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 -fnew-inheriting-ctors @gol
203 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
204 -fno-optional-diags -fpermissive @gol
205 -fno-pretty-templates @gol
206 -frepo -fno-rtti -fsized-deallocation @gol
207 -ftemplate-backtrace-limit=@var{n} @gol
208 -ftemplate-depth=@var{n} @gol
209 -fno-threadsafe-statics -fuse-cxa-atexit @gol
210 -fno-weak -nostdinc++ @gol
211 -fvisibility-inlines-hidden @gol
212 -fvisibility-ms-compat @gol
213 -fext-numeric-literals @gol
214 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
215 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
216 -Wnamespaces -Wnarrowing @gol
217 -Wnoexcept -Wnon-virtual-dtor -Wreorder -Wregister @gol
218 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
219 -Wno-non-template-friend -Wold-style-cast @gol
220 -Woverloaded-virtual -Wno-pmf-conversions @gol
221 -Wsign-promo -Wvirtual-inheritance}
223 @item Objective-C and Objective-C++ Language Options
224 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
225 Objective-C and Objective-C++ Dialects}.
226 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
227 -fgnu-runtime -fnext-runtime @gol
228 -fno-nil-receivers @gol
229 -fobjc-abi-version=@var{n} @gol
230 -fobjc-call-cxx-cdtors @gol
231 -fobjc-direct-dispatch @gol
232 -fobjc-exceptions @gol
235 -fobjc-std=objc1 @gol
236 -fno-local-ivars @gol
237 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
238 -freplace-objc-classes @gol
241 -Wassign-intercept @gol
242 -Wno-protocol -Wselector @gol
243 -Wstrict-selector-match @gol
244 -Wundeclared-selector}
246 @item Diagnostic Message Formatting Options
247 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
248 @gccoptlist{-fmessage-length=@var{n} @gol
249 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
250 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
251 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
252 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch}
254 @item Warning Options
255 @xref{Warning Options,,Options to Request or Suppress Warnings}.
256 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
257 -pedantic-errors @gol
258 -w -Wextra -Wall -Waddress -Waggregate-return @gol
259 -Walloca -Walloca-larger-than=@var{n} @gol
260 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
261 -Wno-attributes -Wbool-compare -Wbool-operation @gol
262 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
263 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
264 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
265 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
266 -Wdelete-incomplete @gol
267 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
268 -Wdisabled-optimization @gol
269 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
270 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
271 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
272 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
273 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-length=@var{n} @gol
274 -Wformat-nonliteral @gol
275 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
276 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
277 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
278 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
279 -Wimplicit-function-declaration -Wimplicit-int @gol
280 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
281 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
282 -Winvalid-pch -Wlarger-than=@var{len} @gol
283 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
284 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
285 -Wmisleading-indentation -Wmissing-braces @gol
286 -Wmissing-field-initializers -Wmissing-include-dirs @gol
287 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
288 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
289 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
290 -Woverride-init-side-effects -Woverlength-strings @gol
291 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
292 -Wparentheses -Wno-pedantic-ms-format @gol
293 -Wplacement-new -Wplacement-new=@var{n} @gol
294 -Wpointer-arith -Wno-pointer-to-int-cast @gol
295 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
296 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
297 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
298 -Wshift-overflow -Wshift-overflow=@var{n} @gol
299 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
300 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
301 -Wno-scalar-storage-order @gol
302 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
303 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
304 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
305 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
306 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
307 -Wmissing-format-attribute -Wsubobject-linkage @gol
308 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
309 -Wswitch-unreachable -Wsync-nand @gol
310 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
311 -Wtype-limits -Wundef @gol
312 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
313 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
314 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
315 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
316 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
317 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
318 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
319 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
320 -Wzero-as-null-pointer-constant -Whsa}
322 @item C and Objective-C-only Warning Options
323 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
324 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
325 -Wold-style-declaration -Wold-style-definition @gol
326 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
327 -Wdeclaration-after-statement -Wpointer-sign}
329 @item Debugging Options
330 @xref{Debugging Options,,Options for Debugging Your Program}.
331 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
332 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
333 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
334 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
335 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
336 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
337 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
338 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
339 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
340 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
341 -fvar-tracking -fvar-tracking-assignments}
343 @item Optimization Options
344 @xref{Optimize Options,,Options that Control Optimization}.
345 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
346 -falign-jumps[=@var{n}] @gol
347 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
348 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
349 -fauto-inc-dec -fbranch-probabilities @gol
350 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
351 -fbtr-bb-exclusive -fcaller-saves @gol
352 -fcombine-stack-adjustments -fconserve-stack @gol
353 -fcompare-elim -fcprop-registers -fcrossjumping @gol
354 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
355 -fcx-limited-range @gol
356 -fdata-sections -fdce -fdelayed-branch @gol
357 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
358 -fdevirtualize-at-ltrans -fdse @gol
359 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
360 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
361 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
362 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
363 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
364 -fif-conversion2 -findirect-inlining @gol
365 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
366 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment -fipa-bit-cp @gol
367 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
368 -fira-algorithm=@var{algorithm} @gol
369 -fira-region=@var{region} -fira-hoist-pressure @gol
370 -fira-loop-pressure -fno-ira-share-save-slots @gol
371 -fno-ira-share-spill-slots @gol
372 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
373 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
374 -fkeep-static-consts -flive-range-shrinkage @gol
375 -floop-block -floop-interchange -floop-strip-mine @gol
376 -floop-unroll-and-jam -floop-nest-optimize @gol
377 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
378 -flto-partition=@var{alg} -fmerge-all-constants @gol
379 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
380 -fmove-loop-invariants -fno-branch-count-reg @gol
381 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
382 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
383 -fno-peephole2 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
384 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
385 -fomit-frame-pointer -foptimize-sibling-calls @gol
386 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
387 -fprefetch-loop-arrays -fprintf-return-value @gol
388 -fprofile-correction @gol
389 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
390 -fprofile-reorder-functions @gol
391 -freciprocal-math -free -frename-registers -freorder-blocks @gol
392 -freorder-blocks-algorithm=@var{algorithm} @gol
393 -freorder-blocks-and-partition -freorder-functions @gol
394 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
395 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
396 -fsched-spec-load -fsched-spec-load-dangerous @gol
397 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
398 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
399 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
400 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
401 -fschedule-fusion @gol
402 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
403 -fselective-scheduling -fselective-scheduling2 @gol
404 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
405 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
406 -fsignaling-nans @gol
407 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
409 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
410 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
411 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
412 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
413 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
414 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting -ftree-loop-if-convert @gol
415 -ftree-loop-if-convert-stores -ftree-loop-im @gol
416 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
417 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
418 -ftree-loop-vectorize @gol
419 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
420 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
421 -ftree-switch-conversion -ftree-tail-merge @gol
422 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
423 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
424 -funsafe-math-optimizations -funswitch-loops @gol
425 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
426 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
427 --param @var{name}=@var{value}
428 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
430 @item Program Instrumentation Options
431 @xref{Instrumentation Options,,Program Instrumentation Options}.
432 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
433 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
434 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
435 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
436 -fsanitize-undefined-trap-on-error -fbounds-check @gol
437 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
438 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
439 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
440 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
441 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
442 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
443 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
444 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
445 -fchkp-use-wrappers @gol
446 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
447 -fstack-protector-explicit -fstack-check @gol
448 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
449 -fno-stack-limit -fsplit-stack @gol
450 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
451 -fvtv-counts -fvtv-debug @gol
452 -finstrument-functions @gol
453 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
454 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
456 @item Preprocessor Options
457 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
458 @gccoptlist{-A@var{question}=@var{answer} @gol
459 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
460 -C -dD -dI -dM -dN @gol
461 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
462 -idirafter @var{dir} @gol
463 -include @var{file} -imacros @var{file} @gol
464 -iprefix @var{file} -iwithprefix @var{dir} @gol
465 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
466 -imultilib @var{dir} -isysroot @var{dir} @gol
467 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
468 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
469 -remap -trigraphs -undef -U@var{macro} @gol
470 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
472 @item Assembler Option
473 @xref{Assembler Options,,Passing Options to the Assembler}.
474 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
477 @xref{Link Options,,Options for Linking}.
478 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
479 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
480 -s -static -static-libgcc -static-libstdc++ @gol
481 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
482 -static-libmpx -static-libmpxwrappers @gol
483 -shared -shared-libgcc -symbolic @gol
484 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
485 -u @var{symbol} -z @var{keyword}}
487 @item Directory Options
488 @xref{Directory Options,,Options for Directory Search}.
489 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
490 -iquote@var{dir} -L@var{dir} -no-canonical-prefixes -I- @gol
491 --sysroot=@var{dir} --no-sysroot-suffix}
493 @item Code Generation Options
494 @xref{Code Gen Options,,Options for Code Generation Conventions}.
495 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
496 -ffixed-@var{reg} -fexceptions @gol
497 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
498 -fasynchronous-unwind-tables @gol
500 -finhibit-size-directive -fno-common -fno-ident @gol
501 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
502 -fno-jump-tables @gol
503 -frecord-gcc-switches @gol
504 -freg-struct-return -fshort-enums -fshort-wchar @gol
505 -fverbose-asm -fpack-struct[=@var{n}] @gol
506 -fleading-underscore -ftls-model=@var{model} @gol
507 -fstack-reuse=@var{reuse_level} @gol
508 -ftrampolines -ftrapv -fwrapv @gol
509 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
510 -fstrict-volatile-bitfields -fsync-libcalls}
512 @item Developer Options
513 @xref{Developer Options,,GCC Developer Options}.
514 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
515 -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
516 -fdbg-cnt=@var{counter-value-list} @gol
517 -fdisable-ipa-@var{pass_name} @gol
518 -fdisable-rtl-@var{pass_name} @gol
519 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
520 -fdisable-tree-@var{pass_name} @gol
521 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
522 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
523 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
524 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
525 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
527 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
528 -fdump-statistics @gol
530 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
531 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
532 -fdump-tree-cfg -fdump-tree-alias @gol
534 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
535 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
536 -fdump-tree-gimple@r{[}-raw@r{]} @gol
537 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
538 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
539 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
540 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
541 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
542 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
543 -fdump-tree-nrv -fdump-tree-vect @gol
544 -fdump-tree-sink @gol
545 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
546 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
547 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
548 -fdump-tree-vtable-verify @gol
549 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
550 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
551 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
552 -fdump-final-insns=@var{file} @gol
553 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
554 -fenable-@var{kind}-@var{pass} @gol
555 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
556 -fira-verbose=@var{n} @gol
557 -flto-report -flto-report-wpa -fmem-report-wpa @gol
558 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
559 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
560 -fprofile-report @gol
561 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
562 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
563 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
564 -fvar-tracking-assignments-toggle -gtoggle @gol
565 -print-file-name=@var{library} -print-libgcc-file-name @gol
566 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
567 -print-prog-name=@var{program} -print-search-dirs -Q @gol
568 -print-sysroot -print-sysroot-headers-suffix @gol
569 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
571 @item Machine-Dependent Options
572 @xref{Submodel Options,,Machine-Dependent Options}.
573 @c This list is ordered alphanumerically by subsection name.
574 @c Try and put the significant identifier (CPU or system) first,
575 @c so users have a clue at guessing where the ones they want will be.
577 @emph{AArch64 Options}
578 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
579 -mgeneral-regs-only @gol
580 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
582 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
583 -mtls-dialect=desc -mtls-dialect=traditional @gol
584 -mtls-size=@var{size} @gol
585 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
586 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
587 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
588 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
589 -mlow-precision-div -mno-low-precision-div @gol
590 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
592 @emph{Adapteva Epiphany Options}
593 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
594 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
595 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
596 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
597 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
598 -msplit-vecmove-early -m1reg-@var{reg}}
601 @gccoptlist{-mbarrel-shifter @gol
602 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
603 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
604 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
605 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
606 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
607 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
608 -mlong-calls -mmedium-calls -msdata @gol
609 -mucb-mcount -mvolatile-cache -mtp-regno=@var{regno} @gol
610 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
611 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
612 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
613 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
614 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
615 -mtune=@var{cpu} -mmultcost=@var{num} @gol
616 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
617 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
620 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
621 -mabi=@var{name} @gol
622 -mapcs-stack-check -mno-apcs-stack-check @gol
623 -mapcs-float -mno-apcs-float @gol
624 -mapcs-reentrant -mno-apcs-reentrant @gol
625 -msched-prolog -mno-sched-prolog @gol
626 -mlittle-endian -mbig-endian @gol
627 -mfloat-abi=@var{name} @gol
628 -mfp16-format=@var{name}
629 -mthumb-interwork -mno-thumb-interwork @gol
630 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
631 -mtune=@var{name} -mprint-tune-info @gol
632 -mstructure-size-boundary=@var{n} @gol
633 -mabort-on-noreturn @gol
634 -mlong-calls -mno-long-calls @gol
635 -msingle-pic-base -mno-single-pic-base @gol
636 -mpic-register=@var{reg} @gol
637 -mnop-fun-dllimport @gol
638 -mpoke-function-name @gol
640 -mtpcs-frame -mtpcs-leaf-frame @gol
641 -mcaller-super-interworking -mcallee-super-interworking @gol
642 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
643 -mword-relocations @gol
644 -mfix-cortex-m3-ldrd @gol
645 -munaligned-access @gol
646 -mneon-for-64bits @gol
647 -mslow-flash-data @gol
648 -masm-syntax-unified @gol
653 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
654 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
655 -mrelax -mrmw -mstrict-X -mtiny-stack -mfract-convert-truncate -nodevicelib @gol
