code for the M32R/X@.
17763 Generate code for the M32R@. This is the default.
17765 @item -mmodel=small
17766 @opindex mmodel=small
17767 Assume all objects live in the lower 16MB of memory (so that their addresses
17768 can be loaded with the @code{ld24} instruction), and assume all subroutines
17769 are reachable with the @code{bl} instruction.
17770 This is the default.
17772 The addressability of a particular object can be set with the
17773 @code{model} attribute.
17775 @item -mmodel=medium
17776 @opindex mmodel=medium
17777 Assume objects may be anywhere in the 32-bit address space (the compiler
17778 generates @code{seth/add3} instructions to load their addresses), and
17779 assume all subroutines are reachable with the @code{bl} instruction.
17781 @item -mmodel=large
17782 @opindex mmodel=large
17783 Assume objects may be anywhere in the 32-bit address space (the compiler
17784 generates @code{seth/add3} instructions to load their addresses), and
17785 assume subroutines may not be reachable with the @code{bl} instruction
17786 (the compiler generates the much slower @code{seth/add3/jl}
17787 instruction sequence).
17790 @opindex msdata=none
17791 Disable use of the small data area. Variables are put into
17792 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
17793 @code{section} attribute has been specified).
17794 This is the default.
17796 The small data area consists of sections @code{.sdata} and @code{.sbss}.
17797 Objects may be explicitly put in the small data area with the
17798 @code{section} attribute using one of these sections.
17800 @item -msdata=sdata
17801 @opindex msdata=sdata
17802 Put small global and static data in the small data area, but do not
17803 generate special code to reference them.
17806 @opindex msdata=use
17807 Put small global and static data in the small data area, and generate
17808 special instructions to reference them.
17812 @cindex smaller data references
17813 Put global and static objects less than or equal to @var{num} bytes
17814 into the small data or BSS sections instead of the normal data or BSS
17815 sections. The default value of @var{num} is 8.
17816 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
17817 for this option to have any effect.
17819 All modules should be compiled with the same @option{-G @var{num}} value.
17820 Compiling with different values of @var{num} may or may not work; if it
17821 doesn't the linker gives an error message---incorrect code is not
17826 Makes the M32R-specific code in the compiler display some statistics
17827 that might help in debugging programs.
17829 @item -malign-loops
17830 @opindex malign-loops
17831 Align all loops to a 32-byte boundary.
17833 @item -mno-align-loops
17834 @opindex mno-align-loops
17835 Do not enforce a 32-byte alignment for loops. This is the default.
17837 @item -missue-rate=@var{number}
17838 @opindex missue-rate=@var{number}
17839 Issue @var{number} instructions per cycle. @var{number} can only be 1
17842 @item -mbranch-cost=@var{number}
17843 @opindex mbranch-cost=@var{number}
17844 @var{number} can only be 1 or 2. If it is 1 then branches are
17845 preferred over conditional code, if it is 2, then the opposite applies.
17847 @item -mflush-trap=@var{number}
17848 @opindex mflush-trap=@var{number}
17849 Specifies the trap number to use to flush the cache. The default is
17850 12. Valid numbers are between 0 and 15 inclusive.
17852 @item -mno-flush-trap
17853 @opindex mno-flush-trap
17854 Specifies that the cache cannot be flushed by using a trap.
17856 @item -mflush-func=@var{name}
17857 @opindex mflush-func=@var{name}
17858 Specifies the name of the operating system function to call to flush
17859 the cache. The default is @samp{_flush_cache}, but a function call
17860 is only used if a trap is not available.
17862 @item -mno-flush-func
17863 @opindex mno-flush-func
17864 Indicates that there is no OS function for flushing the cache.
17868 @node M680x0 Options
17869 @subsection M680x0 Options
17870 @cindex M680x0 options
17872 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
17873 The default settings depend on which architecture was selected when
17874 the compiler was configured; the defaults for the most common choices
17878 @item -march=@var{arch}
17880 Generate code for a specific M680x0 or ColdFire instruction set
17881 architecture. Permissible values of @var{arch} for M680x0
17882 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
17883 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
17884 architectures are selected according to Freescale's ISA classification
17885 and the permissible values are: @samp{isaa}, @samp{isaaplus},
17886 @samp{isab} and @samp{isac}.
17888 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
17889 code for a ColdFire target. The @var{arch} in this macro is one of the
17890 @option{-march} arguments given above.
17892 When used together, @option{-march} and @option{-mtune} select code
17893 that runs on a family of similar processors but that is optimized
17894 for a particular microarchitecture.
17896 @item -mcpu=@var{cpu}
17898 Generate code for a specific M680x0 or ColdFire processor.
17899 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
17900 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
17901 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
17902 below, which also classifies the CPUs into families:
17904 @multitable @columnfractions 0.20 0.80
17905 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
17906 @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}
17907 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
17908 @item @samp{5206e} @tab @samp{5206e}
17909 @item @samp{5208} @tab @samp{5207} @samp{5208}
17910 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
17911 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
17912 @item @samp{5216} @tab @samp{5214} @samp{5216}
17913 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
17914 @item @samp{5225} @tab @samp{5224} @samp{5225}
17915 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17916 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17917 @item @samp{5249} @tab @samp{5249}
17918 @item @samp{5250} @tab @samp{5250}
17919 @item @samp{5271} @tab @samp{5270} @samp{5271}
17920 @item @samp{5272} @tab @samp{5272}
17921 @item @samp{5275} @tab @samp{5274} @samp{5275}
17922 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17923 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17924 @item @samp{5307} @tab @samp{5307}
17925 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17926 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17927 @item @samp{5407} @tab @samp{5407}
17928 @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}
17931 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17932 @var{arch} is compatible with @var{cpu}. Other combinations of
17933 @option{-mcpu} and @option{-march} are rejected.
17935 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17936 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17937 where the value of @var{family} is given by the table above.
17939 @item -mtune=@var{tune}
17941 Tune the code for a particular microarchitecture within the
17942 constraints set by @option{-march} and @option{-mcpu}.
17943 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17944 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17945 and @samp{cpu32}. The ColdFire microarchitectures
17946 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17948 You can also use @option{-mtune=68020-40} for code that needs
17949 to run relatively well on 68020, 68030 and 68040 targets.
17950 @option{-mtune=68020-60} is similar but includes 68060 targets
17951 as well. These two options select the same tuning decisions as
17952 @option{-m68020-40} and @option{-m68020-60} respectively.
17954 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17955 when tuning for 680x0 architecture @var{arch}. It also defines
17956 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17957 option is used. If GCC is tuning for a range of architectures,
17958 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17959 it defines the macros for every architecture in the range.
17961 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17962 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17963 of the arguments given above.
17969 Generate output for a 68000. This is the default
17970 when the compiler is configured for 68000-based systems.
17971 It is equivalent to @option{-march=68000}.
17973 Use this option for microcontrollers with a 68000 or EC000 core,
17974 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17978 Generate output for a 68010. This is the default
17979 when the compiler is configured for 68010-based systems.
17980 It is equivalent to @option{-march=68010}.
17986 Generate output for a 68020. This is the default
17987 when the compiler is configured for 68020-based systems.
17988 It is equivalent to @option{-march=68020}.
17992 Generate output for a 68030. This is the default when the compiler is
17993 configured for 68030-based systems. It is equivalent to
17994 @option{-march=68030}.
17998 Generate output for a 68040. This is the default when the compiler is
17999 configured for 68040-based systems. It is equivalent to
18000 @option{-march=68040}.
18002 This option inhibits the use of 68881/68882 instructions that have to be
18003 emulated by software on the 68040. Use this option if your 68040 does not
18004 have code to emulate those instructions.
18008 Generate output for a 68060. This is the default when the compiler is
18009 configured for 68060-based systems. It is equivalent to
18010 @option{-march=68060}.
18012 This option inhibits the use of 68020 and 68881/68882 instructions that
18013 have to be emulated by software on the 68060. Use this option if your 68060
18014 does not have code to emulate those instructions.
18018 Generate output for a CPU32. This is the default
18019 when the compiler is configured for CPU32-based systems.
18020 It is equivalent to @option{-march=cpu32}.
18022 Use this option for microcontrollers with a
18023 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
18024 68336, 68340, 68341, 68349 and 68360.
18028 Generate output for a 520X ColdFire CPU@. This is the default
18029 when the compiler is configured for 520X-based systems.
18030 It is equivalent to @option{-mcpu=5206}, and is now deprecated
18031 in favor of that option.
18033 Use this option for microcontroller with a 5200 core, including
18034 the MCF5202, MCF5203, MCF5204 and MCF5206.
18038 Generate output for a 5206e ColdFire CPU@. The option is now
18039 deprecated in favor of the equivalent @option{-mcpu=5206e}.
18043 Generate output for a member of the ColdFire 528X family.
18044 The option is now deprecated in favor of the equivalent
18045 @option{-mcpu=528x}.
18049 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
18050 in favor of the equivalent @option{-mcpu=5307}.
18054 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
18055 in favor of the equivalent @option{-mcpu=5407}.
18059 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
18060 This includes use of hardware floating-point instructions.
18061 The option is equivalent to @option{-mcpu=547x}, and is now
18062 deprecated in favor of that option.
18066 Generate output for a 68040, without using any of the new instructions.
18067 This results in code that can run relatively efficiently on either a
18068 68020/68881 or a 68030 or a 68040. The generated code does use the
18069 68881 instructions that are emulated on the 68040.
18071 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
18075 Generate output for a 68060, without using any of the new instructions.
18076 This results in code that can run relatively efficiently on either a
18077 68020/68881 or a 68030 or a 68040. The generated code does use the
18078 68881 instructions that are emulated on the 68060.
18080 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
18084 @opindex mhard-float
18086 Generate floating-point instructions. This is the default for 68020
18087 and above, and for ColdFire devices that have an FPU@. It defines the
18088 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
18089 on ColdFire targets.
18092 @opindex msoft-float
18093 Do not generate floating-point instructions; use library calls instead.
18094 This is the default for 68000, 68010, and 68832 targets. It is also
18095 the default for ColdFire devices that have no FPU.
18101 Generate (do not generate) ColdFire hardware divide and remainder
18102 instructions. If @option{-march} is used without @option{-mcpu},
18103 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
18104 architectures. Otherwise, the default is taken from the target CPU
18105 (either the default CPU, or the one specified by @option{-mcpu}). For
18106 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
18107 @option{-mcpu=5206e}.
18109 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
18113 Consider type @code{int} to be 16 bits wide, like @code{short int}.
18114 Additionally, parameters passed on the stack are also aligned to a
18115 16-bit boundary even on targets whose API mandates promotion to 32-bit.
18119 Do not consider type @code{int} to be 16 bits wide. This is the default.
18122 @itemx -mno-bitfield
18123 @opindex mnobitfield
18124 @opindex mno-bitfield
18125 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
18126 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
18130 Do use the bit-field instructions. The @option{-m68020} option implies
18131 @option{-mbitfield}. This is the default if you use a configuration
18132 designed for a 68020.
18136 Use a different function-calling convention, in which functions
18137 that take a fixed number of arguments return with the @code{rtd}
18138 instruction, which pops their arguments while returning. This
18139 saves one instruction in the caller since there is no need to pop
18140 the arguments there.
18142 This calling convention is incompatible with the one normally
18143 used on Unix, so you cannot use it if you need to call libraries
18144 compiled with the Unix compiler.
18146 Also, you must provide function prototypes for all functions that
18147 take variable numbers of arguments (including @code{printf});
18148 otherwise incorrect code is generated for calls to those
18151 In addition, seriously incorrect code results if you call a
18152 function with too many arguments. (Normally, extra arguments are
18153 harmlessly ignored.)
18155 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
18156 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
18160 Do not use the calling conventions selected by @option{-mrtd}.
18161 This is the default.
18164 @itemx -mno-align-int
18165 @opindex malign-int
18166 @opindex mno-align-int
18167 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
18168 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
18169 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
18170 Aligning variables on 32-bit boundaries produces code that runs somewhat
18171 faster on processors with 32-bit busses at the expense of more memory.
18173 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
18174 aligns structures containing the above types differently than
18175 most published application binary interface specifications for the m68k.
18179 Use the pc-relative addressing mode of the 68000 directly, instead of
18180 using a global offset table. At present, this option implies @option{-fpic},
18181 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
18182 not presently supported with @option{-mpcrel}, though this could be supported for
18183 68020 and higher processors.
18185 @item -mno-strict-align
18186 @itemx -mstrict-align
18187 @opindex mno-strict-align
18188 @opindex mstrict-align
18189 Do not (do) assume that unaligned memory references are handled by
18193 Generate code that allows the data segment to be located in a different
18194 area of memory from the text segment. This allows for execute-in-place in
18195 an environment without virtual memory management. This option implies
18198 @item -mno-sep-data
18199 Generate code that assumes that the data segment follows the text segment.
18200 This is the default.
18202 @item -mid-shared-library
18203 Generate code that supports shared libraries via the library ID method.
18204 This allows for execute-in-place and shared libraries in an environment
18205 without virtual memory management. This option implies @option{-fPIC}.
18207 @item -mno-id-shared-library
18208 Generate code that doesn't assume ID-based shared libraries are being used.
18209 This is the default.
18211 @item -mshared-library-id=n
18212 Specifies the identification number of the ID-based shared library being
18213 compiled. Specifying a value of 0 generates more compact code; specifying
18214 other values forces the allocation of that number to the current
18215 library, but is no more space- or time-efficient than omitting this option.
18221 When generating position-independent code for ColdFire, generate code
18222 that works if the GOT has more than 8192 entries. This code is
18223 larger and slower than code generated without this option. On M680x0
18224 processors, this option is not needed; @option{-fPIC} suffices.
18226 GCC normally uses a single instruction to load values from the GOT@.
18227 While this is relatively efficient, it only works if the GOT
18228 is smaller than about 64k. Anything larger causes the linker
18229 to report an error such as:
18231 @cindex relocation truncated to fit (ColdFire)
18233 relocation truncated to fit: R_68K_GOT16O foobar
18236 If this happens, you should recompile your code with @option{-mxgot}.
18237 It should then work with very large GOTs. However, code generated with
18238 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
18239 the value of a global symbol.
18241 Note that some linkers, including newer versions of the GNU linker,
18242 can create multiple GOTs and sort GOT entries. If you have such a linker,
18243 you should only need to use @option{-mxgot} when compiling a single
18244 object file that accesses more than 8192 GOT entries. Very few do.
18246 These options have no effect unless GCC is generating
18247 position-independent code.
18251 @node MCore Options
18252 @subsection MCore Options
18253 @cindex MCore options
18255 These are the @samp{-m} options defined for the Motorola M*Core
18261 @itemx -mno-hardlit
18263 @opindex mno-hardlit
18264 Inline constants into the code stream if it can be done in two
18265 instructions or less.
18271 Use the divide instruction. (Enabled by default).
18273 @item -mrelax-immediate
18274 @itemx -mno-relax-immediate
18275 @opindex mrelax-immediate
18276 @opindex mno-relax-immediate
18277 Allow arbitrary-sized immediates in bit operations.
18279 @item -mwide-bitfields
18280 @itemx -mno-wide-bitfields
18281 @opindex mwide-bitfields
18282 @opindex mno-wide-bitfields
18283 Always treat bit-fields as @code{int}-sized.
18285 @item -m4byte-functions
18286 @itemx -mno-4byte-functions
18287 @opindex m4byte-functions
18288 @opindex mno-4byte-functions
18289 Force all functions to be aligned to a 4-byte boundary.
18291 @item -mcallgraph-data
18292 @itemx -mno-callgraph-data
18293 @opindex mcallgraph-data
18294 @opindex mno-callgraph-data
18295 Emit callgraph information.
18298 @itemx -mno-slow-bytes
18299 @opindex mslow-bytes
18300 @opindex mno-slow-bytes
18301 Prefer word access when reading byte quantities.
18303 @item -mlittle-endian
18304 @itemx -mbig-endian
18305 @opindex mlittle-endian
18306 @opindex mbig-endian
18307 Generate code for a little-endian target.
18313 Generate code for the 210 processor.
18317 Assume that runtime support has been provided and so omit the
18318 simulator library (@file{libsim.a)} from the linker command line.
18320 @item -mstack-increment=@var{size}
18321 @opindex mstack-increment
18322 Set the maximum amount for a single stack increment operation. Large
18323 values can increase the speed of programs that contain functions
18324 that need a large amount of stack space, but they can also trigger a
18325 segmentation fault if the stack is extended too much. The default
18331 @subsection MeP Options
18332 @cindex MeP options
18338 Enables the @code{abs} instruction, which is the absolute difference
18339 between two registers.
18343 Enables all the optional instructions---average, multiply, divide, bit
18344 operations, leading zero, absolute difference, min/max, clip, and
18350 Enables the @code{ave} instruction, which computes the average of two
18353 @item -mbased=@var{n}
18355 Variables of size @var{n} bytes or smaller are placed in the
18356 @code{.based} section by default. Based variables use the @code{$tp}
18357 register as a base register, and there is a 128-byte limit to the
18358 @code{.based} section.
18362 Enables the bit operation instructions---bit test (@code{btstm}), set
18363 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
18364 test-and-set (@code{tas}).
18366 @item -mc=@var{name}
18368 Selects which section constant data is placed in. @var{name} may
18369 be @samp{tiny}, @samp{near}, or @samp{far}.
18373 Enables the @code{clip} instruction. Note that @option{-mclip} is not
18374 useful unless you also provide @option{-mminmax}.
18376 @item -mconfig=@var{name}
18378 Selects one of the built-in core configurations. Each MeP chip has
18379 one or more modules in it; each module has a core CPU and a variety of
18380 coprocessors, optional instructions, and peripherals. The
18381 @code{MeP-Integrator} tool, not part of GCC, provides these
18382 configurations through this option; using this option is the same as
18383 using all the corresponding command-line options. The default
18384 configuration is @samp{default}.
18388 Enables the coprocessor instructions. By default, this is a 32-bit
18389 coprocessor. Note that the coprocessor is normally enabled via the
18390 @option{-mconfig=} option.
18394 Enables the 32-bit coprocessor's instructions.
18398 Enables the 64-bit coprocessor's instructions.
18402 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
18406 Causes constant variables to be placed in the @code{.near} section.
18410 Enables the @code{div} and @code{divu} instructions.
18414 Generate big-endian code.
18418 Generate little-endian code.
18420 @item -mio-volatile
18421 @opindex mio-volatile
18422 Tells the compiler that any variable marked with the @code{io}
18423 attribute is to be considered volatile.
18427 Causes variables to be assigned to the @code{.far} section by default.
18431 Enables the @code{leadz} (leading zero) instruction.
18435 Causes variables to be assigned to the @code{.near} section by default.
18439 Enables the @code{min} and @code{max} instructions.
18443 Enables the multiplication and multiply-accumulate instructions.
18447 Disables all the optional instructions enabled by @option{-mall-opts}.
18451 Enables the @code{repeat} and @code{erepeat} instructions, used for
18452 low-overhead looping.
18456 Causes all variables to default to the @code{.tiny} section. Note
18457 that there is a 65536-byte limit to this section. Accesses to these
18458 variables use the @code{%gp} base register.
18462 Enables the saturation instructions. Note that the compiler does not
18463 currently generate these itself, but this option is included for
18464 compatibility with other tools, like @code{as}.
18468 Link the SDRAM-based runtime instead of the default ROM-based runtime.
18472 Link the simulator run-time libraries.
18476 Link the simulator runtime libraries, excluding built-in support
18477 for reset and exception vectors and tables.
18481 Causes all functions to default to the @code{.far} section. Without
18482 this option, functions default to the @code{.near} section.
18484 @item -mtiny=@var{n}
18486 Variables that are @var{n} bytes or smaller are allocated to the
18487 @code{.tiny} section. These variables use the @code{$gp} base
18488 register. The default for this option is 4, but note that there's a
18489 65536-byte limit to the @code{.tiny} section.
18493 @node MicroBlaze Options
18494 @subsection MicroBlaze Options
18495 @cindex MicroBlaze Options
18500 @opindex msoft-float
18501 Use software emulation for floating point (default).
18504 @opindex mhard-float
18505 Use hardware floating-point instructions.
18509 Do not optimize block moves, use @code{memcpy}.
18511 @item -mno-clearbss
18512 @opindex mno-clearbss
18513 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
18515 @item -mcpu=@var{cpu-type}
18517 Use features of, and schedule code for, the given CPU.
18518 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
18519 where @var{X} is a major version, @var{YY} is the minor version, and
18520 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
18521 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
18523 @item -mxl-soft-mul
18524 @opindex mxl-soft-mul
18525 Use software multiply emulation (default).
18527 @item -mxl-soft-div
18528 @opindex mxl-soft-div
18529 Use software emulation for divides (default).
18531 @item -mxl-barrel-shift
18532 @opindex mxl-barrel-shift
18533 Use the hardware barrel shifter.
18535 @item -mxl-pattern-compare
18536 @opindex mxl-pattern-compare
18537 Use pattern compare instructions.
18539 @item -msmall-divides
18540 @opindex msmall-divides
18541 Use table lookup optimization for small signed integer divisions.
18543 @item -mxl-stack-check
18544 @opindex mxl-stack-check
18545 This option is deprecated. Use @option{-fstack-check} instead.
18548 @opindex mxl-gp-opt
18549 Use GP-relative @code{.sdata}/@code{.sbss} sections.
18551 @item -mxl-multiply-high
18552 @opindex mxl-multiply-high
18553 Use multiply high instructions for high part of 32x32 multiply.
18555 @item -mxl-float-convert
18556 @opindex mxl-float-convert
18557 Use hardware floating-point conversion instructions.
18559 @item -mxl-float-sqrt
18560 @opindex mxl-float-sqrt
18561 Use hardware floating-point square root instruction.
18564 @opindex mbig-endian
18565 Generate code for a big-endian target.
18567 @item -mlittle-endian
18568 @opindex mlittle-endian
18569 Generate code for a little-endian target.
18572 @opindex mxl-reorder
18573 Use reorder instructions (swap and byte reversed load/store).
18575 @item -mxl-mode-@var{app-model}
18576 Select application model @var{app-model}. Valid models are
18579 normal executable (default), uses startup code @file{crt0.o}.
18582 for use with Xilinx Microprocessor Debugger (XMD) based
18583 software intrusive debug agent called xmdstub. This uses startup file
18584 @file{crt1.o} and sets the start address of the program to 0x800.
18587 for applications that are loaded using a bootloader.
18588 This model uses startup file @file{crt2.o} which does not contain a processor
18589 reset vector handler. This is suitable for transferring control on a
18590 processor reset to the bootloader rather than the application.
18593 for applications that do not require any of the
18594 MicroBlaze vectors. This option may be useful for applications running
18595 within a monitoring application. This model uses @file{crt3.o} as a startup file.
18598 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
18599 @option{-mxl-mode-@var{app-model}}.
18604 @subsection MIPS Options
18605 @cindex MIPS options
18611 Generate big-endian code.
18615 Generate little-endian code. This is the default for @samp{mips*el-*-*}
18618 @item -march=@var{arch}
18620 Generate code that runs on @var{arch}, which can be the name of a
18621 generic MIPS ISA, or the name of a particular processor.
18623 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
18624 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
18625 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
18626 @samp{mips64r5} and @samp{mips64r6}.
18627 The processor names are:
18628 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
18629 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
18630 @samp{5kc}, @samp{5kf},
18632 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
18633 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
18634 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
18635 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
18636 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
18639 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
18641 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
18642 @samp{m5100}, @samp{m5101},
18643 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
18646 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
18647 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
18648 @samp{rm7000}, @samp{rm9000},
18649 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
18652 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
18653 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
18654 @samp{xlr} and @samp{xlp}.
18655 The special value @samp{from-abi} selects the
18656 most compatible architecture for the selected ABI (that is,
18657 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
18659 The native Linux/GNU toolchain also supports the value @samp{native},
18660 which selects the best architecture option for the host processor.
18661 @option{-march=native} has no effect if GCC does not recognize
18664 In processor names, a final @samp{000} can be abbreviated as @samp{k}
18665 (for example, @option{-march=r2k}). Prefixes are optional, and
18666 @samp{vr} may be written @samp{r}.
18668 Names of the form @samp{@var{n}f2_1} refer to processors with
18669 FPUs clocked at half the rate of the core, names of the form
18670 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
18671 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
18672 processors with FPUs clocked a ratio of 3:2 with respect to the core.
18673 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
18674 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
18675 accepted as synonyms for @samp{@var{n}f1_1}.
18677 GCC defines two macros based on the value of this option. The first
18678 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
18679 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
18680 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
18681 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
18682 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
18684 Note that the @code{_MIPS_ARCH} macro uses the processor names given
18685 above. In other words, it has the full prefix and does not
18686 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
18687 the macro names the resolved architecture (either @code{"mips1"} or
18688 @code{"mips3"}). It names the default architecture when no
18689 @option{-march} option is given.
18691 @item -mtune=@var{arch}
18693 Optimize for @var{arch}. Among other things, this option controls
18694 the way instructions are scheduled, and the perceived cost of arithmetic
18695 operations. The list of @var{arch} values is the same as for
18698 When this option is not used, GCC optimizes for the processor
18699 specified by @option{-march}. By using @option{-march} and
18700 @option{-mtune} together, it is possible to generate code that
18701 runs on a family of processors, but optimize the code for one
18702 particular member of that family.
18704 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
18705 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
18706 @option{-march} ones described above.
18710 Equivalent to @option{-march=mips1}.
18714 Equivalent to @option{-march=mips2}.
18718 Equivalent to @option{-march=mips3}.
18722 Equivalent to @option{-march=mips4}.
18726 Equivalent to @option{-march=mips32}.
18730 Equivalent to @option{-march=mips32r3}.
18734 Equivalent to @option{-march=mips32r5}.
18738 Equivalent to @option{-march=mips32r6}.
18742 Equivalent to @option{-march=mips64}.
18746 Equivalent to @option{-march=mips64r2}.
18750 Equivalent to @option{-march=mips64r3}.
18754 Equivalent to @option{-march=mips64r5}.
18758 Equivalent to @option{-march=mips64r6}.
18763 @opindex mno-mips16
18764 Generate (do not generate) MIPS16 code. If GCC is targeting a
18765 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
18767 MIPS16 code generation can also be controlled on a per-function basis
18768 by means of @code{mips16} and @code{nomips16} attributes.
18769 @xref{Function Attributes}, for more information.
18771 @item -mflip-mips16
18772 @opindex mflip-mips16
18773 Generate MIPS16 code on alternating functions. This option is provided
18774 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
18775 not intended for ordinary use in compiling user code.
18777 @item -minterlink-compressed
18778 @item -mno-interlink-compressed
18779 @opindex minterlink-compressed
18780 @opindex mno-interlink-compressed
18781 Require (do not require) that code using the standard (uncompressed) MIPS ISA
18782 be link-compatible with MIPS16 and microMIPS code, and vice versa.
18784 For example, code using the standard ISA encoding cannot jump directly
18785 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
18786 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
18787 knows that the target of the jump is not compressed.
18789 @item -minterlink-mips16
18790 @itemx -mno-interlink-mips16
18791 @opindex minterlink-mips16
18792 @opindex mno-interlink-mips16
18793 Aliases of @option{-minterlink-compressed} and
18794 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
18795 and are retained for backwards compatibility.
18807 Generate code for the given ABI@.
18809 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
18810 generates 64-bit code when you select a 64-bit architecture, but you
18811 can use @option{-mgp32} to get 32-bit code instead.
18813 For information about the O64 ABI, see
18814 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
18816 GCC supports a variant of the o32 ABI in which floating-point registers
18817 are 64 rather than 32 bits wide. You can select this combination with
18818 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
18819 and @code{mfhc1} instructions and is therefore only supported for
18820 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
18822 The register assignments for arguments and return values remain the
18823 same, but each scalar value is passed in a single 64-bit register
18824 rather than a pair of 32-bit registers. For example, scalar
18825 floating-point values are returned in @samp{$f0} only, not a
18826 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
18827 remains the same in that the even-numbered double-precision registers
18830 Two additional variants of the o32 ABI are supported to enable
18831 a transition from 32-bit to 64-bit registers. These are FPXX
18832 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
18833 The FPXX extension mandates that all code must execute correctly
18834 when run using 32-bit or 64-bit registers. The code can be interlinked
18835 with either FP32 or FP64, but not both.
18836 The FP64A extension is similar to the FP64 extension but forbids the
18837 use of odd-numbered single-precision registers. This can be used
18838 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
18839 processors and allows both FP32 and FP64A code to interlink and
18840 run in the same process without changing FPU modes.
18843 @itemx -mno-abicalls
18845 @opindex mno-abicalls
18846 Generate (do not generate) code that is suitable for SVR4-style
18847 dynamic objects. @option{-mabicalls} is the default for SVR4-based
18852 Generate (do not generate) code that is fully position-independent,
18853 and that can therefore be linked into shared libraries. This option
18854 only affects @option{-mabicalls}.
18856 All @option{-mabicalls} code has traditionally been position-independent,
18857 regardless of options like @option{-fPIC} and @option{-fpic}. However,
18858 as an extension, the GNU toolchain allows executables to use absolute
18859 accesses for locally-binding symbols. It can also use shorter GP
18860 initialization sequences and generate direct calls to locally-defined
18861 functions. This mode is selected by @option{-mno-shared}.
18863 @option{-mno-shared} depends on binutils 2.16 or higher and generates
18864 objects that can only be linked by the GNU linker. However, the option
18865 does not affect the ABI of the final executable; it only affects the ABI
18866 of relocatable objects. Using @option{-mno-shared} generally makes
18867 executables both smaller and quicker.
18869 @option{-mshared} is the default.
18875 Assume (do not assume) that the static and dynamic linkers
18876 support PLTs and copy relocations. This option only affects
18877 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
18878 has no effect without @option{-msym32}.
18880 You can make @option{-mplt} the default by configuring
18881 GCC with @option{--with-mips-plt}. The default is
18882 @option{-mno-plt} otherwise.
18888 Lift (do not lift) the usual restrictions on the size of the global
18891 GCC normally uses a single instruction to load values from the GOT@.
18892 While this is relatively efficient, it only works if the GOT
18893 is smaller than about 64k. Anything larger causes the linker
18894 to report an error such as:
18896 @cindex relocation truncated to fit (MIPS)
18898 relocation truncated to fit: R_MIPS_GOT16 foobar
18901 If this happens, you should recompile your code with @option{-mxgot}.
18902 This works with very large GOTs, although the code is also
18903 less efficient, since it takes three instructions to fetch the
18904 value of a global symbol.
18906 Note that some linkers can create multiple GOTs. If you have such a
18907 linker, you should only need to use @option{-mxgot} when a single object
18908 file accesses more than 64k's worth of GOT entries. Very few do.
18910 These options have no effect unless GCC is generating position
18915 Assume that general-purpose registers are 32 bits wide.
18919 Assume that general-purpose registers are 64 bits wide.
18923 Assume that floating-point registers are 32 bits wide.
18927 Assume that floating-point registers are 64 bits wide.
18931 Do not assume the width of floating-point registers.
18934 @opindex mhard-float
18935 Use floating-point coprocessor instructions.
18938 @opindex msoft-float
18939 Do not use floating-point coprocessor instructions. Implement
18940 floating-point calculations using library calls instead.