656 -Waddr-space-convert -Wmisspelled-isr}
658 @emph{Blackfin Options}
659 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
660 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
661 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
662 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
663 -mno-id-shared-library -mshared-library-id=@var{n} @gol
664 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
665 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
666 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
670 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
671 -msim -msdata=@var{sdata-type}}
674 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
675 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
676 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
677 -mstack-align -mdata-align -mconst-align @gol
678 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
679 -melf -maout -melinux -mlinux -sim -sim2 @gol
680 -mmul-bug-workaround -mno-mul-bug-workaround}
683 @gccoptlist{-mmac @gol
684 -mcr16cplus -mcr16c @gol
685 -msim -mint32 -mbit-ops
686 -mdata-model=@var{model}}
688 @emph{Darwin Options}
689 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
690 -arch_only -bind_at_load -bundle -bundle_loader @gol
691 -client_name -compatibility_version -current_version @gol
693 -dependency-file -dylib_file -dylinker_install_name @gol
694 -dynamic -dynamiclib -exported_symbols_list @gol
695 -filelist -flat_namespace -force_cpusubtype_ALL @gol
696 -force_flat_namespace -headerpad_max_install_names @gol
698 -image_base -init -install_name -keep_private_externs @gol
699 -multi_module -multiply_defined -multiply_defined_unused @gol
700 -noall_load -no_dead_strip_inits_and_terms @gol
701 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
702 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
703 -private_bundle -read_only_relocs -sectalign @gol
704 -sectobjectsymbols -whyload -seg1addr @gol
705 -sectcreate -sectobjectsymbols -sectorder @gol
706 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
707 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
708 -segprot -segs_read_only_addr -segs_read_write_addr @gol
709 -single_module -static -sub_library -sub_umbrella @gol
710 -twolevel_namespace -umbrella -undefined @gol
711 -unexported_symbols_list -weak_reference_mismatches @gol
712 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
713 -mkernel -mone-byte-bool}
715 @emph{DEC Alpha Options}
716 @gccoptlist{-mno-fp-regs -msoft-float @gol
717 -mieee -mieee-with-inexact -mieee-conformant @gol
718 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
719 -mtrap-precision=@var{mode} -mbuild-constants @gol
720 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
721 -mbwx -mmax -mfix -mcix @gol
722 -mfloat-vax -mfloat-ieee @gol
723 -mexplicit-relocs -msmall-data -mlarge-data @gol
724 -msmall-text -mlarge-text @gol
725 -mmemory-latency=@var{time}}
728 @gccoptlist{-msmall-model -mno-lsim}
731 @gccoptlist{-msim -mlra -mnodiv}
734 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
735 -mhard-float -msoft-float @gol
736 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
737 -mdouble -mno-double @gol
738 -mmedia -mno-media -mmuladd -mno-muladd @gol
739 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
740 -mlinked-fp -mlong-calls -malign-labels @gol
741 -mlibrary-pic -macc-4 -macc-8 @gol
742 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
743 -moptimize-membar -mno-optimize-membar @gol
744 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
745 -mvliw-branch -mno-vliw-branch @gol
746 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
747 -mno-nested-cond-exec -mtomcat-stats @gol
751 @emph{GNU/Linux Options}
752 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
753 -tno-android-cc -tno-android-ld}
755 @emph{H8/300 Options}
756 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
759 @gccoptlist{-march=@var{architecture-type} @gol
760 -mdisable-fpregs -mdisable-indexing @gol
761 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
762 -mfixed-range=@var{register-range} @gol
763 -mjump-in-delay -mlinker-opt -mlong-calls @gol
764 -mlong-load-store -mno-disable-fpregs @gol
765 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
766 -mno-jump-in-delay -mno-long-load-store @gol
767 -mno-portable-runtime -mno-soft-float @gol
768 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
769 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
770 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
771 -munix=@var{unix-std} -nolibdld -static -threads}
774 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
775 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
776 -mconstant-gp -mauto-pic -mfused-madd @gol
777 -minline-float-divide-min-latency @gol
778 -minline-float-divide-max-throughput @gol
779 -mno-inline-float-divide @gol
780 -minline-int-divide-min-latency @gol
781 -minline-int-divide-max-throughput @gol
782 -mno-inline-int-divide @gol
783 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
784 -mno-inline-sqrt @gol
785 -mdwarf2-asm -mearly-stop-bits @gol
786 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
787 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
788 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
789 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
790 -msched-spec-ldc -msched-spec-control-ldc @gol
791 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
792 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
793 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
794 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
797 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
798 -msign-extend-enabled -muser-enabled}
800 @emph{M32R/D Options}
801 @gccoptlist{-m32r2 -m32rx -m32r @gol
803 -malign-loops -mno-align-loops @gol
804 -missue-rate=@var{number} @gol
805 -mbranch-cost=@var{number} @gol
806 -mmodel=@var{code-size-model-type} @gol
807 -msdata=@var{sdata-type} @gol
808 -mno-flush-func -mflush-func=@var{name} @gol
809 -mno-flush-trap -mflush-trap=@var{number} @gol
813 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
815 @emph{M680x0 Options}
816 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
817 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
818 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
819 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
820 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
821 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
822 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
823 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
827 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
828 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
829 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
830 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
831 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
834 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
835 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
836 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
837 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
840 @emph{MicroBlaze Options}
841 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
842 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
843 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
844 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
845 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
848 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
849 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
850 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
851 -mips16 -mno-mips16 -mflip-mips16 @gol
852 -minterlink-compressed -mno-interlink-compressed @gol
853 -minterlink-mips16 -mno-interlink-mips16 @gol
854 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
855 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
856 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
857 -mno-float -msingle-float -mdouble-float @gol
858 -modd-spreg -mno-odd-spreg @gol
859 -mabs=@var{mode} -mnan=@var{encoding} @gol
860 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
863 -mvirt -mno-virt @gol
865 -mmicromips -mno-micromips @gol
867 -mfpu=@var{fpu-type} @gol
868 -msmartmips -mno-smartmips @gol
869 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
870 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
871 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
872 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
873 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
874 -membedded-data -mno-embedded-data @gol
875 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
876 -mcode-readable=@var{setting} @gol
877 -msplit-addresses -mno-split-addresses @gol
878 -mexplicit-relocs -mno-explicit-relocs @gol
879 -mcheck-zero-division -mno-check-zero-division @gol
880 -mdivide-traps -mdivide-breaks @gol
881 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
882 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
883 -mfix-24k -mno-fix-24k @gol
884 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
885 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
886 -mfix-vr4120 -mno-fix-vr4120 @gol
887 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
888 -mflush-func=@var{func} -mno-flush-func @gol
889 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
890 -mcompact-branches=@var{policy} @gol
891 -mfp-exceptions -mno-fp-exceptions @gol
892 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
893 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
894 -mframe-header-opt -mno-frame-header-opt}
897 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
898 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
899 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
900 -mno-base-addresses -msingle-exit -mno-single-exit}
902 @emph{MN10300 Options}
903 @gccoptlist{-mmult-bug -mno-mult-bug @gol
904 -mno-am33 -mam33 -mam33-2 -mam34 @gol
905 -mtune=@var{cpu-type} @gol
906 -mreturn-pointer-on-d0 @gol
907 -mno-crt0 -mrelax -mliw -msetlb}
910 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
912 @emph{MSP430 Options}
913 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
915 -mcode-region= -mdata-region= @gol
916 -msilicon-errata= -msilicon-errata-warn= @gol
920 @gccoptlist{-mbig-endian -mlittle-endian @gol
921 -mreduced-regs -mfull-regs @gol
922 -mcmov -mno-cmov @gol
923 -mperf-ext -mno-perf-ext @gol
924 -mv3push -mno-v3push @gol
925 -m16bit -mno-16bit @gol
926 -misr-vector-size=@var{num} @gol
927 -mcache-block-size=@var{num} @gol
928 -march=@var{arch} @gol
929 -mcmodel=@var{code-model} @gol
932 @emph{Nios II Options}
933 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
935 -mno-bypass-cache -mbypass-cache @gol
936 -mno-cache-volatile -mcache-volatile @gol
937 -mno-fast-sw-div -mfast-sw-div @gol
938 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
939 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
940 -mcustom-fpu-cfg=@var{name} @gol
941 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
942 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
944 @emph{Nvidia PTX Options}
945 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
947 @emph{PDP-11 Options}
948 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
949 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
950 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
951 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
952 -mbranch-expensive -mbranch-cheap @gol
953 -munix-asm -mdec-asm}
955 @emph{picoChip Options}
956 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
957 -msymbol-as-address -mno-inefficient-warnings}
959 @emph{PowerPC Options}
960 See RS/6000 and PowerPC Options.
963 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
964 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
965 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
967 @emph{RS/6000 and PowerPC Options}
968 @gccoptlist{-mcpu=@var{cpu-type} @gol
969 -mtune=@var{cpu-type} @gol
970 -mcmodel=@var{code-model} @gol
972 -maltivec -mno-altivec @gol
973 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
974 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
975 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
976 -mfprnd -mno-fprnd @gol
977 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
978 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
979 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
980 -malign-power -malign-natural @gol
981 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
982 -msingle-float -mdouble-float -msimple-fpu @gol
983 -mstring -mno-string -mupdate -mno-update @gol
984 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
985 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
986 -mstrict-align -mno-strict-align -mrelocatable @gol
987 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
988 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
989 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
990 -mprioritize-restricted-insns=@var{priority} @gol
991 -msched-costly-dep=@var{dependence_type} @gol
992 -minsert-sched-nops=@var{scheme} @gol
993 -mcall-sysv -mcall-netbsd @gol
994 -maix-struct-return -msvr4-struct-return @gol
995 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
996 -mblock-move-inline-limit=@var{num} @gol
997 -misel -mno-isel @gol
998 -misel=yes -misel=no @gol
1000 -mspe=yes -mspe=no @gol
1002 -mgen-cell-microcode -mwarn-cell-microcode @gol
1003 -mvrsave -mno-vrsave @gol
1004 -mmulhw -mno-mulhw @gol
1005 -mdlmzb -mno-dlmzb @gol
1006 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1007 -mprototype -mno-prototype @gol
1008 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1009 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
1010 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1011 -mno-recip-precision @gol
1012 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1013 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1014 -msave-toc-indirect -mno-save-toc-indirect @gol
1015 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1016 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1017 -mquad-memory -mno-quad-memory @gol
1018 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1019 -mcompat-align-parm -mno-compat-align-parm @gol
1020 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1021 -mupper-regs-di -mno-upper-regs-di @gol
1022 -mupper-regs -mno-upper-regs @gol
1023 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1024 -mgnu-attribute -mno-gnu-attribute @gol
1028 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1030 -mbig-endian-data -mlittle-endian-data @gol
1033 -mas100-syntax -mno-as100-syntax@gol
1035 -mmax-constant-size=@gol
1038 -mallow-string-insns -mno-allow-string-insns@gol
1040 -mno-warn-multiple-fast-interrupts@gol
1041 -msave-acc-in-interrupts}
1043 @emph{S/390 and zSeries Options}
1044 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1045 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1046 -mlong-double-64 -mlong-double-128 @gol
1047 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1048 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1049 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1050 -mhtm -mvx -mzvector @gol
1051 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1052 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1053 -mhotpatch=@var{halfwords},@var{halfwords}}
1055 @emph{Score Options}
1056 @gccoptlist{-meb -mel @gol
1060 -mscore5 -mscore5u -mscore7 -mscore7d}
1063 @gccoptlist{-m1 -m2 -m2e @gol
1064 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1066 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1067 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1068 -mb -ml -mdalign -mrelax @gol
1069 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1070 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1071 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1072 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1073 -maccumulate-outgoing-args @gol
1074 -matomic-model=@var{atomic-model} @gol
1075 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1076 -mcbranch-force-delay-slot @gol
1077 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1078 -mpretend-cmove -mtas}
1080 @emph{Solaris 2 Options}
1081 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1084 @emph{SPARC Options}
1085 @gccoptlist{-mcpu=@var{cpu-type} @gol
1086 -mtune=@var{cpu-type} @gol
1087 -mcmodel=@var{code-model} @gol
1088 -mmemory-model=@var{mem-model} @gol
1089 -m32 -m64 -mapp-regs -mno-app-regs @gol
1090 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1091 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1092 -mhard-quad-float -msoft-quad-float @gol
1093 -mstack-bias -mno-stack-bias @gol
1094 -mstd-struct-return -mno-std-struct-return @gol
1095 -munaligned-doubles -mno-unaligned-doubles @gol
1096 -muser-mode -mno-user-mode @gol
1097 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1098 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1099 -mcbcond -mno-cbcond -mfmaf -mno-fmaf @gol
1100 -mpopc -mno-popc -msubxc -mno-subxc@gol
1101 -mfix-at697f -mfix-ut699}
1104 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1105 -msafe-dma -munsafe-dma @gol
1107 -msmall-mem -mlarge-mem -mstdmain @gol
1108 -mfixed-range=@var{register-range} @gol
1110 -maddress-space-conversion -mno-address-space-conversion @gol
1111 -mcache-size=@var{cache-size} @gol
1112 -matomic-updates -mno-atomic-updates}
1114 @emph{System V Options}
1115 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1117 @emph{TILE-Gx Options}
1118 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1119 -mcmodel=@var{code-model}}
1121 @emph{TILEPro Options}
1122 @gccoptlist{-mcpu=@var{cpu} -m32}
1125 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1126 -mprolog-function -mno-prolog-function -mspace @gol
1127 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1128 -mapp-regs -mno-app-regs @gol
1129 -mdisable-callt -mno-disable-callt @gol
1130 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1131 -mv850e -mv850 -mv850e3v5 @gol
1142 @gccoptlist{-mg -mgnu -munix}
1144 @emph{Visium Options}
1145 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1146 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1149 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1150 -mpointer-size=@var{size}}
1152 @emph{VxWorks Options}
1153 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1154 -Xbind-lazy -Xbind-now}
1157 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1158 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1159 -mfpmath=@var{unit} @gol
1160 -masm=@var{dialect} -mno-fancy-math-387 @gol
1161 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1162 -mno-wide-multiply -mrtd -malign-double @gol
1163 -mpreferred-stack-boundary=@var{num} @gol
1164 -mincoming-stack-boundary=@var{num} @gol
1165 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1166 -mrecip -mrecip=@var{opt} @gol
1167 -mvzeroupper -mprefer-avx128 @gol
1168 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1169 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1170 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1171 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1172 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1173 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1174 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1175 -mpku -mthreads @gol
1176 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1177 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1178 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1179 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1180 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1181 -mregparm=@var{num} -msseregparm @gol
1182 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1183 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1184 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1185 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1186 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1187 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1188 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1189 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1190 -mmitigate-rop -mgeneral-regs-only}
1192 @emph{x86 Windows Options}
1193 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1194 -mnop-fun-dllimport -mthread @gol
1195 -municode -mwin32 -mwindows -fno-set-stack-executable}
1197 @emph{Xstormy16 Options}
1200 @emph{Xtensa Options}
1201 @gccoptlist{-mconst16 -mno-const16 @gol
1202 -mfused-madd -mno-fused-madd @gol
1204 -mserialize-volatile -mno-serialize-volatile @gol
1205 -mtext-section-literals -mno-text-section-literals @gol
1206 -mauto-litpools -mno-auto-litpools @gol
1207 -mtarget-align -mno-target-align @gol
1208 -mlongcalls -mno-longcalls}
1210 @emph{zSeries Options}
1211 See S/390 and zSeries Options.
1215 @node Overall Options
1216 @section Options Controlling the Kind of Output
1218 Compilation can involve up to four stages: preprocessing, compilation
1219 proper, assembly and linking, always in that order. GCC is capable of
1220 preprocessing and compiling several files either into several
1221 assembler input files, or into one assembler input file; then each
1222 assembler input file produces an object file, and linking combines all
1223 the object files (those newly compiled, and those specified as input)
1224 into an executable file.
1226 @cindex file name suffix
1227 For any given input file, the file name suffix determines what kind of
1228 compilation is done:
1232 C source code that must be preprocessed.
1235 C source code that should not be preprocessed.
1238 C++ source code that should not be preprocessed.
1241 Objective-C source code. Note that you must link with the @file{libobjc}
1242 library to make an Objective-C program work.
1245 Objective-C source code that should not be preprocessed.
1249 Objective-C++ source code. Note that you must link with the @file{libobjc}
1250 library to make an Objective-C++ program work. Note that @samp{.M} refers
1251 to a literal capital M@.
1253 @item @var{file}.mii
1254 Objective-C++ source code that should not be preprocessed.
1257 C, C++, Objective-C or Objective-C++ header file to be turned into a
1258 precompiled header (default), or C, C++ header file to be turned into an
1259 Ada spec (via the @option{-fdump-ada-spec} switch).
1262 @itemx @var{file}.cp
1263 @itemx @var{file}.cxx
1264 @itemx @var{file}.cpp
1265 @itemx @var{file}.CPP
1266 @itemx @var{file}.c++
1268 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1269 the last two letters must both be literally @samp{x}. Likewise,
1270 @samp{.C} refers to a literal capital C@.
1274 Objective-C++ source code that must be preprocessed.
1276 @item @var{file}.mii
1277 Objective-C++ source code that should not be preprocessed.
1281 @itemx @var{file}.hp
1282 @itemx @var{file}.hxx
1283 @itemx @var{file}.hpp
1284 @itemx @var{file}.HPP
1285 @itemx @var{file}.h++
1286 @itemx @var{file}.tcc
1287 C++ header file to be turned into a precompiled header or Ada spec.
1290 @itemx @var{file}.for
1291 @itemx @var{file}.ftn
1292 Fixed form Fortran source code that should not be preprocessed.
1295 @itemx @var{file}.FOR
1296 @itemx @var{file}.fpp
1297 @itemx @var{file}.FPP
1298 @itemx @var{file}.FTN
1299 Fixed form Fortran source code that must be preprocessed (with the traditional
1302 @item @var{file}.f90
1303 @itemx @var{file}.f95
1304 @itemx @var{file}.f03
1305 @itemx @var{file}.f08
1306 Free form Fortran source code that should not be preprocessed.
1308 @item @var{file}.F90
1309 @itemx @var{file}.F95
1310 @itemx @var{file}.F03
1311 @itemx @var{file}.F08
1312 Free form Fortran source code that must be preprocessed (with the
1313 traditional preprocessor).
1318 @c FIXME: Descriptions of Java file types.
1324 @item @var{file}.ads
1325 Ada source code file that contains a library unit declaration (a
1326 declaration of a package, subprogram, or generic, or a generic
1327 instantiation), or a library unit renaming declaration (a package,
1328 generic, or subprogram renaming declaration). Such files are also
1331 @item @var{file}.adb
1332 Ada source code file containing a library unit body (a subprogram or
1333 package body). Such files are also called @dfn{bodies}.
1335 @c GCC also knows about some suffixes for languages not yet included:
1346 @itemx @var{file}.sx
1347 Assembler code that must be preprocessed.
1350 An object file to be fed straight into linking.
1351 Any file name with no recognized suffix is treated this way.
1355 You can specify the input language explicitly with the @option{-x} option:
1358 @item -x @var{language}
1359 Specify explicitly the @var{language} for the following input files
1360 (rather than letting the compiler choose a default based on the file
1361 name suffix). This option applies to all following input files until
1362 the next @option{-x} option. Possible values for @var{language} are:
1364 c c-header cpp-output
1365 c++ c++-header c++-cpp-output
1366 objective-c objective-c-header objective-c-cpp-output
1367 objective-c++ objective-c++-header objective-c++-cpp-output
1368 assembler assembler-with-cpp
1370 f77 f77-cpp-input f95 f95-cpp-input
1376 Turn off any specification of a language, so that subsequent files are
1377 handled according to their file name suffixes (as they are if @option{-x}
1378 has not been used at all).
1381 If you only want some of the stages of compilation, you can use
1382 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1383 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1384 @command{gcc} is to stop. Note that some combinations (for example,
1385 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1390 Compile or assemble the source files, but do not link. The linking
1391 stage simply is not done. The ultimate output is in the form of an
1392 object file for each source file.
1394 By default, the object file name for a source file is made by replacing
1395 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1397 Unrecognized input files, not requiring compilation or assembly, are
1402 Stop after the stage of compilation proper; do not assemble. The output
1403 is in the form of an assembler code file for each non-assembler input
1406 By default, the assembler file name for a source file is made by
1407 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1409 Input files that don't require compilation are ignored.
1413 Stop after the preprocessing stage; do not run the compiler proper. The
1414 output is in the form of preprocessed source code, which is sent to the
1417 Input files that don't require preprocessing are ignored.
1419 @cindex output file option
1422 Place output in file @var{file}. This applies to whatever
1423 sort of output is being produced, whether it be an executable file,
1424 an object file, an assembler file or preprocessed C code.
1426 If @option{-o} is not specified, the default is to put an executable
1427 file in @file{a.out}, the object file for
1428 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1429 assembler file in @file{@var{source}.s}, a precompiled header file in
1430 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1435 Print (on standard error output) the commands executed to run the stages
1436 of compilation. Also print the version number of the compiler driver
1437 program and of the preprocessor and the compiler proper.
1441 Like @option{-v} except the commands are not executed and arguments
1442 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1443 This is useful for shell scripts to capture the driver-generated command lines.
1447 Print (on the standard output) a description of the command-line options
1448 understood by @command{gcc}. If the @option{-v} option is also specified
1449 then @option{--help} is also passed on to the various processes
1450 invoked by @command{gcc}, so that they can display the command-line options
1451 they accept. If the @option{-Wextra} option has also been specified
1452 (prior to the @option{--help} option), then command-line options that
1453 have no documentation associated with them are also displayed.
1456 @opindex target-help
1457 Print (on the standard output) a description of target-specific command-line
1458 options for each tool. For some targets extra target-specific
1459 information may also be printed.
1461 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1462 Print (on the standard output) a description of the command-line
1463 options understood by the compiler that fit into all specified classes
1464 and qualifiers. These are the supported classes:
1467 @item @samp{optimizers}
1468 Display all of the optimization options supported by the
1471 @item @samp{warnings}
1472 Display all of the options controlling warning messages
1473 produced by the compiler.
1476 Display target-specific options. Unlike the
1477 @option{--target-help} option however, target-specific options of the
1478 linker and assembler are not displayed. This is because those
1479 tools do not currently support the extended @option{--help=} syntax.
1482 Display the values recognized by the @option{--param}
1485 @item @var{language}
1486 Display the options supported for @var{language}, where
1487 @var{language} is the name of one of the languages supported in this
1491 Display the options that are common to all languages.
1494 These are the supported qualifiers:
1497 @item @samp{undocumented}
1498 Display only those options that are undocumented.
1501 Display options taking an argument that appears after an equal
1502 sign in the same continuous piece of text, such as:
1503 @samp{--help=target}.
1505 @item @samp{separate}
1506 Display options taking an argument that appears as a separate word
1507 following the original option, such as: @samp{-o output-file}.
1510 Thus for example to display all the undocumented target-specific
1511 switches supported by the compiler, use:
1514 --help=target,undocumented
1517 The sense of a qualifier can be inverted by prefixing it with the
1518 @samp{^} character, so for example to display all binary warning
1519 options (i.e., ones that are either on or off and that do not take an
1520 argument) that have a description, use:
1523 --help=warnings,^joined,^undocumented
1526 The argument to @option{--help=} should not consist solely of inverted
1529 Combining several classes is possible, although this usually
1530 restricts the output so much that there is nothing to display. One
1531 case where it does work, however, is when one of the classes is
1532 @var{target}. For example, to display all the target-specific
1533 optimization options, use:
1536 --help=target,optimizers
1539 The @option{--help=} option can be repeated on the command line. Each
1540 successive use displays its requested class of options, skipping
1541 those that have already been displayed.
1543 If the @option{-Q} option appears on the command line before the
1544 @option{--help=} option, then the descriptive text displayed by
1545 @option{--help=} is changed. Instead of describing the displayed
1546 options, an indication is given as to whether the option is enabled,
1547 disabled or set to a specific value (assuming that the compiler
1548 knows this at the point where the @option{--help=} option is used).
1550 Here is a truncated example from the ARM port of @command{gcc}:
1553 % gcc -Q -mabi=2 --help=target -c
1554 The following options are target specific:
1556 -mabort-on-noreturn [disabled]
1560 The output is sensitive to the effects of previous command-line
1561 options, so for example it is possible to find out which optimizations
1562 are enabled at @option{-O2} by using:
1565 -Q -O2 --help=optimizers
1568 Alternatively you can discover which binary optimizations are enabled
1569 by @option{-O3} by using:
1572 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1573 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1574 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1579 Display the version number and copyrights of the invoked GCC@.
1581 @item -pass-exit-codes
1582 @opindex pass-exit-codes
1583 Normally the @command{gcc} program exits with the code of 1 if any
1584 phase of the compiler returns a non-success return code. If you specify
1585 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1586 the numerically highest error produced by any phase returning an error
1587 indication. The C, C++, and Fortran front ends return 4 if an internal
1588 compiler error is encountered.
1592 Use pipes rather than temporary files for communication between the
1593 various stages of compilation. This fails to work on some systems where
1594 the assembler is unable to read from a pipe; but the GNU assembler has
1597 @item -specs=@var{file}
1599 Process @var{file} after the compiler reads in the standard @file{specs}
1600 file, in order to override the defaults which the @command{gcc} driver
1601 program uses when determining what switches to pass to @command{cc1},
1602 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1603 @option{-specs=@var{file}} can be specified on the command line, and they
1604 are processed in order, from left to right. @xref{Spec Files}, for
1605 information about the format of the @var{file}.
1609 Invoke all subcommands under a wrapper program. The name of the
1610 wrapper program and its parameters are passed as a comma separated
1614 gcc -c t.c -wrapper gdb,--args
1618 This invokes all subprograms of @command{gcc} under
1619 @samp{gdb --args}, thus the invocation of @command{cc1} is
1620 @samp{gdb --args cc1 @dots{}}.
1622 @item -fplugin=@var{name}.so
1624 Load the plugin code in file @var{name}.so, assumed to be a
1625 shared object to be dlopen'd by the compiler. The base name of
1626 the shared object file is used to identify the plugin for the
1627 purposes of argument parsing (See
1628 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1629 Each plugin should define the callback functions specified in the
1632 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1633 @opindex fplugin-arg
1634 Define an argument called @var{key} with a value of @var{value}
1635 for the plugin called @var{name}.
1637 @item -fdump-ada-spec@r{[}-slim@r{]}
1638 @opindex fdump-ada-spec
1639 For C and C++ source and include files, generate corresponding Ada specs.
1640 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1641 GNAT User's Guide}, which provides detailed documentation on this feature.
1643 @item -fada-spec-parent=@var{unit}
1644 @opindex fada-spec-parent
1645 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1646 Ada specs as child units of parent @var{unit}.
1648 @item -fdump-go-spec=@var{file}
1649 @opindex fdump-go-spec
1650 For input files in any language, generate corresponding Go
1651 declarations in @var{file}. This generates Go @code{const},
1652 @code{type}, @code{var}, and @code{func} declarations which may be a
1653 useful way to start writing a Go interface to code written in some
1656 @include @value{srcdir}/../libiberty/at-file.texi
1660 @section Compiling C++ Programs
1662 @cindex suffixes for C++ source
1663 @cindex C++ source file suffixes
1664 C++ source files conventionally use one of the suffixes @samp{.C},
1665 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1666 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1667 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1668 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1669 files with these names and compiles them as C++ programs even if you
1670 call the compiler the same way as for compiling C programs (usually
1671 with the name @command{gcc}).
1675 However, the use of @command{gcc} does not add the C++ library.
1676 @command{g++} is a program that calls GCC and automatically specifies linking
1677 against the C++ library. It treats @samp{.c},
1678 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1679 files unless @option{-x} is used. This program is also useful when
1680 precompiling a C header file with a @samp{.h} extension for use in C++
1681 compilations. On many systems, @command{g++} is also installed with
1682 the name @command{c++}.
1684 @cindex invoking @command{g++}
1685 When you compile C++ programs, you may specify many of the same
1686 command-line options that you use for compiling programs in any
1687 language; or command-line options meaningful for C and related
1688 languages; or options that are meaningful only for C++ programs.
1689 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1690 explanations of options for languages related to C@.
1691 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1692 explanations of options that are meaningful only for C++ programs.
1694 @node C Dialect Options
1695 @section Options Controlling C Dialect
1696 @cindex dialect options
1697 @cindex language dialect options
1698 @cindex options, dialect
1700 The following options control the dialect of C (or languages derived
1701 from C, such as C++, Objective-C and Objective-C++) that the compiler
1705 @cindex ANSI support
1709 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1710 equivalent to @option{-std=c++98}.
1712 This turns off certain features of GCC that are incompatible with ISO
1713 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1714 such as the @code{asm} and @code{typeof} keywords, and
1715 predefined macros such as @code{unix} and @code{vax} that identify the
1716 type of system you are using. It also enables the undesirable and
1717 rarely used ISO trigraph feature. For the C compiler,
1718 it disables recognition of C++ style @samp{//} comments as well as
1719 the @code{inline} keyword.
1721 The alternate keywords @code{__asm__}, @code{__extension__},
1722 @code{__inline__} and @code{__typeof__} continue to work despite
1723 @option{-ansi}. You would not want to use them in an ISO C program, of
1724 course, but it is useful to put them in header files that might be included
1725 in compilations done with @option{-ansi}. Alternate predefined macros
1726 such as @code{__unix__} and @code{__vax__} are also available, with or
1727 without @option{-ansi}.
1729 The @option{-ansi} option does not cause non-ISO programs to be
1730 rejected gratuitously. For that, @option{-Wpedantic} is required in
1731 addition to @option{-ansi}. @xref{Warning Options}.
1733 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1734 option is used. Some header files may notice this macro and refrain
1735 from declaring certain functions or defining certain macros that the
1736 ISO standard doesn't call for; this is to avoid interfering with any
1737 programs that might use these names for other things.
1739 Functions that are normally built in but do not have semantics
1740 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1741 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1742 built-in functions provided by GCC}, for details of the functions
1747 Determine the language standard. @xref{Standards,,Language Standards
1748 Supported by GCC}, for details of these standard versions. This option
1749 is currently only supported when compiling C or C++.
1751 The compiler can accept several base standards, such as @samp{c90} or
1752 @samp{c++98}, and GNU dialects of those standards, such as
1753 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1754 compiler accepts all programs following that standard plus those
1755 using GNU extensions that do not contradict it. For example,
1756 @option{-std=c90} turns off certain features of GCC that are
1757 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1758 keywords, but not other GNU extensions that do not have a meaning in
1759 ISO C90, such as omitting the middle term of a @code{?:}
1760 expression. On the other hand, when a GNU dialect of a standard is
1761 specified, all features supported by the compiler are enabled, even when
1762 those features change the meaning of the base standard. As a result, some
1763 strict-conforming programs may be rejected. The particular standard
1764 is used by @option{-Wpedantic} to identify which features are GNU
1765 extensions given that version of the standard. For example
1766 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1767 comments, while @option{-std=gnu99 -Wpedantic} does not.
1769 A value for this option must be provided; possible values are
1775 Support all ISO C90 programs (certain GNU extensions that conflict
1776 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1778 @item iso9899:199409
1779 ISO C90 as modified in amendment 1.
1785 ISO C99. This standard is substantially completely supported, modulo
1786 bugs and floating-point issues
1787 (mainly but not entirely relating to optional C99 features from
1788 Annexes F and G). See
1789 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1790 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1795 ISO C11, the 2011 revision of the ISO C standard. This standard is
1796 substantially completely supported, modulo bugs, floating-point issues
1797 (mainly but not entirely relating to optional C11 features from
1798 Annexes F and G) and the optional Annexes K (Bounds-checking
1799 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1803 GNU dialect of ISO C90 (including some C99 features).
1807 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1811 GNU dialect of ISO C11. This is the default for C code.
1812 The name @samp{gnu1x} is deprecated.
1816 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1817 additional defect reports. Same as @option{-ansi} for C++ code.
1821 GNU dialect of @option{-std=c++98}.
1825 The 2011 ISO C++ standard plus amendments.
1826 The name @samp{c++0x} is deprecated.
1830 GNU dialect of @option{-std=c++11}.
1831 The name @samp{gnu++0x} is deprecated.
1835 The 2014 ISO C++ standard plus amendments.
1836 The name @samp{c++1y} is deprecated.
1840 GNU dialect of @option{-std=c++14}.
1841 This is the default for C++ code.
1842 The name @samp{gnu++1y} is deprecated.
1845 The next revision of the ISO C++ standard, tentatively planned for
1846 2017. Support is highly experimental, and will almost certainly
1847 change in incompatible ways in future releases.
1850 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1851 and will almost certainly change in incompatible ways in future
1855 @item -fgnu89-inline
1856 @opindex fgnu89-inline
1857 The option @option{-fgnu89-inline} tells GCC to use the traditional
1858 GNU semantics for @code{inline} functions when in C99 mode.
1859 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1860 Using this option is roughly equivalent to adding the
1861 @code{gnu_inline} function attribute to all inline functions
1862 (@pxref{Function Attributes}).
1864 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1865 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1866 specifies the default behavior).
1867 This option is not supported in @option{-std=c90} or
1868 @option{-std=gnu90} mode.
1870 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1871 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1872 in effect for @code{inline} functions. @xref{Common Predefined
1873 Macros,,,cpp,The C Preprocessor}.
1875 @item -aux-info @var{filename}
1877 Output to the given filename prototyped declarations for all functions
1878 declared and/or defined in a translation unit, including those in header
1879 files. This option is silently ignored in any language other than C@.
1881 Besides declarations, the file indicates, in comments, the origin of
1882 each declaration (source file and line), whether the declaration was
1883 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1884 @samp{O} for old, respectively, in the first character after the line
1885 number and the colon), and whether it came from a declaration or a
1886 definition (@samp{C} or @samp{F}, respectively, in the following
1887 character). In the case of function definitions, a K&R-style list of
1888 arguments followed by their declarations is also provided, inside
1889 comments, after the declaration.
1891 @item -fallow-parameterless-variadic-functions
1892 @opindex fallow-parameterless-variadic-functions
1893 Accept variadic functions without named parameters.
1895 Although it is possible to define such a function, this is not very
1896 useful as it is not possible to read the arguments. This is only
1897 supported for C as this construct is allowed by C++.
1901 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1902 keyword, so that code can use these words as identifiers. You can use
1903 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1904 instead. @option{-ansi} implies @option{-fno-asm}.
1906 In C++, this switch only affects the @code{typeof} keyword, since
1907 @code{asm} and @code{inline} are standard keywords. You may want to
1908 use the @option{-fno-gnu-keywords} flag instead, which has the same
1909 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1910 switch only affects the @code{asm} and @code{typeof} keywords, since
1911 @code{inline} is a standard keyword in ISO C99.
1914 @itemx -fno-builtin-@var{function}
1915 @opindex fno-builtin
1916 @cindex built-in functions
1917 Don't recognize built-in functions that do not begin with
1918 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1919 functions provided by GCC}, for details of the functions affected,
1920 including those which are not built-in functions when @option{-ansi} or
1921 @option{-std} options for strict ISO C conformance are used because they
1922 do not have an ISO standard meaning.
1924 GCC normally generates special code to handle certain built-in functions
1925 more efficiently; for instance, calls to @code{alloca} may become single
1926 instructions which adjust the stack directly, and calls to @code{memcpy}
1927 may become inline copy loops. The resulting code is often both smaller
1928 and faster, but since the function calls no longer appear as such, you
1929 cannot set a breakpoint on those calls, nor can you change the behavior
1930 of the functions by linking with a different library. In addition,
1931 when a function is recognized as a built-in function, GCC may use
1932 information about that function to warn about problems with calls to
1933 that function, or to generate more efficient code, even if the
1934 resulting code still contains calls to that function. For example,
1935 warnings are given with @option{-Wformat} for bad calls to
1936 @code{printf} when @code{printf} is built in and @code{strlen} is
1937 known not to modify global memory.
1939 With the @option{-fno-builtin-@var{function}} option
1940 only the built-in function @var{function} is
1941 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1942 function is named that is not built-in in this version of GCC, this
1943 option is ignored. There is no corresponding
1944 @option{-fbuiltin-@var{function}} option; if you wish to enable
1945 built-in functions selectively when using @option{-fno-builtin} or
1946 @option{-ffreestanding}, you may define macros such as:
1949 #define abs(n) __builtin_abs ((n))
1950 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1955 @cindex hosted environment
1957 Assert that compilation targets a hosted environment. This implies
1958 @option{-fbuiltin}. A hosted environment is one in which the
1959 entire standard library is available, and in which @code{main} has a return
1960 type of @code{int}. Examples are nearly everything except a kernel.
1961 This is equivalent to @option{-fno-freestanding}.
1963 @item -ffreestanding
1964 @opindex ffreestanding
1965 @cindex hosted environment
1967 Assert that compilation targets a freestanding environment. This
1968 implies @option{-fno-builtin}. A freestanding environment
1969 is one in which the standard library may not exist, and program startup may
1970 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1971 This is equivalent to @option{-fno-hosted}.
1973 @xref{Standards,,Language Standards Supported by GCC}, for details of
1974 freestanding and hosted environments.
1978 @cindex OpenACC accelerator programming
1979 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1980 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1981 compiler generates accelerated code according to the OpenACC Application
1982 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1983 implies @option{-pthread}, and thus is only supported on targets that
1984 have support for @option{-pthread}.
1986 @item -fopenacc-dim=@var{geom}
1987 @opindex fopenacc-dim
1988 @cindex OpenACC accelerator programming
1989 Specify default compute dimensions for parallel offload regions that do
1990 not explicitly specify. The @var{geom} value is a triple of
1991 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
1992 can be omitted, to use a target-specific default value.
1996 @cindex OpenMP parallel
1997 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1998 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1999 compiler generates parallel code according to the OpenMP Application
2000 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
2001 implies @option{-pthread}, and thus is only supported on targets that
2002 have support for @option{-pthread}. @option{-fopenmp} implies
2003 @option{-fopenmp-simd}.
2006 @opindex fopenmp-simd
2009 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2010 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2015 @cindex Enable Cilk Plus
2016 Enable the usage of Cilk Plus language extension features for C/C++.
2017 When the option @option{-fcilkplus} is specified, enable the usage of
2018 the Cilk Plus Language extension features for C/C++. The present
2019 implementation follows ABI version 1.2. This is an experimental
2020 feature that is only partially complete, and whose interface may
2021 change in future versions of GCC as the official specification
2022 changes. Currently, all features but @code{_Cilk_for} have been
2027 When the option @option{-fgnu-tm} is specified, the compiler
2028 generates code for the Linux variant of Intel's current Transactional
2029 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2030 an experimental feature whose interface may change in future versions
2031 of GCC, as the official specification changes. Please note that not
2032 all architectures are supported for this feature.
2034 For more information on GCC's support for transactional memory,
2035 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2036 Transactional Memory Library}.
2038 Note that the transactional memory feature is not supported with
2039 non-call exceptions (@option{-fnon-call-exceptions}).
2041 @item -fms-extensions
2042 @opindex fms-extensions
2043 Accept some non-standard constructs used in Microsoft header files.
2045 In C++ code, this allows member names in structures to be similar
2046 to previous types declarations.
2055 Some cases of unnamed fields in structures and unions are only
2056 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2057 fields within structs/unions}, for details.
2059 Note that this option is off for all targets but x86
2060 targets using ms-abi.