18944 Equivalent to @option{-msoft-float}, but additionally asserts that the
18945 program being compiled does not perform any floating-point operations.
18946 This option is presently supported only by some bare-metal MIPS
18947 configurations, where it may select a special set of libraries
18948 that lack all floating-point support (including, for example, the
18949 floating-point @code{printf} formats).
18950 If code compiled with @option{-mno-float} accidentally contains
18951 floating-point operations, it is likely to suffer a link-time
18952 or run-time failure.
18954 @item -msingle-float
18955 @opindex msingle-float
18956 Assume that the floating-point coprocessor only supports single-precision
18959 @item -mdouble-float
18960 @opindex mdouble-float
18961 Assume that the floating-point coprocessor supports double-precision
18962 operations. This is the default.
18965 @itemx -mno-odd-spreg
18966 @opindex modd-spreg
18967 @opindex mno-odd-spreg
18968 Enable the use of odd-numbered single-precision floating-point registers
18969 for the o32 ABI. This is the default for processors that are known to
18970 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18974 @itemx -mabs=legacy
18976 @opindex mabs=legacy
18977 These options control the treatment of the special not-a-number (NaN)
18978 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18979 @code{neg.@i{fmt}} machine instructions.
18981 By default or when @option{-mabs=legacy} is used the legacy
18982 treatment is selected. In this case these instructions are considered
18983 arithmetic and avoided where correct operation is required and the
18984 input operand might be a NaN. A longer sequence of instructions that
18985 manipulate the sign bit of floating-point datum manually is used
18986 instead unless the @option{-ffinite-math-only} option has also been
18989 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
18990 this case these instructions are considered non-arithmetic and therefore
18991 operating correctly in all cases, including in particular where the
18992 input operand is a NaN. These instructions are therefore always used
18993 for the respective operations.
18996 @itemx -mnan=legacy
18998 @opindex mnan=legacy
18999 These options control the encoding of the special not-a-number (NaN)
19000 IEEE 754 floating-point data.
19002 The @option{-mnan=legacy} option selects the legacy encoding. In this
19003 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
19004 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
19005 by the first bit of their trailing significand field being 1.
19007 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
19008 this case qNaNs are denoted by the first bit of their trailing
19009 significand field being 1, whereas sNaNs are denoted by the first bit of
19010 their trailing significand field being 0.
19012 The default is @option{-mnan=legacy} unless GCC has been configured with
19013 @option{--with-nan=2008}.
19019 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
19020 implement atomic memory built-in functions. When neither option is
19021 specified, GCC uses the instructions if the target architecture
19024 @option{-mllsc} is useful if the runtime environment can emulate the
19025 instructions and @option{-mno-llsc} can be useful when compiling for
19026 nonstandard ISAs. You can make either option the default by
19027 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
19028 respectively. @option{--with-llsc} is the default for some
19029 configurations; see the installation documentation for details.
19035 Use (do not use) revision 1 of the MIPS DSP ASE@.
19036 @xref{MIPS DSP Built-in Functions}. This option defines the
19037 preprocessor macro @code{__mips_dsp}. It also defines
19038 @code{__mips_dsp_rev} to 1.
19044 Use (do not use) revision 2 of the MIPS DSP ASE@.
19045 @xref{MIPS DSP Built-in Functions}. This option defines the
19046 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
19047 It also defines @code{__mips_dsp_rev} to 2.
19050 @itemx -mno-smartmips
19051 @opindex msmartmips
19052 @opindex mno-smartmips
19053 Use (do not use) the MIPS SmartMIPS ASE.
19055 @item -mpaired-single
19056 @itemx -mno-paired-single
19057 @opindex mpaired-single
19058 @opindex mno-paired-single
19059 Use (do not use) paired-single floating-point instructions.
19060 @xref{MIPS Paired-Single Support}. This option requires
19061 hardware floating-point support to be enabled.
19067 Use (do not use) MIPS Digital Media Extension instructions.
19068 This option can only be used when generating 64-bit code and requires
19069 hardware floating-point support to be enabled.
19074 @opindex mno-mips3d
19075 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
19076 The option @option{-mips3d} implies @option{-mpaired-single}.
19079 @itemx -mno-micromips
19080 @opindex mmicromips
19081 @opindex mno-mmicromips
19082 Generate (do not generate) microMIPS code.
19084 MicroMIPS code generation can also be controlled on a per-function basis
19085 by means of @code{micromips} and @code{nomicromips} attributes.
19086 @xref{Function Attributes}, for more information.
19092 Use (do not use) MT Multithreading instructions.
19098 Use (do not use) the MIPS MCU ASE instructions.
19104 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
19110 Use (do not use) the MIPS Virtualization Application Specific instructions.
19116 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
19120 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
19121 an explanation of the default and the way that the pointer size is
19126 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
19128 The default size of @code{int}s, @code{long}s and pointers depends on
19129 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
19130 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
19131 32-bit @code{long}s. Pointers are the same size as @code{long}s,
19132 or the same size as integer registers, whichever is smaller.
19138 Assume (do not assume) that all symbols have 32-bit values, regardless
19139 of the selected ABI@. This option is useful in combination with
19140 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
19141 to generate shorter and faster references to symbolic addresses.
19145 Put definitions of externally-visible data in a small data section
19146 if that data is no bigger than @var{num} bytes. GCC can then generate
19147 more efficient accesses to the data; see @option{-mgpopt} for details.
19149 The default @option{-G} option depends on the configuration.
19151 @item -mlocal-sdata
19152 @itemx -mno-local-sdata
19153 @opindex mlocal-sdata
19154 @opindex mno-local-sdata
19155 Extend (do not extend) the @option{-G} behavior to local data too,
19156 such as to static variables in C@. @option{-mlocal-sdata} is the
19157 default for all configurations.
19159 If the linker complains that an application is using too much small data,
19160 you might want to try rebuilding the less performance-critical parts with
19161 @option{-mno-local-sdata}. You might also want to build large
19162 libraries with @option{-mno-local-sdata}, so that the libraries leave
19163 more room for the main program.
19165 @item -mextern-sdata
19166 @itemx -mno-extern-sdata
19167 @opindex mextern-sdata
19168 @opindex mno-extern-sdata
19169 Assume (do not assume) that externally-defined data is in
19170 a small data section if the size of that data is within the @option{-G} limit.
19171 @option{-mextern-sdata} is the default for all configurations.
19173 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
19174 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
19175 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
19176 is placed in a small data section. If @var{Var} is defined by another
19177 module, you must either compile that module with a high-enough
19178 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
19179 definition. If @var{Var} is common, you must link the application
19180 with a high-enough @option{-G} setting.
19182 The easiest way of satisfying these restrictions is to compile
19183 and link every module with the same @option{-G} option. However,
19184 you may wish to build a library that supports several different
19185 small data limits. You can do this by compiling the library with
19186 the highest supported @option{-G} setting and additionally using
19187 @option{-mno-extern-sdata} to stop the library from making assumptions
19188 about externally-defined data.
19194 Use (do not use) GP-relative accesses for symbols that are known to be
19195 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
19196 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
19199 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
19200 might not hold the value of @code{_gp}. For example, if the code is
19201 part of a library that might be used in a boot monitor, programs that
19202 call boot monitor routines pass an unknown value in @code{$gp}.
19203 (In such situations, the boot monitor itself is usually compiled
19204 with @option{-G0}.)
19206 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
19207 @option{-mno-extern-sdata}.
19209 @item -membedded-data
19210 @itemx -mno-embedded-data
19211 @opindex membedded-data
19212 @opindex mno-embedded-data
19213 Allocate variables to the read-only data section first if possible, then
19214 next in the small data section if possible, otherwise in data. This gives
19215 slightly slower code than the default, but reduces the amount of RAM required
19216 when executing, and thus may be preferred for some embedded systems.
19218 @item -muninit-const-in-rodata
19219 @itemx -mno-uninit-const-in-rodata
19220 @opindex muninit-const-in-rodata
19221 @opindex mno-uninit-const-in-rodata
19222 Put uninitialized @code{const} variables in the read-only data section.
19223 This option is only meaningful in conjunction with @option{-membedded-data}.
19225 @item -mcode-readable=@var{setting}
19226 @opindex mcode-readable
19227 Specify whether GCC may generate code that reads from executable sections.
19228 There are three possible settings:
19231 @item -mcode-readable=yes
19232 Instructions may freely access executable sections. This is the
19235 @item -mcode-readable=pcrel
19236 MIPS16 PC-relative load instructions can access executable sections,
19237 but other instructions must not do so. This option is useful on 4KSc
19238 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
19239 It is also useful on processors that can be configured to have a dual
19240 instruction/data SRAM interface and that, like the M4K, automatically
19241 redirect PC-relative loads to the instruction RAM.
19243 @item -mcode-readable=no
19244 Instructions must not access executable sections. This option can be
19245 useful on targets that are configured to have a dual instruction/data
19246 SRAM interface but that (unlike the M4K) do not automatically redirect
19247 PC-relative loads to the instruction RAM.
19250 @item -msplit-addresses
19251 @itemx -mno-split-addresses
19252 @opindex msplit-addresses
19253 @opindex mno-split-addresses
19254 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
19255 relocation operators. This option has been superseded by
19256 @option{-mexplicit-relocs} but is retained for backwards compatibility.
19258 @item -mexplicit-relocs
19259 @itemx -mno-explicit-relocs
19260 @opindex mexplicit-relocs
19261 @opindex mno-explicit-relocs
19262 Use (do not use) assembler relocation operators when dealing with symbolic
19263 addresses. The alternative, selected by @option{-mno-explicit-relocs},
19264 is to use assembler macros instead.
19266 @option{-mexplicit-relocs} is the default if GCC was configured
19267 to use an assembler that supports relocation operators.
19269 @item -mcheck-zero-division
19270 @itemx -mno-check-zero-division
19271 @opindex mcheck-zero-division
19272 @opindex mno-check-zero-division
19273 Trap (do not trap) on integer division by zero.
19275 The default is @option{-mcheck-zero-division}.
19277 @item -mdivide-traps
19278 @itemx -mdivide-breaks
19279 @opindex mdivide-traps
19280 @opindex mdivide-breaks
19281 MIPS systems check for division by zero by generating either a
19282 conditional trap or a break instruction. Using traps results in
19283 smaller code, but is only supported on MIPS II and later. Also, some
19284 versions of the Linux kernel have a bug that prevents trap from
19285 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
19286 allow conditional traps on architectures that support them and
19287 @option{-mdivide-breaks} to force the use of breaks.
19289 The default is usually @option{-mdivide-traps}, but this can be
19290 overridden at configure time using @option{--with-divide=breaks}.
19291 Divide-by-zero checks can be completely disabled using
19292 @option{-mno-check-zero-division}.
19297 @opindex mno-memcpy
19298 Force (do not force) the use of @code{memcpy} for non-trivial block
19299 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
19300 most constant-sized copies.
19303 @itemx -mno-long-calls
19304 @opindex mlong-calls
19305 @opindex mno-long-calls
19306 Disable (do not disable) use of the @code{jal} instruction. Calling
19307 functions using @code{jal} is more efficient but requires the caller
19308 and callee to be in the same 256 megabyte segment.
19310 This option has no effect on abicalls code. The default is
19311 @option{-mno-long-calls}.
19317 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
19318 instructions, as provided by the R4650 ISA@.
19324 Enable (disable) use of the @code{madd} and @code{msub} integer
19325 instructions. The default is @option{-mimadd} on architectures
19326 that support @code{madd} and @code{msub} except for the 74k
19327 architecture where it was found to generate slower code.
19330 @itemx -mno-fused-madd
19331 @opindex mfused-madd
19332 @opindex mno-fused-madd
19333 Enable (disable) use of the floating-point multiply-accumulate
19334 instructions, when they are available. The default is
19335 @option{-mfused-madd}.
19337 On the R8000 CPU when multiply-accumulate instructions are used,
19338 the intermediate product is calculated to infinite precision
19339 and is not subject to the FCSR Flush to Zero bit. This may be
19340 undesirable in some circumstances. On other processors the result
19341 is numerically identical to the equivalent computation using
19342 separate multiply, add, subtract and negate instructions.
19346 Tell the MIPS assembler to not run its preprocessor over user
19347 assembler files (with a @samp{.s} suffix) when assembling them.
19352 @opindex mno-fix-24k
19353 Work around the 24K E48 (lost data on stores during refill) errata.
19354 The workarounds are implemented by the assembler rather than by GCC@.
19357 @itemx -mno-fix-r4000
19358 @opindex mfix-r4000
19359 @opindex mno-fix-r4000
19360 Work around certain R4000 CPU errata:
19363 A double-word or a variable shift may give an incorrect result if executed
19364 immediately after starting an integer division.
19366 A double-word or a variable shift may give an incorrect result if executed
19367 while an integer multiplication is in progress.
19369 An integer division may give an incorrect result if started in a delay slot
19370 of a taken branch or a jump.
19374 @itemx -mno-fix-r4400
19375 @opindex mfix-r4400
19376 @opindex mno-fix-r4400
19377 Work around certain R4400 CPU errata:
19380 A double-word or a variable shift may give an incorrect result if executed
19381 immediately after starting an integer division.
19385 @itemx -mno-fix-r10000
19386 @opindex mfix-r10000
19387 @opindex mno-fix-r10000
19388 Work around certain R10000 errata:
19391 @code{ll}/@code{sc} sequences may not behave atomically on revisions
19392 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
19395 This option can only be used if the target architecture supports
19396 branch-likely instructions. @option{-mfix-r10000} is the default when
19397 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
19401 @itemx -mno-fix-rm7000
19402 @opindex mfix-rm7000
19403 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
19404 workarounds are implemented by the assembler rather than by GCC@.
19407 @itemx -mno-fix-vr4120
19408 @opindex mfix-vr4120
19409 Work around certain VR4120 errata:
19412 @code{dmultu} does not always produce the correct result.
19414 @code{div} and @code{ddiv} do not always produce the correct result if one
19415 of the operands is negative.
19417 The workarounds for the division errata rely on special functions in
19418 @file{libgcc.a}. At present, these functions are only provided by
19419 the @code{mips64vr*-elf} configurations.
19421 Other VR4120 errata require a NOP to be inserted between certain pairs of
19422 instructions. These errata are handled by the assembler, not by GCC itself.
19425 @opindex mfix-vr4130
19426 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
19427 workarounds are implemented by the assembler rather than by GCC,
19428 although GCC avoids using @code{mflo} and @code{mfhi} if the
19429 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
19430 instructions are available instead.
19433 @itemx -mno-fix-sb1
19435 Work around certain SB-1 CPU core errata.
19436 (This flag currently works around the SB-1 revision 2
19437 ``F1'' and ``F2'' floating-point errata.)
19439 @item -mr10k-cache-barrier=@var{setting}
19440 @opindex mr10k-cache-barrier
19441 Specify whether GCC should insert cache barriers to avoid the
19442 side-effects of speculation on R10K processors.
19444 In common with many processors, the R10K tries to predict the outcome
19445 of a conditional branch and speculatively executes instructions from
19446 the ``taken'' branch. It later aborts these instructions if the
19447 predicted outcome is wrong. However, on the R10K, even aborted
19448 instructions can have side effects.
19450 This problem only affects kernel stores and, depending on the system,
19451 kernel loads. As an example, a speculatively-executed store may load
19452 the target memory into cache and mark the cache line as dirty, even if
19453 the store itself is later aborted. If a DMA operation writes to the
19454 same area of memory before the ``dirty'' line is flushed, the cached
19455 data overwrites the DMA-ed data. See the R10K processor manual
19456 for a full description, including other potential problems.
19458 One workaround is to insert cache barrier instructions before every memory
19459 access that might be speculatively executed and that might have side
19460 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
19461 controls GCC's implementation of this workaround. It assumes that
19462 aborted accesses to any byte in the following regions does not have
19467 the memory occupied by the current function's stack frame;
19470 the memory occupied by an incoming stack argument;
19473 the memory occupied by an object with a link-time-constant address.
19476 It is the kernel's responsibility to ensure that speculative
19477 accesses to these regions are indeed safe.
19479 If the input program contains a function declaration such as:
19485 then the implementation of @code{foo} must allow @code{j foo} and
19486 @code{jal foo} to be executed speculatively. GCC honors this
19487 restriction for functions it compiles itself. It expects non-GCC
19488 functions (such as hand-written assembly code) to do the same.
19490 The option has three forms:
19493 @item -mr10k-cache-barrier=load-store
19494 Insert a cache barrier before a load or store that might be
19495 speculatively executed and that might have side effects even
19498 @item -mr10k-cache-barrier=store
19499 Insert a cache barrier before a store that might be speculatively
19500 executed and that might have side effects even if aborted.
19502 @item -mr10k-cache-barrier=none
19503 Disable the insertion of cache barriers. This is the default setting.
19506 @item -mflush-func=@var{func}
19507 @itemx -mno-flush-func
19508 @opindex mflush-func
19509 Specifies the function to call to flush the I and D caches, or to not
19510 call any such function. If called, the function must take the same
19511 arguments as the common @code{_flush_func}, that is, the address of the
19512 memory range for which the cache is being flushed, the size of the
19513 memory range, and the number 3 (to flush both caches). The default
19514 depends on the target GCC was configured for, but commonly is either
19515 @code{_flush_func} or @code{__cpu_flush}.
19517 @item mbranch-cost=@var{num}
19518 @opindex mbranch-cost
19519 Set the cost of branches to roughly @var{num} ``simple'' instructions.
19520 This cost is only a heuristic and is not guaranteed to produce
19521 consistent results across releases. A zero cost redundantly selects
19522 the default, which is based on the @option{-mtune} setting.
19524 @item -mbranch-likely
19525 @itemx -mno-branch-likely
19526 @opindex mbranch-likely
19527 @opindex mno-branch-likely
19528 Enable or disable use of Branch Likely instructions, regardless of the
19529 default for the selected architecture. By default, Branch Likely
19530 instructions may be generated if they are supported by the selected
19531 architecture. An exception is for the MIPS32 and MIPS64 architectures
19532 and processors that implement those architectures; for those, Branch
19533 Likely instructions are not be generated by default because the MIPS32
19534 and MIPS64 architectures specifically deprecate their use.
19536 @item -mcompact-branches=never
19537 @itemx -mcompact-branches=optimal
19538 @itemx -mcompact-branches=always
19539 @opindex mcompact-branches=never
19540 @opindex mcompact-branches=optimal
19541 @opindex mcompact-branches=always
19542 These options control which form of branches will be generated. The
19543 default is @option{-mcompact-branches=optimal}.
19545 The @option{-mcompact-branches=never} option ensures that compact branch
19546 instructions will never be generated.
19548 The @option{-mcompact-branches=always} option ensures that a compact
19549 branch instruction will be generated if available. If a compact branch
19550 instruction is not available, a delay slot form of the branch will be
19553 This option is supported from MIPS Release 6 onwards.
19555 The @option{-mcompact-branches=optimal} option will cause a delay slot
19556 branch to be used if one is available in the current ISA and the delay
19557 slot is successfully filled. If the delay slot is not filled, a compact
19558 branch will be chosen if one is available.
19560 @item -mfp-exceptions
19561 @itemx -mno-fp-exceptions
19562 @opindex mfp-exceptions
19563 Specifies whether FP exceptions are enabled. This affects how
19564 FP instructions are scheduled for some processors.
19565 The default is that FP exceptions are
19568 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
19569 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
19572 @item -mvr4130-align
19573 @itemx -mno-vr4130-align
19574 @opindex mvr4130-align
19575 The VR4130 pipeline is two-way superscalar, but can only issue two
19576 instructions together if the first one is 8-byte aligned. When this
19577 option is enabled, GCC aligns pairs of instructions that it
19578 thinks should execute in parallel.
19580 This option only has an effect when optimizing for the VR4130.
19581 It normally makes code faster, but at the expense of making it bigger.
19582 It is enabled by default at optimization level @option{-O3}.
19587 Enable (disable) generation of @code{synci} instructions on
19588 architectures that support it. The @code{synci} instructions (if
19589 enabled) are generated when @code{__builtin___clear_cache} is
19592 This option defaults to @option{-mno-synci}, but the default can be
19593 overridden by configuring GCC with @option{--with-synci}.
19595 When compiling code for single processor systems, it is generally safe
19596 to use @code{synci}. However, on many multi-core (SMP) systems, it
19597 does not invalidate the instruction caches on all cores and may lead
19598 to undefined behavior.
19600 @item -mrelax-pic-calls
19601 @itemx -mno-relax-pic-calls
19602 @opindex mrelax-pic-calls
19603 Try to turn PIC calls that are normally dispatched via register
19604 @code{$25} into direct calls. This is only possible if the linker can
19605 resolve the destination at link time and if the destination is within
19606 range for a direct call.
19608 @option{-mrelax-pic-calls} is the default if GCC was configured to use
19609 an assembler and a linker that support the @code{.reloc} assembly
19610 directive and @option{-mexplicit-relocs} is in effect. With
19611 @option{-mno-explicit-relocs}, this optimization can be performed by the
19612 assembler and the linker alone without help from the compiler.
19614 @item -mmcount-ra-address
19615 @itemx -mno-mcount-ra-address
19616 @opindex mmcount-ra-address
19617 @opindex mno-mcount-ra-address
19618 Emit (do not emit) code that allows @code{_mcount} to modify the
19619 calling function's return address. When enabled, this option extends
19620 the usual @code{_mcount} interface with a new @var{ra-address}
19621 parameter, which has type @code{intptr_t *} and is passed in register
19622 @code{$12}. @code{_mcount} can then modify the return address by
19623 doing both of the following:
19626 Returning the new address in register @code{$31}.
19628 Storing the new address in @code{*@var{ra-address}},
19629 if @var{ra-address} is nonnull.
19632 The default is @option{-mno-mcount-ra-address}.
19634 @item -mframe-header-opt
19635 @itemx -mno-frame-header-opt
19636 @opindex mframe-header-opt
19637 Enable (disable) frame header optimization in the o32 ABI. When using the
19638 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
19639 function to write out register arguments. When enabled, this optimization
19640 will suppress the allocation of the frame header if it can be determined that
19643 This optimization is off by default at all optimization levels.
19648 @subsection MMIX Options
19649 @cindex MMIX Options
19651 These options are defined for the MMIX:
19655 @itemx -mno-libfuncs
19657 @opindex mno-libfuncs
19658 Specify that intrinsic library functions are being compiled, passing all
19659 values in registers, no matter the size.
19662 @itemx -mno-epsilon
19664 @opindex mno-epsilon
19665 Generate floating-point comparison instructions that compare with respect
19666 to the @code{rE} epsilon register.
19668 @item -mabi=mmixware
19670 @opindex mabi=mmixware
19672 Generate code that passes function parameters and return values that (in
19673 the called function) are seen as registers @code{$0} and up, as opposed to
19674 the GNU ABI which uses global registers @code{$231} and up.
19676 @item -mzero-extend
19677 @itemx -mno-zero-extend
19678 @opindex mzero-extend
19679 @opindex mno-zero-extend
19680 When reading data from memory in sizes shorter than 64 bits, use (do not
19681 use) zero-extending load instructions by default, rather than
19682 sign-extending ones.
19685 @itemx -mno-knuthdiv
19687 @opindex mno-knuthdiv
19688 Make the result of a division yielding a remainder have the same sign as
19689 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
19690 remainder follows the sign of the dividend. Both methods are
19691 arithmetically valid, the latter being almost exclusively used.
19693 @item -mtoplevel-symbols
19694 @itemx -mno-toplevel-symbols
19695 @opindex mtoplevel-symbols
19696 @opindex mno-toplevel-symbols
19697 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
19698 code can be used with the @code{PREFIX} assembly directive.
19702 Generate an executable in the ELF format, rather than the default
19703 @samp{mmo} format used by the @command{mmix} simulator.
19705 @item -mbranch-predict
19706 @itemx -mno-branch-predict
19707 @opindex mbranch-predict
19708 @opindex mno-branch-predict
19709 Use (do not use) the probable-branch instructions, when static branch
19710 prediction indicates a probable branch.
19712 @item -mbase-addresses
19713 @itemx -mno-base-addresses
19714 @opindex mbase-addresses
19715 @opindex mno-base-addresses
19716 Generate (do not generate) code that uses @emph{base addresses}. Using a
19717 base address automatically generates a request (handled by the assembler
19718 and the linker) for a constant to be set up in a global register. The
19719 register is used for one or more base address requests within the range 0
19720 to 255 from the value held in the register. The generally leads to short
19721 and fast code, but the number of different data items that can be
19722 addressed is limited. This means that a program that uses lots of static
19723 data may require @option{-mno-base-addresses}.
19725 @item -msingle-exit
19726 @itemx -mno-single-exit
19727 @opindex msingle-exit
19728 @opindex mno-single-exit
19729 Force (do not force) generated code to have a single exit point in each
19733 @node MN10300 Options
19734 @subsection MN10300 Options
19735 @cindex MN10300 options
19737 These @option{-m} options are defined for Matsushita MN10300 architectures:
19742 Generate code to avoid bugs in the multiply instructions for the MN10300
19743 processors. This is the default.
19745 @item -mno-mult-bug
19746 @opindex mno-mult-bug
19747 Do not generate code to avoid bugs in the multiply instructions for the
19748 MN10300 processors.
19752 Generate code using features specific to the AM33 processor.
19756 Do not generate code using features specific to the AM33 processor. This
19761 Generate code using features specific to the AM33/2.0 processor.
19765 Generate code using features specific to the AM34 processor.
19767 @item -mtune=@var{cpu-type}
19769 Use the timing characteristics of the indicated CPU type when
19770 scheduling instructions. This does not change the targeted processor
19771 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
19772 @samp{am33-2} or @samp{am34}.
19774 @item -mreturn-pointer-on-d0
19775 @opindex mreturn-pointer-on-d0
19776 When generating a function that returns a pointer, return the pointer
19777 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
19778 only in @code{a0}, and attempts to call such functions without a prototype
19779 result in errors. Note that this option is on by default; use
19780 @option{-mno-return-pointer-on-d0} to disable it.
19784 Do not link in the C run-time initialization object file.
19788 Indicate to the linker that it should perform a relaxation optimization pass
19789 to shorten branches, calls and absolute memory addresses. This option only
19790 has an effect when used on the command line for the final link step.
19792 This option makes symbolic debugging impossible.
19796 Allow the compiler to generate @emph{Long Instruction Word}
19797 instructions if the target is the @samp{AM33} or later. This is the
19798 default. This option defines the preprocessor macro @code{__LIW__}.
19802 Do not allow the compiler to generate @emph{Long Instruction Word}
19803 instructions. This option defines the preprocessor macro
19808 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
19809 instructions if the target is the @samp{AM33} or later. This is the
19810 default. This option defines the preprocessor macro @code{__SETLB__}.
19814 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
19815 instructions. This option defines the preprocessor macro
19816 @code{__NO_SETLB__}.
19820 @node Moxie Options
19821 @subsection Moxie Options
19822 @cindex Moxie Options
19828 Generate big-endian code. This is the default for @samp{moxie-*-*}
19833 Generate little-endian code.
19837 Generate mul.x and umul.x instructions. This is the default for
19838 @samp{moxiebox-*-*} configurations.
19842 Do not link in the C run-time initialization object file.
19846 @node MSP430 Options
19847 @subsection MSP430 Options
19848 @cindex MSP430 Options
19850 These options are defined for the MSP430:
19856 Force assembly output to always use hex constants. Normally such
19857 constants are signed decimals, but this option is available for
19858 testsuite and/or aesthetic purposes.
19862 Select the MCU to target. This is used to create a C preprocessor
19863 symbol based upon the MCU name, converted to upper case and pre- and
19864 post-fixed with @samp{__}. This in turn is used by the
19865 @file{msp430.h} header file to select an MCU-specific supplementary
19868 The option also sets the ISA to use. If the MCU name is one that is
19869 known to only support the 430 ISA then that is selected, otherwise the
19870 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
19871 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
19872 name selects the 430X ISA.
19874 In addition an MCU-specific linker script is added to the linker
19875 command line. The script's name is the name of the MCU with
19876 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
19877 command line defines the C preprocessor symbol @code{__XXX__} and
19878 cause the linker to search for a script called @file{xxx.ld}.
19880 This option is also passed on to the assembler.
19883 @itemx -mno-warn-mcu
19885 @opindex mno-warn-mcu
19886 This option enables or disables warnings about conflicts between the
19887 MCU name specified by the @option{-mmcu} option and the ISA set by the
19888 @option{-mcpu} option and/or the hardware multiply support set by the
19889 @option{-mhwmult} option. It also toggles warnings about unrecognized
19890 MCU names. This option is on by default.
19894 Specifies the ISA to use. Accepted values are @samp{msp430},
19895 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
19896 @option{-mmcu=} option should be used to select the ISA.
19900 Link to the simulator runtime libraries and linker script. Overrides
19901 any scripts that would be selected by the @option{-mmcu=} option.
19905 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
19909 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
19913 This option is passed to the assembler and linker, and allows the
19914 linker to perform certain optimizations that cannot be done until
19919 Describes the type of hardware multiply supported by the target.
19920 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19921 for the original 16-bit-only multiply supported by early MCUs.
19922 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19923 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19924 A value of @samp{auto} can also be given. This tells GCC to deduce
19925 the hardware multiply support based upon the MCU name provided by the
19926 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
19927 the MCU name is not recognized then no hardware multiply support is
19928 assumed. @code{auto} is the default setting.
19930 Hardware multiplies are normally performed by calling a library
19931 routine. This saves space in the generated code. When compiling at
19932 @option{-O3} or higher however the hardware multiplier is invoked
19933 inline. This makes for bigger, but faster code.
19935 The hardware multiply routines disable interrupts whilst running and
19936 restore the previous interrupt state when they finish. This makes
19937 them safe to use inside interrupt handlers as well as in normal code.
19941 Enable the use of a minimum runtime environment - no static
19942 initializers or constructors. This is intended for memory-constrained
19943 devices. The compiler includes special symbols in some objects
19944 that tell the linker and runtime which code fragments are required.
19946 @item -mcode-region=
19947 @itemx -mdata-region=
19948 @opindex mcode-region
19949 @opindex mdata-region
19950 These options tell the compiler where to place functions and data that
19951 do not have one of the @code{lower}, @code{upper}, @code{either} or
19952 @code{section} attributes. Possible values are @code{lower},
19953 @code{upper}, @code{either} or @code{any}. The first three behave
19954 like the corresponding attribute. The fourth possible value -
19955 @code{any} - is the default. It leaves placement entirely up to the
19956 linker script and how it assigns the standard sections
19957 (@code{.text}, @code{.data}, etc) to the memory regions.
19959 @item -msilicon-errata=
19960 @opindex msilicon-errata
19961 This option passes on a request to assembler to enable the fixes for
19962 the named silicon errata.
19964 @item -msilicon-errata-warn=
19965 @opindex msilicon-errata-warn
19966 This option passes on a request to the assembler to enable warning
19967 messages when a silicon errata might need to be applied.