2062 @item -fplan9-extensions
2063 @opindex fplan9-extensions
2064 Accept some non-standard constructs used in Plan 9 code.
2066 This enables @option{-fms-extensions}, permits passing pointers to
2067 structures with anonymous fields to functions that expect pointers to
2068 elements of the type of the field, and permits referring to anonymous
2069 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2070 struct/union fields within structs/unions}, for details. This is only
2071 supported for C, not C++.
2075 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2076 options for strict ISO C conformance) implies @option{-trigraphs}.
2078 @cindex traditional C language
2079 @cindex C language, traditional
2081 @itemx -traditional-cpp
2082 @opindex traditional-cpp
2083 @opindex traditional
2084 Formerly, these options caused GCC to attempt to emulate a pre-standard
2085 C compiler. They are now only supported with the @option{-E} switch.
2086 The preprocessor continues to support a pre-standard mode. See the GNU
2087 CPP manual for details.
2089 @item -fcond-mismatch
2090 @opindex fcond-mismatch
2091 Allow conditional expressions with mismatched types in the second and
2092 third arguments. The value of such an expression is void. This option
2093 is not supported for C++.
2095 @item -flax-vector-conversions
2096 @opindex flax-vector-conversions
2097 Allow implicit conversions between vectors with differing numbers of
2098 elements and/or incompatible element types. This option should not be
2101 @item -funsigned-char
2102 @opindex funsigned-char
2103 Let the type @code{char} be unsigned, like @code{unsigned char}.
2105 Each kind of machine has a default for what @code{char} should
2106 be. It is either like @code{unsigned char} by default or like
2107 @code{signed char} by default.
2109 Ideally, a portable program should always use @code{signed char} or
2110 @code{unsigned char} when it depends on the signedness of an object.
2111 But many programs have been written to use plain @code{char} and
2112 expect it to be signed, or expect it to be unsigned, depending on the
2113 machines they were written for. This option, and its inverse, let you
2114 make such a program work with the opposite default.
2116 The type @code{char} is always a distinct type from each of
2117 @code{signed char} or @code{unsigned char}, even though its behavior
2118 is always just like one of those two.
2121 @opindex fsigned-char
2122 Let the type @code{char} be signed, like @code{signed char}.
2124 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2125 the negative form of @option{-funsigned-char}. Likewise, the option
2126 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2128 @item -fsigned-bitfields
2129 @itemx -funsigned-bitfields
2130 @itemx -fno-signed-bitfields
2131 @itemx -fno-unsigned-bitfields
2132 @opindex fsigned-bitfields
2133 @opindex funsigned-bitfields
2134 @opindex fno-signed-bitfields
2135 @opindex fno-unsigned-bitfields
2136 These options control whether a bit-field is signed or unsigned, when the
2137 declaration does not use either @code{signed} or @code{unsigned}. By
2138 default, such a bit-field is signed, because this is consistent: the
2139 basic integer types such as @code{int} are signed types.
2141 @item -fsso-struct=@var{endianness}
2142 @opindex fsso-struct
2143 Set the default scalar storage order of structures and unions to the
2144 specified endianness. The accepted values are @samp{big-endian} and
2145 @samp{little-endian}. If the option is not passed, the compiler uses
2146 the native endianness of the target. This option is not supported for C++.
2148 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2149 code that is not binary compatible with code generated without it if the
2150 specified endianness is not the native endianness of the target.
2153 @node C++ Dialect Options
2154 @section Options Controlling C++ Dialect
2156 @cindex compiler options, C++
2157 @cindex C++ options, command-line
2158 @cindex options, C++
2159 This section describes the command-line options that are only meaningful
2160 for C++ programs. You can also use most of the GNU compiler options
2161 regardless of what language your program is in. For example, you
2162 might compile a file @file{firstClass.C} like this:
2165 g++ -g -fstrict-enums -O -c firstClass.C
2169 In this example, only @option{-fstrict-enums} is an option meant
2170 only for C++ programs; you can use the other options with any
2171 language supported by GCC@.
2173 Some options for compiling C programs, such as @option{-std}, are also
2174 relevant for C++ programs.
2175 @xref{C Dialect Options,,Options Controlling C Dialect}.
2177 Here is a list of options that are @emph{only} for compiling C++ programs:
2181 @item -fabi-version=@var{n}
2182 @opindex fabi-version
2183 Use version @var{n} of the C++ ABI@. The default is version 0.
2185 Version 0 refers to the version conforming most closely to
2186 the C++ ABI specification. Therefore, the ABI obtained using version 0
2187 will change in different versions of G++ as ABI bugs are fixed.
2189 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2191 Version 2 is the version of the C++ ABI that first appeared in G++
2192 3.4, and was the default through G++ 4.9.
2194 Version 3 corrects an error in mangling a constant address as a
2197 Version 4, which first appeared in G++ 4.5, implements a standard
2198 mangling for vector types.
2200 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2201 attribute const/volatile on function pointer types, decltype of a
2202 plain decl, and use of a function parameter in the declaration of
2205 Version 6, which first appeared in G++ 4.7, corrects the promotion
2206 behavior of C++11 scoped enums and the mangling of template argument
2207 packs, const/static_cast, prefix ++ and --, and a class scope function
2208 used as a template argument.
2210 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2211 builtin type and corrects the mangling of lambdas in default argument
2214 Version 8, which first appeared in G++ 4.9, corrects the substitution
2215 behavior of function types with function-cv-qualifiers.
2217 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2220 Version 10, which first appeared in G++ 6.1, adds mangling of
2221 attributes that affect type identity, such as ia32 calling convention
2222 attributes (e.g. @samp{stdcall}).
2224 Version 11, which first appeared in G++ 7, corrects the mangling of
2225 sizeof... expressions. It also implies
2226 @option{-fnew-inheriting-ctors}.
2228 See also @option{-Wabi}.
2230 @item -fabi-compat-version=@var{n}
2231 @opindex fabi-compat-version
2232 On targets that support strong aliases, G++
2233 works around mangling changes by creating an alias with the correct
2234 mangled name when defining a symbol with an incorrect mangled name.
2235 This switch specifies which ABI version to use for the alias.
2237 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2238 compatibility). If another ABI version is explicitly selected, this
2239 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2240 use @option{-fabi-compat-version=2}.
2242 If this option is not provided but @option{-Wabi=@var{n}} is, that
2243 version is used for compatibility aliases. If this option is provided
2244 along with @option{-Wabi} (without the version), the version from this
2245 option is used for the warning.
2247 @item -fno-access-control
2248 @opindex fno-access-control
2249 Turn off all access checking. This switch is mainly useful for working
2250 around bugs in the access control code.
2253 @opindex faligned-new
2254 Enable support for C++17 @code{new} of types that require more
2255 alignment than @code{void* ::operator new(std::size_t)} provides. A
2256 numeric argument such as @code{-faligned-new=32} can be used to
2257 specify how much alignment (in bytes) is provided by that function,
2258 but few users will need to override the default of
2259 @code{alignof(std::max_align_t)}.
2263 Check that the pointer returned by @code{operator new} is non-null
2264 before attempting to modify the storage allocated. This check is
2265 normally unnecessary because the C++ standard specifies that
2266 @code{operator new} only returns @code{0} if it is declared
2267 @code{throw()}, in which case the compiler always checks the
2268 return value even without this option. In all other cases, when
2269 @code{operator new} has a non-empty exception specification, memory
2270 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2271 @samp{new (nothrow)}.
2275 Enable support for the C++ Extensions for Concepts Technical
2276 Specification, ISO 19217 (2015), which allows code like
2279 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2280 template <Addable T> T add (T a, T b) @{ return a + b; @}
2283 @item -fconstexpr-depth=@var{n}
2284 @opindex fconstexpr-depth
2285 Set the maximum nested evaluation depth for C++11 constexpr functions
2286 to @var{n}. A limit is needed to detect endless recursion during
2287 constant expression evaluation. The minimum specified by the standard
2290 @item -fconstexpr-loop-limit=@var{n}
2291 @opindex fconstexpr-loop-limit
2292 Set the maximum number of iterations for a loop in C++14 constexpr functions
2293 to @var{n}. A limit is needed to detect infinite loops during
2294 constant expression evaluation. The default is 262144 (1<<18).
2296 @item -fdeduce-init-list
2297 @opindex fdeduce-init-list
2298 Enable deduction of a template type parameter as
2299 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2302 template <class T> auto forward(T t) -> decltype (realfn (t))
2309 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2313 This deduction was implemented as a possible extension to the
2314 originally proposed semantics for the C++11 standard, but was not part
2315 of the final standard, so it is disabled by default. This option is
2316 deprecated, and may be removed in a future version of G++.
2318 @item -ffriend-injection
2319 @opindex ffriend-injection
2320 Inject friend functions into the enclosing namespace, so that they are
2321 visible outside the scope of the class in which they are declared.
2322 Friend functions were documented to work this way in the old Annotated
2323 C++ Reference Manual.
2324 However, in ISO C++ a friend function that is not declared
2325 in an enclosing scope can only be found using argument dependent
2326 lookup. GCC defaults to the standard behavior.
2328 This option is for compatibility, and may be removed in a future
2331 @item -fno-elide-constructors
2332 @opindex fno-elide-constructors
2333 The C++ standard allows an implementation to omit creating a temporary
2334 that is only used to initialize another object of the same type.
2335 Specifying this option disables that optimization, and forces G++ to
2336 call the copy constructor in all cases. This option also causes G++
2337 to call trivial member functions which otherwise would be expanded inline.
2339 In C++17, the compiler is required to omit these temporaries, but this
2340 option still affects trivial member functions.
2342 @item -fno-enforce-eh-specs
2343 @opindex fno-enforce-eh-specs
2344 Don't generate code to check for violation of exception specifications
2345 at run time. This option violates the C++ standard, but may be useful
2346 for reducing code size in production builds, much like defining
2347 @code{NDEBUG}. This does not give user code permission to throw
2348 exceptions in violation of the exception specifications; the compiler
2349 still optimizes based on the specifications, so throwing an
2350 unexpected exception results in undefined behavior at run time.
2352 @item -fextern-tls-init
2353 @itemx -fno-extern-tls-init
2354 @opindex fextern-tls-init
2355 @opindex fno-extern-tls-init
2356 The C++11 and OpenMP standards allow @code{thread_local} and
2357 @code{threadprivate} variables to have dynamic (runtime)
2358 initialization. To support this, any use of such a variable goes
2359 through a wrapper function that performs any necessary initialization.
2360 When the use and definition of the variable are in the same
2361 translation unit, this overhead can be optimized away, but when the
2362 use is in a different translation unit there is significant overhead
2363 even if the variable doesn't actually need dynamic initialization. If
2364 the programmer can be sure that no use of the variable in a
2365 non-defining TU needs to trigger dynamic initialization (either
2366 because the variable is statically initialized, or a use of the
2367 variable in the defining TU will be executed before any uses in
2368 another TU), they can avoid this overhead with the
2369 @option{-fno-extern-tls-init} option.
2371 On targets that support symbol aliases, the default is
2372 @option{-fextern-tls-init}. On targets that do not support symbol
2373 aliases, the default is @option{-fno-extern-tls-init}.
2376 @itemx -fno-for-scope
2378 @opindex fno-for-scope
2379 If @option{-ffor-scope} is specified, the scope of variables declared in
2380 a @i{for-init-statement} is limited to the @code{for} loop itself,
2381 as specified by the C++ standard.
2382 If @option{-fno-for-scope} is specified, the scope of variables declared in
2383 a @i{for-init-statement} extends to the end of the enclosing scope,
2384 as was the case in old versions of G++, and other (traditional)
2385 implementations of C++.
2387 If neither flag is given, the default is to follow the standard,
2388 but to allow and give a warning for old-style code that would
2389 otherwise be invalid, or have different behavior.
2391 @item -fno-gnu-keywords
2392 @opindex fno-gnu-keywords
2393 Do not recognize @code{typeof} as a keyword, so that code can use this
2394 word as an identifier. You can use the keyword @code{__typeof__} instead.
2395 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2396 @option{-std=c++98}, @option{-std=c++11}, etc.
2398 @item -fno-implicit-templates
2399 @opindex fno-implicit-templates
2400 Never emit code for non-inline templates that are instantiated
2401 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2402 @xref{Template Instantiation}, for more information.
2404 @item -fno-implicit-inline-templates
2405 @opindex fno-implicit-inline-templates
2406 Don't emit code for implicit instantiations of inline templates, either.
2407 The default is to handle inlines differently so that compiles with and
2408 without optimization need the same set of explicit instantiations.
2410 @item -fno-implement-inlines
2411 @opindex fno-implement-inlines
2412 To save space, do not emit out-of-line copies of inline functions
2413 controlled by @code{#pragma implementation}. This causes linker
2414 errors if these functions are not inlined everywhere they are called.
2416 @item -fms-extensions
2417 @opindex fms-extensions
2418 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2419 int and getting a pointer to member function via non-standard syntax.
2421 @item -fnew-inheriting-ctors
2422 @opindex fnew-inheriting-ctors
2423 Enable the P0136 adjustment to the semantics of C++11 constructor
2424 inheritance. This is part of C++17 but also considered to be a Defect
2425 Report against C++11 and C++14. This flag is enabled by default
2426 unless @option{-fabi-version=10} or lower is specified.
2428 @item -fno-nonansi-builtins
2429 @opindex fno-nonansi-builtins
2430 Disable built-in declarations of functions that are not mandated by
2431 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2432 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2435 @opindex fnothrow-opt
2436 Treat a @code{throw()} exception specification as if it were a
2437 @code{noexcept} specification to reduce or eliminate the text size
2438 overhead relative to a function with no exception specification. If
2439 the function has local variables of types with non-trivial
2440 destructors, the exception specification actually makes the
2441 function smaller because the EH cleanups for those variables can be
2442 optimized away. The semantic effect is that an exception thrown out of
2443 a function with such an exception specification results in a call
2444 to @code{terminate} rather than @code{unexpected}.
2446 @item -fno-operator-names
2447 @opindex fno-operator-names
2448 Do not treat the operator name keywords @code{and}, @code{bitand},
2449 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2450 synonyms as keywords.
2452 @item -fno-optional-diags
2453 @opindex fno-optional-diags
2454 Disable diagnostics that the standard says a compiler does not need to
2455 issue. Currently, the only such diagnostic issued by G++ is the one for
2456 a name having multiple meanings within a class.
2459 @opindex fpermissive
2460 Downgrade some diagnostics about nonconformant code from errors to
2461 warnings. Thus, using @option{-fpermissive} allows some
2462 nonconforming code to compile.
2464 @item -fno-pretty-templates
2465 @opindex fno-pretty-templates
2466 When an error message refers to a specialization of a function
2467 template, the compiler normally prints the signature of the
2468 template followed by the template arguments and any typedefs or
2469 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2470 rather than @code{void f(int)}) so that it's clear which template is
2471 involved. When an error message refers to a specialization of a class
2472 template, the compiler omits any template arguments that match
2473 the default template arguments for that template. If either of these
2474 behaviors make it harder to understand the error message rather than
2475 easier, you can use @option{-fno-pretty-templates} to disable them.
2479 Enable automatic template instantiation at link time. This option also
2480 implies @option{-fno-implicit-templates}. @xref{Template
2481 Instantiation}, for more information.
2485 Disable generation of information about every class with virtual
2486 functions for use by the C++ run-time type identification features
2487 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2488 of the language, you can save some space by using this flag. Note that
2489 exception handling uses the same information, but G++ generates it as
2490 needed. The @code{dynamic_cast} operator can still be used for casts that
2491 do not require run-time type information, i.e.@: casts to @code{void *} or to
2492 unambiguous base classes.
2494 @item -fsized-deallocation
2495 @opindex fsized-deallocation
2496 Enable the built-in global declarations
2498 void operator delete (void *, std::size_t) noexcept;
2499 void operator delete[] (void *, std::size_t) noexcept;
2501 as introduced in C++14. This is useful for user-defined replacement
2502 deallocation functions that, for example, use the size of the object
2503 to make deallocation faster. Enabled by default under
2504 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2505 warns about places that might want to add a definition.
2507 @item -fstrict-enums
2508 @opindex fstrict-enums
2509 Allow the compiler to optimize using the assumption that a value of
2510 enumerated type can only be one of the values of the enumeration (as
2511 defined in the C++ standard; basically, a value that can be
2512 represented in the minimum number of bits needed to represent all the
2513 enumerators). This assumption may not be valid if the program uses a
2514 cast to convert an arbitrary integer value to the enumerated type.
2516 @item -fstrong-eval-order
2517 @opindex fstrong-eval-order
2518 Evaluate member access, array subscripting, and shift expressions in
2519 left-to-right order, and evaluate assignment in right-to-left order,
2520 as adopted for C++17. Enabled by default with @option{-std=c++1z}.
2521 @option{-fstrong-eval-order=some} enables just the ordering of member
2522 access and shift expressions, and is the default without
2523 @option{-std=c++1z}.
2525 @item -ftemplate-backtrace-limit=@var{n}
2526 @opindex ftemplate-backtrace-limit
2527 Set the maximum number of template instantiation notes for a single
2528 warning or error to @var{n}. The default value is 10.
2530 @item -ftemplate-depth=@var{n}
2531 @opindex ftemplate-depth
2532 Set the maximum instantiation depth for template classes to @var{n}.
2533 A limit on the template instantiation depth is needed to detect
2534 endless recursions during template class instantiation. ANSI/ISO C++
2535 conforming programs must not rely on a maximum depth greater than 17
2536 (changed to 1024 in C++11). The default value is 900, as the compiler
2537 can run out of stack space before hitting 1024 in some situations.
2539 @item -fno-threadsafe-statics
2540 @opindex fno-threadsafe-statics
2541 Do not emit the extra code to use the routines specified in the C++
2542 ABI for thread-safe initialization of local statics. You can use this
2543 option to reduce code size slightly in code that doesn't need to be
2546 @item -fuse-cxa-atexit
2547 @opindex fuse-cxa-atexit
2548 Register destructors for objects with static storage duration with the
2549 @code{__cxa_atexit} function rather than the @code{atexit} function.
2550 This option is required for fully standards-compliant handling of static
2551 destructors, but only works if your C library supports
2552 @code{__cxa_atexit}.
2554 @item -fno-use-cxa-get-exception-ptr
2555 @opindex fno-use-cxa-get-exception-ptr
2556 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2557 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2558 if the runtime routine is not available.
2560 @item -fvisibility-inlines-hidden
2561 @opindex fvisibility-inlines-hidden
2562 This switch declares that the user does not attempt to compare
2563 pointers to inline functions or methods where the addresses of the two functions
2564 are taken in different shared objects.
2566 The effect of this is that GCC may, effectively, mark inline methods with
2567 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2568 appear in the export table of a DSO and do not require a PLT indirection
2569 when used within the DSO@. Enabling this option can have a dramatic effect
2570 on load and link times of a DSO as it massively reduces the size of the
2571 dynamic export table when the library makes heavy use of templates.
2573 The behavior of this switch is not quite the same as marking the
2574 methods as hidden directly, because it does not affect static variables
2575 local to the function or cause the compiler to deduce that
2576 the function is defined in only one shared object.
2578 You may mark a method as having a visibility explicitly to negate the
2579 effect of the switch for that method. For example, if you do want to
2580 compare pointers to a particular inline method, you might mark it as
2581 having default visibility. Marking the enclosing class with explicit
2582 visibility has no effect.
2584 Explicitly instantiated inline methods are unaffected by this option
2585 as their linkage might otherwise cross a shared library boundary.
2586 @xref{Template Instantiation}.
2588 @item -fvisibility-ms-compat
2589 @opindex fvisibility-ms-compat
2590 This flag attempts to use visibility settings to make GCC's C++
2591 linkage model compatible with that of Microsoft Visual Studio.
2593 The flag makes these changes to GCC's linkage model:
2597 It sets the default visibility to @code{hidden}, like
2598 @option{-fvisibility=hidden}.
2601 Types, but not their members, are not hidden by default.
2604 The One Definition Rule is relaxed for types without explicit
2605 visibility specifications that are defined in more than one
2606 shared object: those declarations are permitted if they are
2607 permitted when this option is not used.
2610 In new code it is better to use @option{-fvisibility=hidden} and
2611 export those classes that are intended to be externally visible.
2612 Unfortunately it is possible for code to rely, perhaps accidentally,
2613 on the Visual Studio behavior.
2615 Among the consequences of these changes are that static data members
2616 of the same type with the same name but defined in different shared
2617 objects are different, so changing one does not change the other;
2618 and that pointers to function members defined in different shared
2619 objects may not compare equal. When this flag is given, it is a
2620 violation of the ODR to define types with the same name differently.
2624 Do not use weak symbol support, even if it is provided by the linker.
2625 By default, G++ uses weak symbols if they are available. This
2626 option exists only for testing, and should not be used by end-users;
2627 it results in inferior code and has no benefits. This option may
2628 be removed in a future release of G++.
2632 Do not search for header files in the standard directories specific to
2633 C++, but do still search the other standard directories. (This option
2634 is used when building the C++ library.)
2637 In addition, these optimization, warning, and code generation options
2638 have meanings only for C++ programs:
2641 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2644 Warn when G++ it generates code that is probably not compatible with
2645 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2646 ABI with each major release, normally @option{-Wabi} will warn only if
2647 there is a check added later in a release series for an ABI issue
2648 discovered since the initial release. @option{-Wabi} will warn about
2649 more things if an older ABI version is selected (with
2650 @option{-fabi-version=@var{n}}).
2652 @option{-Wabi} can also be used with an explicit version number to
2653 warn about compatibility with a particular @option{-fabi-version}
2654 level, e.g. @option{-Wabi=2} to warn about changes relative to
2655 @option{-fabi-version=2}.
2657 If an explicit version number is provided and
2658 @option{-fabi-compat-version} is not specified, the version number
2659 from this option is used for compatibility aliases. If no explicit
2660 version number is provided with this option, but
2661 @option{-fabi-compat-version} is specified, that version number is
2662 used for ABI warnings.
2664 Although an effort has been made to warn about
2665 all such cases, there are probably some cases that are not warned about,
2666 even though G++ is generating incompatible code. There may also be
2667 cases where warnings are emitted even though the code that is generated
2670 You should rewrite your code to avoid these warnings if you are
2671 concerned about the fact that code generated by G++ may not be binary
2672 compatible with code generated by other compilers.
2674 Known incompatibilities in @option{-fabi-version=2} (which was the
2675 default from GCC 3.4 to 4.9) include:
2680 A template with a non-type template parameter of reference type was
2681 mangled incorrectly:
2684 template <int &> struct S @{@};
2688 This was fixed in @option{-fabi-version=3}.
2691 SIMD vector types declared using @code{__attribute ((vector_size))} were
2692 mangled in a non-standard way that does not allow for overloading of
2693 functions taking vectors of different sizes.
2695 The mangling was changed in @option{-fabi-version=4}.
2698 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2699 qualifiers, and @code{decltype} of a plain declaration was folded away.
2701 These mangling issues were fixed in @option{-fabi-version=5}.
2704 Scoped enumerators passed as arguments to a variadic function are
2705 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2706 On most targets this does not actually affect the parameter passing
2707 ABI, as there is no way to pass an argument smaller than @code{int}.
2709 Also, the ABI changed the mangling of template argument packs,
2710 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2711 a class scope function used as a template argument.
2713 These issues were corrected in @option{-fabi-version=6}.
2716 Lambdas in default argument scope were mangled incorrectly, and the
2717 ABI changed the mangling of @code{nullptr_t}.
2719 These issues were corrected in @option{-fabi-version=7}.
2722 When mangling a function type with function-cv-qualifiers, the
2723 un-qualified function type was incorrectly treated as a substitution
2726 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2729 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2730 unaligned accesses. Note that this did not affect the ABI of a
2731 function with a @code{nullptr_t} parameter, as parameters have a
2734 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2737 Target-specific attributes that affect the identity of a type, such as
2738 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2739 did not affect the mangled name, leading to name collisions when
2740 function pointers were used as template arguments.
2742 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2746 It also warns about psABI-related changes. The known psABI changes at this
2752 For SysV/x86-64, unions with @code{long double} members are
2753 passed in memory as specified in psABI. For example:
2763 @code{union U} is always passed in memory.
2767 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2770 Warn when a type with an ABI tag is used in a context that does not
2771 have that ABI tag. See @ref{C++ Attributes} for more information
2774 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2775 @opindex Wctor-dtor-privacy
2776 @opindex Wno-ctor-dtor-privacy
2777 Warn when a class seems unusable because all the constructors or
2778 destructors in that class are private, and it has neither friends nor
2779 public static member functions. Also warn if there are no non-private
2780 methods, and there's at least one private member function that isn't
2781 a constructor or destructor.
2783 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2784 @opindex Wdelete-non-virtual-dtor
2785 @opindex Wno-delete-non-virtual-dtor
2786 Warn when @code{delete} is used to destroy an instance of a class that
2787 has virtual functions and non-virtual destructor. It is unsafe to delete
2788 an instance of a derived class through a pointer to a base class if the
2789 base class does not have a virtual destructor. This warning is enabled
2792 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2793 @opindex Wliteral-suffix
2794 @opindex Wno-literal-suffix
2795 Warn when a string or character literal is followed by a ud-suffix which does
2796 not begin with an underscore. As a conforming extension, GCC treats such
2797 suffixes as separate preprocessing tokens in order to maintain backwards
2798 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2802 #define __STDC_FORMAT_MACROS
2803 #include <inttypes.h>
2808 printf("My int64: %" PRId64"\n", i64);
2812 In this case, @code{PRId64} is treated as a separate preprocessing token.
2814 This warning is enabled by default.
2816 @item -Wlto-type-mismatch
2817 @opindex Wlto-type-mismatch
2818 @opindex Wno-lto-type-mismatch
2820 During the link-time optimization warn about type mismatches in
2821 global declarations from different compilation units.
2822 Requires @option{-flto} to be enabled. Enabled by default.
2824 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2826 @opindex Wno-narrowing
2827 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2828 conversion prohibited by C++11 occurs within
2832 int i = @{ 2.2 @}; // error: narrowing from double to int
2835 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2837 When a later standard is in effect, e.g. when using @option{-std=c++11},
2838 narrowing conversions are diagnosed by default, as required by the standard.
2839 A narrowing conversion from a constant produces an error,
2840 and a narrowing conversion from a non-constant produces a warning,
2841 but @option{-Wno-narrowing} suppresses the diagnostic.
2842 Note that this does not affect the meaning of well-formed code;
2843 narrowing conversions are still considered ill-formed in SFINAE contexts.
2845 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2847 @opindex Wno-noexcept
2848 Warn when a noexcept-expression evaluates to false because of a call
2849 to a function that does not have a non-throwing exception
2850 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2851 the compiler to never throw an exception.
2853 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2854 @opindex Wnon-virtual-dtor
2855 @opindex Wno-non-virtual-dtor
2856 Warn when a class has virtual functions and an accessible non-virtual
2857 destructor itself or in an accessible polymorphic base class, in which
2858 case it is possible but unsafe to delete an instance of a derived
2859 class through a pointer to the class itself or base class. This
2860 warning is automatically enabled if @option{-Weffc++} is specified.
2862 @item -Wregister @r{(C++ and Objective-C++ only)}
2864 @opindex Wno-register
2865 Warn on uses of the @code{register} storage class specifier, except
2866 when it is part of the GNU @ref{Explicit Register Variables} extension.
2867 The use of the @code{register} keyword as storage class specifier has
2868 been deprecated in C++11 and removed in C++17.
2869 Enabled by default with @option{-std=c++1z}.
2871 @item -Wreorder @r{(C++ and Objective-C++ only)}
2873 @opindex Wno-reorder
2874 @cindex reordering, warning
2875 @cindex warning for reordering of member initializers
2876 Warn when the order of member initializers given in the code does not
2877 match the order in which they must be executed. For instance:
2883 A(): j (0), i (1) @{ @}
2888 The compiler rearranges the member initializers for @code{i}
2889 and @code{j} to match the declaration order of the members, emitting
2890 a warning to that effect. This warning is enabled by @option{-Wall}.
2892 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2893 @opindex fext-numeric-literals
2894 @opindex fno-ext-numeric-literals
2895 Accept imaginary, fixed-point, or machine-defined
2896 literal number suffixes as GNU extensions.
2897 When this option is turned off these suffixes are treated
2898 as C++11 user-defined literal numeric suffixes.
2899 This is on by default for all pre-C++11 dialects and all GNU dialects:
2900 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2901 @option{-std=gnu++14}.
2902 This option is off by default
2903 for ISO C++11 onwards (@option{-std=c++11}, ...).
2906 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2909 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2912 Warn about violations of the following style guidelines from Scott Meyers'
2913 @cite{Effective C++} series of books:
2917 Define a copy constructor and an assignment operator for classes
2918 with dynamically-allocated memory.
2921 Prefer initialization to assignment in constructors.
2924 Have @code{operator=} return a reference to @code{*this}.
2927 Don't try to return a reference when you must return an object.
2930 Distinguish between prefix and postfix forms of increment and
2931 decrement operators.
2934 Never overload @code{&&}, @code{||}, or @code{,}.
2938 This option also enables @option{-Wnon-virtual-dtor}, which is also
2939 one of the effective C++ recommendations. However, the check is
2940 extended to warn about the lack of virtual destructor in accessible
2941 non-polymorphic bases classes too.
2943 When selecting this option, be aware that the standard library
2944 headers do not obey all of these guidelines; use @samp{grep -v}
2945 to filter out those warnings.
2947 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2948 @opindex Wstrict-null-sentinel
2949 @opindex Wno-strict-null-sentinel
2950 Warn about the use of an uncasted @code{NULL} as sentinel. When
2951 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2952 to @code{__null}. Although it is a null pointer constant rather than a
2953 null pointer, it is guaranteed to be of the same size as a pointer.
2954 But this use is not portable across different compilers.
2956 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2957 @opindex Wno-non-template-friend
2958 @opindex Wnon-template-friend
2959 Disable warnings when non-templatized friend functions are declared
2960 within a template. Since the advent of explicit template specification
2961 support in G++, if the name of the friend is an unqualified-id (i.e.,
2962 @samp{friend foo(int)}), the C++ language specification demands that the
2963 friend declare or define an ordinary, nontemplate function. (Section
2964 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2965 could be interpreted as a particular specialization of a templatized
2966 function. Because this non-conforming behavior is no longer the default
2967 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2968 check existing code for potential trouble spots and is on by default.
2969 This new compiler behavior can be turned off with
2970 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2971 but disables the helpful warning.
2973 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2974 @opindex Wold-style-cast
2975 @opindex Wno-old-style-cast
2976 Warn if an old-style (C-style) cast to a non-void type is used within
2977 a C++ program. The new-style casts (@code{dynamic_cast},
2978 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2979 less vulnerable to unintended effects and much easier to search for.
2981 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2982 @opindex Woverloaded-virtual
2983 @opindex Wno-overloaded-virtual
2984 @cindex overloaded virtual function, warning
2985 @cindex warning for overloaded virtual function
2986 Warn when a function declaration hides virtual functions from a
2987 base class. For example, in:
2994 struct B: public A @{
2999 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3010 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3011 @opindex Wno-pmf-conversions
3012 @opindex Wpmf-conversions
3013 Disable the diagnostic for converting a bound pointer to member function
3016 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3017 @opindex Wsign-promo
3018 @opindex Wno-sign-promo
3019 Warn when overload resolution chooses a promotion from unsigned or
3020 enumerated type to a signed type, over a conversion to an unsigned type of
3021 the same size. Previous versions of G++ tried to preserve
3022 unsignedness, but the standard mandates the current behavior.
3024 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3026 Warn when a primary template declaration is encountered. Some coding
3027 rules disallow templates, and this may be used to enforce that rule.
3028 The warning is inactive inside a system header file, such as the STL, so
3029 one can still use the STL. One may also instantiate or specialize
3032 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3033 @opindex Wmultiple-inheritance
3034 Warn when a class is defined with multiple direct base classes. Some
3035 coding rules disallow multiple inheritance, and this may be used to
3036 enforce that rule. The warning is inactive inside a system header file,
3037 such as the STL, so one can still use the STL. One may also define
3038 classes that indirectly use multiple inheritance.
3040 @item -Wvirtual-inheritance
3041 @opindex Wvirtual-inheritance
3042 Warn when a class is defined with a virtual direct base classe. Some
3043 coding rules disallow multiple inheritance, and this may be used to
3044 enforce that rule. The warning is inactive inside a system header file,
3045 such as the STL, so one can still use the STL. One may also define
3046 classes that indirectly use virtual inheritance.
3049 @opindex Wnamespaces
3050 Warn when a namespace definition is opened. Some coding rules disallow
3051 namespaces, and this may be used to enforce that rule. The warning is
3052 inactive inside a system header file, such as the STL, so one can still
3053 use the STL. One may also use using directives and qualified names.
3055 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3057 @opindex Wno-terminate
3058 Disable the warning about a throw-expression that will immediately
3059 result in a call to @code{terminate}.
3062 @node Objective-C and Objective-C++ Dialect Options
3063 @section Options Controlling Objective-C and Objective-C++ Dialects
3065 @cindex compiler options, Objective-C and Objective-C++
3066 @cindex Objective-C and Objective-C++ options, command-line
3067 @cindex options, Objective-C and Objective-C++
3068 (NOTE: This manual does not describe the Objective-C and Objective-C++
3069 languages themselves. @xref{Standards,,Language Standards
3070 Supported by GCC}, for references.)
3072 This section describes the command-line options that are only meaningful
3073 for Objective-C and Objective-C++ programs. You can also use most of
3074 the language-independent GNU compiler options.
3075 For example, you might compile a file @file{some_class.m} like this:
3078 gcc -g -fgnu-runtime -O -c some_class.m
3082 In this example, @option{-fgnu-runtime} is an option meant only for
3083 Objective-C and Objective-C++ programs; you can use the other options with
3084 any language supported by GCC@.
3086 Note that since Objective-C is an extension of the C language, Objective-C
3087 compilations may also use options specific to the C front-end (e.g.,
3088 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3089 C++-specific options (e.g., @option{-Wabi}).
3091 Here is a list of options that are @emph{only} for compiling Objective-C
3092 and Objective-C++ programs:
3095 @item -fconstant-string-class=@var{class-name}
3096 @opindex fconstant-string-class
3097 Use @var{class-name} as the name of the class to instantiate for each
3098 literal string specified with the syntax @code{@@"@dots{}"}. The default
3099 class name is @code{NXConstantString} if the GNU runtime is being used, and
3100 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3101 @option{-fconstant-cfstrings} option, if also present, overrides the
3102 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3103 to be laid out as constant CoreFoundation strings.
3106 @opindex fgnu-runtime
3107 Generate object code compatible with the standard GNU Objective-C
3108 runtime. This is the default for most types of systems.
3110 @item -fnext-runtime
3111 @opindex fnext-runtime
3112 Generate output compatible with the NeXT runtime. This is the default
3113 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3114 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3117 @item -fno-nil-receivers
3118 @opindex fno-nil-receivers
3119 Assume that all Objective-C message dispatches (@code{[receiver
3120 message:arg]}) in this translation unit ensure that the receiver is
3121 not @code{nil}. This allows for more efficient entry points in the
3122 runtime to be used. This option is only available in conjunction with
3123 the NeXT runtime and ABI version 0 or 1.