19971 @node NDS32 Options
19972 @subsection NDS32 Options
19973 @cindex NDS32 Options
19975 These options are defined for NDS32 implementations:
19980 @opindex mbig-endian
19981 Generate code in big-endian mode.
19983 @item -mlittle-endian
19984 @opindex mlittle-endian
19985 Generate code in little-endian mode.
19987 @item -mreduced-regs
19988 @opindex mreduced-regs
19989 Use reduced-set registers for register allocation.
19992 @opindex mfull-regs
19993 Use full-set registers for register allocation.
19997 Generate conditional move instructions.
20001 Do not generate conditional move instructions.
20005 Generate performance extension instructions.
20007 @item -mno-perf-ext
20008 @opindex mno-perf-ext
20009 Do not generate performance extension instructions.
20013 Generate v3 push25/pop25 instructions.
20016 @opindex mno-v3push
20017 Do not generate v3 push25/pop25 instructions.
20021 Generate 16-bit instructions.
20024 @opindex mno-16-bit
20025 Do not generate 16-bit instructions.
20027 @item -misr-vector-size=@var{num}
20028 @opindex misr-vector-size
20029 Specify the size of each interrupt vector, which must be 4 or 16.
20031 @item -mcache-block-size=@var{num}
20032 @opindex mcache-block-size
20033 Specify the size of each cache block,
20034 which must be a power of 2 between 4 and 512.
20036 @item -march=@var{arch}
20038 Specify the name of the target architecture.
20040 @item -mcmodel=@var{code-model}
20042 Set the code model to one of
20045 All the data and read-only data segments must be within 512KB addressing space.
20046 The text segment must be within 16MB addressing space.
20047 @item @samp{medium}
20048 The data segment must be within 512KB while the read-only data segment can be
20049 within 4GB addressing space. The text segment should be still within 16MB
20052 All the text and data segments can be within 4GB addressing space.
20056 @opindex mctor-dtor
20057 Enable constructor/destructor feature.
20061 Guide linker to relax instructions.
20065 @node Nios II Options
20066 @subsection Nios II Options
20067 @cindex Nios II options
20068 @cindex Altera Nios II options
20070 These are the options defined for the Altera Nios II processor.
20076 @cindex smaller data references
20077 Put global and static objects less than or equal to @var{num} bytes
20078 into the small data or BSS sections instead of the normal data or BSS
20079 sections. The default value of @var{num} is 8.
20081 @item -mgpopt=@var{option}
20086 Generate (do not generate) GP-relative accesses. The following
20087 @var{option} names are recognized:
20092 Do not generate GP-relative accesses.
20095 Generate GP-relative accesses for small data objects that are not
20096 external, weak, or uninitialized common symbols.
20097 Also use GP-relative addressing for objects that
20098 have been explicitly placed in a small data section via a @code{section}
20102 As for @samp{local}, but also generate GP-relative accesses for
20103 small data objects that are external, weak, or common. If you use this option,
20104 you must ensure that all parts of your program (including libraries) are
20105 compiled with the same @option{-G} setting.
20108 Generate GP-relative accesses for all data objects in the program. If you
20109 use this option, the entire data and BSS segments
20110 of your program must fit in 64K of memory and you must use an appropriate
20111 linker script to allocate them within the addressable range of the
20115 Generate GP-relative addresses for function pointers as well as data
20116 pointers. If you use this option, the entire text, data, and BSS segments
20117 of your program must fit in 64K of memory and you must use an appropriate
20118 linker script to allocate them within the addressable range of the
20123 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
20124 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
20126 The default is @option{-mgpopt} except when @option{-fpic} or
20127 @option{-fPIC} is specified to generate position-independent code.
20128 Note that the Nios II ABI does not permit GP-relative accesses from
20131 You may need to specify @option{-mno-gpopt} explicitly when building
20132 programs that include large amounts of small data, including large
20133 GOT data sections. In this case, the 16-bit offset for GP-relative
20134 addressing may not be large enough to allow access to the entire
20135 small data section.
20141 Generate little-endian (default) or big-endian (experimental) code,
20144 @item -march=@var{arch}
20146 This specifies the name of the target Nios II architecture. GCC uses this
20147 name to determine what kind of instructions it can emit when generating
20148 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
20150 The preprocessor macro @code{__nios2_arch__} is available to programs,
20151 with value 1 or 2, indicating the targeted ISA level.
20153 @item -mbypass-cache
20154 @itemx -mno-bypass-cache
20155 @opindex mno-bypass-cache
20156 @opindex mbypass-cache
20157 Force all load and store instructions to always bypass cache by
20158 using I/O variants of the instructions. The default is not to
20161 @item -mno-cache-volatile
20162 @itemx -mcache-volatile
20163 @opindex mcache-volatile
20164 @opindex mno-cache-volatile
20165 Volatile memory access bypass the cache using the I/O variants of
20166 the load and store instructions. The default is not to bypass the cache.
20168 @item -mno-fast-sw-div
20169 @itemx -mfast-sw-div
20170 @opindex mno-fast-sw-div
20171 @opindex mfast-sw-div
20172 Do not use table-based fast divide for small numbers. The default
20173 is to use the fast divide at @option{-O3} and above.
20177 @itemx -mno-hw-mulx
20181 @opindex mno-hw-mul
20183 @opindex mno-hw-mulx
20185 @opindex mno-hw-div
20187 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
20188 instructions by the compiler. The default is to emit @code{mul}
20189 and not emit @code{div} and @code{mulx}.
20195 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
20196 CDX (code density) instructions. Enabling these instructions also
20197 requires @option{-march=r2}. Since these instructions are optional
20198 extensions to the R2 architecture, the default is not to emit them.
20200 @item -mcustom-@var{insn}=@var{N}
20201 @itemx -mno-custom-@var{insn}
20202 @opindex mcustom-@var{insn}
20203 @opindex mno-custom-@var{insn}
20204 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
20205 custom instruction with encoding @var{N} when generating code that uses
20206 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
20207 instruction 253 for single-precision floating-point add operations instead
20208 of the default behavior of using a library call.
20210 The following values of @var{insn} are supported. Except as otherwise
20211 noted, floating-point operations are expected to be implemented with
20212 normal IEEE 754 semantics and correspond directly to the C operators or the
20213 equivalent GCC built-in functions (@pxref{Other Builtins}).
20215 Single-precision floating point:
20218 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
20219 Binary arithmetic operations.
20225 Unary absolute value.
20227 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
20228 Comparison operations.
20230 @item @samp{fmins}, @samp{fmaxs}
20231 Floating-point minimum and maximum. These instructions are only
20232 generated if @option{-ffinite-math-only} is specified.
20234 @item @samp{fsqrts}
20235 Unary square root operation.
20237 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
20238 Floating-point trigonometric and exponential functions. These instructions
20239 are only generated if @option{-funsafe-math-optimizations} is also specified.
20243 Double-precision floating point:
20246 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
20247 Binary arithmetic operations.
20253 Unary absolute value.
20255 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
20256 Comparison operations.
20258 @item @samp{fmind}, @samp{fmaxd}
20259 Double-precision minimum and maximum. These instructions are only
20260 generated if @option{-ffinite-math-only} is specified.
20262 @item @samp{fsqrtd}
20263 Unary square root operation.
20265 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
20266 Double-precision trigonometric and exponential functions. These instructions
20267 are only generated if @option{-funsafe-math-optimizations} is also specified.
20273 @item @samp{fextsd}
20274 Conversion from single precision to double precision.
20276 @item @samp{ftruncds}
20277 Conversion from double precision to single precision.
20279 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
20280 Conversion from floating point to signed or unsigned integer types, with
20281 truncation towards zero.
20284 Conversion from single-precision floating point to signed integer,
20285 rounding to the nearest integer and ties away from zero.
20286 This corresponds to the @code{__builtin_lroundf} function when
20287 @option{-fno-math-errno} is used.
20289 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
20290 Conversion from signed or unsigned integer types to floating-point types.
20294 In addition, all of the following transfer instructions for internal
20295 registers X and Y must be provided to use any of the double-precision
20296 floating-point instructions. Custom instructions taking two
20297 double-precision source operands expect the first operand in the
20298 64-bit register X. The other operand (or only operand of a unary
20299 operation) is given to the custom arithmetic instruction with the
20300 least significant half in source register @var{src1} and the most
20301 significant half in @var{src2}. A custom instruction that returns a
20302 double-precision result returns the most significant 32 bits in the
20303 destination register and the other half in 32-bit register Y.
20304 GCC automatically generates the necessary code sequences to write
20305 register X and/or read register Y when double-precision floating-point
20306 instructions are used.
20311 Write @var{src1} into the least significant half of X and @var{src2} into
20312 the most significant half of X.
20315 Write @var{src1} into Y.
20317 @item @samp{frdxhi}, @samp{frdxlo}
20318 Read the most or least (respectively) significant half of X and store it in
20322 Read the value of Y and store it into @var{dest}.
20325 Note that you can gain more local control over generation of Nios II custom
20326 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
20327 and @code{target("no-custom-@var{insn}")} function attributes
20328 (@pxref{Function Attributes})
20329 or pragmas (@pxref{Function Specific Option Pragmas}).
20331 @item -mcustom-fpu-cfg=@var{name}
20332 @opindex mcustom-fpu-cfg
20334 This option enables a predefined, named set of custom instruction encodings
20335 (see @option{-mcustom-@var{insn}} above).
20336 Currently, the following sets are defined:
20338 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
20339 @gccoptlist{-mcustom-fmuls=252 @gol
20340 -mcustom-fadds=253 @gol
20341 -mcustom-fsubs=254 @gol
20342 -fsingle-precision-constant}
20344 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
20345 @gccoptlist{-mcustom-fmuls=252 @gol
20346 -mcustom-fadds=253 @gol
20347 -mcustom-fsubs=254 @gol
20348 -mcustom-fdivs=255 @gol
20349 -fsingle-precision-constant}
20351 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
20352 @gccoptlist{-mcustom-floatus=243 @gol
20353 -mcustom-fixsi=244 @gol
20354 -mcustom-floatis=245 @gol
20355 -mcustom-fcmpgts=246 @gol
20356 -mcustom-fcmples=249 @gol
20357 -mcustom-fcmpeqs=250 @gol
20358 -mcustom-fcmpnes=251 @gol
20359 -mcustom-fmuls=252 @gol
20360 -mcustom-fadds=253 @gol
20361 -mcustom-fsubs=254 @gol
20362 -mcustom-fdivs=255 @gol
20363 -fsingle-precision-constant}
20365 Custom instruction assignments given by individual
20366 @option{-mcustom-@var{insn}=} options override those given by
20367 @option{-mcustom-fpu-cfg=}, regardless of the
20368 order of the options on the command line.
20370 Note that you can gain more local control over selection of a FPU
20371 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
20372 function attribute (@pxref{Function Attributes})
20373 or pragma (@pxref{Function Specific Option Pragmas}).
20377 These additional @samp{-m} options are available for the Altera Nios II
20378 ELF (bare-metal) target:
20384 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
20385 startup and termination code, and is typically used in conjunction with
20386 @option{-msys-crt0=} to specify the location of the alternate startup code
20387 provided by the HAL BSP.
20391 Link with a limited version of the C library, @option{-lsmallc}, rather than
20394 @item -msys-crt0=@var{startfile}
20396 @var{startfile} is the file name of the startfile (crt0) to use
20397 when linking. This option is only useful in conjunction with @option{-mhal}.
20399 @item -msys-lib=@var{systemlib}
20401 @var{systemlib} is the library name of the library that provides
20402 low-level system calls required by the C library,
20403 e.g. @code{read} and @code{write}.
20404 This option is typically used to link with a library provided by a HAL BSP.
20408 @node Nvidia PTX Options
20409 @subsection Nvidia PTX Options
20410 @cindex Nvidia PTX options
20411 @cindex nvptx options
20413 These options are defined for Nvidia PTX:
20421 Generate code for 32-bit or 64-bit ABI.
20424 @opindex mmainkernel
20425 Link in code for a __main kernel. This is for stand-alone instead of
20426 offloading execution.
20430 Apply partitioned execution optimizations. This is the default when any
20431 level of optimization is selected.
20435 @node PDP-11 Options
20436 @subsection PDP-11 Options
20437 @cindex PDP-11 Options
20439 These options are defined for the PDP-11:
20444 Use hardware FPP floating point. This is the default. (FIS floating
20445 point on the PDP-11/40 is not supported.)
20448 @opindex msoft-float
20449 Do not use hardware floating point.
20453 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
20457 Return floating-point results in memory. This is the default.
20461 Generate code for a PDP-11/40.
20465 Generate code for a PDP-11/45. This is the default.
20469 Generate code for a PDP-11/10.
20471 @item -mbcopy-builtin
20472 @opindex mbcopy-builtin
20473 Use inline @code{movmemhi} patterns for copying memory. This is the
20478 Do not use inline @code{movmemhi} patterns for copying memory.
20484 Use 16-bit @code{int}. This is the default.
20490 Use 32-bit @code{int}.
20493 @itemx -mno-float32
20495 @opindex mno-float32
20496 Use 64-bit @code{float}. This is the default.
20499 @itemx -mno-float64
20501 @opindex mno-float64
20502 Use 32-bit @code{float}.
20506 Use @code{abshi2} pattern. This is the default.
20510 Do not use @code{abshi2} pattern.
20512 @item -mbranch-expensive
20513 @opindex mbranch-expensive
20514 Pretend that branches are expensive. This is for experimenting with
20515 code generation only.
20517 @item -mbranch-cheap
20518 @opindex mbranch-cheap
20519 Do not pretend that branches are expensive. This is the default.
20523 Use Unix assembler syntax. This is the default when configured for
20524 @samp{pdp11-*-bsd}.
20528 Use DEC assembler syntax. This is the default when configured for any
20529 PDP-11 target other than @samp{pdp11-*-bsd}.
20532 @node picoChip Options
20533 @subsection picoChip Options
20534 @cindex picoChip options
20536 These @samp{-m} options are defined for picoChip implementations:
20540 @item -mae=@var{ae_type}
20542 Set the instruction set, register set, and instruction scheduling
20543 parameters for array element type @var{ae_type}. Supported values
20544 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
20546 @option{-mae=ANY} selects a completely generic AE type. Code
20547 generated with this option runs on any of the other AE types. The
20548 code is not as efficient as it would be if compiled for a specific
20549 AE type, and some types of operation (e.g., multiplication) do not
20550 work properly on all types of AE.
20552 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
20553 for compiled code, and is the default.
20555 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
20556 option may suffer from poor performance of byte (char) manipulation,
20557 since the DSP AE does not provide hardware support for byte load/stores.
20559 @item -msymbol-as-address
20560 Enable the compiler to directly use a symbol name as an address in a
20561 load/store instruction, without first loading it into a
20562 register. Typically, the use of this option generates larger
20563 programs, which run faster than when the option isn't used. However, the
20564 results vary from program to program, so it is left as a user option,
20565 rather than being permanently enabled.
20567 @item -mno-inefficient-warnings
20568 Disables warnings about the generation of inefficient code. These
20569 warnings can be generated, for example, when compiling code that
20570 performs byte-level memory operations on the MAC AE type. The MAC AE has
20571 no hardware support for byte-level memory operations, so all byte
20572 load/stores must be synthesized from word load/store operations. This is
20573 inefficient and a warning is generated to indicate
20574 that you should rewrite the code to avoid byte operations, or to target
20575 an AE type that has the necessary hardware support. This option disables
20580 @node PowerPC Options
20581 @subsection PowerPC Options
20582 @cindex PowerPC options
20584 These are listed under @xref{RS/6000 and PowerPC Options}.
20587 @subsection RL78 Options
20588 @cindex RL78 Options
20594 Links in additional target libraries to support operation within a
20603 Specifies the type of hardware multiplication and division support to
20604 be used. The simplest is @code{none}, which uses software for both
20605 multiplication and division. This is the default. The @code{g13}
20606 value is for the hardware multiply/divide peripheral found on the
20607 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
20608 the multiplication and division instructions supported by the RL78/G14
20609 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
20610 the value @code{mg10} is an alias for @code{none}.
20612 In addition a C preprocessor macro is defined, based upon the setting
20613 of this option. Possible values are: @code{__RL78_MUL_NONE__},
20614 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
20621 Specifies the RL78 core to target. The default is the G14 core, also
20622 known as an S3 core or just RL78. The G13 or S2 core does not have
20623 multiply or divide instructions, instead it uses a hardware peripheral
20624 for these operations. The G10 or S1 core does not have register
20625 banks, so it uses a different calling convention.
20627 If this option is set it also selects the type of hardware multiply
20628 support to use, unless this is overridden by an explicit
20629 @option{-mmul=none} option on the command line. Thus specifying
20630 @option{-mcpu=g13} enables the use of the G13 hardware multiply
20631 peripheral and specifying @option{-mcpu=g10} disables the use of
20632 hardware multiplications altogether.
20634 Note, although the RL78/G14 core is the default target, specifying
20635 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
20636 change the behavior of the toolchain since it also enables G14
20637 hardware multiply support. If these options are not specified on the
20638 command line then software multiplication routines will be used even
20639 though the code targets the RL78 core. This is for backwards
20640 compatibility with older toolchains which did not have hardware
20641 multiply and divide support.
20643 In addition a C preprocessor macro is defined, based upon the setting
20644 of this option. Possible values are: @code{__RL78_G10__},
20645 @code{__RL78_G13__} or @code{__RL78_G14__}.
20655 These are aliases for the corresponding @option{-mcpu=} option. They
20656 are provided for backwards compatibility.
20660 Allow the compiler to use all of the available registers. By default
20661 registers @code{r24..r31} are reserved for use in interrupt handlers.
20662 With this option enabled these registers can be used in ordinary
20665 @item -m64bit-doubles
20666 @itemx -m32bit-doubles
20667 @opindex m64bit-doubles
20668 @opindex m32bit-doubles
20669 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20670 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20671 @option{-m32bit-doubles}.
20673 @item -msave-mduc-in-interrupts
20674 @item -mno-save-mduc-in-interrupts
20675 @opindex msave-mduc-in-interrupts
20676 @opindex mno-save-mduc-in-interrupts
20677 Specifies that interrupt handler functions should preserve the
20678 MDUC registers. This is only necessary if normal code might use
20679 the MDUC registers, for example because it performs multiplication
20680 and division operations. The default is to ignore the MDUC registers
20681 as this makes the interrupt handlers faster. The target option -mg13
20682 needs to be passed for this to work as this feature is only available
20683 on the G13 target (S2 core). The MDUC registers will only be saved
20684 if the interrupt handler performs a multiplication or division
20685 operation or it calls another function.
20689 @node RS/6000 and PowerPC Options
20690 @subsection IBM RS/6000 and PowerPC Options
20691 @cindex RS/6000 and PowerPC Options
20692 @cindex IBM RS/6000 and PowerPC Options
20694 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
20696 @item -mpowerpc-gpopt
20697 @itemx -mno-powerpc-gpopt
20698 @itemx -mpowerpc-gfxopt
20699 @itemx -mno-powerpc-gfxopt
20702 @itemx -mno-powerpc64
20706 @itemx -mno-popcntb
20708 @itemx -mno-popcntd
20717 @itemx -mno-hard-dfp
20718 @opindex mpowerpc-gpopt
20719 @opindex mno-powerpc-gpopt
20720 @opindex mpowerpc-gfxopt
20721 @opindex mno-powerpc-gfxopt
20722 @opindex mpowerpc64
20723 @opindex mno-powerpc64
20727 @opindex mno-popcntb
20729 @opindex mno-popcntd
20735 @opindex mno-mfpgpr
20737 @opindex mno-hard-dfp
20738 You use these options to specify which instructions are available on the
20739 processor you are using. The default value of these options is
20740 determined when configuring GCC@. Specifying the
20741 @option{-mcpu=@var{cpu_type}} overrides the specification of these
20742 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
20743 rather than the options listed above.
20745 Specifying @option{-mpowerpc-gpopt} allows
20746 GCC to use the optional PowerPC architecture instructions in the
20747 General Purpose group, including floating-point square root. Specifying
20748 @option{-mpowerpc-gfxopt} allows GCC to
20749 use the optional PowerPC architecture instructions in the Graphics
20750 group, including floating-point select.
20752 The @option{-mmfcrf} option allows GCC to generate the move from
20753 condition register field instruction implemented on the POWER4
20754 processor and other processors that support the PowerPC V2.01
20756 The @option{-mpopcntb} option allows GCC to generate the popcount and
20757 double-precision FP reciprocal estimate instruction implemented on the
20758 POWER5 processor and other processors that support the PowerPC V2.02
20760 The @option{-mpopcntd} option allows GCC to generate the popcount
20761 instruction implemented on the POWER7 processor and other processors
20762 that support the PowerPC V2.06 architecture.
20763 The @option{-mfprnd} option allows GCC to generate the FP round to
20764 integer instructions implemented on the POWER5+ processor and other
20765 processors that support the PowerPC V2.03 architecture.
20766 The @option{-mcmpb} option allows GCC to generate the compare bytes
20767 instruction implemented on the POWER6 processor and other processors
20768 that support the PowerPC V2.05 architecture.
20769 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
20770 general-purpose register instructions implemented on the POWER6X
20771 processor and other processors that support the extended PowerPC V2.05
20773 The @option{-mhard-dfp} option allows GCC to generate the decimal
20774 floating-point instructions implemented on some POWER processors.
20776 The @option{-mpowerpc64} option allows GCC to generate the additional
20777 64-bit instructions that are found in the full PowerPC64 architecture
20778 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
20779 @option{-mno-powerpc64}.
20781 @item -mcpu=@var{cpu_type}
20783 Set architecture type, register usage, and
20784 instruction scheduling parameters for machine type @var{cpu_type}.
20785 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
20786 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
20787 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
20788 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
20789 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
20790 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
20791 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
20792 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
20793 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
20794 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
20795 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
20798 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
20799 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
20800 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
20801 architecture machine types, with an appropriate, generic processor
20802 model assumed for scheduling purposes.
20804 The other options specify a specific processor. Code generated under
20805 those options runs best on that processor, and may not run at all on
20808 The @option{-mcpu} options automatically enable or disable the
20811 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
20812 -mpopcntb -mpopcntd -mpowerpc64 @gol
20813 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
20814 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
20815 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
20816 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
20818 The particular options set for any particular CPU varies between
20819 compiler versions, depending on what setting seems to produce optimal
20820 code for that CPU; it doesn't necessarily reflect the actual hardware's
20821 capabilities. If you wish to set an individual option to a particular
20822 value, you may specify it after the @option{-mcpu} option, like
20823 @option{-mcpu=970 -mno-altivec}.
20825 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
20826 not enabled or disabled by the @option{-mcpu} option at present because
20827 AIX does not have full support for these options. You may still
20828 enable or disable them individually if you're sure it'll work in your
20831 @item -mtune=@var{cpu_type}
20833 Set the instruction scheduling parameters for machine type
20834 @var{cpu_type}, but do not set the architecture type or register usage,
20835 as @option{-mcpu=@var{cpu_type}} does. The same
20836 values for @var{cpu_type} are used for @option{-mtune} as for
20837 @option{-mcpu}. If both are specified, the code generated uses the
20838 architecture and registers set by @option{-mcpu}, but the
20839 scheduling parameters set by @option{-mtune}.
20841 @item -mcmodel=small
20842 @opindex mcmodel=small
20843 Generate PowerPC64 code for the small model: The TOC is limited to
20846 @item -mcmodel=medium
20847 @opindex mcmodel=medium
20848 Generate PowerPC64 code for the medium model: The TOC and other static
20849 data may be up to a total of 4G in size.
20851 @item -mcmodel=large
20852 @opindex mcmodel=large
20853 Generate PowerPC64 code for the large model: The TOC may be up to 4G
20854 in size. Other data and code is only limited by the 64-bit address
20858 @itemx -mno-altivec
20860 @opindex mno-altivec
20861 Generate code that uses (does not use) AltiVec instructions, and also
20862 enable the use of built-in functions that allow more direct access to
20863 the AltiVec instruction set. You may also need to set
20864 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
20867 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
20868 @option{-maltivec=be}, the element order for AltiVec intrinsics such
20869 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
20870 match array element order corresponding to the endianness of the
20871 target. That is, element zero identifies the leftmost element in a
20872 vector register when targeting a big-endian platform, and identifies
20873 the rightmost element in a vector register when targeting a
20874 little-endian platform.
20877 @opindex maltivec=be
20878 Generate AltiVec instructions using big-endian element order,
20879 regardless of whether the target is big- or little-endian. This is
20880 the default when targeting a big-endian platform.
20882 The element order is used to interpret element numbers in AltiVec
20883 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20884 @code{vec_insert}. By default, these match array element order
20885 corresponding to the endianness for the target.
20888 @opindex maltivec=le
20889 Generate AltiVec instructions using little-endian element order,
20890 regardless of whether the target is big- or little-endian. This is
20891 the default when targeting a little-endian platform. This option is
20892 currently ignored when targeting a big-endian platform.
20894 The element order is used to interpret element numbers in AltiVec
20895 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20896 @code{vec_insert}. By default, these match array element order
20897 corresponding to the endianness for the target.
20902 @opindex mno-vrsave
20903 Generate VRSAVE instructions when generating AltiVec code.
20905 @item -mgen-cell-microcode
20906 @opindex mgen-cell-microcode
20907 Generate Cell microcode instructions.
20909 @item -mwarn-cell-microcode
20910 @opindex mwarn-cell-microcode
20911 Warn when a Cell microcode instruction is emitted. An example
20912 of a Cell microcode instruction is a variable shift.
20915 @opindex msecure-plt
20916 Generate code that allows @command{ld} and @command{ld.so}
20917 to build executables and shared
20918 libraries with non-executable @code{.plt} and @code{.got} sections.
20920 32-bit SYSV ABI option.
20924 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
20926 requires @code{.plt} and @code{.got}
20927 sections that are both writable and executable.
20928 This is a PowerPC 32-bit SYSV ABI option.
20934 This switch enables or disables the generation of ISEL instructions.
20936 @item -misel=@var{yes/no}
20937 This switch has been deprecated. Use @option{-misel} and
20938 @option{-mno-isel} instead.
20942 Enable Local Register Allocation. This is still experimental for PowerPC,
20943 so by default the compiler uses standard reload
20944 (i.e. @option{-mno-lra}).
20950 This switch enables or disables the generation of SPE simd
20956 @opindex mno-paired
20957 This switch enables or disables the generation of PAIRED simd
20960 @item -mspe=@var{yes/no}
20961 This option has been deprecated. Use @option{-mspe} and
20962 @option{-mno-spe} instead.
20968 Generate code that uses (does not use) vector/scalar (VSX)
20969 instructions, and also enable the use of built-in functions that allow
20970 more direct access to the VSX instruction set.
20975 @opindex mno-crypto
20976 Enable the use (disable) of the built-in functions that allow direct
20977 access to the cryptographic instructions that were added in version
20978 2.07 of the PowerPC ISA.
20980 @item -mdirect-move
20981 @itemx -mno-direct-move
20982 @opindex mdirect-move
20983 @opindex mno-direct-move
20984 Generate code that uses (does not use) the instructions to move data
20985 between the general purpose registers and the vector/scalar (VSX)
20986 registers that were added in version 2.07 of the PowerPC ISA.
20992 Enable (disable) the use of the built-in functions that allow direct
20993 access to the Hardware Transactional Memory (HTM) instructions that
20994 were added in version 2.07 of the PowerPC ISA.
20996 @item -mpower8-fusion
20997 @itemx -mno-power8-fusion
20998 @opindex mpower8-fusion
20999 @opindex mno-power8-fusion
21000 Generate code that keeps (does not keeps) some integer operations
21001 adjacent so that the instructions can be fused together on power8 and
21004 @item -mpower8-vector
21005 @itemx -mno-power8-vector
21006 @opindex mpower8-vector
21007 @opindex mno-power8-vector
21008 Generate code that uses (does not use) the vector and scalar
21009 instructions that were added in version 2.07 of the PowerPC ISA. Also
21010 enable the use of built-in functions that allow more direct access to
21011 the vector instructions.
21013 @item -mquad-memory
21014 @itemx -mno-quad-memory
21015 @opindex mquad-memory
21016 @opindex mno-quad-memory
21017 Generate code that uses (does not use) the non-atomic quad word memory
21018 instructions. The @option{-mquad-memory} option requires use of
21021 @item -mquad-memory-atomic
21022 @itemx -mno-quad-memory-atomic
21023 @opindex mquad-memory-atomic
21024 @opindex mno-quad-memory-atomic
21025 Generate code that uses (does not use) the atomic quad word memory
21026 instructions. The @option{-mquad-memory-atomic} option requires use of
21029 @item -mupper-regs-di
21030 @itemx -mno-upper-regs-di
21031 @opindex mupper-regs-di
21032 @opindex mno-upper-regs-di
21033 Generate code that uses (does not use) the scalar instructions that
21034 target all 64 registers in the vector/scalar floating point register
21035 set that were added in version 2.06 of the PowerPC ISA when processing
21036 integers. @option{-mupper-regs-di} is turned on by default if you use
21037 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21038 @option{-mcpu=power9}, or @option{-mvsx} options.
21040 @item -mupper-regs-df
21041 @itemx -mno-upper-regs-df
21042 @opindex mupper-regs-df
21043 @opindex mno-upper-regs-df
21044 Generate code that uses (does not use) the scalar double precision
21045 instructions that target all 64 registers in the vector/scalar
21046 floating point register set that were added in version 2.06 of the
21047 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
21048 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21049 @option{-mcpu=power9}, or @option{-mvsx} options.
21051 @item -mupper-regs-sf
21052 @itemx -mno-upper-regs-sf
21053 @opindex mupper-regs-sf
21054 @opindex mno-upper-regs-sf
21055 Generate code that uses (does not use) the scalar single precision
21056 instructions that target all 64 registers in the vector/scalar
21057 floating point register set that were added in version 2.07 of the
21058 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
21059 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
21060 @option{-mcpu=power9} options.
21063 @itemx -mno-upper-regs
21064 @opindex mupper-regs
21065 @opindex mno-upper-regs
21066 Generate code that uses (does not use) the scalar
21067 instructions that target all 64 registers in the vector/scalar
21068 floating point register set, depending on the model of the machine.
21070 If the @option{-mno-upper-regs} option is used, it turns off both
21071 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
21074 @itemx -mno-float128
21076 @opindex mno-float128
21077 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
21078 and use either software emulation for IEEE 128-bit floating point or
21079 hardware instructions.
21081 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
21082 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
21083 option. The @option{-mfloat128} option only works on PowerPC 64-bit
21086 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
21087 @option{-mfloat128} option will also enable the generation of ISA 3.0
21088 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
21089 floating point will be done with software emulation.
21091 @item -mfloat128-hardware
21092 @itemx -mno-float128-hardware
21093 @opindex mfloat128-hardware
21094 @opindex mno-float128-hardware
21095 Enable/disable using ISA 3.0 hardware instructions to support the
21096 @var{__float128} data type.
21098 If you use @option{-mfloat128-hardware}, it will enable the option
21099 @option{-mfloat128} as well.