3125 @item -fobjc-abi-version=@var{n}
3126 @opindex fobjc-abi-version
3127 Use version @var{n} of the Objective-C ABI for the selected runtime.
3128 This option is currently supported only for the NeXT runtime. In that
3129 case, Version 0 is the traditional (32-bit) ABI without support for
3130 properties and other Objective-C 2.0 additions. Version 1 is the
3131 traditional (32-bit) ABI with support for properties and other
3132 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3133 nothing is specified, the default is Version 0 on 32-bit target
3134 machines, and Version 2 on 64-bit target machines.
3136 @item -fobjc-call-cxx-cdtors
3137 @opindex fobjc-call-cxx-cdtors
3138 For each Objective-C class, check if any of its instance variables is a
3139 C++ object with a non-trivial default constructor. If so, synthesize a
3140 special @code{- (id) .cxx_construct} instance method which runs
3141 non-trivial default constructors on any such instance variables, in order,
3142 and then return @code{self}. Similarly, check if any instance variable
3143 is a C++ object with a non-trivial destructor, and if so, synthesize a
3144 special @code{- (void) .cxx_destruct} method which runs
3145 all such default destructors, in reverse order.
3147 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3148 methods thusly generated only operate on instance variables
3149 declared in the current Objective-C class, and not those inherited
3150 from superclasses. It is the responsibility of the Objective-C
3151 runtime to invoke all such methods in an object's inheritance
3152 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3153 by the runtime immediately after a new object instance is allocated;
3154 the @code{- (void) .cxx_destruct} methods are invoked immediately
3155 before the runtime deallocates an object instance.
3157 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3158 support for invoking the @code{- (id) .cxx_construct} and
3159 @code{- (void) .cxx_destruct} methods.
3161 @item -fobjc-direct-dispatch
3162 @opindex fobjc-direct-dispatch
3163 Allow fast jumps to the message dispatcher. On Darwin this is
3164 accomplished via the comm page.
3166 @item -fobjc-exceptions
3167 @opindex fobjc-exceptions
3168 Enable syntactic support for structured exception handling in
3169 Objective-C, similar to what is offered by C++ and Java. This option
3170 is required to use the Objective-C keywords @code{@@try},
3171 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3172 @code{@@synchronized}. This option is available with both the GNU
3173 runtime and the NeXT runtime (but not available in conjunction with
3174 the NeXT runtime on Mac OS X 10.2 and earlier).
3178 Enable garbage collection (GC) in Objective-C and Objective-C++
3179 programs. This option is only available with the NeXT runtime; the
3180 GNU runtime has a different garbage collection implementation that
3181 does not require special compiler flags.
3183 @item -fobjc-nilcheck
3184 @opindex fobjc-nilcheck
3185 For the NeXT runtime with version 2 of the ABI, check for a nil
3186 receiver in method invocations before doing the actual method call.
3187 This is the default and can be disabled using
3188 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3189 checked for nil in this way no matter what this flag is set to.
3190 Currently this flag does nothing when the GNU runtime, or an older
3191 version of the NeXT runtime ABI, is used.
3193 @item -fobjc-std=objc1
3195 Conform to the language syntax of Objective-C 1.0, the language
3196 recognized by GCC 4.0. This only affects the Objective-C additions to
3197 the C/C++ language; it does not affect conformance to C/C++ standards,
3198 which is controlled by the separate C/C++ dialect option flags. When
3199 this option is used with the Objective-C or Objective-C++ compiler,
3200 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3201 This is useful if you need to make sure that your Objective-C code can
3202 be compiled with older versions of GCC@.
3204 @item -freplace-objc-classes
3205 @opindex freplace-objc-classes
3206 Emit a special marker instructing @command{ld(1)} not to statically link in
3207 the resulting object file, and allow @command{dyld(1)} to load it in at
3208 run time instead. This is used in conjunction with the Fix-and-Continue
3209 debugging mode, where the object file in question may be recompiled and
3210 dynamically reloaded in the course of program execution, without the need
3211 to restart the program itself. Currently, Fix-and-Continue functionality
3212 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3217 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3218 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3219 compile time) with static class references that get initialized at load time,
3220 which improves run-time performance. Specifying the @option{-fzero-link} flag
3221 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3222 to be retained. This is useful in Zero-Link debugging mode, since it allows
3223 for individual class implementations to be modified during program execution.
3224 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3225 regardless of command-line options.
3227 @item -fno-local-ivars
3228 @opindex fno-local-ivars
3229 @opindex flocal-ivars
3230 By default instance variables in Objective-C can be accessed as if
3231 they were local variables from within the methods of the class they're
3232 declared in. This can lead to shadowing between instance variables
3233 and other variables declared either locally inside a class method or
3234 globally with the same name. Specifying the @option{-fno-local-ivars}
3235 flag disables this behavior thus avoiding variable shadowing issues.
3237 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3238 @opindex fivar-visibility
3239 Set the default instance variable visibility to the specified option
3240 so that instance variables declared outside the scope of any access
3241 modifier directives default to the specified visibility.
3245 Dump interface declarations for all classes seen in the source file to a
3246 file named @file{@var{sourcename}.decl}.
3248 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3249 @opindex Wassign-intercept
3250 @opindex Wno-assign-intercept
3251 Warn whenever an Objective-C assignment is being intercepted by the
3254 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3255 @opindex Wno-protocol
3257 If a class is declared to implement a protocol, a warning is issued for
3258 every method in the protocol that is not implemented by the class. The
3259 default behavior is to issue a warning for every method not explicitly
3260 implemented in the class, even if a method implementation is inherited
3261 from the superclass. If you use the @option{-Wno-protocol} option, then
3262 methods inherited from the superclass are considered to be implemented,
3263 and no warning is issued for them.
3265 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3267 @opindex Wno-selector
3268 Warn if multiple methods of different types for the same selector are
3269 found during compilation. The check is performed on the list of methods
3270 in the final stage of compilation. Additionally, a check is performed
3271 for each selector appearing in a @code{@@selector(@dots{})}
3272 expression, and a corresponding method for that selector has been found
3273 during compilation. Because these checks scan the method table only at
3274 the end of compilation, these warnings are not produced if the final
3275 stage of compilation is not reached, for example because an error is
3276 found during compilation, or because the @option{-fsyntax-only} option is
3279 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3280 @opindex Wstrict-selector-match
3281 @opindex Wno-strict-selector-match
3282 Warn if multiple methods with differing argument and/or return types are
3283 found for a given selector when attempting to send a message using this
3284 selector to a receiver of type @code{id} or @code{Class}. When this flag
3285 is off (which is the default behavior), the compiler omits such warnings
3286 if any differences found are confined to types that share the same size
3289 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3290 @opindex Wundeclared-selector
3291 @opindex Wno-undeclared-selector
3292 Warn if a @code{@@selector(@dots{})} expression referring to an
3293 undeclared selector is found. A selector is considered undeclared if no
3294 method with that name has been declared before the
3295 @code{@@selector(@dots{})} expression, either explicitly in an
3296 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3297 an @code{@@implementation} section. This option always performs its
3298 checks as soon as a @code{@@selector(@dots{})} expression is found,
3299 while @option{-Wselector} only performs its checks in the final stage of
3300 compilation. This also enforces the coding style convention
3301 that methods and selectors must be declared before being used.
3303 @item -print-objc-runtime-info
3304 @opindex print-objc-runtime-info
3305 Generate C header describing the largest structure that is passed by
3310 @node Diagnostic Message Formatting Options
3311 @section Options to Control Diagnostic Messages Formatting
3312 @cindex options to control diagnostics formatting
3313 @cindex diagnostic messages
3314 @cindex message formatting
3316 Traditionally, diagnostic messages have been formatted irrespective of
3317 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3318 options described below
3319 to control the formatting algorithm for diagnostic messages,
3320 e.g.@: how many characters per line, how often source location
3321 information should be reported. Note that some language front ends may not
3322 honor these options.
3325 @item -fmessage-length=@var{n}
3326 @opindex fmessage-length
3327 Try to format error messages so that they fit on lines of about
3328 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3329 done; each error message appears on a single line. This is the
3330 default for all front ends.
3332 @item -fdiagnostics-show-location=once
3333 @opindex fdiagnostics-show-location
3334 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3335 reporter to emit source location information @emph{once}; that is, in
3336 case the message is too long to fit on a single physical line and has to
3337 be wrapped, the source location won't be emitted (as prefix) again,
3338 over and over, in subsequent continuation lines. This is the default
3341 @item -fdiagnostics-show-location=every-line
3342 Only meaningful in line-wrapping mode. Instructs the diagnostic
3343 messages reporter to emit the same source location information (as
3344 prefix) for physical lines that result from the process of breaking
3345 a message which is too long to fit on a single line.
3347 @item -fdiagnostics-color[=@var{WHEN}]
3348 @itemx -fno-diagnostics-color
3349 @opindex fdiagnostics-color
3350 @cindex highlight, color
3351 @vindex GCC_COLORS @r{environment variable}
3352 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3353 or @samp{auto}. The default depends on how the compiler has been configured,
3354 it can be any of the above @var{WHEN} options or also @samp{never}
3355 if @env{GCC_COLORS} environment variable isn't present in the environment,
3356 and @samp{auto} otherwise.
3357 @samp{auto} means to use color only when the standard error is a terminal.
3358 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3359 aliases for @option{-fdiagnostics-color=always} and
3360 @option{-fdiagnostics-color=never}, respectively.
3362 The colors are defined by the environment variable @env{GCC_COLORS}.
3363 Its value is a colon-separated list of capabilities and Select Graphic
3364 Rendition (SGR) substrings. SGR commands are interpreted by the
3365 terminal or terminal emulator. (See the section in the documentation
3366 of your text terminal for permitted values and their meanings as
3367 character attributes.) These substring values are integers in decimal
3368 representation and can be concatenated with semicolons.
3369 Common values to concatenate include
3371 @samp{4} for underline,
3373 @samp{7} for inverse,
3374 @samp{39} for default foreground color,
3375 @samp{30} to @samp{37} for foreground colors,
3376 @samp{90} to @samp{97} for 16-color mode foreground colors,
3377 @samp{38;5;0} to @samp{38;5;255}
3378 for 88-color and 256-color modes foreground colors,
3379 @samp{49} for default background color,
3380 @samp{40} to @samp{47} for background colors,
3381 @samp{100} to @samp{107} for 16-color mode background colors,
3382 and @samp{48;5;0} to @samp{48;5;255}
3383 for 88-color and 256-color modes background colors.
3385 The default @env{GCC_COLORS} is
3387 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:quote=01:\
3388 fixit-insert=32:fixit-delete=31:\
3389 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32
3392 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3393 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3394 @samp{01} is bold, and @samp{31} is red.
3395 Setting @env{GCC_COLORS} to the empty string disables colors.
3396 Supported capabilities are as follows.
3400 @vindex error GCC_COLORS @r{capability}
3401 SGR substring for error: markers.
3404 @vindex warning GCC_COLORS @r{capability}
3405 SGR substring for warning: markers.
3408 @vindex note GCC_COLORS @r{capability}
3409 SGR substring for note: markers.
3412 @vindex range1 GCC_COLORS @r{capability}
3413 SGR substring for first additional range.
3416 @vindex range2 GCC_COLORS @r{capability}
3417 SGR substring for second additional range.
3420 @vindex locus GCC_COLORS @r{capability}
3421 SGR substring for location information, @samp{file:line} or
3422 @samp{file:line:column} etc.
3425 @vindex quote GCC_COLORS @r{capability}
3426 SGR substring for information printed within quotes.
3429 @vindex fixit-insert GCC_COLORS @r{capability}
3430 SGR substring for fix-it hints suggesting text to
3431 be inserted or replaced.
3434 @vindex fixit-delete GCC_COLORS @r{capability}
3435 SGR substring for fix-it hints suggesting text to
3438 @item diff-filename=
3439 @vindex diff-filename GCC_COLORS @r{capability}
3440 SGR substring for filename headers within generated patches.
3443 @vindex diff-hunk GCC_COLORS @r{capability}
3444 SGR substring for the starts of hunks within generated patches.
3447 @vindex diff-delete GCC_COLORS @r{capability}
3448 SGR substring for deleted lines within generated patches.
3451 @vindex diff-insert GCC_COLORS @r{capability}
3452 SGR substring for inserted lines within generated patches.
3455 @item -fno-diagnostics-show-option
3456 @opindex fno-diagnostics-show-option
3457 @opindex fdiagnostics-show-option
3458 By default, each diagnostic emitted includes text indicating the
3459 command-line option that directly controls the diagnostic (if such an
3460 option is known to the diagnostic machinery). Specifying the
3461 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3463 @item -fno-diagnostics-show-caret
3464 @opindex fno-diagnostics-show-caret
3465 @opindex fdiagnostics-show-caret
3466 By default, each diagnostic emitted includes the original source line
3467 and a caret @samp{^} indicating the column. This option suppresses this
3468 information. The source line is truncated to @var{n} characters, if
3469 the @option{-fmessage-length=n} option is given. When the output is done
3470 to the terminal, the width is limited to the width given by the
3471 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3473 @item -fdiagnostics-parseable-fixits
3474 @opindex fdiagnostics-parseable-fixits
3475 Emit fix-it hints in a machine-parseable format, suitable for consumption
3476 by IDEs. For each fix-it, a line will be printed after the relevant
3477 diagnostic, starting with the string ``fix-it:''. For example:
3480 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3483 The location is expressed as a half-open range, expressed as a count of
3484 bytes, starting at byte 1 for the initial column. In the above example,
3485 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3489 00000000011111111112222222222
3490 12345678901234567890123456789
3491 gtk_widget_showall (dlg);
3496 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3497 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3498 (e.g. vertical tab as ``\013'').
3500 An empty replacement string indicates that the given range is to be removed.
3501 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3502 be inserted at the given position.
3504 @item -fdiagnostics-generate-patch
3505 @opindex fdiagnostics-generate-patch
3506 Print fix-it hints to stderr in unified diff format, after any diagnostics
3507 are printed. For example:
3514 void show_cb(GtkDialog *dlg)
3516 - gtk_widget_showall(dlg);
3517 + gtk_widget_show_all(dlg);
3522 The diff may or may not be colorized, following the same rules
3523 as for diagnostics (see @option{-fdiagnostics-color}).
3527 @node Warning Options
3528 @section Options to Request or Suppress Warnings
3529 @cindex options to control warnings
3530 @cindex warning messages
3531 @cindex messages, warning
3532 @cindex suppressing warnings
3534 Warnings are diagnostic messages that report constructions that
3535 are not inherently erroneous but that are risky or suggest there
3536 may have been an error.
3538 The following language-independent options do not enable specific
3539 warnings but control the kinds of diagnostics produced by GCC@.
3542 @cindex syntax checking
3544 @opindex fsyntax-only
3545 Check the code for syntax errors, but don't do anything beyond that.
3547 @item -fmax-errors=@var{n}
3548 @opindex fmax-errors
3549 Limits the maximum number of error messages to @var{n}, at which point
3550 GCC bails out rather than attempting to continue processing the source
3551 code. If @var{n} is 0 (the default), there is no limit on the number
3552 of error messages produced. If @option{-Wfatal-errors} is also
3553 specified, then @option{-Wfatal-errors} takes precedence over this
3558 Inhibit all warning messages.
3563 Make all warnings into errors.
3568 Make the specified warning into an error. The specifier for a warning
3569 is appended; for example @option{-Werror=switch} turns the warnings
3570 controlled by @option{-Wswitch} into errors. This switch takes a
3571 negative form, to be used to negate @option{-Werror} for specific
3572 warnings; for example @option{-Wno-error=switch} makes
3573 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3576 The warning message for each controllable warning includes the
3577 option that controls the warning. That option can then be used with
3578 @option{-Werror=} and @option{-Wno-error=} as described above.
3579 (Printing of the option in the warning message can be disabled using the
3580 @option{-fno-diagnostics-show-option} flag.)
3582 Note that specifying @option{-Werror=}@var{foo} automatically implies
3583 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3586 @item -Wfatal-errors
3587 @opindex Wfatal-errors
3588 @opindex Wno-fatal-errors
3589 This option causes the compiler to abort compilation on the first error
3590 occurred rather than trying to keep going and printing further error
3595 You can request many specific warnings with options beginning with
3596 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3597 implicit declarations. Each of these specific warning options also
3598 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3599 example, @option{-Wno-implicit}. This manual lists only one of the
3600 two forms, whichever is not the default. For further
3601 language-specific options also refer to @ref{C++ Dialect Options} and
3602 @ref{Objective-C and Objective-C++ Dialect Options}.
3604 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3605 options, such as @option{-Wunused}, which may turn on further options,
3606 such as @option{-Wunused-value}. The combined effect of positive and
3607 negative forms is that more specific options have priority over less
3608 specific ones, independently of their position in the command-line. For
3609 options of the same specificity, the last one takes effect. Options
3610 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3611 as if they appeared at the end of the command-line.
3613 When an unrecognized warning option is requested (e.g.,
3614 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3615 that the option is not recognized. However, if the @option{-Wno-} form
3616 is used, the behavior is slightly different: no diagnostic is
3617 produced for @option{-Wno-unknown-warning} unless other diagnostics
3618 are being produced. This allows the use of new @option{-Wno-} options
3619 with old compilers, but if something goes wrong, the compiler
3620 warns that an unrecognized option is present.
3627 Issue all the warnings demanded by strict ISO C and ISO C++;
3628 reject all programs that use forbidden extensions, and some other
3629 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3630 version of the ISO C standard specified by any @option{-std} option used.
3632 Valid ISO C and ISO C++ programs should compile properly with or without
3633 this option (though a rare few require @option{-ansi} or a
3634 @option{-std} option specifying the required version of ISO C)@. However,
3635 without this option, certain GNU extensions and traditional C and C++
3636 features are supported as well. With this option, they are rejected.
3638 @option{-Wpedantic} does not cause warning messages for use of the
3639 alternate keywords whose names begin and end with @samp{__}. Pedantic
3640 warnings are also disabled in the expression that follows
3641 @code{__extension__}. However, only system header files should use
3642 these escape routes; application programs should avoid them.
3643 @xref{Alternate Keywords}.
3645 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3646 C conformance. They soon find that it does not do quite what they want:
3647 it finds some non-ISO practices, but not all---only those for which
3648 ISO C @emph{requires} a diagnostic, and some others for which
3649 diagnostics have been added.
3651 A feature to report any failure to conform to ISO C might be useful in
3652 some instances, but would require considerable additional work and would
3653 be quite different from @option{-Wpedantic}. We don't have plans to
3654 support such a feature in the near future.
3656 Where the standard specified with @option{-std} represents a GNU
3657 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3658 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3659 extended dialect is based. Warnings from @option{-Wpedantic} are given
3660 where they are required by the base standard. (It does not make sense
3661 for such warnings to be given only for features not in the specified GNU
3662 C dialect, since by definition the GNU dialects of C include all
3663 features the compiler supports with the given option, and there would be
3664 nothing to warn about.)
3666 @item -pedantic-errors
3667 @opindex pedantic-errors
3668 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3669 requires a diagnostic, in some cases where there is undefined behavior
3670 at compile-time and in some other cases that do not prevent compilation
3671 of programs that are valid according to the standard. This is not
3672 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3673 by this option and not enabled by the latter and vice versa.
3678 This enables all the warnings about constructions that some users
3679 consider questionable, and that are easy to avoid (or modify to
3680 prevent the warning), even in conjunction with macros. This also
3681 enables some language-specific warnings described in @ref{C++ Dialect
3682 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3684 @option{-Wall} turns on the following warning flags:
3686 @gccoptlist{-Waddress @gol
3687 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3689 -Wbool-operation @gol
3690 -Wc++11-compat -Wc++14-compat@gol
3691 -Wchar-subscripts @gol
3693 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3694 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3696 -Wint-in-bool-context @gol
3697 -Wimplicit @r{(C and Objective-C only)} @gol
3698 -Wimplicit-int @r{(C and Objective-C only)} @gol
3699 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3700 -Winit-self @r{(only for C++)} @gol
3701 -Wlogical-not-parentheses
3702 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3703 -Wmaybe-uninitialized @gol
3704 -Wmemset-elt-size @gol
3705 -Wmemset-transposed-args @gol
3706 -Wmisleading-indentation @r{(only for C/C++)} @gol
3707 -Wmissing-braces @r{(only for C/ObjC)} @gol
3708 -Wnarrowing @r{(only for C++)} @gol
3710 -Wnonnull-compare @gol
3716 -Wsequence-point @gol
3717 -Wsign-compare @r{(only in C++)} @gol
3718 -Wsizeof-pointer-memaccess @gol
3719 -Wstrict-aliasing @gol
3720 -Wstrict-overflow=1 @gol
3722 -Wtautological-compare @gol
3724 -Wuninitialized @gol
3725 -Wunknown-pragmas @gol
3726 -Wunused-function @gol
3729 -Wunused-variable @gol
3730 -Wvolatile-register-var @gol
3733 Note that some warning flags are not implied by @option{-Wall}. Some of
3734 them warn about constructions that users generally do not consider
3735 questionable, but which occasionally you might wish to check for;
3736 others warn about constructions that are necessary or hard to avoid in
3737 some cases, and there is no simple way to modify the code to suppress
3738 the warning. Some of them are enabled by @option{-Wextra} but many of
3739 them must be enabled individually.
3745 This enables some extra warning flags that are not enabled by
3746 @option{-Wall}. (This option used to be called @option{-W}. The older
3747 name is still supported, but the newer name is more descriptive.)
3749 @gccoptlist{-Wclobbered @gol
3751 -Wignored-qualifiers @gol
3752 -Wimplicit-fallthrough=3 @gol
3753 -Wmissing-field-initializers @gol
3754 -Wmissing-parameter-type @r{(C only)} @gol
3755 -Wold-style-declaration @r{(C only)} @gol
3756 -Woverride-init @gol
3757 -Wsign-compare @r{(C only)} @gol
3759 -Wuninitialized @gol
3760 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3761 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3762 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3765 The option @option{-Wextra} also prints warning messages for the
3771 A pointer is compared against integer zero with @code{<}, @code{<=},
3772 @code{>}, or @code{>=}.
3775 (C++ only) An enumerator and a non-enumerator both appear in a
3776 conditional expression.
3779 (C++ only) Ambiguous virtual bases.
3782 (C++ only) Subscripting an array that has been declared @code{register}.
3785 (C++ only) Taking the address of a variable that has been declared
3789 (C++ only) A base class is not initialized in the copy constructor
3794 @item -Wchar-subscripts
3795 @opindex Wchar-subscripts
3796 @opindex Wno-char-subscripts
3797 Warn if an array subscript has type @code{char}. This is a common cause
3798 of error, as programmers often forget that this type is signed on some
3800 This warning is enabled by @option{-Wall}.
3804 @opindex Wno-comment
3805 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3806 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3807 This warning is enabled by @option{-Wall}.
3809 @item -Wno-coverage-mismatch
3810 @opindex Wno-coverage-mismatch
3811 Warn if feedback profiles do not match when using the
3812 @option{-fprofile-use} option.
3813 If a source file is changed between compiling with @option{-fprofile-gen} and
3814 with @option{-fprofile-use}, the files with the profile feedback can fail
3815 to match the source file and GCC cannot use the profile feedback
3816 information. By default, this warning is enabled and is treated as an
3817 error. @option{-Wno-coverage-mismatch} can be used to disable the
3818 warning or @option{-Wno-error=coverage-mismatch} can be used to
3819 disable the error. Disabling the error for this warning can result in
3820 poorly optimized code and is useful only in the
3821 case of very minor changes such as bug fixes to an existing code-base.
3822 Completely disabling the warning is not recommended.
3825 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3827 Suppress warning messages emitted by @code{#warning} directives.
3829 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3830 @opindex Wdouble-promotion
3831 @opindex Wno-double-promotion
3832 Give a warning when a value of type @code{float} is implicitly
3833 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3834 floating-point unit implement @code{float} in hardware, but emulate
3835 @code{double} in software. On such a machine, doing computations
3836 using @code{double} values is much more expensive because of the
3837 overhead required for software emulation.
3839 It is easy to accidentally do computations with @code{double} because
3840 floating-point literals are implicitly of type @code{double}. For
3844 float area(float radius)
3846 return 3.14159 * radius * radius;
3850 the compiler performs the entire computation with @code{double}
3851 because the floating-point literal is a @code{double}.
3853 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3854 @opindex Wduplicate-decl-specifier
3855 @opindex Wno-duplicate-decl-specifier
3856 Warn if a declaration has duplicate @code{const}, @code{volatile},
3857 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3861 @itemx -Wformat=@var{n}
3864 @opindex ffreestanding
3865 @opindex fno-builtin
3867 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3868 the arguments supplied have types appropriate to the format string
3869 specified, and that the conversions specified in the format string make
3870 sense. This includes standard functions, and others specified by format
3871 attributes (@pxref{Function Attributes}), in the @code{printf},
3872 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3873 not in the C standard) families (or other target-specific families).
3874 Which functions are checked without format attributes having been
3875 specified depends on the standard version selected, and such checks of
3876 functions without the attribute specified are disabled by
3877 @option{-ffreestanding} or @option{-fno-builtin}.
3879 The formats are checked against the format features supported by GNU
3880 libc version 2.2. These include all ISO C90 and C99 features, as well
3881 as features from the Single Unix Specification and some BSD and GNU
3882 extensions. Other library implementations may not support all these
3883 features; GCC does not support warning about features that go beyond a
3884 particular library's limitations. However, if @option{-Wpedantic} is used
3885 with @option{-Wformat}, warnings are given about format features not
3886 in the selected standard version (but not for @code{strfmon} formats,
3887 since those are not in any version of the C standard). @xref{C Dialect
3888 Options,,Options Controlling C Dialect}.
3895 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3896 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3897 @option{-Wformat} also checks for null format arguments for several
3898 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3899 aspects of this level of format checking can be disabled by the
3900 options: @option{-Wno-format-contains-nul},
3901 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3902 @option{-Wformat} is enabled by @option{-Wall}.
3904 @item -Wno-format-contains-nul
3905 @opindex Wno-format-contains-nul
3906 @opindex Wformat-contains-nul
3907 If @option{-Wformat} is specified, do not warn about format strings that
3910 @item -Wno-format-extra-args
3911 @opindex Wno-format-extra-args
3912 @opindex Wformat-extra-args
3913 If @option{-Wformat} is specified, do not warn about excess arguments to a
3914 @code{printf} or @code{scanf} format function. The C standard specifies
3915 that such arguments are ignored.
3917 Where the unused arguments lie between used arguments that are
3918 specified with @samp{$} operand number specifications, normally
3919 warnings are still given, since the implementation could not know what
3920 type to pass to @code{va_arg} to skip the unused arguments. However,
3921 in the case of @code{scanf} formats, this option suppresses the
3922 warning if the unused arguments are all pointers, since the Single
3923 Unix Specification says that such unused arguments are allowed.
3925 @item -Wformat-length
3926 @itemx -Wformat-length=@var{level}
3927 @opindex Wformat-length
3928 @opindex Wno-format-length
3929 Warn about calls to formatted input/output functions such as @code{sprintf}
3930 that might overflow the destination buffer, or about bounded functions such
3931 as @code{snprintf} that might result in output truncation. When the exact
3932 number of bytes written by a format directive cannot be determined at
3933 compile-time it is estimated based on heuristics that depend on the
3934 @var{level} argument and on optimization. While enabling optimization
3935 will in most cases improve the accuracy of the warning, it may also
3936 result in false positives.
3939 @item -Wformat-length
3940 @item -Wformat-length=1
3941 @opindex Wformat-length
3942 @opindex Wno-format-length
3943 Level @var{1} of @option{-Wformat-length} enabled by @option{-Wformat}
3944 employs a conservative approach that warns only about calls that most
3945 likely overflow the buffer or result in output truncation. At this
3946 level, numeric arguments to format directives with unknown values are
3947 assumed to have the value of one, and strings of unknown length to be
3948 empty. Numeric arguments that are known to be bounded to a subrange
3949 of their type, or string arguments whose output is bounded either by
3950 their directive's precision or by a finite set of string literals, are
3951 assumed to take on the value within the range that results in the most
3952 bytes on output. For example, the call to @code{sprintf} below is
3953 diagnosed because even with both @var{a} and @var{b} equal to zero,
3954 the terminating NUL character (@code{'\0'}) appended by the function
3955 to the destination buffer will be written past its end. Increasing
3956 the size of the buffer by a single byte is sufficient to avoid the
3957 warning, though it may not be sufficient to avoid the overflow.
3960 void f (int a, int b)
3963 sprintf (buf, "a = %i, b = %i\n", a, b);
3967 @item -Wformat-length=2
3968 Level @var{2} warns also about calls that might overflow the destination
3969 buffer or result in truncation given an argument of sufficient length
3970 or magnitude. At level @var{2}, unknown numeric arguments are assumed
3971 to have the minimum representable value for signed types with a precision
3972 greater than 1, and the maximum representable value otherwise. Unknown
3973 string arguments whose length cannot be assumed to be bounded either by
3974 the directive's precision, or by a finite set of string literals they
3975 may evaluate to, or the character array they may point to, are assumed
3976 to be 1 character long.
3978 At level @var{2}, the call in the example above is again diagnosed, but
3979 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
3980 @code{%i} directive will write some of its digits beyond the end of
3981 the destination buffer. To make the call safe regardless of the values
3982 of the two variables, the size of the destination buffer must be increased
3983 to at least 34 bytes. GCC includes the minimum size of the buffer in
3984 an informational note following the warning.
3986 An alternative to increasing the size of the destination buffer is to
3987 constrain the range of formatted values. The maximum length of string
3988 arguments can be bounded by specifying the precision in the format
3989 directive. When numeric arguments of format directives can be assumed
3990 to be bounded by less than the precision of their type, choosing
3991 an appropriate length modifier to the format specifier will reduce
3992 the required buffer size. For example, if @var{a} and @var{b} in the
3993 example above can be assumed to be within the precision of
3994 the @code{short int} type then using either the @code{%hi} format
3995 directive or casting the argument to @code{short} reduces the maximum
3996 required size of the buffer to 24 bytes.
3999 void f (int a, int b)
4002 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4007 @item -Wno-format-zero-length
4008 @opindex Wno-format-zero-length
4009 @opindex Wformat-zero-length
4010 If @option{-Wformat} is specified, do not warn about zero-length formats.
4011 The C standard specifies that zero-length formats are allowed.
4016 Enable @option{-Wformat} plus additional format checks. Currently
4017 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4020 @item -Wformat-nonliteral
4021 @opindex Wformat-nonliteral
4022 @opindex Wno-format-nonliteral
4023 If @option{-Wformat} is specified, also warn if the format string is not a
4024 string literal and so cannot be checked, unless the format function
4025 takes its format arguments as a @code{va_list}.
4027 @item -Wformat-security
4028 @opindex Wformat-security
4029 @opindex Wno-format-security
4030 If @option{-Wformat} is specified, also warn about uses of format
4031 functions that represent possible security problems. At present, this
4032 warns about calls to @code{printf} and @code{scanf} functions where the
4033 format string is not a string literal and there are no format arguments,
4034 as in @code{printf (foo);}. This may be a security hole if the format
4035 string came from untrusted input and contains @samp{%n}. (This is
4036 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4037 in future warnings may be added to @option{-Wformat-security} that are not
4038 included in @option{-Wformat-nonliteral}.)
4040 @item -Wformat-signedness
4041 @opindex Wformat-signedness
4042 @opindex Wno-format-signedness
4043 If @option{-Wformat} is specified, also warn if the format string
4044 requires an unsigned argument and the argument is signed and vice versa.
4047 @opindex Wformat-y2k
4048 @opindex Wno-format-y2k
4049 If @option{-Wformat} is specified, also warn about @code{strftime}
4050 formats that may yield only a two-digit year.
4055 @opindex Wno-nonnull
4056 Warn about passing a null pointer for arguments marked as
4057 requiring a non-null value by the @code{nonnull} function attribute.
4059 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4060 can be disabled with the @option{-Wno-nonnull} option.
4062 @item -Wnonnull-compare
4063 @opindex Wnonnull-compare
4064 @opindex Wno-nonnull-compare
4065 Warn when comparing an argument marked with the @code{nonnull}
4066 function attribute against null inside the function.
4068 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4069 can be disabled with the @option{-Wno-nonnull-compare} option.
4071 @item -Wnull-dereference
4072 @opindex Wnull-dereference
4073 @opindex Wno-null-dereference
4074 Warn if the compiler detects paths that trigger erroneous or
4075 undefined behavior due to dereferencing a null pointer. This option
4076 is only active when @option{-fdelete-null-pointer-checks} is active,
4077 which is enabled by optimizations in most targets. The precision of
4078 the warnings depends on the optimization options used.
4080 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4082 @opindex Wno-init-self
4083 Warn about uninitialized variables that are initialized with themselves.
4084 Note this option can only be used with the @option{-Wuninitialized} option.
4086 For example, GCC warns about @code{i} being uninitialized in the
4087 following snippet only when @option{-Winit-self} has been specified:
4098 This warning is enabled by @option{-Wall} in C++.
4100 @item -Wimplicit-int @r{(C and Objective-C only)}
4101 @opindex Wimplicit-int
4102 @opindex Wno-implicit-int
4103 Warn when a declaration does not specify a type.
4104 This warning is enabled by @option{-Wall}.
4106 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4107 @opindex Wimplicit-function-declaration
4108 @opindex Wno-implicit-function-declaration
4109 Give a warning whenever a function is used before being declared. In
4110 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4111 enabled by default and it is made into an error by
4112 @option{-pedantic-errors}. This warning is also enabled by
4115 @item -Wimplicit @r{(C and Objective-C only)}
4117 @opindex Wno-implicit
4118 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4119 This warning is enabled by @option{-Wall}.
4121 @item -Wimplicit-fallthrough
4122 @opindex Wimplicit-fallthrough
4123 @opindex Wno-implicit-fallthrough
4124 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4125 and @option{-Wno-implicit-fallthrough} is the same as
4126 @option{-Wimplicit-fallthrough=0}.
4128 @item -Wimplicit-fallthrough=@var{n}
4129 @opindex Wimplicit-fallthrough=
4130 Warn when a switch case falls through. For example:
4148 This warning does not warn when the last statement of a case cannot
4149 fall through, e.g. when there is a return statement or a call to function
4150 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4151 also takes into account control flow statements, such as ifs, and only
4152 warns when appropriate. E.g.@:
4162 @} else if (i < 1) @{
4172 Since there are occasions where a switch case fall through is desirable,
4173 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4174 to be used along with a null statement to suppress this warning that
4175 would normally occur:
4183 __attribute__ ((fallthrough));
4190 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4191 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4192 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4193 Instead of the these attributes, it is also possible to add a fallthrough
4194 comment to silence the warning. The whole body of the C or C++ style comment
4195 should match the given regular expressions listed below. The option argument
4196 @var{n} specifies what kind of comments are accepted:
4200 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4202 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4203 expression, any comment is used as fallthrough comment.