21101 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
21102 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
21103 the IEEE 128-bit floating point support will not be enabled.
21105 @item -mfloat-gprs=@var{yes/single/double/no}
21106 @itemx -mfloat-gprs
21107 @opindex mfloat-gprs
21108 This switch enables or disables the generation of floating-point
21109 operations on the general-purpose registers for architectures that
21112 The argument @samp{yes} or @samp{single} enables the use of
21113 single-precision floating-point operations.
21115 The argument @samp{double} enables the use of single and
21116 double-precision floating-point operations.
21118 The argument @samp{no} disables floating-point operations on the
21119 general-purpose registers.
21121 This option is currently only available on the MPC854x.
21127 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
21128 targets (including GNU/Linux). The 32-bit environment sets int, long
21129 and pointer to 32 bits and generates code that runs on any PowerPC
21130 variant. The 64-bit environment sets int to 32 bits and long and
21131 pointer to 64 bits, and generates code for PowerPC64, as for
21132 @option{-mpowerpc64}.
21135 @itemx -mno-fp-in-toc
21136 @itemx -mno-sum-in-toc
21137 @itemx -mminimal-toc
21139 @opindex mno-fp-in-toc
21140 @opindex mno-sum-in-toc
21141 @opindex mminimal-toc
21142 Modify generation of the TOC (Table Of Contents), which is created for
21143 every executable file. The @option{-mfull-toc} option is selected by
21144 default. In that case, GCC allocates at least one TOC entry for
21145 each unique non-automatic variable reference in your program. GCC
21146 also places floating-point constants in the TOC@. However, only
21147 16,384 entries are available in the TOC@.
21149 If you receive a linker error message that saying you have overflowed
21150 the available TOC space, you can reduce the amount of TOC space used
21151 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
21152 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
21153 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
21154 generate code to calculate the sum of an address and a constant at
21155 run time instead of putting that sum into the TOC@. You may specify one
21156 or both of these options. Each causes GCC to produce very slightly
21157 slower and larger code at the expense of conserving TOC space.
21159 If you still run out of space in the TOC even when you specify both of
21160 these options, specify @option{-mminimal-toc} instead. This option causes
21161 GCC to make only one TOC entry for every file. When you specify this
21162 option, GCC produces code that is slower and larger but which
21163 uses extremely little TOC space. You may wish to use this option
21164 only on files that contain less frequently-executed code.
21170 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
21171 @code{long} type, and the infrastructure needed to support them.
21172 Specifying @option{-maix64} implies @option{-mpowerpc64},
21173 while @option{-maix32} disables the 64-bit ABI and
21174 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
21177 @itemx -mno-xl-compat
21178 @opindex mxl-compat
21179 @opindex mno-xl-compat
21180 Produce code that conforms more closely to IBM XL compiler semantics
21181 when using AIX-compatible ABI@. Pass floating-point arguments to
21182 prototyped functions beyond the register save area (RSA) on the stack
21183 in addition to argument FPRs. Do not assume that most significant
21184 double in 128-bit long double value is properly rounded when comparing
21185 values and converting to double. Use XL symbol names for long double
21188 The AIX calling convention was extended but not initially documented to
21189 handle an obscure K&R C case of calling a function that takes the
21190 address of its arguments with fewer arguments than declared. IBM XL
21191 compilers access floating-point arguments that do not fit in the
21192 RSA from the stack when a subroutine is compiled without
21193 optimization. Because always storing floating-point arguments on the
21194 stack is inefficient and rarely needed, this option is not enabled by
21195 default and only is necessary when calling subroutines compiled by IBM
21196 XL compilers without optimization.
21200 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
21201 application written to use message passing with special startup code to
21202 enable the application to run. The system must have PE installed in the
21203 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
21204 must be overridden with the @option{-specs=} option to specify the
21205 appropriate directory location. The Parallel Environment does not
21206 support threads, so the @option{-mpe} option and the @option{-pthread}
21207 option are incompatible.
21209 @item -malign-natural
21210 @itemx -malign-power
21211 @opindex malign-natural
21212 @opindex malign-power
21213 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
21214 @option{-malign-natural} overrides the ABI-defined alignment of larger
21215 types, such as floating-point doubles, on their natural size-based boundary.
21216 The option @option{-malign-power} instructs GCC to follow the ABI-specified
21217 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
21219 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
21223 @itemx -mhard-float
21224 @opindex msoft-float
21225 @opindex mhard-float
21226 Generate code that does not use (uses) the floating-point register set.
21227 Software floating-point emulation is provided if you use the
21228 @option{-msoft-float} option, and pass the option to GCC when linking.
21230 @item -msingle-float
21231 @itemx -mdouble-float
21232 @opindex msingle-float
21233 @opindex mdouble-float
21234 Generate code for single- or double-precision floating-point operations.
21235 @option{-mdouble-float} implies @option{-msingle-float}.
21238 @opindex msimple-fpu
21239 Do not generate @code{sqrt} and @code{div} instructions for hardware
21240 floating-point unit.
21242 @item -mfpu=@var{name}
21244 Specify type of floating-point unit. Valid values for @var{name} are
21245 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
21246 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
21247 @samp{sp_full} (equivalent to @option{-msingle-float}),
21248 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
21251 @opindex mxilinx-fpu
21252 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
21255 @itemx -mno-multiple
21257 @opindex mno-multiple
21258 Generate code that uses (does not use) the load multiple word
21259 instructions and the store multiple word instructions. These
21260 instructions are generated by default on POWER systems, and not
21261 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
21262 PowerPC systems, since those instructions do not work when the
21263 processor is in little-endian mode. The exceptions are PPC740 and
21264 PPC750 which permit these instructions in little-endian mode.
21269 @opindex mno-string
21270 Generate code that uses (does not use) the load string instructions
21271 and the store string word instructions to save multiple registers and
21272 do small block moves. These instructions are generated by default on
21273 POWER systems, and not generated on PowerPC systems. Do not use
21274 @option{-mstring} on little-endian PowerPC systems, since those
21275 instructions do not work when the processor is in little-endian mode.
21276 The exceptions are PPC740 and PPC750 which permit these instructions
21277 in little-endian mode.
21282 @opindex mno-update
21283 Generate code that uses (does not use) the load or store instructions
21284 that update the base register to the address of the calculated memory
21285 location. These instructions are generated by default. If you use
21286 @option{-mno-update}, there is a small window between the time that the
21287 stack pointer is updated and the address of the previous frame is
21288 stored, which means code that walks the stack frame across interrupts or
21289 signals may get corrupted data.
21291 @item -mavoid-indexed-addresses
21292 @itemx -mno-avoid-indexed-addresses
21293 @opindex mavoid-indexed-addresses
21294 @opindex mno-avoid-indexed-addresses
21295 Generate code that tries to avoid (not avoid) the use of indexed load
21296 or store instructions. These instructions can incur a performance
21297 penalty on Power6 processors in certain situations, such as when
21298 stepping through large arrays that cross a 16M boundary. This option
21299 is enabled by default when targeting Power6 and disabled otherwise.
21302 @itemx -mno-fused-madd
21303 @opindex mfused-madd
21304 @opindex mno-fused-madd
21305 Generate code that uses (does not use) the floating-point multiply and
21306 accumulate instructions. These instructions are generated by default
21307 if hardware floating point is used. The machine-dependent
21308 @option{-mfused-madd} option is now mapped to the machine-independent
21309 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21310 mapped to @option{-ffp-contract=off}.
21316 Generate code that uses (does not use) the half-word multiply and
21317 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
21318 These instructions are generated by default when targeting those
21325 Generate code that uses (does not use) the string-search @samp{dlmzb}
21326 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
21327 generated by default when targeting those processors.
21329 @item -mno-bit-align
21331 @opindex mno-bit-align
21332 @opindex mbit-align
21333 On System V.4 and embedded PowerPC systems do not (do) force structures
21334 and unions that contain bit-fields to be aligned to the base type of the
21337 For example, by default a structure containing nothing but 8
21338 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
21339 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
21340 the structure is aligned to a 1-byte boundary and is 1 byte in
21343 @item -mno-strict-align
21344 @itemx -mstrict-align
21345 @opindex mno-strict-align
21346 @opindex mstrict-align
21347 On System V.4 and embedded PowerPC systems do not (do) assume that
21348 unaligned memory references are handled by the system.
21350 @item -mrelocatable
21351 @itemx -mno-relocatable
21352 @opindex mrelocatable
21353 @opindex mno-relocatable
21354 Generate code that allows (does not allow) a static executable to be
21355 relocated to a different address at run time. A simple embedded
21356 PowerPC system loader should relocate the entire contents of
21357 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
21358 a table of 32-bit addresses generated by this option. For this to
21359 work, all objects linked together must be compiled with
21360 @option{-mrelocatable} or @option{-mrelocatable-lib}.
21361 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
21363 @item -mrelocatable-lib
21364 @itemx -mno-relocatable-lib
21365 @opindex mrelocatable-lib
21366 @opindex mno-relocatable-lib
21367 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
21368 @code{.fixup} section to allow static executables to be relocated at
21369 run time, but @option{-mrelocatable-lib} does not use the smaller stack
21370 alignment of @option{-mrelocatable}. Objects compiled with
21371 @option{-mrelocatable-lib} may be linked with objects compiled with
21372 any combination of the @option{-mrelocatable} options.
21378 On System V.4 and embedded PowerPC systems do not (do) assume that
21379 register 2 contains a pointer to a global area pointing to the addresses
21380 used in the program.
21383 @itemx -mlittle-endian
21385 @opindex mlittle-endian
21386 On System V.4 and embedded PowerPC systems compile code for the
21387 processor in little-endian mode. The @option{-mlittle-endian} option is
21388 the same as @option{-mlittle}.
21391 @itemx -mbig-endian
21393 @opindex mbig-endian
21394 On System V.4 and embedded PowerPC systems compile code for the
21395 processor in big-endian mode. The @option{-mbig-endian} option is
21396 the same as @option{-mbig}.
21398 @item -mdynamic-no-pic
21399 @opindex mdynamic-no-pic
21400 On Darwin and Mac OS X systems, compile code so that it is not
21401 relocatable, but that its external references are relocatable. The
21402 resulting code is suitable for applications, but not shared
21405 @item -msingle-pic-base
21406 @opindex msingle-pic-base
21407 Treat the register used for PIC addressing as read-only, rather than
21408 loading it in the prologue for each function. The runtime system is
21409 responsible for initializing this register with an appropriate value
21410 before execution begins.
21412 @item -mprioritize-restricted-insns=@var{priority}
21413 @opindex mprioritize-restricted-insns
21414 This option controls the priority that is assigned to
21415 dispatch-slot restricted instructions during the second scheduling
21416 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
21417 or @samp{2} to assign no, highest, or second-highest (respectively)
21418 priority to dispatch-slot restricted
21421 @item -msched-costly-dep=@var{dependence_type}
21422 @opindex msched-costly-dep
21423 This option controls which dependences are considered costly
21424 by the target during instruction scheduling. The argument
21425 @var{dependence_type} takes one of the following values:
21429 No dependence is costly.
21432 All dependences are costly.
21434 @item @samp{true_store_to_load}
21435 A true dependence from store to load is costly.
21437 @item @samp{store_to_load}
21438 Any dependence from store to load is costly.
21441 Any dependence for which the latency is greater than or equal to
21442 @var{number} is costly.
21445 @item -minsert-sched-nops=@var{scheme}
21446 @opindex minsert-sched-nops
21447 This option controls which NOP insertion scheme is used during
21448 the second scheduling pass. The argument @var{scheme} takes one of the
21456 Pad with NOPs any dispatch group that has vacant issue slots,
21457 according to the scheduler's grouping.
21459 @item @samp{regroup_exact}
21460 Insert NOPs to force costly dependent insns into
21461 separate groups. Insert exactly as many NOPs as needed to force an insn
21462 to a new group, according to the estimated processor grouping.
21465 Insert NOPs to force costly dependent insns into
21466 separate groups. Insert @var{number} NOPs to force an insn to a new group.
21470 @opindex mcall-sysv
21471 On System V.4 and embedded PowerPC systems compile code using calling
21472 conventions that adhere to the March 1995 draft of the System V
21473 Application Binary Interface, PowerPC processor supplement. This is the
21474 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
21476 @item -mcall-sysv-eabi
21478 @opindex mcall-sysv-eabi
21479 @opindex mcall-eabi
21480 Specify both @option{-mcall-sysv} and @option{-meabi} options.
21482 @item -mcall-sysv-noeabi
21483 @opindex mcall-sysv-noeabi
21484 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
21486 @item -mcall-aixdesc
21488 On System V.4 and embedded PowerPC systems compile code for the AIX
21492 @opindex mcall-linux
21493 On System V.4 and embedded PowerPC systems compile code for the
21494 Linux-based GNU system.
21496 @item -mcall-freebsd
21497 @opindex mcall-freebsd
21498 On System V.4 and embedded PowerPC systems compile code for the
21499 FreeBSD operating system.
21501 @item -mcall-netbsd
21502 @opindex mcall-netbsd
21503 On System V.4 and embedded PowerPC systems compile code for the
21504 NetBSD operating system.
21506 @item -mcall-openbsd
21507 @opindex mcall-netbsd
21508 On System V.4 and embedded PowerPC systems compile code for the
21509 OpenBSD operating system.
21511 @item -maix-struct-return
21512 @opindex maix-struct-return
21513 Return all structures in memory (as specified by the AIX ABI)@.
21515 @item -msvr4-struct-return
21516 @opindex msvr4-struct-return
21517 Return structures smaller than 8 bytes in registers (as specified by the
21520 @item -mabi=@var{abi-type}
21522 Extend the current ABI with a particular extension, or remove such extension.
21523 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
21524 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
21525 @samp{elfv1}, @samp{elfv2}@.
21529 Extend the current ABI with SPE ABI extensions. This does not change
21530 the default ABI, instead it adds the SPE ABI extensions to the current
21534 @opindex mabi=no-spe
21535 Disable Book-E SPE ABI extensions for the current ABI@.
21537 @item -mabi=ibmlongdouble
21538 @opindex mabi=ibmlongdouble
21539 Change the current ABI to use IBM extended-precision long double.
21540 This is a PowerPC 32-bit SYSV ABI option.
21542 @item -mabi=ieeelongdouble
21543 @opindex mabi=ieeelongdouble
21544 Change the current ABI to use IEEE extended-precision long double.
21545 This is a PowerPC 32-bit Linux ABI option.
21548 @opindex mabi=elfv1
21549 Change the current ABI to use the ELFv1 ABI.
21550 This is the default ABI for big-endian PowerPC 64-bit Linux.
21551 Overriding the default ABI requires special system support and is
21552 likely to fail in spectacular ways.
21555 @opindex mabi=elfv2
21556 Change the current ABI to use the ELFv2 ABI.
21557 This is the default ABI for little-endian PowerPC 64-bit Linux.
21558 Overriding the default ABI requires special system support and is
21559 likely to fail in spectacular ways.
21561 @item -mgnu-attribute
21562 @itemx -mno-gnu-attribute
21563 @opindex mgnu-attribute
21564 @opindex mno-gnu-attribute
21565 Emit .gnu_attribute assembly directives to set tag/value pairs in a
21566 .gnu.attributes section that specify ABI variations in function
21567 parameters or return values.
21570 @itemx -mno-prototype
21571 @opindex mprototype
21572 @opindex mno-prototype
21573 On System V.4 and embedded PowerPC systems assume that all calls to
21574 variable argument functions are properly prototyped. Otherwise, the
21575 compiler must insert an instruction before every non-prototyped call to
21576 set or clear bit 6 of the condition code register (@code{CR}) to
21577 indicate whether floating-point values are passed in the floating-point
21578 registers in case the function takes variable arguments. With
21579 @option{-mprototype}, only calls to prototyped variable argument functions
21580 set or clear the bit.
21584 On embedded PowerPC systems, assume that the startup module is called
21585 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
21586 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
21591 On embedded PowerPC systems, assume that the startup module is called
21592 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
21597 On embedded PowerPC systems, assume that the startup module is called
21598 @file{crt0.o} and the standard C libraries are @file{libads.a} and
21601 @item -myellowknife
21602 @opindex myellowknife
21603 On embedded PowerPC systems, assume that the startup module is called
21604 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
21609 On System V.4 and embedded PowerPC systems, specify that you are
21610 compiling for a VxWorks system.
21614 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
21615 header to indicate that @samp{eabi} extended relocations are used.
21621 On System V.4 and embedded PowerPC systems do (do not) adhere to the
21622 Embedded Applications Binary Interface (EABI), which is a set of
21623 modifications to the System V.4 specifications. Selecting @option{-meabi}
21624 means that the stack is aligned to an 8-byte boundary, a function
21625 @code{__eabi} is called from @code{main} to set up the EABI
21626 environment, and the @option{-msdata} option can use both @code{r2} and
21627 @code{r13} to point to two separate small data areas. Selecting
21628 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
21629 no EABI initialization function is called from @code{main}, and the
21630 @option{-msdata} option only uses @code{r13} to point to a single
21631 small data area. The @option{-meabi} option is on by default if you
21632 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
21635 @opindex msdata=eabi
21636 On System V.4 and embedded PowerPC systems, put small initialized
21637 @code{const} global and static data in the @code{.sdata2} section, which
21638 is pointed to by register @code{r2}. Put small initialized
21639 non-@code{const} global and static data in the @code{.sdata} section,
21640 which is pointed to by register @code{r13}. Put small uninitialized
21641 global and static data in the @code{.sbss} section, which is adjacent to
21642 the @code{.sdata} section. The @option{-msdata=eabi} option is
21643 incompatible with the @option{-mrelocatable} option. The
21644 @option{-msdata=eabi} option also sets the @option{-memb} option.
21647 @opindex msdata=sysv
21648 On System V.4 and embedded PowerPC systems, put small global and static
21649 data in the @code{.sdata} section, which is pointed to by register
21650 @code{r13}. Put small uninitialized global and static data in the
21651 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
21652 The @option{-msdata=sysv} option is incompatible with the
21653 @option{-mrelocatable} option.
21655 @item -msdata=default
21657 @opindex msdata=default
21659 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
21660 compile code the same as @option{-msdata=eabi}, otherwise compile code the
21661 same as @option{-msdata=sysv}.
21664 @opindex msdata=data
21665 On System V.4 and embedded PowerPC systems, put small global
21666 data in the @code{.sdata} section. Put small uninitialized global
21667 data in the @code{.sbss} section. Do not use register @code{r13}
21668 to address small data however. This is the default behavior unless
21669 other @option{-msdata} options are used.
21673 @opindex msdata=none
21675 On embedded PowerPC systems, put all initialized global and static data
21676 in the @code{.data} section, and all uninitialized data in the
21677 @code{.bss} section.
21679 @item -mblock-move-inline-limit=@var{num}
21680 @opindex mblock-move-inline-limit
21681 Inline all block moves (such as calls to @code{memcpy} or structure
21682 copies) less than or equal to @var{num} bytes. The minimum value for
21683 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
21684 targets. The default value is target-specific.
21688 @cindex smaller data references (PowerPC)
21689 @cindex .sdata/.sdata2 references (PowerPC)
21690 On embedded PowerPC systems, put global and static items less than or
21691 equal to @var{num} bytes into the small data or BSS sections instead of
21692 the normal data or BSS section. By default, @var{num} is 8. The
21693 @option{-G @var{num}} switch is also passed to the linker.
21694 All modules should be compiled with the same @option{-G @var{num}} value.
21697 @itemx -mno-regnames
21699 @opindex mno-regnames
21700 On System V.4 and embedded PowerPC systems do (do not) emit register
21701 names in the assembly language output using symbolic forms.
21704 @itemx -mno-longcall
21706 @opindex mno-longcall
21707 By default assume that all calls are far away so that a longer and more
21708 expensive calling sequence is required. This is required for calls
21709 farther than 32 megabytes (33,554,432 bytes) from the current location.
21710 A short call is generated if the compiler knows
21711 the call cannot be that far away. This setting can be overridden by
21712 the @code{shortcall} function attribute, or by @code{#pragma
21715 Some linkers are capable of detecting out-of-range calls and generating
21716 glue code on the fly. On these systems, long calls are unnecessary and
21717 generate slower code. As of this writing, the AIX linker can do this,
21718 as can the GNU linker for PowerPC/64. It is planned to add this feature
21719 to the GNU linker for 32-bit PowerPC systems as well.
21721 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
21722 callee, L42}, plus a @dfn{branch island} (glue code). The two target
21723 addresses represent the callee and the branch island. The
21724 Darwin/PPC linker prefers the first address and generates a @code{bl
21725 callee} if the PPC @code{bl} instruction reaches the callee directly;
21726 otherwise, the linker generates @code{bl L42} to call the branch
21727 island. The branch island is appended to the body of the
21728 calling function; it computes the full 32-bit address of the callee
21731 On Mach-O (Darwin) systems, this option directs the compiler emit to
21732 the glue for every direct call, and the Darwin linker decides whether
21733 to use or discard it.
21735 In the future, GCC may ignore all longcall specifications
21736 when the linker is known to generate glue.
21738 @item -mtls-markers
21739 @itemx -mno-tls-markers
21740 @opindex mtls-markers
21741 @opindex mno-tls-markers
21742 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
21743 specifying the function argument. The relocation allows the linker to
21744 reliably associate function call with argument setup instructions for
21745 TLS optimization, which in turn allows GCC to better schedule the
21750 Adds support for multithreading with the @dfn{pthreads} library.
21751 This option sets flags for both the preprocessor and linker.
21756 This option enables use of the reciprocal estimate and
21757 reciprocal square root estimate instructions with additional
21758 Newton-Raphson steps to increase precision instead of doing a divide or
21759 square root and divide for floating-point arguments. You should use
21760 the @option{-ffast-math} option when using @option{-mrecip} (or at
21761 least @option{-funsafe-math-optimizations},
21762 @option{-ffinite-math-only}, @option{-freciprocal-math} and
21763 @option{-fno-trapping-math}). Note that while the throughput of the
21764 sequence is generally higher than the throughput of the non-reciprocal
21765 instruction, the precision of the sequence can be decreased by up to 2
21766 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
21769 @item -mrecip=@var{opt}
21770 @opindex mrecip=opt
21771 This option controls which reciprocal estimate instructions
21772 may be used. @var{opt} is a comma-separated list of options, which may
21773 be preceded by a @code{!} to invert the option:
21778 Enable all estimate instructions.
21781 Enable the default instructions, equivalent to @option{-mrecip}.
21784 Disable all estimate instructions, equivalent to @option{-mno-recip}.
21787 Enable the reciprocal approximation instructions for both
21788 single and double precision.
21791 Enable the single-precision reciprocal approximation instructions.
21794 Enable the double-precision reciprocal approximation instructions.
21797 Enable the reciprocal square root approximation instructions for both
21798 single and double precision.
21801 Enable the single-precision reciprocal square root approximation instructions.
21804 Enable the double-precision reciprocal square root approximation instructions.
21808 So, for example, @option{-mrecip=all,!rsqrtd} enables
21809 all of the reciprocal estimate instructions, except for the
21810 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
21811 which handle the double-precision reciprocal square root calculations.
21813 @item -mrecip-precision
21814 @itemx -mno-recip-precision
21815 @opindex mrecip-precision
21816 Assume (do not assume) that the reciprocal estimate instructions
21817 provide higher-precision estimates than is mandated by the PowerPC
21818 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
21819 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
21820 The double-precision square root estimate instructions are not generated by
21821 default on low-precision machines, since they do not provide an
21822 estimate that converges after three steps.
21824 @item -mveclibabi=@var{type}
21825 @opindex mveclibabi
21826 Specifies the ABI type to use for vectorizing intrinsics using an
21827 external library. The only type supported at present is @samp{mass},
21828 which specifies to use IBM's Mathematical Acceleration Subsystem
21829 (MASS) libraries for vectorizing intrinsics using external libraries.
21830 GCC currently emits calls to @code{acosd2}, @code{acosf4},
21831 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
21832 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
21833 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
21834 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
21835 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
21836 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
21837 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
21838 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
21839 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
21840 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
21841 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
21842 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
21843 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
21844 for power7. Both @option{-ftree-vectorize} and
21845 @option{-funsafe-math-optimizations} must also be enabled. The MASS
21846 libraries must be specified at link time.
21851 Generate (do not generate) the @code{friz} instruction when the
21852 @option{-funsafe-math-optimizations} option is used to optimize
21853 rounding of floating-point values to 64-bit integer and back to floating
21854 point. The @code{friz} instruction does not return the same value if
21855 the floating-point number is too large to fit in an integer.
21857 @item -mpointers-to-nested-functions
21858 @itemx -mno-pointers-to-nested-functions
21859 @opindex mpointers-to-nested-functions
21860 Generate (do not generate) code to load up the static chain register
21861 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
21862 systems where a function pointer points to a 3-word descriptor giving
21863 the function address, TOC value to be loaded in register @code{r2}, and
21864 static chain value to be loaded in register @code{r11}. The
21865 @option{-mpointers-to-nested-functions} is on by default. You cannot
21866 call through pointers to nested functions or pointers
21867 to functions compiled in other languages that use the static chain if
21868 you use @option{-mno-pointers-to-nested-functions}.
21870 @item -msave-toc-indirect
21871 @itemx -mno-save-toc-indirect
21872 @opindex msave-toc-indirect
21873 Generate (do not generate) code to save the TOC value in the reserved
21874 stack location in the function prologue if the function calls through
21875 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
21876 saved in the prologue, it is saved just before the call through the
21877 pointer. The @option{-mno-save-toc-indirect} option is the default.
21879 @item -mcompat-align-parm
21880 @itemx -mno-compat-align-parm
21881 @opindex mcompat-align-parm
21882 Generate (do not generate) code to pass structure parameters with a
21883 maximum alignment of 64 bits, for compatibility with older versions
21886 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
21887 structure parameter on a 128-bit boundary when that structure contained
21888 a member requiring 128-bit alignment. This is corrected in more
21889 recent versions of GCC. This option may be used to generate code
21890 that is compatible with functions compiled with older versions of
21893 The @option{-mno-compat-align-parm} option is the default.
21897 @subsection RX Options
21900 These command-line options are defined for RX targets:
21903 @item -m64bit-doubles
21904 @itemx -m32bit-doubles
21905 @opindex m64bit-doubles
21906 @opindex m32bit-doubles
21907 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21908 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21909 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
21910 works on 32-bit values, which is why the default is
21911 @option{-m32bit-doubles}.
21917 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
21918 floating-point hardware. The default is enabled for the RX600
21919 series and disabled for the RX200 series.
21921 Floating-point instructions are only generated for 32-bit floating-point
21922 values, however, so the FPU hardware is not used for doubles if the
21923 @option{-m64bit-doubles} option is used.
21925 @emph{Note} If the @option{-fpu} option is enabled then
21926 @option{-funsafe-math-optimizations} is also enabled automatically.
21927 This is because the RX FPU instructions are themselves unsafe.
21929 @item -mcpu=@var{name}
21931 Selects the type of RX CPU to be targeted. Currently three types are
21932 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
21933 the specific @samp{RX610} CPU. The default is @samp{RX600}.
21935 The only difference between @samp{RX600} and @samp{RX610} is that the
21936 @samp{RX610} does not support the @code{MVTIPL} instruction.
21938 The @samp{RX200} series does not have a hardware floating-point unit
21939 and so @option{-nofpu} is enabled by default when this type is
21942 @item -mbig-endian-data
21943 @itemx -mlittle-endian-data
21944 @opindex mbig-endian-data
21945 @opindex mlittle-endian-data
21946 Store data (but not code) in the big-endian format. The default is
21947 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
21950 @item -msmall-data-limit=@var{N}
21951 @opindex msmall-data-limit
21952 Specifies the maximum size in bytes of global and static variables
21953 which can be placed into the small data area. Using the small data
21954 area can lead to smaller and faster code, but the size of area is
21955 limited and it is up to the programmer to ensure that the area does
21956 not overflow. Also when the small data area is used one of the RX's
21957 registers (usually @code{r13}) is reserved for use pointing to this
21958 area, so it is no longer available for use by the compiler. This
21959 could result in slower and/or larger code if variables are pushed onto
21960 the stack instead of being held in this register.
21962 Note, common variables (variables that have not been initialized) and
21963 constants are not placed into the small data area as they are assigned
21964 to other sections in the output executable.
21966 The default value is zero, which disables this feature. Note, this
21967 feature is not enabled by default with higher optimization levels
21968 (@option{-O2} etc) because of the potentially detrimental effects of
21969 reserving a register. It is up to the programmer to experiment and
21970 discover whether this feature is of benefit to their program. See the
21971 description of the @option{-mpid} option for a description of how the
21972 actual register to hold the small data area pointer is chosen.
21978 Use the simulator runtime. The default is to use the libgloss
21979 board-specific runtime.
21981 @item -mas100-syntax
21982 @itemx -mno-as100-syntax
21983 @opindex mas100-syntax
21984 @opindex mno-as100-syntax
21985 When generating assembler output use a syntax that is compatible with
21986 Renesas's AS100 assembler. This syntax can also be handled by the GAS
21987 assembler, but it has some restrictions so it is not generated by default.
21989 @item -mmax-constant-size=@var{N}
21990 @opindex mmax-constant-size
21991 Specifies the maximum size, in bytes, of a constant that can be used as
21992 an operand in a RX instruction. Although the RX instruction set does
21993 allow constants of up to 4 bytes in length to be used in instructions,
21994 a longer value equates to a longer instruction. Thus in some
21995 circumstances it can be beneficial to restrict the size of constants
21996 that are used in instructions. Constants that are too big are instead
21997 placed into a constant pool and referenced via register indirection.
21999 The value @var{N} can be between 0 and 4. A value of 0 (the default)
22000 or 4 means that constants of any size are allowed.
22004 Enable linker relaxation. Linker relaxation is a process whereby the
22005 linker attempts to reduce the size of a program by finding shorter
22006 versions of various instructions. Disabled by default.
22008 @item -mint-register=@var{N}
22009 @opindex mint-register
22010 Specify the number of registers to reserve for fast interrupt handler
22011 functions. The value @var{N} can be between 0 and 4. A value of 1
22012 means that register @code{r13} is reserved for the exclusive use
22013 of fast interrupt handlers. A value of 2 reserves @code{r13} and
22014 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
22015 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
22016 A value of 0, the default, does not reserve any registers.
22018 @item -msave-acc-in-interrupts
22019 @opindex msave-acc-in-interrupts
22020 Specifies that interrupt handler functions should preserve the
22021 accumulator register. This is only necessary if normal code might use
22022 the accumulator register, for example because it performs 64-bit
22023 multiplications. The default is to ignore the accumulator as this
22024 makes the interrupt handlers faster.
22030 Enables the generation of position independent data. When enabled any
22031 access to constant data is done via an offset from a base address
22032 held in a register. This allows the location of constant data to be
22033 determined at run time without requiring the executable to be
22034 relocated, which is a benefit to embedded applications with tight
22035 memory constraints. Data that can be modified is not affected by this
22038 Note, using this feature reserves a register, usually @code{r13}, for
22039 the constant data base address. This can result in slower and/or
22040 larger code, especially in complicated functions.
22042 The actual register chosen to hold the constant data base address
22043 depends upon whether the @option{-msmall-data-limit} and/or the
22044 @option{-mint-register} command-line options are enabled. Starting
22045 with register @code{r13} and proceeding downwards, registers are
22046 allocated first to satisfy the requirements of @option{-mint-register},
22047 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
22048 is possible for the small data area register to be @code{r8} if both
22049 @option{-mint-register=4} and @option{-mpid} are specified on the
22052 By default this feature is not enabled. The default can be restored
22053 via the @option{-mno-pid} command-line option.
22055 @item -mno-warn-multiple-fast-interrupts
22056 @itemx -mwarn-multiple-fast-interrupts
22057 @opindex mno-warn-multiple-fast-interrupts
22058 @opindex mwarn-multiple-fast-interrupts
22059 Prevents GCC from issuing a warning message if it finds more than one
22060 fast interrupt handler when it is compiling a file. The default is to
22061 issue a warning for each extra fast interrupt handler found, as the RX
22062 only supports one such interrupt.
22064 @item -mallow-string-insns
22065 @itemx -mno-allow-string-insns
22066 @opindex mallow-string-insns
22067 @opindex mno-allow-string-insns
22068 Enables or disables the use of the string manipulation instructions
22069 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
22070 @code{SWHILE} and also the @code{RMPA} instruction. These
22071 instructions may prefetch data, which is not safe to do if accessing
22072 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
22073 for more information).
22075 The default is to allow these instructions, but it is not possible for
22076 GCC to reliably detect all circumstances where a string instruction
22077 might be used to access an I/O register, so their use cannot be
22078 disabled automatically. Instead it is reliant upon the programmer to
22079 use the @option{-mno-allow-string-insns} option if their program
22080 accesses I/O space.
22082 When the instructions are enabled GCC defines the C preprocessor
22083 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
22084 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
22090 Use only (or not only) @code{JSR} instructions to access functions.
22091 This option can be used when code size exceeds the range of @code{BSR}
22092 instructions. Note that @option{-mno-jsr} does not mean to not use
22093 @code{JSR} but instead means that any type of branch may be used.
22096 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
22097 has special significance to the RX port when used with the
22098 @code{interrupt} function attribute. This attribute indicates a
22099 function intended to process fast interrupts. GCC ensures
22100 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
22101 and/or @code{r13} and only provided that the normal use of the
22102 corresponding registers have been restricted via the
22103 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
22106 @node S/390 and zSeries Options
22107 @subsection S/390 and zSeries Options
22108 @cindex S/390 and zSeries Options
22110 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
22114 @itemx -msoft-float
22115 @opindex mhard-float
22116 @opindex msoft-float
22117 Use (do not use) the hardware floating-point instructions and registers
22118 for floating-point operations. When @option{-msoft-float} is specified,
22119 functions in @file{libgcc.a} are used to perform floating-point
22120 operations. When @option{-mhard-float} is specified, the compiler
22121 generates IEEE floating-point instructions. This is the default.
22124 @itemx -mno-hard-dfp
22126 @opindex mno-hard-dfp
22127 Use (do not use) the hardware decimal-floating-point instructions for
22128 decimal-floating-point operations. When @option{-mno-hard-dfp} is
22129 specified, functions in @file{libgcc.a} are used to perform
22130 decimal-floating-point operations. When @option{-mhard-dfp} is
22131 specified, the compiler generates decimal-floating-point hardware
22132 instructions. This is the default for @option{-march=z9-ec} or higher.
22134 @item -mlong-double-64
22135 @itemx -mlong-double-128
22136 @opindex mlong-double-64
22137 @opindex mlong-double-128
22138 These switches control the size of @code{long double} type. A size
22139 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22140 type. This is the default.
22143 @itemx -mno-backchain
22144 @opindex mbackchain
22145 @opindex mno-backchain
22146 Store (do not store) the address of the caller's frame as backchain pointer
22147 into the callee's stack frame.
22148 A backchain may be needed to allow debugging using tools that do not understand
22149 DWARF call frame information.
22150 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
22151 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
22152 the backchain is placed into the topmost word of the 96/160 byte register
22155 In general, code compiled with @option{-mbackchain} is call-compatible with
22156 code compiled with @option{-mmo-backchain}; however, use of the backchain
22157 for debugging purposes usually requires that the whole binary is built with
22158 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
22159 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22160 to build a linux kernel use @option{-msoft-float}.
22162 The default is to not maintain the backchain.
22164 @item -mpacked-stack
22165 @itemx -mno-packed-stack
22166 @opindex mpacked-stack
22167 @opindex mno-packed-stack
22168 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
22169 specified, the compiler uses the all fields of the 96/160 byte register save
22170 area only for their default purpose; unused fields still take up stack space.
22171 When @option{-mpacked-stack} is specified, register save slots are densely
22172 packed at the top of the register save area; unused space is reused for other
22173 purposes, allowing for more efficient use of the available stack space.
22174 However, when @option{-mbackchain} is also in effect, the topmost word of
22175 the save area is always used to store the backchain, and the return address
22176 register is always saved two words below the backchain.
22178 As long as the stack frame backchain is not used, code generated with
22179 @option{-mpacked-stack} is call-compatible with code generated with
22180 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
22181 S/390 or zSeries generated code that uses the stack frame backchain at run
22182 time, not just for debugging purposes. Such code is not call-compatible
22183 with code compiled with @option{-mpacked-stack}. Also, note that the
22184 combination of @option{-mbackchain},
22185 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22186 to build a linux kernel use @option{-msoft-float}.
22188 The default is to not use the packed stack layout.
22191 @itemx -mno-small-exec
22192 @opindex msmall-exec
22193 @opindex mno-small-exec
22194 Generate (or do not generate) code using the @code{bras} instruction
22195 to do subroutine calls.
22196 This only works reliably if the total executable size does not
22197 exceed 64k. The default is to use the @code{basr} instruction instead,
22198 which does not have this limitation.
22204 When @option{-m31} is specified, generate code compliant to the
22205 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
22206 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
22207 particular to generate 64-bit instructions. For the @samp{s390}
22208 targets, the default is @option{-m31}, while the @samp{s390x}
22209 targets default to @option{-m64}.
22215 When @option{-mzarch} is specified, generate code using the
22216 instructions available on z/Architecture.
22217 When @option{-mesa} is specified, generate code using the
22218 instructions available on ESA/390. Note that @option{-mesa} is
22219 not possible with @option{-m64}.
22220 When generating code compliant to the GNU/Linux for S/390 ABI,
22221 the default is @option{-mesa}. When generating code compliant
22222 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
22228 The @option{-mhtm} option enables a set of builtins making use of
22229 instructions available with the transactional execution facility
22230 introduced with the IBM zEnterprise EC12 machine generation
22231 @ref{S/390 System z Built-in Functions}.
22232 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
22238 When @option{-mvx} is specified, generate code using the instructions
22239 available with the vector extension facility introduced with the IBM
22240 z13 machine generation.
22241 This option changes the ABI for some vector type values with regard to
22242 alignment and calling conventions. In case vector type values are
22243 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
22244 command will be added to mark the resulting binary with the ABI used.
22245 @option{-mvx} is enabled by default when using @option{-march=z13}.
22248 @itemx -mno-zvector
22250 @opindex mno-zvector
22251 The @option{-mzvector} option enables vector language extensions and
22252 builtins using instructions available with the vector extension
22253 facility introduced with the IBM z13 machine generation.
22254 This option adds support for @samp{vector} to be used as a keyword to
22255 define vector type variables and arguments. @samp{vector} is only
22256 available when GNU extensions are enabled. It will not be expanded
22257 when requesting strict standard compliance e.g. with @option{-std=c99}.
22258 In addition to the GCC low-level builtins @option{-mzvector} enables
22259 a set of builtins added for compatibility with AltiVec-style
22260 implementations like Power and Cell. In order to make use of these
22261 builtins the header file @file{vecintrin.h} needs to be included.
22262 @option{-mzvector} is disabled by default.
22268 Generate (or do not generate) code using the @code{mvcle} instruction
22269 to perform block moves. When @option{-mno-mvcle} is specified,
22270 use a @code{mvc} loop instead. This is the default unless optimizing for
22277 Print (or do not print) additional debug information when compiling.
22278 The default is to not print debug information.
22280 @item -march=@var{cpu-type}
22282 Generate code that runs on @var{cpu-type}, which is the name of a
22283 system representing a certain processor type. Possible values for
22284 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
22285 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
22286 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
22289 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
22290 @samp{g6} are deprecated and will be removed with future releases.
22292 Specifying @samp{native} as cpu type can be used to select the best
22293 architecture option for the host processor.
22294 @option{-march=native} has no effect if GCC does not recognize the
22297 @item -mtune=@var{cpu-type}
22299 Tune to @var{cpu-type} everything applicable about the generated code,
22300 except for the ABI and the set of available instructions.
22301 The list of @var{cpu-type} values is the same as for @option{-march}.
22302 The default is the value used for @option{-march}.
22305 @itemx -mno-tpf-trace
22306 @opindex mtpf-trace
22307 @opindex mno-tpf-trace
22308 Generate code that adds (does not add) in TPF OS specific branches to trace
22309 routines in the operating system. This option is off by default, even
22310 when compiling for the TPF OS@.
22313 @itemx -mno-fused-madd
22314 @opindex mfused-madd
22315 @opindex mno-fused-madd
22316 Generate code that uses (does not use) the floating-point multiply and
22317 accumulate instructions. These instructions are generated by default if
22318 hardware floating point is used.
22320 @item -mwarn-framesize=@var{framesize}
22321 @opindex mwarn-framesize
22322 Emit a warning if the current function exceeds the given frame size. Because
22323 this is a compile-time check it doesn't need to be a real problem when the program
22324 runs. It is intended to identify functions that most probably cause
22325 a stack overflow. It is useful to be used in an environment with limited stack
22326 size e.g.@: the linux kernel.
22328 @item -mwarn-dynamicstack
22329 @opindex mwarn-dynamicstack
22330 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
22331 arrays. This is generally a bad idea with a limited stack size.
22333 @item -mstack-guard=@var{stack-guard}
22334 @itemx -mstack-size=@var{stack-size}
22335 @opindex mstack-guard
22336 @opindex mstack-size
22337 If these options are provided the S/390 back end emits additional instructions in
22338 the function prologue that trigger a trap if the stack size is @var{stack-guard}
22339 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
22340 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
22341 the frame size of the compiled function is chosen.
22342 These options are intended to be used to help debugging stack overflow problems.
22343 The additionally emitted code causes only little overhead and hence can also be
22344 used in production-like systems without greater performance degradation. The given
22345 values have to be exact powers of 2 and @var{stack-size} has to be greater than
22346 @var{stack-guard} without exceeding 64k.
22347 In order to be efficient the extra code makes the assumption that the stack starts
22348 at an address aligned to the value given by @var{stack-size}.
22349 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
22351 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
22353 If the hotpatch option is enabled, a ``hot-patching'' function
22354 prologue is generated for all functions in the compilation unit.
22355 The funtion label is prepended with the given number of two-byte
22356 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
22357 the label, 2 * @var{post-halfwords} bytes are appended, using the
22358 largest NOP like instructions the architecture allows (maximum
22361 If both arguments are zero, hotpatching is disabled.
22363 This option can be overridden for individual functions with the
22364 @code{hotpatch} attribute.
22367 @node Score Options
22368 @subsection Score Options
22369 @cindex Score Options
22371 These options are defined for Score implementations:
22376 Compile code for big-endian mode. This is the default.
22380 Compile code for little-endian mode.
22384 Disable generation of @code{bcnz} instructions.
22388 Enable generation of unaligned load and store instructions.
22392 Enable the use of multiply-accumulate instructions. Disabled by default.
22396 Specify the SCORE5 as the target architecture.
22400 Specify the SCORE5U of the target architecture.
22404 Specify the SCORE7 as the target architecture. This is the default.
22408 Specify the SCORE7D as the target architecture.
22412 @subsection SH Options
22414 These @samp{-m} options are defined for the SH implementations:
22419 Generate code for the SH1.
22423 Generate code for the SH2.
22426 Generate code for the SH2e.
22430 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
22431 that the floating-point unit is not used.
22433 @item -m2a-single-only
22434 @opindex m2a-single-only
22435 Generate code for the SH2a-FPU, in such a way that no double-precision
22436 floating-point operations are used.
22439 @opindex m2a-single
22440 Generate code for the SH2a-FPU assuming the floating-point unit is in
22441 single-precision mode by default.
22445 Generate code for the SH2a-FPU assuming the floating-point unit is in
22446 double-precision mode by default.
22450 Generate code for the SH3.
22454 Generate code for the SH3e.
22458 Generate code for the SH4 without a floating-point unit.
22460 @item -m4-single-only
22461 @opindex m4-single-only
22462 Generate code for the SH4 with a floating-point unit that only
22463 supports single-precision arithmetic.
22467 Generate code for the SH4 assuming the floating-point unit is in
22468 single-precision mode by default.
22472 Generate code for the SH4.
22476 Generate code for SH4-100.
22478 @item -m4-100-nofpu
22479 @opindex m4-100-nofpu
22480 Generate code for SH4-100 in such a way that the
22481 floating-point unit is not used.
22483 @item -m4-100-single
22484 @opindex m4-100-single
22485 Generate code for SH4-100 assuming the floating-point unit is in
22486 single-precision mode by default.
22488 @item -m4-100-single-only
22489 @opindex m4-100-single-only
22490 Generate code for SH4-100 in such a way that no double-precision
22491 floating-point operations are used.
22495 Generate code for SH4-200.
22497 @item -m4-200-nofpu
22498 @opindex m4-200-nofpu
22499 Generate code for SH4-200 without in such a way that the
22500 floating-point unit is not used.
22502 @item -m4-200-single
22503 @opindex m4-200-single
22504 Generate code for SH4-200 assuming the floating-point unit is in
22505 single-precision mode by default.
22507 @item -m4-200-single-only
22508 @opindex m4-200-single-only
22509 Generate code for SH4-200 in such a way that no double-precision
22510 floating-point operations are used.
22514 Generate code for SH4-300.
22516 @item -m4-300-nofpu
22517 @opindex m4-300-nofpu
22518 Generate code for SH4-300 without in such a way that the
22519 floating-point unit is not used.
22521 @item -m4-300-single
22522 @opindex m4-300-single
22523 Generate code for SH4-300 in such a way that no double-precision
22524 floating-point operations are used.
22526 @item -m4-300-single-only
22527 @opindex m4-300-single-only
22528 Generate code for SH4-300 in such a way that no double-precision
22529 floating-point operations are used.
22533 Generate code for SH4-340 (no MMU, no FPU).
22537 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
22542 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
22543 floating-point unit is not used.
22545 @item -m4a-single-only
22546 @opindex m4a-single-only
22547 Generate code for the SH4a, in such a way that no double-precision
22548 floating-point operations are used.
22551 @opindex m4a-single
22552 Generate code for the SH4a assuming the floating-point unit is in
22553 single-precision mode by default.
22557 Generate code for the SH4a.
22561 Same as @option{-m4a-nofpu}, except that it implicitly passes
22562 @option{-dsp} to the assembler. GCC doesn't generate any DSP
22563 instructions at the moment.
22567 Compile code for the processor in big-endian mode.
22571 Compile code for the processor in little-endian mode.
22575 Align doubles at 64-bit boundaries. Note that this changes the calling
22576 conventions, and thus some functions from the standard C library do
22577 not work unless you recompile it first with @option{-mdalign}.
22581 Shorten some address references at link time, when possible; uses the
22582 linker option @option{-relax}.
22586 Use 32-bit offsets in @code{switch} tables. The default is to use
22591 Enable the use of bit manipulation instructions on SH2A.
22595 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
22596 alignment constraints.
22600 Comply with the calling conventions defined by Renesas.
22603 @opindex mno-renesas
22604 Comply with the calling conventions defined for GCC before the Renesas
22605 conventions were available. This option is the default for all
22606 targets of the SH toolchain.
22609 @opindex mnomacsave
22610 Mark the @code{MAC} register as call-clobbered, even if
22611 @option{-mrenesas} is given.
22617 Control the IEEE compliance of floating-point comparisons, which affects the
22618 handling of cases where the result of a comparison is unordered. By default
22619 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
22620 enabled @option{-mno-ieee} is implicitly set, which results in faster
22621 floating-point greater-equal and less-equal comparisons. The implicit settings
22622 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
22624 @item -minline-ic_invalidate
22625 @opindex minline-ic_invalidate
22626 Inline code to invalidate instruction cache entries after setting up
22627 nested function trampolines.
22628 This option has no effect if @option{-musermode} is in effect and the selected
22629 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
22631 If the selected code generation option does not allow the use of the @code{icbi}
22632 instruction, and @option{-musermode} is not in effect, the inlined code
22633 manipulates the instruction cache address array directly with an associative
22634 write. This not only requires privileged mode at run time, but it also
22635 fails if the cache line had been mapped via the TLB and has become unmapped.
22639 Dump instruction size and location in the assembly code.
22642 @opindex mpadstruct
22643 This option is deprecated. It pads structures to multiple of 4 bytes,
22644 which is incompatible with the SH ABI@.
22646 @item -matomic-model=@var{model}
22647 @opindex matomic-model=@var{model}
22648 Sets the model of atomic operations and additional parameters as a comma
22649 separated list. For details on the atomic built-in functions see
22650 @ref{__atomic Builtins}. The following models and parameters are supported:
22655 Disable compiler generated atomic sequences and emit library calls for atomic
22656 operations. This is the default if the target is not @code{sh*-*-linux*}.
22659 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
22660 built-in functions. The generated atomic sequences require additional support
22661 from the interrupt/exception handling code of the system and are only suitable
22662 for SH3* and SH4* single-core systems. This option is enabled by default when
22663 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
22664 this option also partially utilizes the hardware atomic instructions
22665 @code{movli.l} and @code{movco.l} to create more efficient code, unless
22666 @samp{strict} is specified.
22669 Generate software atomic sequences that use a variable in the thread control
22670 block. This is a variation of the gUSA sequences which can also be used on
22671 SH1* and SH2* targets. The generated atomic sequences require additional
22672 support from the interrupt/exception handling code of the system and are only
22673 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
22674 parameter has to be specified as well.
22677 Generate software atomic sequences that temporarily disable interrupts by
22678 setting @code{SR.IMASK = 1111}. This model works only when the program runs
22679 in privileged mode and is only suitable for single-core systems. Additional
22680 support from the interrupt/exception handling code of the system is not
22681 required. This model is enabled by default when the target is
22682 @code{sh*-*-linux*} and SH1* or SH2*.
22685 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
22686 instructions only. This is only available on SH4A and is suitable for
22687 multi-core systems. Since the hardware instructions support only 32 bit atomic
22688 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
22689 Code compiled with this option is also compatible with other software
22690 atomic model interrupt/exception handling systems if executed on an SH4A
22691 system. Additional support from the interrupt/exception handling code of the
22692 system is not required for this model.
22695 This parameter specifies the offset in bytes of the variable in the thread
22696 control block structure that should be used by the generated atomic sequences
22697 when the @samp{soft-tcb} model has been selected. For other models this
22698 parameter is ignored. The specified value must be an integer multiple of four
22699 and in the range 0-1020.
22702 This parameter prevents mixed usage of multiple atomic models, even if they
22703 are compatible, and makes the compiler generate atomic sequences of the
22704 specified model only.
22710 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
22711 Notice that depending on the particular hardware and software configuration
22712 this can degrade overall performance due to the operand cache line flushes
22713 that are implied by the @code{tas.b} instruction. On multi-core SH4A
22714 processors the @code{tas.b} instruction must be used with caution since it
22715 can result in data corruption for certain cache configurations.
22718 @opindex mprefergot
22719 When generating position-independent code, emit function calls using
22720 the Global Offset Table instead of the Procedure Linkage Table.
22723 @itemx -mno-usermode
22725 @opindex mno-usermode
22726 Don't allow (allow) the compiler generating privileged mode code. Specifying
22727 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
22728 inlined code would not work in user mode. @option{-musermode} is the default
22729 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
22730 @option{-musermode} has no effect, since there is no user mode.
22732 @item -multcost=@var{number}
22733 @opindex multcost=@var{number}
22734 Set the cost to assume for a multiply insn.
22736 @item -mdiv=@var{strategy}
22737 @opindex mdiv=@var{strategy}
22738 Set the division strategy to be used for integer division operations.
22739 @var{strategy} can be one of:
22744 Calls a library function that uses the single-step division instruction
22745 @code{div1} to perform the operation. Division by zero calculates an
22746 unspecified result and does not trap. This is the default except for SH4,
22747 SH2A and SHcompact.
22750 Calls a library function that performs the operation in double precision
22751 floating point. Division by zero causes a floating-point exception. This is
22752 the default for SHcompact with FPU. Specifying this for targets that do not
22753 have a double precision FPU defaults to @code{call-div1}.
22756 Calls a library function that uses a lookup table for small divisors and
22757 the @code{div1} instruction with case distinction for larger divisors. Division
22758 by zero calculates an unspecified result and does not trap. This is the default
22759 for SH4. Specifying this for targets that do not have dynamic shift
22760 instructions defaults to @code{call-div1}.
22764 When a division strategy has not been specified the default strategy is
22765 selected based on the current target. For SH2A the default strategy is to
22766 use the @code{divs} and @code{divu} instructions instead of library function
22769 @item -maccumulate-outgoing-args
22770 @opindex maccumulate-outgoing-args
22771 Reserve space once for outgoing arguments in the function prologue rather
22772 than around each call. Generally beneficial for performance and size. Also
22773 needed for unwinding to avoid changing the stack frame around conditional code.
22775 @item -mdivsi3_libfunc=@var{name}
22776 @opindex mdivsi3_libfunc=@var{name}
22777 Set the name of the library function used for 32-bit signed division to
22779 This only affects the name used in the @samp{call} division strategies, and
22780 the compiler still expects the same sets of input/output/clobbered registers as
22781 if this option were not present.
22783 @item -mfixed-range=@var{register-range}
22784 @opindex mfixed-range
22785 Generate code treating the given register range as fixed registers.
22786 A fixed register is one that the register allocator can not use. This is
22787 useful when compiling kernel code. A register range is specified as
22788 two registers separated by a dash. Multiple register ranges can be
22789 specified separated by a comma.
22791 @item -mbranch-cost=@var{num}
22792 @opindex mbranch-cost=@var{num}
22793 Assume @var{num} to be the cost for a branch instruction. Higher numbers
22794 make the compiler try to generate more branch-free code if possible.
22795 If not specified the value is selected depending on the processor type that
22796 is being compiled for.
22799 @itemx -mno-zdcbranch
22800 @opindex mzdcbranch
22801 @opindex mno-zdcbranch
22802 Assume (do not assume) that zero displacement conditional branch instructions
22803 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
22804 compiler prefers zero displacement branch code sequences. This is
22805 enabled by default when generating code for SH4 and SH4A. It can be explicitly
22806 disabled by specifying @option{-mno-zdcbranch}.
22808 @item -mcbranch-force-delay-slot
22809 @opindex mcbranch-force-delay-slot
22810 Force the usage of delay slots for conditional branches, which stuffs the delay
22811 slot with a @code{nop} if a suitable instruction can't be found. By default
22812 this option is disabled. It can be enabled to work around hardware bugs as
22813 found in the original SH7055.
22816 @itemx -mno-fused-madd
22817 @opindex mfused-madd
22818 @opindex mno-fused-madd
22819 Generate code that uses (does not use) the floating-point multiply and
22820 accumulate instructions. These instructions are generated by default
22821 if hardware floating point is used. The machine-dependent
22822 @option{-mfused-madd} option is now mapped to the machine-independent
22823 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22824 mapped to @option{-ffp-contract=off}.
22830 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
22831 and cosine approximations. The option @option{-mfsca} must be used in
22832 combination with @option{-funsafe-math-optimizations}. It is enabled by default
22833 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
22834 approximations even if @option{-funsafe-math-optimizations} is in effect.
22840 Allow or disallow the compiler to emit the @code{fsrra} instruction for
22841 reciprocal square root approximations. The option @option{-mfsrra} must be used
22842 in combination with @option{-funsafe-math-optimizations} and
22843 @option{-ffinite-math-only}. It is enabled by default when generating code for
22844 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
22845 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
22848 @item -mpretend-cmove
22849 @opindex mpretend-cmove
22850 Prefer zero-displacement conditional branches for conditional move instruction
22851 patterns. This can result in faster code on the SH4 processor.
22855 Generate code using the FDPIC ABI.
22859 @node Solaris 2 Options
22860 @subsection Solaris 2 Options
22861 @cindex Solaris 2 options
22863 These @samp{-m} options are supported on Solaris 2:
22866 @item -mclear-hwcap
22867 @opindex mclear-hwcap
22868 @option{-mclear-hwcap} tells the compiler to remove the hardware
22869 capabilities generated by the Solaris assembler. This is only necessary
22870 when object files use ISA extensions not supported by the current
22871 machine, but check at runtime whether or not to use them.
22873 @item -mimpure-text
22874 @opindex mimpure-text
22875 @option{-mimpure-text}, used in addition to @option{-shared}, tells
22876 the compiler to not pass @option{-z text} to the linker when linking a
22877 shared object. Using this option, you can link position-dependent
22878 code into a shared object.
22880 @option{-mimpure-text} suppresses the ``relocations remain against
22881 allocatable but non-writable sections'' linker error message.
22882 However, the necessary relocations trigger copy-on-write, and the
22883 shared object is not actually shared across processes. Instead of
22884 using @option{-mimpure-text}, you should compile all source code with
22885 @option{-fpic} or @option{-fPIC}.
22889 These switches are supported in addition to the above on Solaris 2:
22894 Add support for multithreading using the POSIX threads library. This
22895 option sets flags for both the preprocessor and linker. This option does
22896 not affect the thread safety of object code produced by the compiler or
22897 that of libraries supplied with it.
22901 This is a synonym for @option{-pthreads}.
22904 @node SPARC Options
22905 @subsection SPARC Options
22906 @cindex SPARC options
22908 These @samp{-m} options are supported on the SPARC:
22911 @item -mno-app-regs
22913 @opindex mno-app-regs
22915 Specify @option{-mapp-regs} to generate output using the global registers
22916 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
22917 global register 1, each global register 2 through 4 is then treated as an
22918 allocable register that is clobbered by function calls. This is the default.
22920 To be fully SVR4 ABI-compliant at the cost of some performance loss,
22921 specify @option{-mno-app-regs}. You should compile libraries and system
22922 software with this option.
22928 With @option{-mflat}, the compiler does not generate save/restore instructions
22929 and uses a ``flat'' or single register window model. This model is compatible
22930 with the regular register window model. The local registers and the input
22931 registers (0--5) are still treated as ``call-saved'' registers and are
22932 saved on the stack as needed.
22934 With @option{-mno-flat} (the default), the compiler generates save/restore
22935 instructions (except for leaf functions). This is the normal operating mode.
22938 @itemx -mhard-float
22940 @opindex mhard-float
22941 Generate output containing floating-point instructions. This is the
22945 @itemx -msoft-float
22947 @opindex msoft-float
22948 Generate output containing library calls for floating point.
22949 @strong{Warning:} the requisite libraries are not available for all SPARC
22950 targets. Normally the facilities of the machine's usual C compiler are
22951 used, but this cannot be done directly in cross-compilation. You must make
22952 your own arrangements to provide suitable library functions for
22953 cross-compilation. The embedded targets @samp{sparc-*-aout} and
22954 @samp{sparclite-*-*} do provide software floating-point support.
22956 @option{-msoft-float} changes the calling convention in the output file;
22957 therefore, it is only useful if you compile @emph{all} of a program with
22958 this option. In particular, you need to compile @file{libgcc.a}, the
22959 library that comes with GCC, with @option{-msoft-float} in order for
22962 @item -mhard-quad-float
22963 @opindex mhard-quad-float
22964 Generate output containing quad-word (long double) floating-point
22967 @item -msoft-quad-float
22968 @opindex msoft-quad-float
22969 Generate output containing library calls for quad-word (long double)
22970 floating-point instructions. The functions called are those specified
22971 in the SPARC ABI@. This is the default.
22973 As of this writing, there are no SPARC implementations that have hardware
22974 support for the quad-word floating-point instructions. They all invoke
22975 a trap handler for one of these instructions, and then the trap handler
22976 emulates the effect of the instruction. Because of the trap handler overhead,
22977 this is much slower than calling the ABI library routines. Thus the
22978 @option{-msoft-quad-float} option is the default.
22980 @item -mno-unaligned-doubles
22981 @itemx -munaligned-doubles
22982 @opindex mno-unaligned-doubles
22983 @opindex munaligned-doubles
22984 Assume that doubles have 8-byte alignment. This is the default.
22986 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
22987 alignment only if they are contained in another type, or if they have an
22988 absolute address. Otherwise, it assumes they have 4-byte alignment.
22989 Specifying this option avoids some rare compatibility problems with code
22990 generated by other compilers. It is not the default because it results
22991 in a performance loss, especially for floating-point code.
22994 @itemx -mno-user-mode
22995 @opindex muser-mode
22996 @opindex mno-user-mode
22997 Do not generate code that can only run in supervisor mode. This is relevant
22998 only for the @code{casa} instruction emitted for the LEON3 processor. This
23001 @item -mfaster-structs
23002 @itemx -mno-faster-structs
23003 @opindex mfaster-structs
23004 @opindex mno-faster-structs
23005 With @option{-mfaster-structs}, the compiler assumes that structures
23006 should have 8-byte alignment. This enables the use of pairs of
23007 @code{ldd} and @code{std} instructions for copies in structure
23008 assignment, in place of twice as many @code{ld} and @code{st} pairs.
23009 However, the use of this changed alignment directly violates the SPARC
23010 ABI@. Thus, it's intended only for use on targets where the developer
23011 acknowledges that their resulting code is not directly in line with
23012 the rules of the ABI@.
23014 @item -mstd-struct-return
23015 @itemx -mno-std-struct-return
23016 @opindex mstd-struct-return
23017 @opindex mno-std-struct-return
23018 With @option{-mstd-struct-return}, the compiler generates checking code
23019 in functions returning structures or unions to detect size mismatches
23020 between the two sides of function calls, as per the 32-bit ABI@.
23022 The default is @option{-mno-std-struct-return}. This option has no effect
23025 @item -mcpu=@var{cpu_type}
23027 Set the instruction set, register set, and instruction scheduling parameters
23028 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23029 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
23030 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
23031 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
23032 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
23033 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
23035 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
23036 which selects the best architecture option for the host processor.
23037 @option{-mcpu=native} has no effect if GCC does not recognize
23040 Default instruction scheduling parameters are used for values that select
23041 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
23042 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
23044 Here is a list of each supported architecture and their supported
23052 supersparc, hypersparc, leon, leon3
23055 f930, f934, sparclite86x
23061 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
23064 By default (unless configured otherwise), GCC generates code for the V7
23065 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
23066 additionally optimizes it for the Cypress CY7C602 chip, as used in the
23067 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
23068 SPARCStation 1, 2, IPX etc.
23070 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
23071 architecture. The only difference from V7 code is that the compiler emits
23072 the integer multiply and integer divide instructions which exist in SPARC-V8
23073 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
23074 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
23077 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
23078 the SPARC architecture. This adds the integer multiply, integer divide step
23079 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
23080 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
23081 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
23082 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
23083 MB86934 chip, which is the more recent SPARClite with FPU@.
23085 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
23086 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
23087 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
23088 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
23089 optimizes it for the TEMIC SPARClet chip.
23091 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
23092 architecture. This adds 64-bit integer and floating-point move instructions,
23093 3 additional floating-point condition code registers and conditional move
23094 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
23095 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
23096 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
23097 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
23098 @option{-mcpu=niagara}, the compiler additionally optimizes it for
23099 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
23100 additionally optimizes it for Sun UltraSPARC T2 chips. With
23101 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
23102 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
23103 additionally optimizes it for Sun UltraSPARC T4 chips. With
23104 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
23105 Oracle SPARC M7 chips.
23107 @item -mtune=@var{cpu_type}
23109 Set the instruction scheduling parameters for machine type
23110 @var{cpu_type}, but do not set the instruction set or register set that the
23111 option @option{-mcpu=@var{cpu_type}} does.
23113 The same values for @option{-mcpu=@var{cpu_type}} can be used for
23114 @option{-mtune=@var{cpu_type}}, but the only useful values are those
23115 that select a particular CPU implementation. Those are
23116 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
23117 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
23118 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
23119 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
23120 @samp{niagara4} and @samp{niagara7}. With native Solaris and
23121 GNU/Linux toolchains, @samp{native} can also be used.
23126 @opindex mno-v8plus
23127 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
23128 difference from the V8 ABI is that the global and out registers are
23129 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
23130 mode for all SPARC-V9 processors.
23136 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
23137 Visual Instruction Set extensions. The default is @option{-mno-vis}.
23143 With @option{-mvis2}, GCC generates code that takes advantage of
23144 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
23145 default is @option{-mvis2} when targeting a cpu that supports such
23146 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
23147 also sets @option{-mvis}.
23153 With @option{-mvis3}, GCC generates code that takes advantage of
23154 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
23155 default is @option{-mvis3} when targeting a cpu that supports such
23156 instructions, such as niagara-3 and later. Setting @option{-mvis3}
23157 also sets @option{-mvis2} and @option{-mvis}.
23163 With @option{-mvis4}, GCC generates code that takes advantage of
23164 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
23165 default is @option{-mvis4} when targeting a cpu that supports such
23166 instructions, such as niagara-7 and later. Setting @option{-mvis4}
23167 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
23172 @opindex mno-cbcond
23173 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
23174 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
23175 when targeting a CPU that supports such instructions, such as Niagara-4 and
23182 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
23183 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
23184 when targeting a CPU that supports such instructions, such as Niagara-3 and
23191 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
23192 Population Count instruction. The default is @option{-mpopc}
23193 when targeting a CPU that supports such an instruction, such as Niagara-2 and
23200 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
23201 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
23202 when targeting a CPU that supports such an instruction, such as Niagara-7 and
23206 @opindex mfix-at697f
23207 Enable the documented workaround for the single erratum of the Atmel AT697F
23208 processor (which corresponds to erratum #13 of the AT697E processor).
23211 @opindex mfix-ut699
23212 Enable the documented workarounds for the floating-point errata and the data
23213 cache nullify errata of the UT699 processor.
23216 These @samp{-m} options are supported in addition to the above
23217 on SPARC-V9 processors in 64-bit environments:
23224 Generate code for a 32-bit or 64-bit environment.
23225 The 32-bit environment sets int, long and pointer to 32 bits.
23226 The 64-bit environment sets int to 32 bits and long and pointer
23229 @item -mcmodel=@var{which}
23231 Set the code model to one of
23235 The Medium/Low code model: 64-bit addresses, programs
23236 must be linked in the low 32 bits of memory. Programs can be statically
23237 or dynamically linked.
23240 The Medium/Middle code model: 64-bit addresses, programs
23241 must be linked in the low 44 bits of memory, the text and data segments must
23242 be less than 2GB in size and the data segment must be located within 2GB of
23246 The Medium/Anywhere code model: 64-bit addresses, programs
23247 may be linked anywhere in memory, the text and data segments must be less
23248 than 2GB in size and the data segment must be located within 2GB of the
23252 The Medium/Anywhere code model for embedded systems:
23253 64-bit addresses, the text and data segments must be less than 2GB in
23254 size, both starting anywhere in memory (determined at link time). The
23255 global register %g4 points to the base of the data segment. Programs
23256 are statically linked and PIC is not supported.
23259 @item -mmemory-model=@var{mem-model}
23260 @opindex mmemory-model
23261 Set the memory model in force on the processor to one of
23265 The default memory model for the processor and operating system.
23268 Relaxed Memory Order
23271 Partial Store Order
23277 Sequential Consistency
23280 These memory models are formally defined in Appendix D of the Sparc V9
23281 architecture manual, as set in the processor's @code{PSTATE.MM} field.
23284 @itemx -mno-stack-bias
23285 @opindex mstack-bias
23286 @opindex mno-stack-bias
23287 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
23288 frame pointer if present, are offset by @minus{}2047 which must be added back
23289 when making stack frame references. This is the default in 64-bit mode.
23290 Otherwise, assume no such offset is present.
23294 @subsection SPU Options
23295 @cindex SPU options
23297 These @samp{-m} options are supported on the SPU:
23301 @itemx -merror-reloc
23302 @opindex mwarn-reloc
23303 @opindex merror-reloc
23305 The loader for SPU does not handle dynamic relocations. By default, GCC
23306 gives an error when it generates code that requires a dynamic
23307 relocation. @option{-mno-error-reloc} disables the error,
23308 @option{-mwarn-reloc} generates a warning instead.
23311 @itemx -munsafe-dma
23313 @opindex munsafe-dma
23315 Instructions that initiate or test completion of DMA must not be
23316 reordered with respect to loads and stores of the memory that is being
23318 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
23319 memory accesses, but that can lead to inefficient code in places where the
23320 memory is known to not change. Rather than mark the memory as volatile,
23321 you can use @option{-msafe-dma} to tell the compiler to treat
23322 the DMA instructions as potentially affecting all memory.
23324 @item -mbranch-hints
23325 @opindex mbranch-hints
23327 By default, GCC generates a branch hint instruction to avoid
23328 pipeline stalls for always-taken or probably-taken branches. A hint
23329 is not generated closer than 8 instructions away from its branch.
23330 There is little reason to disable them, except for debugging purposes,
23331 or to make an object a little bit smaller.
23335 @opindex msmall-mem
23336 @opindex mlarge-mem
23338 By default, GCC generates code assuming that addresses are never larger
23339 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
23340 a full 32-bit address.
23345 By default, GCC links against startup code that assumes the SPU-style
23346 main function interface (which has an unconventional parameter list).
23347 With @option{-mstdmain}, GCC links your program against startup
23348 code that assumes a C99-style interface to @code{main}, including a
23349 local copy of @code{argv} strings.
23351 @item -mfixed-range=@var{register-range}
23352 @opindex mfixed-range
23353 Generate code treating the given register range as fixed registers.
23354 A fixed register is one that the register allocator cannot use. This is
23355 useful when compiling kernel code. A register range is specified as
23356 two registers separated by a dash. Multiple register ranges can be
23357 specified separated by a comma.
23363 Compile code assuming that pointers to the PPU address space accessed
23364 via the @code{__ea} named address space qualifier are either 32 or 64
23365 bits wide. The default is 32 bits. As this is an ABI-changing option,
23366 all object code in an executable must be compiled with the same setting.
23368 @item -maddress-space-conversion
23369 @itemx -mno-address-space-conversion
23370 @opindex maddress-space-conversion
23371 @opindex mno-address-space-conversion
23372 Allow/disallow treating the @code{__ea} address space as superset
23373 of the generic address space. This enables explicit type casts
23374 between @code{__ea} and generic pointer as well as implicit
23375 conversions of generic pointers to @code{__ea} pointers. The
23376 default is to allow address space pointer conversions.
23378 @item -mcache-size=@var{cache-size}
23379 @opindex mcache-size
23380 This option controls the version of libgcc that the compiler links to an
23381 executable and selects a software-managed cache for accessing variables
23382 in the @code{__ea} address space with a particular cache size. Possible
23383 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
23384 and @samp{128}. The default cache size is 64KB.
23386 @item -matomic-updates
23387 @itemx -mno-atomic-updates
23388 @opindex matomic-updates
23389 @opindex mno-atomic-updates
23390 This option controls the version of libgcc that the compiler links to an
23391 executable and selects whether atomic updates to the software-managed
23392 cache of PPU-side variables are used. If you use atomic updates, changes
23393 to a PPU variable from SPU code using the @code{__ea} named address space
23394 qualifier do not interfere with changes to other PPU variables residing
23395 in the same cache line from PPU code. If you do not use atomic updates,
23396 such interference may occur; however, writing back cache lines is
23397 more efficient. The default behavior is to use atomic updates.
23400 @itemx -mdual-nops=@var{n}
23401 @opindex mdual-nops
23402 By default, GCC inserts NOPs to increase dual issue when it expects
23403 it to increase performance. @var{n} can be a value from 0 to 10. A
23404 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
23405 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
23407 @item -mhint-max-nops=@var{n}
23408 @opindex mhint-max-nops
23409 Maximum number of NOPs to insert for a branch hint. A branch hint must
23410 be at least 8 instructions away from the branch it is affecting. GCC
23411 inserts up to @var{n} NOPs to enforce this, otherwise it does not
23412 generate the branch hint.
23414 @item -mhint-max-distance=@var{n}
23415 @opindex mhint-max-distance
23416 The encoding of the branch hint instruction limits the hint to be within
23417 256 instructions of the branch it is affecting. By default, GCC makes
23418 sure it is within 125.
23421 @opindex msafe-hints
23422 Work around a hardware bug that causes the SPU to stall indefinitely.
23423 By default, GCC inserts the @code{hbrp} instruction to make sure
23424 this stall won't happen.
23428 @node System V Options
23429 @subsection Options for System V
23431 These additional options are available on System V Release 4 for
23432 compatibility with other compilers on those systems:
23437 Create a shared object.
23438 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
23442 Identify the versions of each tool used by the compiler, in a
23443 @code{.ident} assembler directive in the output.
23447 Refrain from adding @code{.ident} directives to the output file (this is
23450 @item -YP,@var{dirs}
23452 Search the directories @var{dirs}, and no others, for libraries
23453 specified with @option{-l}.
23455 @item -Ym,@var{dir}
23457 Look in the directory @var{dir} to find the M4 preprocessor.
23458 The assembler uses this option.
23459 @c This is supposed to go with a -Yd for predefined M4 macro files, but
23460 @c the generic assembler that comes with Solaris takes just -Ym.
23463 @node TILE-Gx Options
23464 @subsection TILE-Gx Options
23465 @cindex TILE-Gx options
23467 These @samp{-m} options are supported on the TILE-Gx:
23470 @item -mcmodel=small
23471 @opindex mcmodel=small
23472 Generate code for the small model. The distance for direct calls is
23473 limited to 500M in either direction. PC-relative addresses are 32
23474 bits. Absolute addresses support the full address range.
23476 @item -mcmodel=large
23477 @opindex mcmodel=large
23478 Generate code for the large model. There is no limitation on call
23479 distance, pc-relative addresses, or absolute addresses.
23481 @item -mcpu=@var{name}
23483 Selects the type of CPU to be targeted. Currently the only supported
23484 type is @samp{tilegx}.
23490 Generate code for a 32-bit or 64-bit environment. The 32-bit
23491 environment sets int, long, and pointer to 32 bits. The 64-bit
23492 environment sets int to 32 bits and long and pointer to 64 bits.
23495 @itemx -mlittle-endian
23496 @opindex mbig-endian
23497 @opindex mlittle-endian
23498 Generate code in big/little endian mode, respectively.
23501 @node TILEPro Options
23502 @subsection TILEPro Options
23503 @cindex TILEPro options
23505 These @samp{-m} options are supported on the TILEPro:
23508 @item -mcpu=@var{name}
23510 Selects the type of CPU to be targeted. Currently the only supported
23511 type is @samp{tilepro}.
23515 Generate code for a 32-bit environment, which sets int, long, and
23516 pointer to 32 bits. This is the only supported behavior so the flag
23517 is essentially ignored.
23521 @subsection V850 Options
23522 @cindex V850 Options
23524 These @samp{-m} options are defined for V850 implementations:
23528 @itemx -mno-long-calls
23529 @opindex mlong-calls
23530 @opindex mno-long-calls
23531 Treat all calls as being far away (near). If calls are assumed to be
23532 far away, the compiler always loads the function's address into a
23533 register, and calls indirect through the pointer.
23539 Do not optimize (do optimize) basic blocks that use the same index
23540 pointer 4 or more times to copy pointer into the @code{ep} register, and
23541 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
23542 option is on by default if you optimize.
23544 @item -mno-prolog-function
23545 @itemx -mprolog-function
23546 @opindex mno-prolog-function
23547 @opindex mprolog-function
23548 Do not use (do use) external functions to save and restore registers
23549 at the prologue and epilogue of a function. The external functions
23550 are slower, but use less code space if more than one function saves
23551 the same number of registers. The @option{-mprolog-function} option
23552 is on by default if you optimize.
23556 Try to make the code as small as possible. At present, this just turns
23557 on the @option{-mep} and @option{-mprolog-function} options.
23559 @item -mtda=@var{n}
23561 Put static or global variables whose size is @var{n} bytes or less into
23562 the tiny data area that register @code{ep} points to. The tiny data
23563 area can hold up to 256 bytes in total (128 bytes for byte references).
23565 @item -msda=@var{n}
23567 Put static or global variables whose size is @var{n} bytes or less into
23568 the small data area that register @code{gp} points to. The small data
23569 area can hold up to 64 kilobytes.
23571 @item -mzda=@var{n}
23573 Put static or global variables whose size is @var{n} bytes or less into
23574 the first 32 kilobytes of memory.
23578 Specify that the target processor is the V850.
23582 Specify that the target processor is the V850E3V5. The preprocessor
23583 constant @code{__v850e3v5__} is defined if this option is used.
23587 Specify that the target processor is the V850E3V5. This is an alias for
23588 the @option{-mv850e3v5} option.
23592 Specify that the target processor is the V850E2V3. The preprocessor
23593 constant @code{__v850e2v3__} is defined if this option is used.
23597 Specify that the target processor is the V850E2. The preprocessor
23598 constant @code{__v850e2__} is defined if this option is used.
23602 Specify that the target processor is the V850E1. The preprocessor
23603 constants @code{__v850e1__} and @code{__v850e__} are defined if
23604 this option is used.
23608 Specify that the target processor is the V850ES. This is an alias for
23609 the @option{-mv850e1} option.
23613 Specify that the target processor is the V850E@. The preprocessor
23614 constant @code{__v850e__} is defined if this option is used.
23616 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
23617 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
23618 are defined then a default target processor is chosen and the
23619 relevant @samp{__v850*__} preprocessor constant is defined.
23621 The preprocessor constants @code{__v850} and @code{__v851__} are always
23622 defined, regardless of which processor variant is the target.
23624 @item -mdisable-callt
23625 @itemx -mno-disable-callt
23626 @opindex mdisable-callt
23627 @opindex mno-disable-callt
23628 This option suppresses generation of the @code{CALLT} instruction for the
23629 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
23632 This option is enabled by default when the RH850 ABI is
23633 in use (see @option{-mrh850-abi}), and disabled by default when the
23634 GCC ABI is in use. If @code{CALLT} instructions are being generated
23635 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
23641 Pass on (or do not pass on) the @option{-mrelax} command-line option
23645 @itemx -mno-long-jumps
23646 @opindex mlong-jumps
23647 @opindex mno-long-jumps
23648 Disable (or re-enable) the generation of PC-relative jump instructions.
23651 @itemx -mhard-float
23652 @opindex msoft-float
23653 @opindex mhard-float
23654 Disable (or re-enable) the generation of hardware floating point
23655 instructions. This option is only significant when the target
23656 architecture is @samp{V850E2V3} or higher. If hardware floating point
23657 instructions are being generated then the C preprocessor symbol
23658 @code{__FPU_OK__} is defined, otherwise the symbol
23659 @code{__NO_FPU__} is defined.
23663 Enables the use of the e3v5 LOOP instruction. The use of this
23664 instruction is not enabled by default when the e3v5 architecture is
23665 selected because its use is still experimental.
23669 @opindex mrh850-abi
23671 Enables support for the RH850 version of the V850 ABI. This is the
23672 default. With this version of the ABI the following rules apply:
23676 Integer sized structures and unions are returned via a memory pointer
23677 rather than a register.
23680 Large structures and unions (more than 8 bytes in size) are passed by
23684 Functions are aligned to 16-bit boundaries.
23687 The @option{-m8byte-align} command-line option is supported.
23690 The @option{-mdisable-callt} command-line option is enabled by
23691 default. The @option{-mno-disable-callt} command-line option is not
23695 When this version of the ABI is enabled the C preprocessor symbol
23696 @code{__V850_RH850_ABI__} is defined.
23700 Enables support for the old GCC version of the V850 ABI. With this
23701 version of the ABI the following rules apply:
23705 Integer sized structures and unions are returned in register @code{r10}.
23708 Large structures and unions (more than 8 bytes in size) are passed by
23712 Functions are aligned to 32-bit boundaries, unless optimizing for
23716 The @option{-m8byte-align} command-line option is not supported.
23719 The @option{-mdisable-callt} command-line option is supported but not
23720 enabled by default.
23723 When this version of the ABI is enabled the C preprocessor symbol
23724 @code{__V850_GCC_ABI__} is defined.
23726 @item -m8byte-align
23727 @itemx -mno-8byte-align
23728 @opindex m8byte-align
23729 @opindex mno-8byte-align
23730 Enables support for @code{double} and @code{long long} types to be
23731 aligned on 8-byte boundaries. The default is to restrict the
23732 alignment of all objects to at most 4-bytes. When
23733 @option{-m8byte-align} is in effect the C preprocessor symbol
23734 @code{__V850_8BYTE_ALIGN__} is defined.
23737 @opindex mbig-switch
23738 Generate code suitable for big switch tables. Use this option only if
23739 the assembler/linker complain about out of range branches within a switch
23744 This option causes r2 and r5 to be used in the code generated by
23745 the compiler. This setting is the default.
23747 @item -mno-app-regs
23748 @opindex mno-app-regs
23749 This option causes r2 and r5 to be treated as fixed registers.
23754 @subsection VAX Options
23755 @cindex VAX options
23757 These @samp{-m} options are defined for the VAX:
23762 Do not output certain jump instructions (@code{aobleq} and so on)
23763 that the Unix assembler for the VAX cannot handle across long
23768 Do output those jump instructions, on the assumption that the
23769 GNU assembler is being used.
23773 Output code for G-format floating-point numbers instead of D-format.
23776 @node Visium Options
23777 @subsection Visium Options
23778 @cindex Visium options
23784 A program which performs file I/O and is destined to run on an MCM target
23785 should be linked with this option. It causes the libraries libc.a and
23786 libdebug.a to be linked. The program should be run on the target under
23787 the control of the GDB remote debugging stub.
23791 A program which performs file I/O and is destined to run on the simulator
23792 should be linked with option. This causes libraries libc.a and libsim.a to
23796 @itemx -mhard-float
23798 @opindex mhard-float
23799 Generate code containing floating-point instructions. This is the
23803 @itemx -msoft-float
23805 @opindex msoft-float
23806 Generate code containing library calls for floating-point.
23808 @option{-msoft-float} changes the calling convention in the output file;
23809 therefore, it is only useful if you compile @emph{all} of a program with
23810 this option. In particular, you need to compile @file{libgcc.a}, the
23811 library that comes with GCC, with @option{-msoft-float} in order for
23814 @item -mcpu=@var{cpu_type}
23816 Set the instruction set, register set, and instruction scheduling parameters
23817 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23818 @samp{mcm}, @samp{gr5} and @samp{gr6}.
23820 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
23822 By default (unless configured otherwise), GCC generates code for the GR5
23823 variant of the Visium architecture.
23825 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
23826 architecture. The only difference from GR5 code is that the compiler will
23827 generate block move instructions.
23829 @item -mtune=@var{cpu_type}
23831 Set the instruction scheduling parameters for machine type @var{cpu_type},
23832 but do not set the instruction set or register set that the option
23833 @option{-mcpu=@var{cpu_type}} would.
23837 Generate code for the supervisor mode, where there are no restrictions on
23838 the access to general registers. This is the default.
23841 @opindex muser-mode
23842 Generate code for the user mode, where the access to some general registers
23843 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
23844 mode; on the GR6, only registers r29 to r31 are affected.
23848 @subsection VMS Options
23850 These @samp{-m} options are defined for the VMS implementations:
23853 @item -mvms-return-codes
23854 @opindex mvms-return-codes
23855 Return VMS condition codes from @code{main}. The default is to return POSIX-style
23856 condition (e.g.@ error) codes.
23858 @item -mdebug-main=@var{prefix}
23859 @opindex mdebug-main=@var{prefix}
23860 Flag the first routine whose name starts with @var{prefix} as the main
23861 routine for the debugger.
23865 Default to 64-bit memory allocation routines.
23867 @item -mpointer-size=@var{size}
23868 @opindex mpointer-size=@var{size}
23869 Set the default size of pointers. Possible options for @var{size} are
23870 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
23871 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
23872 The later option disables @code{pragma pointer_size}.
23875 @node VxWorks Options
23876 @subsection VxWorks Options
23877 @cindex VxWorks Options
23879 The options in this section are defined for all VxWorks targets.
23880 Options specific to the target hardware are listed with the other
23881 options for that target.
23886 GCC can generate code for both VxWorks kernels and real time processes
23887 (RTPs). This option switches from the former to the latter. It also
23888 defines the preprocessor macro @code{__RTP__}.
23891 @opindex non-static
23892 Link an RTP executable against shared libraries rather than static
23893 libraries. The options @option{-static} and @option{-shared} can
23894 also be used for RTPs (@pxref{Link Options}); @option{-static}
23901 These options are passed down to the linker. They are defined for
23902 compatibility with Diab.
23905 @opindex Xbind-lazy
23906 Enable lazy binding of function calls. This option is equivalent to
23907 @option{-Wl,-z,now} and is defined for compatibility with Diab.
23911 Disable lazy binding of function calls. This option is the default and
23912 is defined for compatibility with Diab.
23916 @subsection x86 Options
23917 @cindex x86 Options
23919 These @samp{-m} options are defined for the x86 family of computers.
23923 @item -march=@var{cpu-type}
23925 Generate instructions for the machine type @var{cpu-type}. In contrast to
23926 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
23927 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
23928 to generate code that may not run at all on processors other than the one
23929 indicated. Specifying @option{-march=@var{cpu-type}} implies
23930 @option{-mtune=@var{cpu-type}}.
23932 The choices for @var{cpu-type} are:
23936 This selects the CPU to generate code for at compilation time by determining
23937 the processor type of the compiling machine. Using @option{-march=native}
23938 enables all instruction subsets supported by the local machine (hence
23939 the result might not run on different machines). Using @option{-mtune=native}
23940 produces code optimized for the local machine under the constraints
23941 of the selected instruction set.
23944 Original Intel i386 CPU@.
23947 Intel i486 CPU@. (No scheduling is implemented for this chip.)
23951 Intel Pentium CPU with no MMX support.
23954 Intel Lakemont MCU, based on Intel Pentium CPU.
23957 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
23960 Intel Pentium Pro CPU@.
23963 When used with @option{-march}, the Pentium Pro
23964 instruction set is used, so the code runs on all i686 family chips.
23965 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
23968 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
23973 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
23977 Intel Pentium M; low-power version of Intel Pentium III CPU
23978 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
23982 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
23985 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
23989 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
23990 SSE2 and SSE3 instruction set support.
23993 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
23994 instruction set support.
23997 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23998 SSE4.1, SSE4.2 and POPCNT instruction set support.
24001 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24002 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
24005 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24006 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
24009 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24010 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
24011 instruction set support.
24014 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24015 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24016 BMI, BMI2 and F16C instruction set support.
24019 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24020 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24021 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
24024 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24025 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24026 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
24027 XSAVES instruction set support.
24030 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
24031 instruction set support.
24034 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24035 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
24038 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24039 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24040 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
24041 AVX512CD instruction set support.
24043 @item skylake-avx512
24044 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24045 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24046 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
24047 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
24050 AMD K6 CPU with MMX instruction set support.
24054 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
24057 @itemx athlon-tbird
24058 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
24064 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
24065 instruction set support.
24071 Processors based on the AMD K8 core with x86-64 instruction set support,
24072 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
24073 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
24074 instruction set extensions.)
24077 @itemx opteron-sse3
24078 @itemx athlon64-sse3
24079 Improved versions of AMD K8 cores with SSE3 instruction set support.
24083 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
24084 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
24085 instruction set extensions.)
24088 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
24089 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
24090 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
24092 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24093 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
24094 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
24097 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24098 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
24099 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
24100 64-bit instruction set extensions.
24102 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24103 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
24104 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
24105 SSE4.2, ABM and 64-bit instruction set extensions.
24108 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
24109 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
24110 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
24111 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
24112 instruction set extensions.
24115 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
24116 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
24117 instruction set extensions.)
24120 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
24121 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
24122 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
24125 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
24129 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
24130 instruction set support.
24133 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
24134 (No scheduling is implemented for this chip.)
24137 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
24138 (No scheduling is implemented for this chip.)
24141 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24142 (No scheduling is implemented for this chip.)
24145 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
24146 (No scheduling is implemented for this chip.)
24149 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
24150 (No scheduling is implemented for this chip.)
24153 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24154 (No scheduling is implemented for this chip.)
24157 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
24158 (No scheduling is implemented for this chip.)
24161 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
24162 AVX and AVX2 instruction set support.
24163 (No scheduling is implemented for this chip.)
24166 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24167 instruction set support.
24168 (No scheduling is implemented for this chip.)
24171 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24172 instruction set support.
24173 (No scheduling is implemented for this chip.)
24176 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24177 instruction set support.
24178 (No scheduling is implemented for this chip.)
24181 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24182 instruction set support.
24183 (No scheduling is implemented for this chip.)
24186 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24187 instruction set support.
24188 (No scheduling is implemented for this chip.)
24191 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24192 instruction set support.
24193 (No scheduling is implemented for this chip.)
24196 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
24199 @item -mtune=@var{cpu-type}
24201 Tune to @var{cpu-type} everything applicable about the generated code, except
24202 for the ABI and the set of available instructions.
24203 While picking a specific @var{cpu-type} schedules things appropriately
24204 for that particular chip, the compiler does not generate any code that
24205 cannot run on the default machine type unless you use a
24206 @option{-march=@var{cpu-type}} option.
24207 For example, if GCC is configured for i686-pc-linux-gnu
24208 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
24209 but still runs on i686 machines.
24211 The choices for @var{cpu-type} are the same as for @option{-march}.
24212 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
24216 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
24217 If you know the CPU on which your code will run, then you should use
24218 the corresponding @option{-mtune} or @option{-march} option instead of
24219 @option{-mtune=generic}. But, if you do not know exactly what CPU users
24220 of your application will have, then you should use this option.
24222 As new processors are deployed in the marketplace, the behavior of this
24223 option will change. Therefore, if you upgrade to a newer version of
24224 GCC, code generation controlled by this option will change to reflect
24226 that are most common at the time that version of GCC is released.
24228 There is no @option{-march=generic} option because @option{-march}
24229 indicates the instruction set the compiler can use, and there is no
24230 generic instruction set applicable to all processors. In contrast,
24231 @option{-mtune} indicates the processor (or, in this case, collection of
24232 processors) for which the code is optimized.
24235 Produce code optimized for the most current Intel processors, which are
24236 Haswell and Silvermont for this version of GCC. If you know the CPU
24237 on which your code will run, then you should use the corresponding
24238 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
24239 But, if you want your application performs better on both Haswell and
24240 Silvermont, then you should use this option.
24242 As new Intel processors are deployed in the marketplace, the behavior of
24243 this option will change. Therefore, if you upgrade to a newer version of
24244 GCC, code generation controlled by this option will change to reflect
24245 the most current Intel processors at the time that version of GCC is
24248 There is no @option{-march=intel} option because @option{-march} indicates
24249 the instruction set the compiler can use, and there is no common
24250 instruction set applicable to all processors. In contrast,
24251 @option{-mtune} indicates the processor (or, in this case, collection of
24252 processors) for which the code is optimized.
24255 @item -mcpu=@var{cpu-type}
24257 A deprecated synonym for @option{-mtune}.
24259 @item -mfpmath=@var{unit}
24261 Generate floating-point arithmetic for selected unit @var{unit}. The choices
24262 for @var{unit} are:
24266 Use the standard 387 floating-point coprocessor present on the majority of chips and
24267 emulated otherwise. Code compiled with this option runs almost everywhere.
24268 The temporary results are computed in 80-bit precision instead of the precision
24269 specified by the type, resulting in slightly different results compared to most
24270 of other chips. See @option{-ffloat-store} for more detailed description.
24272 This is the default choice for x86-32 targets.
24275 Use scalar floating-point instructions present in the SSE instruction set.
24276 This instruction set is supported by Pentium III and newer chips,
24277 and in the AMD line
24278 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
24279 instruction set supports only single-precision arithmetic, thus the double and
24280 extended-precision arithmetic are still done using 387. A later version, present
24281 only in Pentium 4 and AMD x86-64 chips, supports double-precision
24284 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
24285 or @option{-msse2} switches to enable SSE extensions and make this option
24286 effective. For the x86-64 compiler, these extensions are enabled by default.
24288 The resulting code should be considerably faster in the majority of cases and avoid
24289 the numerical instability problems of 387 code, but may break some existing
24290 code that expects temporaries to be 80 bits.
24292 This is the default choice for the x86-64 compiler.
24297 Attempt to utilize both instruction sets at once. This effectively doubles the
24298 amount of available registers, and on chips with separate execution units for
24299 387 and SSE the execution resources too. Use this option with care, as it is
24300 still experimental, because the GCC register allocator does not model separate
24301 functional units well, resulting in unstable performance.
24304 @item -masm=@var{dialect}
24305 @opindex masm=@var{dialect}
24306 Output assembly instructions using selected @var{dialect}. Also affects
24307 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
24308 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
24309 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
24310 not support @samp{intel}.
24313 @itemx -mno-ieee-fp
24315 @opindex mno-ieee-fp
24316 Control whether or not the compiler uses IEEE floating-point
24317 comparisons. These correctly handle the case where the result of a
24318 comparison is unordered.
24323 @opindex mhard-float
24324 Generate output containing 80387 instructions for floating point.
24329 @opindex msoft-float
24330 Generate output containing library calls for floating point.
24332 @strong{Warning:} the requisite libraries are not part of GCC@.
24333 Normally the facilities of the machine's usual C compiler are used, but
24334 this can't be done directly in cross-compilation. You must make your
24335 own arrangements to provide suitable library functions for
24338 On machines where a function returns floating-point results in the 80387
24339 register stack, some floating-point opcodes may be emitted even if
24340 @option{-msoft-float} is used.
24342 @item -mno-fp-ret-in-387
24343 @opindex mno-fp-ret-in-387
24344 Do not use the FPU registers for return values of functions.
24346 The usual calling convention has functions return values of types
24347 @code{float} and @code{double} in an FPU register, even if there
24348 is no FPU@. The idea is that the operating system should emulate
24351 The option @option{-mno-fp-ret-in-387} causes such values to be returned
24352 in ordinary CPU registers instead.
24354 @item -mno-fancy-math-387
24355 @opindex mno-fancy-math-387
24356 Some 387 emulators do not support the @code{sin}, @code{cos} and
24357 @code{sqrt} instructions for the 387. Specify this option to avoid
24358 generating those instructions. This option is the default on
24359 OpenBSD and NetBSD@. This option is overridden when @option{-march}
24360 indicates that the target CPU always has an FPU and so the
24361 instruction does not need emulation. These
24362 instructions are not generated unless you also use the
24363 @option{-funsafe-math-optimizations} switch.
24365 @item -malign-double
24366 @itemx -mno-align-double
24367 @opindex malign-double
24368 @opindex mno-align-double
24369 Control whether GCC aligns @code{double}, @code{long double}, and
24370 @code{long long} variables on a two-word boundary or a one-word
24371 boundary. Aligning @code{double} variables on a two-word boundary
24372 produces code that runs somewhat faster on a Pentium at the
24373 expense of more memory.
24375 On x86-64, @option{-malign-double} is enabled by default.
24377 @strong{Warning:} if you use the @option{-malign-double} switch,
24378 structures containing the above types are aligned differently than
24379 the published application binary interface specifications for the x86-32
24380 and are not binary compatible with structures in code compiled
24381 without that switch.
24383 @item -m96bit-long-double
24384 @itemx -m128bit-long-double
24385 @opindex m96bit-long-double
24386 @opindex m128bit-long-double
24387 These switches control the size of @code{long double} type. The x86-32
24388 application binary interface specifies the size to be 96 bits,
24389 so @option{-m96bit-long-double} is the default in 32-bit mode.
24391 Modern architectures (Pentium and newer) prefer @code{long double}
24392 to be aligned to an 8- or 16-byte boundary. In arrays or structures
24393 conforming to the ABI, this is not possible. So specifying
24394 @option{-m128bit-long-double} aligns @code{long double}
24395 to a 16-byte boundary by padding the @code{long double} with an additional
24398 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
24399 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
24401 Notice that neither of these options enable any extra precision over the x87
24402 standard of 80 bits for a @code{long double}.
24404 @strong{Warning:} if you override the default value for your target ABI, this
24405 changes the size of
24406 structures and arrays containing @code{long double} variables,
24407 as well as modifying the function calling convention for functions taking
24408 @code{long double}. Hence they are not binary-compatible
24409 with code compiled without that switch.
24411 @item -mlong-double-64
24412 @itemx -mlong-double-80
24413 @itemx -mlong-double-128
24414 @opindex mlong-double-64
24415 @opindex mlong-double-80
24416 @opindex mlong-double-128
24417 These switches control the size of @code{long double} type. A size
24418 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24419 type. This is the default for 32-bit Bionic C library. A size
24420 of 128 bits makes the @code{long double} type equivalent to the
24421 @code{__float128} type. This is the default for 64-bit Bionic C library.
24423 @strong{Warning:} if you override the default value for your target ABI, this
24424 changes the size of
24425 structures and arrays containing @code{long double} variables,
24426 as well as modifying the function calling convention for functions taking
24427 @code{long double}. Hence they are not binary-compatible
24428 with code compiled without that switch.
24430 @item -malign-data=@var{type}
24431 @opindex malign-data
24432 Control how GCC aligns variables. Supported values for @var{type} are
24433 @samp{compat} uses increased alignment value compatible uses GCC 4.8
24434 and earlier, @samp{abi} uses alignment value as specified by the
24435 psABI, and @samp{cacheline} uses increased alignment value to match
24436 the cache line size. @samp{compat} is the default.
24438 @item -mlarge-data-threshold=@var{threshold}
24439 @opindex mlarge-data-threshold
24440 When @option{-mcmodel=medium} is specified, data objects larger than
24441 @var{threshold} are placed in the large data section. This value must be the
24442 same across all objects linked into the binary, and defaults to 65535.
24446 Use a different function-calling convention, in which functions that
24447 take a fixed number of arguments return with the @code{ret @var{num}}
24448 instruction, which pops their arguments while returning. This saves one
24449 instruction in the caller since there is no need to pop the arguments
24452 You can specify that an individual function is called with this calling
24453 sequence with the function attribute @code{stdcall}. You can also
24454 override the @option{-mrtd} option by using the function attribute
24455 @code{cdecl}. @xref{Function Attributes}.
24457 @strong{Warning:} this calling convention is incompatible with the one
24458 normally used on Unix, so you cannot use it if you need to call
24459 libraries compiled with the Unix compiler.
24461 Also, you must provide function prototypes for all functions that
24462 take variable numbers of arguments (including @code{printf});
24463 otherwise incorrect code is generated for calls to those
24466 In addition, seriously incorrect code results if you call a
24467 function with too many arguments. (Normally, extra arguments are
24468 harmlessly ignored.)
24470 @item -mregparm=@var{num}
24472 Control how many registers are used to pass integer arguments. By
24473 default, no registers are used to pass arguments, and at most 3
24474 registers can be used. You can control this behavior for a specific
24475 function by using the function attribute @code{regparm}.
24476 @xref{Function Attributes}.
24478 @strong{Warning:} if you use this switch, and
24479 @var{num} is nonzero, then you must build all modules with the same
24480 value, including any libraries. This includes the system libraries and
24484 @opindex msseregparm
24485 Use SSE register passing conventions for float and double arguments
24486 and return values. You can control this behavior for a specific
24487 function by using the function attribute @code{sseregparm}.
24488 @xref{Function Attributes}.
24490 @strong{Warning:} if you use this switch then you must build all
24491 modules with the same value, including any libraries. This includes
24492 the system libraries and startup modules.
24494 @item -mvect8-ret-in-mem
24495 @opindex mvect8-ret-in-mem
24496 Return 8-byte vectors in memory instead of MMX registers. This is the
24497 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
24498 Studio compilers until version 12. Later compiler versions (starting
24499 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
24500 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
24501 you need to remain compatible with existing code produced by those
24502 previous compiler versions or older versions of GCC@.
24511 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
24512 is specified, the significands of results of floating-point operations are
24513 rounded to 24 bits (single precision); @option{-mpc64} rounds the
24514 significands of results of floating-point operations to 53 bits (double
24515 precision) and @option{-mpc80} rounds the significands of results of
24516 floating-point operations to 64 bits (extended double precision), which is
24517 the default. When this option is used, floating-point operations in higher
24518 precisions are not available to the programmer without setting the FPU
24519 control word explicitly.
24521 Setting the rounding of floating-point operations to less than the default
24522 80 bits can speed some programs by 2% or more. Note that some mathematical
24523 libraries assume that extended-precision (80-bit) floating-point operations
24524 are enabled by default; routines in such libraries could suffer significant
24525 loss of accuracy, typically through so-called ``catastrophic cancellation'',
24526 when this option is used to set the precision to less than extended precision.
24528 @item -mstackrealign
24529 @opindex mstackrealign
24530 Realign the stack at entry. On the x86, the @option{-mstackrealign}
24531 option generates an alternate prologue and epilogue that realigns the
24532 run-time stack if necessary. This supports mixing legacy codes that keep
24533 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
24534 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
24535 applicable to individual functions.
24537 @item -mpreferred-stack-boundary=@var{num}
24538 @opindex mpreferred-stack-boundary
24539 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24540 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24541 the default is 4 (16 bytes or 128 bits).
24543 @strong{Warning:} When generating code for the x86-64 architecture with
24544 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
24545 used to keep the stack boundary aligned to 8 byte boundary. Since
24546 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
24547 intended to be used in controlled environment where stack space is
24548 important limitation. This option leads to wrong code when functions
24549 compiled with 16 byte stack alignment (such as functions from a standard
24550 library) are called with misaligned stack. In this case, SSE
24551 instructions may lead to misaligned memory access traps. In addition,
24552 variable arguments are handled incorrectly for 16 byte aligned
24553 objects (including x87 long double and __int128), leading to wrong
24554 results. You must build all modules with
24555 @option{-mpreferred-stack-boundary=3}, including any libraries. This
24556 includes the system libraries and startup modules.
24558 @item -mincoming-stack-boundary=@var{num}
24559 @opindex mincoming-stack-boundary
24560 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
24561 boundary. If @option{-mincoming-stack-boundary} is not specified,
24562 the one specified by @option{-mpreferred-stack-boundary} is used.
24564 On Pentium and Pentium Pro, @code{double} and @code{long double} values
24565 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
24566 suffer significant run time performance penalties. On Pentium III, the
24567 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
24568 properly if it is not 16-byte aligned.
24570 To ensure proper alignment of this values on the stack, the stack boundary
24571 must be as aligned as that required by any value stored on the stack.
24572 Further, every function must be generated such that it keeps the stack
24573 aligned. Thus calling a function compiled with a higher preferred
24574 stack boundary from a function compiled with a lower preferred stack
24575 boundary most likely misaligns the stack. It is recommended that
24576 libraries that use callbacks always use the default setting.
24578 This extra alignment does consume extra stack space, and generally
24579 increases code size. Code that is sensitive to stack space usage, such
24580 as embedded systems and operating system kernels, may want to reduce the
24581 preferred alignment to @option{-mpreferred-stack-boundary=2}.
24638 @itemx -mavx512ifma
24639 @opindex mavx512ifma
24641 @itemx -mavx512vbmi
24642 @opindex mavx512vbmi
24654 @opindex mclfushopt
24671 @itemx -mprefetchwt1
24672 @opindex mprefetchwt1
24728 These switches enable the use of instructions in the MMX, SSE,
24729 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
24730 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
24731 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
24732 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
24733 extended instruction sets. Each has a corresponding @option{-mno-} option
24734 to disable use of these instructions.
24736 These extensions are also available as built-in functions: see
24737 @ref{x86 Built-in Functions}, for details of the functions enabled and
24738 disabled by these switches.
24740 To generate SSE/SSE2 instructions automatically from floating-point
24741 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
24743 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
24744 generates new AVX instructions or AVX equivalence for all SSEx instructions
24747 These options enable GCC to use these extended instructions in
24748 generated code, even without @option{-mfpmath=sse}. Applications that
24749 perform run-time CPU detection must compile separate files for each
24750 supported architecture, using the appropriate flags. In particular,
24751 the file containing the CPU detection code should be compiled without
24754 @item -mdump-tune-features
24755 @opindex mdump-tune-features
24756 This option instructs GCC to dump the names of the x86 performance
24757 tuning features and default settings. The names can be used in
24758 @option{-mtune-ctrl=@var{feature-list}}.
24760 @item -mtune-ctrl=@var{feature-list}
24761 @opindex mtune-ctrl=@var{feature-list}
24762 This option is used to do fine grain control of x86 code generation features.
24763 @var{feature-list} is a comma separated list of @var{feature} names. See also
24764 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
24765 on if it is not preceded with @samp{^}, otherwise, it is turned off.
24766 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
24767 developers. Using it may lead to code paths not covered by testing and can
24768 potentially result in compiler ICEs or runtime errors.
24771 @opindex mno-default
24772 This option instructs GCC to turn off all tunable features. See also
24773 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
24777 This option instructs GCC to emit a @code{cld} instruction in the prologue
24778 of functions that use string instructions. String instructions depend on
24779 the DF flag to select between autoincrement or autodecrement mode. While the
24780 ABI specifies the DF flag to be cleared on function entry, some operating
24781 systems violate this specification by not clearing the DF flag in their
24782 exception dispatchers. The exception handler can be invoked with the DF flag
24783 set, which leads to wrong direction mode when string instructions are used.
24784 This option can be enabled by default on 32-bit x86 targets by configuring
24785 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
24786 instructions can be suppressed with the @option{-mno-cld} compiler option
24790 @opindex mvzeroupper
24791 This option instructs GCC to emit a @code{vzeroupper} instruction
24792 before a transfer of control flow out of the function to minimize
24793 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
24796 @item -mprefer-avx128
24797 @opindex mprefer-avx128
24798 This option instructs GCC to use 128-bit AVX instructions instead of
24799 256-bit AVX instructions in the auto-vectorizer.
24803 This option enables GCC to generate @code{CMPXCHG16B} instructions.
24804 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
24805 (or oword) data types.
24806 This is useful for high-resolution counters that can be updated
24807 by multiple processors (or cores). This instruction is generated as part of
24808 atomic built-in functions: see @ref{__sync Builtins} or
24809 @ref{__atomic Builtins} for details.
24813 This option enables generation of @code{SAHF} instructions in 64-bit code.
24814 Early Intel Pentium 4 CPUs with Intel 64 support,
24815 prior to the introduction of Pentium 4 G1 step in December 2005,
24816 lacked the @code{LAHF} and @code{SAHF} instructions
24817 which are supported by AMD64.
24818 These are load and store instructions, respectively, for certain status flags.
24819 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
24820 @code{drem}, and @code{remainder} built-in functions;
24821 see @ref{Other Builtins} for details.
24825 This option enables use of the @code{movbe} instruction to implement
24826 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
24830 This option enables built-in functions @code{__builtin_ia32_crc32qi},
24831 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
24832 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
24836 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
24837 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
24838 with an additional Newton-Raphson step
24839 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
24840 (and their vectorized
24841 variants) for single-precision floating-point arguments. These instructions
24842 are generated only when @option{-funsafe-math-optimizations} is enabled
24843 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
24844 Note that while the throughput of the sequence is higher than the throughput
24845 of the non-reciprocal instruction, the precision of the sequence can be
24846 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
24848 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
24849 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
24850 combination), and doesn't need @option{-mrecip}.
24852 Also note that GCC emits the above sequence with additional Newton-Raphson step
24853 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
24854 already with @option{-ffast-math} (or the above option combination), and
24855 doesn't need @option{-mrecip}.
24857 @item -mrecip=@var{opt}
24858 @opindex mrecip=opt
24859 This option controls which reciprocal estimate instructions
24860 may be used. @var{opt} is a comma-separated list of options, which may
24861 be preceded by a @samp{!} to invert the option:
24865 Enable all estimate instructions.
24868 Enable the default instructions, equivalent to @option{-mrecip}.
24871 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24874 Enable the approximation for scalar division.
24877 Enable the approximation for vectorized division.
24880 Enable the approximation for scalar square root.
24883 Enable the approximation for vectorized square root.
24886 So, for example, @option{-mrecip=all,!sqrt} enables
24887 all of the reciprocal approximations, except for square root.
24889 @item -mveclibabi=@var{type}
24890 @opindex mveclibabi
24891 Specifies the ABI type to use for vectorizing intrinsics using an
24892 external library. Supported values for @var{type} are @samp{svml}
24893 for the Intel short
24894 vector math library and @samp{acml} for the AMD math core library.
24895 To use this option, both @option{-ftree-vectorize} and
24896 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
24897 ABI-compatible library must be specified at link time.
24899 GCC currently emits calls to @code{vmldExp2},
24900 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
24901 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
24902 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
24903 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
24904 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
24905 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
24906 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
24907 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
24908 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
24909 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
24910 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
24911 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
24912 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
24913 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
24914 when @option{-mveclibabi=acml} is used.
24916 @item -mabi=@var{name}
24918 Generate code for the specified calling convention. Permissible values
24919 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
24920 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
24921 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
24922 You can control this behavior for specific functions by
24923 using the function attributes @code{ms_abi} and @code{sysv_abi}.
24924 @xref{Function Attributes}.
24926 @item -mtls-dialect=@var{type}
24927 @opindex mtls-dialect
24928 Generate code to access thread-local storage using the @samp{gnu} or
24929 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
24930 @samp{gnu2} is more efficient, but it may add compile- and run-time
24931 requirements that cannot be satisfied on all systems.
24934 @itemx -mno-push-args
24935 @opindex mpush-args
24936 @opindex mno-push-args
24937 Use PUSH operations to store outgoing parameters. This method is shorter
24938 and usually equally fast as method using SUB/MOV operations and is enabled
24939 by default. In some cases disabling it may improve performance because of
24940 improved scheduling and reduced dependencies.
24942 @item -maccumulate-outgoing-args
24943 @opindex maccumulate-outgoing-args
24944 If enabled, the maximum amount of space required for outgoing arguments is
24945 computed in the function prologue. This is faster on most modern CPUs
24946 because of reduced dependencies, improved scheduling and reduced stack usage
24947 when the preferred stack boundary is not equal to 2. The drawback is a notable
24948 increase in code size. This switch implies @option{-mno-push-args}.
24952 Support thread-safe exception handling on MinGW. Programs that rely
24953 on thread-safe exception handling must compile and link all code with the
24954 @option{-mthreads} option. When compiling, @option{-mthreads} defines
24955 @option{-D_MT}; when linking, it links in a special thread helper library
24956 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
24958 @item -mms-bitfields
24959 @itemx -mno-ms-bitfields
24960 @opindex mms-bitfields
24961 @opindex mno-ms-bitfields
24963 Enable/disable bit-field layout compatible with the native Microsoft
24966 If @code{packed} is used on a structure, or if bit-fields are used,
24967 it may be that the Microsoft ABI lays out the structure differently
24968 than the way GCC normally does. Particularly when moving packed
24969 data between functions compiled with GCC and the native Microsoft compiler
24970 (either via function call or as data in a file), it may be necessary to access
24973 This option is enabled by default for Microsoft Windows
24974 targets. This behavior can also be controlled locally by use of variable
24975 or type attributes. For more information, see @ref{x86 Variable Attributes}
24976 and @ref{x86 Type Attributes}.
24978 The Microsoft structure layout algorithm is fairly simple with the exception
24979 of the bit-field packing.
24980 The padding and alignment of members of structures and whether a bit-field
24981 can straddle a storage-unit boundary are determine by these rules:
24984 @item Structure members are stored sequentially in the order in which they are
24985 declared: the first member has the lowest memory address and the last member
24988 @item Every data object has an alignment requirement. The alignment requirement
24989 for all data except structures, unions, and arrays is either the size of the
24990 object or the current packing size (specified with either the
24991 @code{aligned} attribute or the @code{pack} pragma),
24992 whichever is less. For structures, unions, and arrays,
24993 the alignment requirement is the largest alignment requirement of its members.
24994 Every object is allocated an offset so that:
24997 offset % alignment_requirement == 0
25000 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
25001 unit if the integral types are the same size and if the next bit-field fits
25002 into the current allocation unit without crossing the boundary imposed by the
25003 common alignment requirements of the bit-fields.
25006 MSVC interprets zero-length bit-fields in the following ways:
25009 @item If a zero-length bit-field is inserted between two bit-fields that
25010 are normally coalesced, the bit-fields are not coalesced.
25017 unsigned long bf_1 : 12;
25019 unsigned long bf_2 : 12;
25024 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
25025 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
25027 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
25028 alignment of the zero-length bit-field is greater than the member that follows it,
25029 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
25050 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
25051 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
25052 bit-field does not affect the alignment of @code{bar} or, as a result, the size
25055 Taking this into account, it is important to note the following:
25058 @item If a zero-length bit-field follows a normal bit-field, the type of the
25059 zero-length bit-field may affect the alignment of the structure as whole. For
25060 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
25061 normal bit-field, and is of type short.
25063 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
25064 still affect the alignment of the structure:
25075 Here, @code{t4} takes up 4 bytes.
25078 @item Zero-length bit-fields following non-bit-field members are ignored:
25090 Here, @code{t5} takes up 2 bytes.
25094 @item -mno-align-stringops
25095 @opindex mno-align-stringops
25096 Do not align the destination of inlined string operations. This switch reduces
25097 code size and improves performance in case the destination is already aligned,
25098 but GCC doesn't know about it.
25100 @item -minline-all-stringops
25101 @opindex minline-all-stringops
25102 By default GCC inlines string operations only when the destination is
25103 known to be aligned to least a 4-byte boundary.
25104 This enables more inlining and increases code
25105 size, but may improve performance of code that depends on fast
25106 @code{memcpy}, @code{strlen},
25107 and @code{memset} for short lengths.
25109 @item -minline-stringops-dynamically
25110 @opindex minline-stringops-dynamically
25111 For string operations of unknown size, use run-time checks with
25112 inline code for small blocks and a library call for large blocks.
25114 @item -mstringop-strategy=@var{alg}
25115 @opindex mstringop-strategy=@var{alg}
25116 Override the internal decision heuristic for the particular algorithm to use
25117 for inlining string operations. The allowed values for @var{alg} are:
25123 Expand using i386 @code{rep} prefix of the specified size.
25127 @itemx unrolled_loop
25128 Expand into an inline loop.
25131 Always use a library call.
25134 @item -mmemcpy-strategy=@var{strategy}
25135 @opindex mmemcpy-strategy=@var{strategy}
25136 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
25137 should be inlined and what inline algorithm to use when the expected size
25138 of the copy operation is known. @var{strategy}
25139 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
25140 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
25141 the max byte size with which inline algorithm @var{alg} is allowed. For the last
25142 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
25143 in the list must be specified in increasing order. The minimal byte size for
25144 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
25147 @item -mmemset-strategy=@var{strategy}
25148 @opindex mmemset-strategy=@var{strategy}
25149 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
25150 @code{__builtin_memset} expansion.
25152 @item -momit-leaf-frame-pointer
25153 @opindex momit-leaf-frame-pointer
25154 Don't keep the frame pointer in a register for leaf functions. This
25155 avoids the instructions to save, set up, and restore frame pointers and
25156 makes an extra register available in leaf functions. The option
25157 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
25158 which might make debugging harder.
25160 @item -mtls-direct-seg-refs
25161 @itemx -mno-tls-direct-seg-refs
25162 @opindex mtls-direct-seg-refs
25163 Controls whether TLS variables may be accessed with offsets from the
25164 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
25165 or whether the thread base pointer must be added. Whether or not this
25166 is valid depends on the operating system, and whether it maps the
25167 segment to cover the entire TLS area.
25169 For systems that use the GNU C Library, the default is on.
25172 @itemx -mno-sse2avx
25174 Specify that the assembler should encode SSE instructions with VEX
25175 prefix. The option @option{-mavx} turns this on by default.
25180 If profiling is active (@option{-pg}), put the profiling
25181 counter call before the prologue.
25182 Note: On x86 architectures the attribute @code{ms_hook_prologue}
25183 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
25185 @item -mrecord-mcount
25186 @itemx -mno-record-mcount
25187 @opindex mrecord-mcount
25188 If profiling is active (@option{-pg}), generate a __mcount_loc section
25189 that contains pointers to each profiling call. This is useful for
25190 automatically patching and out calls.
25193 @itemx -mno-nop-mcount
25194 @opindex mnop-mcount
25195 If profiling is active (@option{-pg}), generate the calls to
25196 the profiling functions as NOPs. This is useful when they
25197 should be patched in later dynamically. This is likely only
25198 useful together with @option{-mrecord-mcount}.
25200 @item -mskip-rax-setup
25201 @itemx -mno-skip-rax-setup
25202 @opindex mskip-rax-setup
25203 When generating code for the x86-64 architecture with SSE extensions
25204 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
25205 register when there are no variable arguments passed in vector registers.
25207 @strong{Warning:} Since RAX register is used to avoid unnecessarily
25208 saving vector registers on stack when passing variable arguments, the
25209 impacts of this option are callees may waste some stack space,
25210 misbehave or jump to a random location. GCC 4.4 or newer don't have
25211 those issues, regardless the RAX register value.
25214 @itemx -mno-8bit-idiv
25215 @opindex m8bit-idiv
25216 On some processors, like Intel Atom, 8-bit unsigned integer divide is
25217 much faster than 32-bit/64-bit integer divide. This option generates a
25218 run-time check. If both dividend and divisor are within range of 0
25219 to 255, 8-bit unsigned integer divide is used instead of
25220 32-bit/64-bit integer divide.
25222 @item -mavx256-split-unaligned-load
25223 @itemx -mavx256-split-unaligned-store
25224 @opindex mavx256-split-unaligned-load
25225 @opindex mavx256-split-unaligned-store
25226 Split 32-byte AVX unaligned load and store.
25228 @item -mstack-protector-guard=@var{guard}
25229 @opindex mstack-protector-guard=@var{guard}
25230 Generate stack protection code using canary at @var{guard}. Supported
25231 locations are @samp{global} for global canary or @samp{tls} for per-thread
25232 canary in the TLS block (the default). This option has effect only when
25233 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
25235 @item -mmitigate-rop
25236 @opindex mmitigate-rop
25237 Try to avoid generating code sequences that contain unintended return
25238 opcodes, to mitigate against certain forms of attack. At the moment,
25239 this option is limited in what it can do and should not be relied
25240 on to provide serious protection.
25242 @item -mgeneral-regs-only
25243 @opindex mgeneral-regs-only
25244 Generate code that uses only the general-purpose registers. This
25245 prevents the compiler from using floating-point, vector, mask and bound
25250 These @samp{-m} switches are supported in addition to the above
25251 on x86-64 processors in 64-bit environments.
25264 Generate code for a 16-bit, 32-bit or 64-bit environment.
25265 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
25267 generates code that runs on any i386 system.
25269 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
25270 types to 64 bits, and generates code for the x86-64 architecture.
25271 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
25272 and @option{-mdynamic-no-pic} options.
25274 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
25276 generates code for the x86-64 architecture.
25278 The @option{-m16} option is the same as @option{-m32}, except for that
25279 it outputs the @code{.code16gcc} assembly directive at the beginning of
25280 the assembly output so that the binary can run in 16-bit mode.
25282 The @option{-miamcu} option generates code which conforms to Intel MCU
25283 psABI. It requires the @option{-m32} option to be turned on.
25285 @item -mno-red-zone
25286 @opindex mno-red-zone
25287 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
25288 by the x86-64 ABI; it is a 128-byte area beyond the location of the
25289 stack pointer that is not modified by signal or interrupt handlers
25290 and therefore can be used for temporary data without adjusting the stack
25291 pointer. The flag @option{-mno-red-zone} disables this red zone.
25293 @item -mcmodel=small
25294 @opindex mcmodel=small
25295 Generate code for the small code model: the program and its symbols must
25296 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
25297 Programs can be statically or dynamically linked. This is the default
25300 @item -mcmodel=kernel
25301 @opindex mcmodel=kernel
25302 Generate code for the kernel code model. The kernel runs in the
25303 negative 2 GB of the address space.
25304 This model has to be used for Linux kernel code.
25306 @item -mcmodel=medium
25307 @opindex mcmodel=medium
25308 Generate code for the medium model: the program is linked in the lower 2
25309 GB of the address space. Small symbols are also placed there. Symbols
25310 with sizes larger than @option{-mlarge-data-threshold} are put into
25311 large data or BSS sections and can be located above 2GB. Programs can
25312 be statically or dynamically linked.
25314 @item -mcmodel=large
25315 @opindex mcmodel=large
25316 Generate code for the large model. This model makes no assumptions
25317 about addresses and sizes of sections.
25319 @item -maddress-mode=long
25320 @opindex maddress-mode=long
25321 Generate code for long address mode. This is only supported for 64-bit
25322 and x32 environments. It is the default address mode for 64-bit
25325 @item -maddress-mode=short
25326 @opindex maddress-mode=short
25327 Generate code for short address mode. This is only supported for 32-bit
25328 and x32 environments. It is the default address mode for 32-bit and
25332 @node x86 Windows Options
25333 @subsection x86 Windows Options
25334 @cindex x86 Windows Options
25335 @cindex Windows Options for x86
25337 These additional options are available for Microsoft Windows targets:
25343 specifies that a console application is to be generated, by
25344 instructing the linker to set the PE header subsystem type
25345 required for console applications.
25346 This option is available for Cygwin and MinGW targets and is
25347 enabled by default on those targets.
25351 This option is available for Cygwin and MinGW targets. It
25352 specifies that a DLL---a dynamic link library---is to be
25353 generated, enabling the selection of the required runtime
25354 startup object and entry point.
25356 @item -mnop-fun-dllimport
25357 @opindex mnop-fun-dllimport
25358 This option is available for Cygwin and MinGW targets. It
25359 specifies that the @code{dllimport} attribute should be ignored.
25363 This option is available for MinGW targets. It specifies
25364 that MinGW-specific thread support is to be used.
25368 This option is available for MinGW-w64 targets. It causes
25369 the @code{UNICODE} preprocessor macro to be predefined, and
25370 chooses Unicode-capable runtime startup code.
25374 This option is available for Cygwin and MinGW targets. It
25375 specifies that the typical Microsoft Windows predefined macros are to
25376 be set in the pre-processor, but does not influence the choice
25377 of runtime library/startup code.
25381 This option is available for Cygwin and MinGW targets. It
25382 specifies that a GUI application is to be generated by
25383 instructing the linker to set the PE header subsystem type
25386 @item -fno-set-stack-executable
25387 @opindex fno-set-stack-executable
25388 This option is available for MinGW targets. It specifies that
25389 the executable flag for the stack used by nested functions isn't
25390 set. This is necessary for binaries running in kernel mode of
25391 Microsoft Windows, as there the User32 API, which is used to set executable
25392 privileges, isn't available.
25394 @item -fwritable-relocated-rdata
25395 @opindex fno-writable-relocated-rdata
25396 This option is available for MinGW and Cygwin targets. It specifies
25397 that relocated-data in read-only section is put into the @code{.data}
25398 section. This is a necessary for older runtimes not supporting
25399 modification of @code{.rdata} sections for pseudo-relocation.
25401 @item -mpe-aligned-commons
25402 @opindex mpe-aligned-commons
25403 This option is available for Cygwin and MinGW targets. It
25404 specifies that the GNU extension to the PE file format that
25405 permits the correct alignment of COMMON variables should be
25406 used when generating code. It is enabled by default if
25407 GCC detects that the target assembler found during configuration
25408 supports the feature.
25411 See also under @ref{x86 Options} for standard options.
25413 @node Xstormy16 Options
25414 @subsection Xstormy16 Options
25415 @cindex Xstormy16 Options
25417 These options are defined for Xstormy16:
25422 Choose startup files and linker script suitable for the simulator.
25425 @node Xtensa Options
25426 @subsection Xtensa Options
25427 @cindex Xtensa Options
25429 These options are supported for Xtensa targets:
25433 @itemx -mno-const16
25435 @opindex mno-const16
25436 Enable or disable use of @code{CONST16} instructions for loading
25437 constant values. The @code{CONST16} instruction is currently not a
25438 standard option from Tensilica. When enabled, @code{CONST16}
25439 instructions are always used in place of the standard @code{L32R}
25440 instructions. The use of @code{CONST16} is enabled by default only if
25441 the @code{L32R} instruction is not available.
25444 @itemx -mno-fused-madd
25445 @opindex mfused-madd
25446 @opindex mno-fused-madd
25447 Enable or disable use of fused multiply/add and multiply/subtract
25448 instructions in the floating-point option. This has no effect if the
25449 floating-point option is not also enabled. Disabling fused multiply/add
25450 and multiply/subtract instructions forces the compiler to use separate
25451 instructions for the multiply and add/subtract operations. This may be
25452 desirable in some cases where strict IEEE 754-compliant results are
25453 required: the fused multiply add/subtract instructions do not round the
25454 intermediate result, thereby producing results with @emph{more} bits of
25455 precision than specified by the IEEE standard. Disabling fused multiply
25456 add/subtract instructions also ensures that the program output is not
25457 sensitive to the compiler's ability to combine multiply and add/subtract
25460 @item -mserialize-volatile
25461 @itemx -mno-serialize-volatile
25462 @opindex mserialize-volatile
25463 @opindex mno-serialize-volatile
25464 When this option is enabled, GCC inserts @code{MEMW} instructions before
25465 @code{volatile} memory references to guarantee sequential consistency.
25466 The default is @option{-mserialize-volatile}. Use
25467 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
25469 @item -mforce-no-pic
25470 @opindex mforce-no-pic
25471 For targets, like GNU/Linux, where all user-mode Xtensa code must be
25472 position-independent code (PIC), this option disables PIC for compiling
25475 @item -mtext-section-literals
25476 @itemx -mno-text-section-literals
25477 @opindex mtext-section-literals
25478 @opindex mno-text-section-literals
25479 These options control the treatment of literal pools. The default is
25480 @option{-mno-text-section-literals}, which places literals in a separate
25481 section in the output file. This allows the literal pool to be placed
25482 in a data RAM/ROM, and it also allows the linker to combine literal
25483 pools from separate object files to remove redundant literals and
25484 improve code size. With @option{-mtext-section-literals}, the literals
25485 are interspersed in the text section in order to keep them as close as
25486 possible to their references. This may be necessary for large assembly
25487 files. Literals for each function are placed right before that function.
25489 @item -mauto-litpools
25490 @itemx -mno-auto-litpools
25491 @opindex mauto-litpools
25492 @opindex mno-auto-litpools
25493 These options control the treatment of literal pools. The default is
25494 @option{-mno-auto-litpools}, which places literals in a separate
25495 section in the output file unless @option{-mtext-section-literals} is
25496 used. With @option{-mauto-litpools} the literals are interspersed in
25497 the text section by the assembler. Compiler does not produce explicit
25498 @code{.literal} directives and loads literals into registers with
25499 @code{MOVI} instructions instead of @code{L32R} to let the assembler
25500 do relaxation and place literals as necessary. This option allows
25501 assembler to create several literal pools per function and assemble
25502 very big functions, which may not be possible with
25503 @option{-mtext-section-literals}.
25505 @item -mtarget-align
25506 @itemx -mno-target-align
25507 @opindex mtarget-align
25508 @opindex mno-target-align
25509 When this option is enabled, GCC instructs the assembler to
25510 automatically align instructions to reduce branch penalties at the
25511 expense of some code density. The assembler attempts to widen density
25512 instructions to align branch targets and the instructions following call
25513 instructions. If there are not enough preceding safe density
25514 instructions to align a target, no widening is performed. The
25515 default is @option{-mtarget-align}. These options do not affect the
25516 treatment of auto-aligned instructions like @code{LOOP}, which the
25517 assembler always aligns, either by widening density instructions or
25518 by inserting NOP instructions.
25521 @itemx -mno-longcalls
25522 @opindex mlongcalls
25523 @opindex mno-longcalls
25524 When this option is enabled, GCC instructs the assembler to translate
25525 direct calls to indirect calls unless it can determine that the target
25526 of a direct call is in the range allowed by the call instruction. This
25527 translation typically occurs for calls to functions in other source
25528 files. Specifically, the assembler translates a direct @code{CALL}
25529 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
25530 The default is @option{-mno-longcalls}. This option should be used in
25531 programs where the call target can potentially be out of range. This
25532 option is implemented in the assembler, not the compiler, so the
25533 assembly code generated by GCC still shows direct call
25534 instructions---look at the disassembled object code to see the actual
25535 instructions. Note that the assembler uses an indirect call for
25536 every cross-file call, not just those that really are out of range.
25539 @node zSeries Options
25540 @subsection zSeries Options
25541 @cindex zSeries options
25543 These are listed under @xref{S/390 and zSeries Options}.
25549 @section Specifying Subprocesses and the Switches to Pass to Them
25552 @command{gcc} is a driver program. It performs its job by invoking a
25553 sequence of other programs to do the work of compiling, assembling and
25554 linking. GCC interprets its command-line parameters and uses these to
25555 deduce which programs it should invoke, and which command-line options
25556 it ought to place on their command lines. This behavior is controlled
25557 by @dfn{spec strings}. In most cases there is one spec string for each
25558 program that GCC can invoke, but a few programs have multiple spec
25559 strings to control their behavior. The spec strings built into GCC can
25560 be overridden by using the @option{-specs=} command-line switch to specify
25563 @dfn{Spec files} are plain-text files that are used to construct spec
25564 strings. They consist of a sequence of directives separated by blank
25565 lines. The type of directive is determined by the first non-whitespace
25566 character on the line, which can be one of the following:
25569 @item %@var{command}
25570 Issues a @var{command} to the spec file processor. The commands that can
25574 @item %include <@var{file}>
25575 @cindex @code{%include}
25576 Search for @var{file} and insert its text at the current point in the
25579 @item %include_noerr <@var{file}>
25580 @cindex @code{%include_noerr}
25581 Just like @samp{%include}, but do not generate an error message if the include
25582 file cannot be found.
25584 @item %rename @var{old_name} @var{new_name}
25585 @cindex @code{%rename}
25586 Rename the spec string @var{old_name} to @var{new_name}.
25590 @item *[@var{spec_name}]:
25591 This tells the compiler to create, override or delete the named spec
25592 string. All lines after this directive up to the next directive or
25593 blank line are considered to be the text for the spec string. If this
25594 results in an empty string then the spec is deleted. (Or, if the
25595 spec did not exist, then nothing happens.) Otherwise, if the spec
25596 does not currently exist a new spec is created. If the spec does
25597 exist then its contents are overridden by the text of this
25598 directive, unless the first character of that text is the @samp{+}
25599 character, in which case the text is appended to the spec.
25601 @item [@var{suffix}]:
25602 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
25603 and up to the next directive or blank line are considered to make up the
25604 spec string for the indicated suffix. When the compiler encounters an
25605 input file with the named suffix, it processes the spec string in
25606 order to work out how to compile that file. For example:
25610 z-compile -input %i
25613 This says that any input file whose name ends in @samp{.ZZ} should be
25614 passed to the program @samp{z-compile}, which should be invoked with the
25615 command-line switch @option{-input} and with the result of performing the
25616 @samp{%i} substitution. (See below.)
25618 As an alternative to providing a spec string, the text following a
25619 suffix directive can be one of the following:
25622 @item @@@var{language}
25623 This says that the suffix is an alias for a known @var{language}. This is
25624 similar to using the @option{-x} command-line switch to GCC to specify a
25625 language explicitly. For example:
25632 Says that .ZZ files are, in fact, C++ source files.
25635 This causes an error messages saying:
25638 @var{name} compiler not installed on this system.
25642 GCC already has an extensive list of suffixes built into it.
25643 This directive adds an entry to the end of the list of suffixes, but
25644 since the list is searched from the end backwards, it is effectively
25645 possible to override earlier entries using this technique.
25649 GCC has the following spec strings built into it. Spec files can
25650 override these strings or create their own. Note that individual
25651 targets can also add their own spec strings to this list.
25654 asm Options to pass to the assembler
25655 asm_final Options to pass to the assembler post-processor
25656 cpp Options to pass to the C preprocessor
25657 cc1 Options to pass to the C compiler
25658 cc1plus Options to pass to the C++ compiler
25659 endfile Object files to include at the end of the link
25660 link Options to pass to the linker
25661 lib Libraries to include on the command line to the linker
25662 libgcc Decides which GCC support library to pass to the linker
25663 linker Sets the name of the linker
25664 predefines Defines to be passed to the C preprocessor
25665 signed_char Defines to pass to CPP to say whether @code{char} is signed
25667 startfile Object files to include at the start of the link
25670 Here is a small example of a spec file:
25673 %rename lib old_lib
25676 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
25679 This example renames the spec called @samp{lib} to @samp{old_lib} and
25680 then overrides the previous definition of @samp{lib} with a new one.
25681 The new definition adds in some extra command-line options before
25682 including the text of the old definition.
25684 @dfn{Spec strings} are a list of command-line options to be passed to their
25685 corresponding program. In addition, the spec strings can contain
25686 @samp{%}-prefixed sequences to substitute variable text or to
25687 conditionally insert text into the command line. Using these constructs
25688 it is possible to generate quite complex command lines.
25690 Here is a table of all defined @samp{%}-sequences for spec
25691 strings. Note that spaces are not generated automatically around the
25692 results of expanding these sequences. Therefore you can concatenate them
25693 together or combine them with constant text in a single argument.
25697 Substitute one @samp{%} into the program name or argument.
25700 Substitute the name of the input file being processed.
25703 Substitute the basename of the input file being processed.
25704 This is the substring up to (and not including) the last period
25705 and not including the directory.
25708 This is the same as @samp{%b}, but include the file suffix (text after
25712 Marks the argument containing or following the @samp{%d} as a
25713 temporary file name, so that that file is deleted if GCC exits
25714 successfully. Unlike @samp{%g}, this contributes no text to the
25717 @item %g@var{suffix}
25718 Substitute a file name that has suffix @var{suffix} and is chosen
25719 once per compilation, and mark the argument in the same way as
25720 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
25721 name is now chosen in a way that is hard to predict even when previously
25722 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
25723 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
25724 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
25725 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
25726 was simply substituted with a file name chosen once per compilation,
25727 without regard to any appended suffix (which was therefore treated
25728 just like ordinary text), making such attacks more likely to succeed.
25730 @item %u@var{suffix}
25731 Like @samp{%g}, but generates a new temporary file name
25732 each time it appears instead of once per compilation.
25734 @item %U@var{suffix}
25735 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
25736 new one if there is no such last file name. In the absence of any
25737 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
25738 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
25739 involves the generation of two distinct file names, one
25740 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
25741 simply substituted with a file name chosen for the previous @samp{%u},
25742 without regard to any appended suffix.
25744 @item %j@var{suffix}
25745 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
25746 writable, and if @option{-save-temps} is not used;
25747 otherwise, substitute the name
25748 of a temporary file, just like @samp{%u}. This temporary file is not
25749 meant for communication between processes, but rather as a junk
25750 disposal mechanism.
25752 @item %|@var{suffix}
25753 @itemx %m@var{suffix}
25754 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
25755 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
25756 all. These are the two most common ways to instruct a program that it
25757 should read from standard input or write to standard output. If you
25758 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
25759 construct: see for example @file{f/lang-specs.h}.
25761 @item %.@var{SUFFIX}
25762 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
25763 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
25764 terminated by the next space or %.
25767 Marks the argument containing or following the @samp{%w} as the
25768 designated output file of this compilation. This puts the argument
25769 into the sequence of arguments that @samp{%o} substitutes.
25772 Substitutes the names of all the output files, with spaces
25773 automatically placed around them. You should write spaces
25774 around the @samp{%o} as well or the results are undefined.
25775 @samp{%o} is for use in the specs for running the linker.
25776 Input files whose names have no recognized suffix are not compiled
25777 at all, but they are included among the output files, so they are
25781 Substitutes the suffix for object files. Note that this is
25782 handled specially when it immediately follows @samp{%g, %u, or %U},
25783 because of the need for those to form complete file names. The
25784 handling is such that @samp{%O} is treated exactly as if it had already
25785 been substituted, except that @samp{%g, %u, and %U} do not currently
25786 support additional @var{suffix} characters following @samp{%O} as they do
25787 following, for example, @samp{.o}.
25790 Substitutes the standard macro predefinitions for the
25791 current target machine. Use this when running @command{cpp}.
25794 Like @samp{%p}, but puts @samp{__} before and after the name of each
25795 predefined macro, except for macros that start with @samp{__} or with
25796 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
25800 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
25801 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
25802 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
25803 and @option{-imultilib} as necessary.
25806 Current argument is the name of a library or startup file of some sort.
25807 Search for that file in a standard list of directories and substitute
25808 the full name found. The current working directory is included in the
25809 list of directories scanned.
25812 Current argument is the name of a linker script. Search for that file
25813 in the current list of directories to scan for libraries. If the file
25814 is located insert a @option{--script} option into the command line
25815 followed by the full path name found. If the file is not found then
25816 generate an error message. Note: the current working directory is not
25820 Print @var{str} as an error message. @var{str} is terminated by a newline.
25821 Use this when inconsistent options are detected.
25823 @item %(@var{name})
25824 Substitute the contents of spec string @var{name} at this point.
25826 @item %x@{@var{option}@}
25827 Accumulate an option for @samp{%X}.
25830 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
25834 Output the accumulated assembler options specified by @option{-Wa}.
25837 Output the accumulated preprocessor options specified by @option{-Wp}.
25840 Process the @code{asm} spec. This is used to compute the
25841 switches to be passed to the assembler.
25844 Process the @code{asm_final} spec. This is a spec string for
25845 passing switches to an assembler post-processor, if such a program is
25849 Process the @code{link} spec. This is the spec for computing the
25850 command line passed to the linker. Typically it makes use of the
25851 @samp{%L %G %S %D and %E} sequences.
25854 Dump out a @option{-L} option for each directory that GCC believes might
25855 contain startup files. If the target supports multilibs then the
25856 current multilib directory is prepended to each of these paths.
25859 Process the @code{lib} spec. This is a spec string for deciding which
25860 libraries are included on the command line to the linker.
25863 Process the @code{libgcc} spec. This is a spec string for deciding
25864 which GCC support library is included on the command line to the linker.
25867 Process the @code{startfile} spec. This is a spec for deciding which
25868 object files are the first ones passed to the linker. Typically
25869 this might be a file named @file{crt0.o}.
25872 Process the @code{endfile} spec. This is a spec string that specifies
25873 the last object files that are passed to the linker.
25876 Process the @code{cpp} spec. This is used to construct the arguments
25877 to be passed to the C preprocessor.
25880 Process the @code{cc1} spec. This is used to construct the options to be
25881 passed to the actual C compiler (@command{cc1}).
25884 Process the @code{cc1plus} spec. This is used to construct the options to be
25885 passed to the actual C++ compiler (@command{cc1plus}).
25888 Substitute the variable part of a matched option. See below.
25889 Note that each comma in the substituted string is replaced by
25893 Remove all occurrences of @code{-S} from the command line. Note---this
25894 command is position dependent. @samp{%} commands in the spec string
25895 before this one see @code{-S}, @samp{%} commands in the spec string
25896 after this one do not.
25898 @item %:@var{function}(@var{args})
25899 Call the named function @var{function}, passing it @var{args}.
25900 @var{args} is first processed as a nested spec string, then split
25901 into an argument vector in the usual fashion. The function returns
25902 a string which is processed as if it had appeared literally as part
25903 of the current spec.
25905 The following built-in spec functions are provided:
25908 @item @code{getenv}
25909 The @code{getenv} spec function takes two arguments: an environment
25910 variable name and a string. If the environment variable is not
25911 defined, a fatal error is issued. Otherwise, the return value is the
25912 value of the environment variable concatenated with the string. For
25913 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
25916 %:getenv(TOPDIR /include)
25919 expands to @file{/path/to/top/include}.
25921 @item @code{if-exists}
25922 The @code{if-exists} spec function takes one argument, an absolute
25923 pathname to a file. If the file exists, @code{if-exists} returns the
25924 pathname. Here is a small example of its usage:
25928 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
25931 @item @code{if-exists-else}
25932 The @code{if-exists-else} spec function is similar to the @code{if-exists}
25933 spec function, except that it takes two arguments. The first argument is
25934 an absolute pathname to a file. If the file exists, @code{if-exists-else}
25935 returns the pathname. If it does not exist, it returns the second argument.
25936 This way, @code{if-exists-else} can be used to select one file or another,
25937 based on the existence of the first. Here is a small example of its usage:
25941 crt0%O%s %:if-exists(crti%O%s) \
25942 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
25945 @item @code{replace-outfile}
25946 The @code{replace-outfile} spec function takes two arguments. It looks for the
25947 first argument in the outfiles array and replaces it with the second argument. Here
25948 is a small example of its usage:
25951 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
25954 @item @code{remove-outfile}
25955 The @code{remove-outfile} spec function takes one argument. It looks for the
25956 first argument in the outfiles array and removes it. Here is a small example
25960 %:remove-outfile(-lm)
25963 @item @code{pass-through-libs}
25964 The @code{pass-through-libs} spec function takes any number of arguments. It
25965 finds any @option{-l} options and any non-options ending in @file{.a} (which it
25966 assumes are the names of linker input library archive files) and returns a
25967 result containing all the found arguments each prepended by
25968 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
25969 intended to be passed to the LTO linker plugin.
25972 %:pass-through-libs(%G %L %G)
25975 @item @code{print-asm-header}
25976 The @code{print-asm-header} function takes no arguments and simply
25977 prints a banner like:
25983 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
25986 It is used to separate compiler options from assembler options
25987 in the @option{--target-help} output.
25990 @item %@{@code{S}@}
25991 Substitutes the @code{-S} switch, if that switch is given to GCC@.
25992 If that switch is not specified, this substitutes nothing. Note that
25993 the leading dash is omitted when specifying this option, and it is
25994 automatically inserted if the substitution is performed. Thus the spec
25995 string @samp{%@{foo@}} matches the command-line option @option{-foo}
25996 and outputs the command-line option @option{-foo}.
25998 @item %W@{@code{S}@}
25999 Like %@{@code{S}@} but mark last argument supplied within as a file to be
26000 deleted on failure.
26002 @item %@{@code{S}*@}
26003 Substitutes all the switches specified to GCC whose names start
26004 with @code{-S}, but which also take an argument. This is used for
26005 switches like @option{-o}, @option{-D}, @option{-I}, etc.
26006 GCC considers @option{-o foo} as being
26007 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
26008 text, including the space. Thus two arguments are generated.
26010 @item %@{@code{S}*&@code{T}*@}
26011 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
26012 (the order of @code{S} and @code{T} in the spec is not significant).
26013 There can be any number of ampersand-separated variables; for each the
26014 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
26016 @item %@{@code{S}:@code{X}@}
26017 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
26019 @item %@{!@code{S}:@code{X}@}
26020 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
26022 @item %@{@code{S}*:@code{X}@}
26023 Substitutes @code{X} if one or more switches whose names start with
26024 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
26025 once, no matter how many such switches appeared. However, if @code{%*}
26026 appears somewhere in @code{X}, then @code{X} is substituted once
26027 for each matching switch, with the @code{%*} replaced by the part of
26028 that switch matching the @code{*}.
26030 If @code{%*} appears as the last part of a spec sequence then a space
26031 is added after the end of the last substitution. If there is more
26032 text in the sequence, however, then a space is not generated. This
26033 allows the @code{%*} substitution to be used as part of a larger
26034 string. For example, a spec string like this:
26037 %@{mcu=*:--script=%*/memory.ld@}
26041 when matching an option like @option{-mcu=newchip} produces:
26044 --script=newchip/memory.ld
26047 @item %@{.@code{S}:@code{X}@}
26048 Substitutes @code{X}, if processing a file with suffix @code{S}.
26050 @item %@{!.@code{S}:@code{X}@}
26051 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
26053 @item %@{,@code{S}:@code{X}@}
26054 Substitutes @code{X}, if processing a file for language @code{S}.
26056 @item %@{!,@code{S}:@code{X}@}
26057 Substitutes @code{X}, if not processing a file for language @code{S}.
26059 @item %@{@code{S}|@code{P}:@code{X}@}
26060 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
26061 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
26062 @code{*} sequences as well, although they have a stronger binding than
26063 the @samp{|}. If @code{%*} appears in @code{X}, all of the
26064 alternatives must be starred, and only the first matching alternative
26067 For example, a spec string like this:
26070 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
26074 outputs the following command-line options from the following input
26075 command-line options:
26080 -d fred.c -foo -baz -boggle
26081 -d jim.d -bar -baz -boggle
26084 @item %@{S:X; T:Y; :D@}
26086 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
26087 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
26088 be as many clauses as you need. This may be combined with @code{.},
26089 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
26094 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
26095 construct may contain other nested @samp{%} constructs or spaces, or
26096 even newlines. They are processed as usual, as described above.
26097 Trailing white space in @code{X} is ignored. White space may also
26098 appear anywhere on the left side of the colon in these constructs,
26099 except between @code{.} or @code{*} and the corresponding word.
26101 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
26102 handled specifically in these constructs. If another value of
26103 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
26104 @option{-W} switch is found later in the command line, the earlier
26105 switch value is ignored, except with @{@code{S}*@} where @code{S} is
26106 just one letter, which passes all matching options.
26108 The character @samp{|} at the beginning of the predicate text is used to
26109 indicate that a command should be piped to the following command, but
26110 only if @option{-pipe} is specified.
26112 It is built into GCC which switches take arguments and which do not.
26113 (You might think it would be useful to generalize this to allow each
26114 compiler's spec to say which switches take arguments. But this cannot
26115 be done in a consistent fashion. GCC cannot even decide which input
26116 files have been specified without knowing which switches take arguments,
26117 and it must know which input files to compile in order to tell which
26120 GCC also knows implicitly that arguments starting in @option{-l} are to be
26121 treated as compiler output files, and passed to the linker in their
26122 proper position among the other output files.
26124 @node Environment Variables
26125 @section Environment Variables Affecting GCC
26126 @cindex environment variables
26128 @c man begin ENVIRONMENT
26129 This section describes several environment variables that affect how GCC
26130 operates. Some of them work by specifying directories or prefixes to use
26131 when searching for various kinds of files. Some are used to specify other
26132 aspects of the compilation environment.
26134 Note that you can also specify places to search using options such as
26135 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
26136 take precedence over places specified using environment variables, which
26137 in turn take precedence over those specified by the configuration of GCC@.
26138 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
26139 GNU Compiler Collection (GCC) Internals}.
26144 @c @itemx LC_COLLATE
26146 @c @itemx LC_MONETARY
26147 @c @itemx LC_NUMERIC
26152 @c @findex LC_COLLATE
26153 @findex LC_MESSAGES
26154 @c @findex LC_MONETARY
26155 @c @findex LC_NUMERIC
26159 These environment variables control the way that GCC uses
26160 localization information which allows GCC to work with different
26161 national conventions. GCC inspects the locale categories
26162 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
26163 so. These locale categories can be set to any value supported by your
26164 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
26165 Kingdom encoded in UTF-8.
26167 The @env{LC_CTYPE} environment variable specifies character
26168 classification. GCC uses it to determine the character boundaries in
26169 a string; this is needed for some multibyte encodings that contain quote
26170 and escape characters that are otherwise interpreted as a string
26173 The @env{LC_MESSAGES} environment variable specifies the language to
26174 use in diagnostic messages.
26176 If the @env{LC_ALL} environment variable is set, it overrides the value
26177 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
26178 and @env{LC_MESSAGES} default to the value of the @env{LANG}
26179 environment variable. If none of these variables are set, GCC
26180 defaults to traditional C English behavior.
26184 If @env{TMPDIR} is set, it specifies the directory to use for temporary
26185 files. GCC uses temporary files to hold the output of one stage of
26186 compilation which is to be used as input to the next stage: for example,
26187 the output of the preprocessor, which is the input to the compiler
26190 @item GCC_COMPARE_DEBUG
26191 @findex GCC_COMPARE_DEBUG
26192 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
26193 @option{-fcompare-debug} to the compiler driver. See the documentation
26194 of this option for more details.
26196 @item GCC_EXEC_PREFIX
26197 @findex GCC_EXEC_PREFIX
26198 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
26199 names of the subprograms executed by the compiler. No slash is added
26200 when this prefix is combined with the name of a subprogram, but you can
26201 specify a prefix that ends with a slash if you wish.
26203 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
26204 an appropriate prefix to use based on the pathname it is invoked with.
26206 If GCC cannot find the subprogram using the specified prefix, it
26207 tries looking in the usual places for the subprogram.
26209 The default value of @env{GCC_EXEC_PREFIX} is
26210 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
26211 the installed compiler. In many cases @var{prefix} is the value
26212 of @code{prefix} when you ran the @file{configure} script.
26214 Other prefixes specified with @option{-B} take precedence over this prefix.
26216 This prefix is also used for finding files such as @file{crt0.o} that are
26219 In addition, the prefix is used in an unusual way in finding the
26220 directories to search for header files. For each of the standard
26221 directories whose name normally begins with @samp{/usr/local/lib/gcc}
26222 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
26223 replacing that beginning with the specified prefix to produce an
26224 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
26225 @file{foo/bar} just before it searches the standard directory
26226 @file{/usr/local/lib/bar}.
26227 If a standard directory begins with the configured
26228 @var{prefix} then the value of @var{prefix} is replaced by
26229 @env{GCC_EXEC_PREFIX} when looking for header files.
26231 @item COMPILER_PATH
26232 @findex COMPILER_PATH
26233 The value of @env{COMPILER_PATH} is a colon-separated list of
26234 directories, much like @env{PATH}. GCC tries the directories thus
26235 specified when searching for subprograms, if it can't find the
26236 subprograms using @env{GCC_EXEC_PREFIX}.
26239 @findex LIBRARY_PATH
26240 The value of @env{LIBRARY_PATH} is a colon-separated list of
26241 directories, much like @env{PATH}. When configured as a native compiler,
26242 GCC tries the directories thus specified when searching for special
26243 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
26244 using GCC also uses these directories when searching for ordinary
26245 libraries for the @option{-l} option (but directories specified with
26246 @option{-L} come first).
26250 @cindex locale definition
26251 This variable is used to pass locale information to the compiler. One way in
26252 which this information is used is to determine the character set to be used
26253 when character literals, string literals and comments are parsed in C and C++.
26254 When the compiler is configured to allow multibyte characters,
26255 the following values for @env{LANG} are recognized:
26259 Recognize JIS characters.
26261 Recognize SJIS characters.
26263 Recognize EUCJP characters.
26266 If @env{LANG} is not defined, or if it has some other value, then the
26267 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
26268 recognize and translate multibyte characters.
26272 Some additional environment variables affect the behavior of the
26275 @include cppenv.texi
26279 @node Precompiled Headers
26280 @section Using Precompiled Headers
26281 @cindex precompiled headers
26282 @cindex speed of compilation
26284 Often large projects have many header files that are included in every
26285 source file. The time the compiler takes to process these header files
26286 over and over again can account for nearly all of the time required to
26287 build the project. To make builds faster, GCC allows you to
26288 @dfn{precompile} a header file.
26290 To create a precompiled header file, simply compile it as you would any
26291 other file, if necessary using the @option{-x} option to make the driver
26292 treat it as a C or C++ header file. You may want to use a
26293 tool like @command{make} to keep the precompiled header up-to-date when
26294 the headers it contains change.
26296 A precompiled header file is searched for when @code{#include} is
26297 seen in the compilation. As it searches for the included file
26298 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
26299 compiler looks for a precompiled header in each directory just before it
26300 looks for the include file in that directory. The name searched for is
26301 the name specified in the @code{#include} with @samp{.gch} appended. If
26302 the precompiled header file can't be used, it is ignored.
26304 For instance, if you have @code{#include "all.h"}, and you have
26305 @file{all.h.gch} in the same directory as @file{all.h}, then the
26306 precompiled header file is used if possible, and the original
26307 header is used otherwise.
26309 Alternatively, you might decide to put the precompiled header file in a
26310 directory and use @option{-I} to ensure that directory is searched
26311 before (or instead of) the directory containing the original header.
26312 Then, if you want to check that the precompiled header file is always
26313 used, you can put a file of the same name as the original header in this
26314 directory containing an @code{#error} command.
26316 This also works with @option{-include}. So yet another way to use
26317 precompiled headers, good for projects not designed with precompiled
26318 header files in mind, is to simply take most of the header files used by
26319 a project, include them from another header file, precompile that header
26320 file, and @option{-include} the precompiled header. If the header files
26321 have guards against multiple inclusion, they are skipped because
26322 they've already been included (in the precompiled header).
26324 If you need to precompile the same header file for different
26325 languages, targets, or compiler options, you can instead make a
26326 @emph{directory} named like @file{all.h.gch}, and put each precompiled
26327 header in the directory, perhaps using @option{-o}. It doesn't matter
26328 what you call the files in the directory; every precompiled header in
26329 the directory is considered. The first precompiled header
26330 encountered in the directory that is valid for this compilation is
26331 used; they're searched in no particular order.
26333 There are many other possibilities, limited only by your imagination,
26334 good sense, and the constraints of your build system.
26336 A precompiled header file can be used only when these conditions apply:
26340 Only one precompiled header can be used in a particular compilation.
26343 A precompiled header can't be used once the first C token is seen. You
26344 can have preprocessor directives before a precompiled header; you cannot
26345 include a precompiled header from inside another header.
26348 The precompiled header file must be produced for the same language as
26349 the current compilation. You can't use a C precompiled header for a C++
26353 The precompiled header file must have been produced by the same compiler
26354 binary as the current compilation is using.
26357 Any macros defined before the precompiled header is included must
26358 either be defined in the same way as when the precompiled header was
26359 generated, or must not affect the precompiled header, which usually
26360 means that they don't appear in the precompiled header at all.
26362 The @option{-D} option is one way to define a macro before a
26363 precompiled header is included; using a @code{#define} can also do it.
26364 There are also some options that define macros implicitly, like
26365 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
26368 @item If debugging information is output when using the precompiled
26369 header, using @option{-g} or similar, the same kind of debugging information
26370 must have been output when building the precompiled header. However,
26371 a precompiled header built using @option{-g} can be used in a compilation
26372 when no debugging information is being output.
26374 @item The same @option{-m} options must generally be used when building
26375 and using the precompiled header. @xref{Submodel Options},
26376 for any cases where this rule is relaxed.
26378 @item Each of the following options must be the same when building and using
26379 the precompiled header:
26381 @gccoptlist{-fexceptions}
26384 Some other command-line options starting with @option{-f},
26385 @option{-p}, or @option{-O} must be defined in the same way as when
26386 the precompiled header was generated. At present, it's not clear
26387 which options are safe to change and which are not; the safest choice
26388 is to use exactly the same options when generating and using the
26389 precompiled header. The following are known to be safe:
26391 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
26392 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
26393 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
26398 For all of these except the last, the compiler automatically
26399 ignores the precompiled header if the conditions aren't met. If you
26400 find an option combination that doesn't work and doesn't cause the
26401 precompiled header to be ignored, please consider filing a bug report,
26404 If you do use differing options when generating and using the
26405 precompiled header, the actual behavior is a mixture of the
26406 behavior for the options. For instance, if you use @option{-g} to
26407 generate the precompiled header but not when using it, you may or may
26408 not get debugging information for routines in the precompiled header.