4205 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4206 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4208 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4209 following regular expressions:
4213 @item @code{-fallthrough}
4215 @item @code{@@fallthrough@@}
4217 @item @code{lint -fallthrough[ \t]*}
4219 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4221 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4223 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4227 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4228 following regular expressions:
4232 @item @code{-fallthrough}
4234 @item @code{@@fallthrough@@}
4236 @item @code{lint -fallthrough[ \t]*}
4238 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4242 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4243 fallthrough comments, only attributes disable the warning.
4247 The comment needs to be followed after optional whitespace and other comments
4248 by @code{case} or @code{default} keywords or by a user label that preceeds some
4249 @code{case} or @code{default} label.
4264 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4266 @item -Wignored-qualifiers @r{(C and C++ only)}
4267 @opindex Wignored-qualifiers
4268 @opindex Wno-ignored-qualifiers
4269 Warn if the return type of a function has a type qualifier
4270 such as @code{const}. For ISO C such a type qualifier has no effect,
4271 since the value returned by a function is not an lvalue.
4272 For C++, the warning is only emitted for scalar types or @code{void}.
4273 ISO C prohibits qualified @code{void} return types on function
4274 definitions, so such return types always receive a warning
4275 even without this option.
4277 This warning is also enabled by @option{-Wextra}.
4279 @item -Wignored-attributes @r{(C and C++ only)}
4280 @opindex Wignored-attributes
4281 @opindex Wno-ignored-attributes
4282 Warn when an attribute is ignored. This is different from the
4283 @option{-Wattributes} option in that it warns whenever the compiler decides
4284 to drop an attribute, not that the attribute is either unknown, used in a
4285 wrong place, etc. This warning is enabled by default.
4290 Warn if the type of @code{main} is suspicious. @code{main} should be
4291 a function with external linkage, returning int, taking either zero
4292 arguments, two, or three arguments of appropriate types. This warning
4293 is enabled by default in C++ and is enabled by either @option{-Wall}
4294 or @option{-Wpedantic}.
4296 @item -Wmisleading-indentation @r{(C and C++ only)}
4297 @opindex Wmisleading-indentation
4298 @opindex Wno-misleading-indentation
4299 Warn when the indentation of the code does not reflect the block structure.
4300 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4301 @code{for} clauses with a guarded statement that does not use braces,
4302 followed by an unguarded statement with the same indentation.
4304 In the following example, the call to ``bar'' is misleadingly indented as
4305 if it were guarded by the ``if'' conditional.
4308 if (some_condition ())
4310 bar (); /* Gotcha: this is not guarded by the "if". */
4313 In the case of mixed tabs and spaces, the warning uses the
4314 @option{-ftabstop=} option to determine if the statements line up
4317 The warning is not issued for code involving multiline preprocessor logic
4318 such as the following example.
4323 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4329 The warning is not issued after a @code{#line} directive, since this
4330 typically indicates autogenerated code, and no assumptions can be made
4331 about the layout of the file that the directive references.
4333 This warning is enabled by @option{-Wall} in C and C++.
4335 @item -Wmissing-braces
4336 @opindex Wmissing-braces
4337 @opindex Wno-missing-braces
4338 Warn if an aggregate or union initializer is not fully bracketed. In
4339 the following example, the initializer for @code{a} is not fully
4340 bracketed, but that for @code{b} is fully bracketed. This warning is
4341 enabled by @option{-Wall} in C.
4344 int a[2][2] = @{ 0, 1, 2, 3 @};
4345 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4348 This warning is enabled by @option{-Wall}.
4350 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4351 @opindex Wmissing-include-dirs
4352 @opindex Wno-missing-include-dirs
4353 Warn if a user-supplied include directory does not exist.
4356 @opindex Wparentheses
4357 @opindex Wno-parentheses
4358 Warn if parentheses are omitted in certain contexts, such
4359 as when there is an assignment in a context where a truth value
4360 is expected, or when operators are nested whose precedence people
4361 often get confused about.
4363 Also warn if a comparison like @code{x<=y<=z} appears; this is
4364 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4365 interpretation from that of ordinary mathematical notation.
4367 Also warn for dangerous uses of the GNU extension to
4368 @code{?:} with omitted middle operand. When the condition
4369 in the @code{?}: operator is a boolean expression, the omitted value is
4370 always 1. Often programmers expect it to be a value computed
4371 inside the conditional expression instead.
4373 This warning is enabled by @option{-Wall}.
4375 @item -Wsequence-point
4376 @opindex Wsequence-point
4377 @opindex Wno-sequence-point
4378 Warn about code that may have undefined semantics because of violations
4379 of sequence point rules in the C and C++ standards.
4381 The C and C++ standards define the order in which expressions in a C/C++
4382 program are evaluated in terms of @dfn{sequence points}, which represent
4383 a partial ordering between the execution of parts of the program: those
4384 executed before the sequence point, and those executed after it. These
4385 occur after the evaluation of a full expression (one which is not part
4386 of a larger expression), after the evaluation of the first operand of a
4387 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4388 function is called (but after the evaluation of its arguments and the
4389 expression denoting the called function), and in certain other places.
4390 Other than as expressed by the sequence point rules, the order of
4391 evaluation of subexpressions of an expression is not specified. All
4392 these rules describe only a partial order rather than a total order,
4393 since, for example, if two functions are called within one expression
4394 with no sequence point between them, the order in which the functions
4395 are called is not specified. However, the standards committee have
4396 ruled that function calls do not overlap.
4398 It is not specified when between sequence points modifications to the
4399 values of objects take effect. Programs whose behavior depends on this
4400 have undefined behavior; the C and C++ standards specify that ``Between
4401 the previous and next sequence point an object shall have its stored
4402 value modified at most once by the evaluation of an expression.
4403 Furthermore, the prior value shall be read only to determine the value
4404 to be stored.''. If a program breaks these rules, the results on any
4405 particular implementation are entirely unpredictable.
4407 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4408 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4409 diagnosed by this option, and it may give an occasional false positive
4410 result, but in general it has been found fairly effective at detecting
4411 this sort of problem in programs.
4413 The C++17 standard will define the order of evaluation of operands in
4414 more cases: in particular it requires that the right-hand side of an
4415 assignment be evaluated before the left-hand side, so the above
4416 examples are no longer undefined. But this warning will still warn
4417 about them, to help people avoid writing code that is undefined in C
4418 and earlier revisions of C++.
4420 The standard is worded confusingly, therefore there is some debate
4421 over the precise meaning of the sequence point rules in subtle cases.
4422 Links to discussions of the problem, including proposed formal
4423 definitions, may be found on the GCC readings page, at
4424 @uref{http://gcc.gnu.org/@/readings.html}.
4426 This warning is enabled by @option{-Wall} for C and C++.
4428 @item -Wno-return-local-addr
4429 @opindex Wno-return-local-addr
4430 @opindex Wreturn-local-addr
4431 Do not warn about returning a pointer (or in C++, a reference) to a
4432 variable that goes out of scope after the function returns.
4435 @opindex Wreturn-type
4436 @opindex Wno-return-type
4437 Warn whenever a function is defined with a return type that defaults
4438 to @code{int}. Also warn about any @code{return} statement with no
4439 return value in a function whose return type is not @code{void}
4440 (falling off the end of the function body is considered returning
4443 For C only, warn about a @code{return} statement with an expression in a
4444 function whose return type is @code{void}, unless the expression type is
4445 also @code{void}. As a GNU extension, the latter case is accepted
4446 without a warning unless @option{-Wpedantic} is used.
4448 For C++, a function without return type always produces a diagnostic
4449 message, even when @option{-Wno-return-type} is specified. The only
4450 exceptions are @code{main} and functions defined in system headers.
4452 This warning is enabled by @option{-Wall}.
4454 @item -Wshift-count-negative
4455 @opindex Wshift-count-negative
4456 @opindex Wno-shift-count-negative
4457 Warn if shift count is negative. This warning is enabled by default.
4459 @item -Wshift-count-overflow
4460 @opindex Wshift-count-overflow
4461 @opindex Wno-shift-count-overflow
4462 Warn if shift count >= width of type. This warning is enabled by default.
4464 @item -Wshift-negative-value
4465 @opindex Wshift-negative-value
4466 @opindex Wno-shift-negative-value
4467 Warn if left shifting a negative value. This warning is enabled by
4468 @option{-Wextra} in C99 and C++11 modes (and newer).
4470 @item -Wshift-overflow
4471 @itemx -Wshift-overflow=@var{n}
4472 @opindex Wshift-overflow
4473 @opindex Wno-shift-overflow
4474 Warn about left shift overflows. This warning is enabled by
4475 default in C99 and C++11 modes (and newer).
4478 @item -Wshift-overflow=1
4479 This is the warning level of @option{-Wshift-overflow} and is enabled
4480 by default in C99 and C++11 modes (and newer). This warning level does
4481 not warn about left-shifting 1 into the sign bit. (However, in C, such
4482 an overflow is still rejected in contexts where an integer constant expression
4485 @item -Wshift-overflow=2
4486 This warning level also warns about left-shifting 1 into the sign bit,
4487 unless C++14 mode is active.
4493 Warn whenever a @code{switch} statement has an index of enumerated type
4494 and lacks a @code{case} for one or more of the named codes of that
4495 enumeration. (The presence of a @code{default} label prevents this
4496 warning.) @code{case} labels outside the enumeration range also
4497 provoke warnings when this option is used (even if there is a
4498 @code{default} label).
4499 This warning is enabled by @option{-Wall}.
4501 @item -Wswitch-default
4502 @opindex Wswitch-default
4503 @opindex Wno-switch-default
4504 Warn whenever a @code{switch} statement does not have a @code{default}
4508 @opindex Wswitch-enum
4509 @opindex Wno-switch-enum
4510 Warn whenever a @code{switch} statement has an index of enumerated type
4511 and lacks a @code{case} for one or more of the named codes of that
4512 enumeration. @code{case} labels outside the enumeration range also
4513 provoke warnings when this option is used. The only difference
4514 between @option{-Wswitch} and this option is that this option gives a
4515 warning about an omitted enumeration code even if there is a
4516 @code{default} label.
4519 @opindex Wswitch-bool
4520 @opindex Wno-switch-bool
4521 Warn whenever a @code{switch} statement has an index of boolean type
4522 and the case values are outside the range of a boolean type.
4523 It is possible to suppress this warning by casting the controlling
4524 expression to a type other than @code{bool}. For example:
4527 switch ((int) (a == 4))
4533 This warning is enabled by default for C and C++ programs.
4535 @item -Wswitch-unreachable
4536 @opindex Wswitch-unreachable
4537 @opindex Wno-switch-unreachable
4538 Warn whenever a @code{switch} statement contains statements between the
4539 controlling expression and the first case label, which will never be
4540 executed. For example:
4552 @option{-Wswitch-unreachable} does not warn if the statement between the
4553 controlling expression and the first case label is just a declaration:
4566 This warning is enabled by default for C and C++ programs.
4568 @item -Wsync-nand @r{(C and C++ only)}
4570 @opindex Wno-sync-nand
4571 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4572 built-in functions are used. These functions changed semantics in GCC 4.4.
4576 @opindex Wno-trigraphs
4577 Warn if any trigraphs are encountered that might change the meaning of
4578 the program (trigraphs within comments are not warned about).
4579 This warning is enabled by @option{-Wall}.
4581 @item -Wunused-but-set-parameter
4582 @opindex Wunused-but-set-parameter
4583 @opindex Wno-unused-but-set-parameter
4584 Warn whenever a function parameter is assigned to, but otherwise unused
4585 (aside from its declaration).
4587 To suppress this warning use the @code{unused} attribute
4588 (@pxref{Variable Attributes}).
4590 This warning is also enabled by @option{-Wunused} together with
4593 @item -Wunused-but-set-variable
4594 @opindex Wunused-but-set-variable
4595 @opindex Wno-unused-but-set-variable
4596 Warn whenever a local variable is assigned to, but otherwise unused
4597 (aside from its declaration).
4598 This warning is enabled by @option{-Wall}.
4600 To suppress this warning use the @code{unused} attribute
4601 (@pxref{Variable Attributes}).
4603 This warning is also enabled by @option{-Wunused}, which is enabled
4606 @item -Wunused-function
4607 @opindex Wunused-function
4608 @opindex Wno-unused-function
4609 Warn whenever a static function is declared but not defined or a
4610 non-inline static function is unused.
4611 This warning is enabled by @option{-Wall}.
4613 @item -Wunused-label
4614 @opindex Wunused-label
4615 @opindex Wno-unused-label
4616 Warn whenever a label is declared but not used.
4617 This warning is enabled by @option{-Wall}.
4619 To suppress this warning use the @code{unused} attribute
4620 (@pxref{Variable Attributes}).
4622 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4623 @opindex Wunused-local-typedefs
4624 Warn when a typedef locally defined in a function is not used.
4625 This warning is enabled by @option{-Wall}.
4627 @item -Wunused-parameter
4628 @opindex Wunused-parameter
4629 @opindex Wno-unused-parameter
4630 Warn whenever a function parameter is unused aside from its declaration.
4632 To suppress this warning use the @code{unused} attribute
4633 (@pxref{Variable Attributes}).
4635 @item -Wno-unused-result
4636 @opindex Wunused-result
4637 @opindex Wno-unused-result
4638 Do not warn if a caller of a function marked with attribute
4639 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4640 its return value. The default is @option{-Wunused-result}.
4642 @item -Wunused-variable
4643 @opindex Wunused-variable
4644 @opindex Wno-unused-variable
4645 Warn whenever a local or static variable is unused aside from its
4646 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4647 but not for C++. This warning is enabled by @option{-Wall}.
4649 To suppress this warning use the @code{unused} attribute
4650 (@pxref{Variable Attributes}).
4652 @item -Wunused-const-variable
4653 @itemx -Wunused-const-variable=@var{n}
4654 @opindex Wunused-const-variable
4655 @opindex Wno-unused-const-variable
4656 Warn whenever a constant static variable is unused aside from its declaration.
4657 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4658 for C, but not for C++. In C this declares variable storage, but in C++ this
4659 is not an error since const variables take the place of @code{#define}s.
4661 To suppress this warning use the @code{unused} attribute
4662 (@pxref{Variable Attributes}).
4665 @item -Wunused-const-variable=1
4666 This is the warning level that is enabled by @option{-Wunused-variable} for
4667 C. It warns only about unused static const variables defined in the main
4668 compilation unit, but not about static const variables declared in any
4671 @item -Wunused-const-variable=2
4672 This warning level also warns for unused constant static variables in
4673 headers (excluding system headers). This is the warning level of
4674 @option{-Wunused-const-variable} and must be explicitly requested since
4675 in C++ this isn't an error and in C it might be harder to clean up all
4679 @item -Wunused-value
4680 @opindex Wunused-value
4681 @opindex Wno-unused-value
4682 Warn whenever a statement computes a result that is explicitly not
4683 used. To suppress this warning cast the unused expression to
4684 @code{void}. This includes an expression-statement or the left-hand
4685 side of a comma expression that contains no side effects. For example,
4686 an expression such as @code{x[i,j]} causes a warning, while
4687 @code{x[(void)i,j]} does not.
4689 This warning is enabled by @option{-Wall}.
4694 All the above @option{-Wunused} options combined.
4696 In order to get a warning about an unused function parameter, you must
4697 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4698 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4700 @item -Wuninitialized
4701 @opindex Wuninitialized
4702 @opindex Wno-uninitialized
4703 Warn if an automatic variable is used without first being initialized
4704 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4705 warn if a non-static reference or non-static @code{const} member
4706 appears in a class without constructors.
4708 If you want to warn about code that uses the uninitialized value of the
4709 variable in its own initializer, use the @option{-Winit-self} option.
4711 These warnings occur for individual uninitialized or clobbered
4712 elements of structure, union or array variables as well as for
4713 variables that are uninitialized or clobbered as a whole. They do
4714 not occur for variables or elements declared @code{volatile}. Because
4715 these warnings depend on optimization, the exact variables or elements
4716 for which there are warnings depends on the precise optimization
4717 options and version of GCC used.
4719 Note that there may be no warning about a variable that is used only
4720 to compute a value that itself is never used, because such
4721 computations may be deleted by data flow analysis before the warnings
4724 @item -Winvalid-memory-model
4725 @opindex Winvalid-memory-model
4726 @opindex Wno-invalid-memory-model
4727 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4728 and the C11 atomic generic functions with a memory consistency argument
4729 that is either invalid for the operation or outside the range of values
4730 of the @code{memory_order} enumeration. For example, since the
4731 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4732 defined for the relaxed, release, and sequentially consistent memory
4733 orders the following code is diagnosed:
4738 __atomic_store_n (i, 0, memory_order_consume);
4742 @option{-Winvalid-memory-model} is enabled by default.
4744 @item -Wmaybe-uninitialized
4745 @opindex Wmaybe-uninitialized
4746 @opindex Wno-maybe-uninitialized
4747 For an automatic variable, if there exists a path from the function
4748 entry to a use of the variable that is initialized, but there exist
4749 some other paths for which the variable is not initialized, the compiler
4750 emits a warning if it cannot prove the uninitialized paths are not
4751 executed at run time. These warnings are made optional because GCC is
4752 not smart enough to see all the reasons why the code might be correct
4753 in spite of appearing to have an error. Here is one example of how
4774 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4775 always initialized, but GCC doesn't know this. To suppress the
4776 warning, you need to provide a default case with assert(0) or
4779 @cindex @code{longjmp} warnings
4780 This option also warns when a non-volatile automatic variable might be
4781 changed by a call to @code{longjmp}. These warnings as well are possible
4782 only in optimizing compilation.
4784 The compiler sees only the calls to @code{setjmp}. It cannot know
4785 where @code{longjmp} will be called; in fact, a signal handler could
4786 call it at any point in the code. As a result, you may get a warning
4787 even when there is in fact no problem because @code{longjmp} cannot
4788 in fact be called at the place that would cause a problem.
4790 Some spurious warnings can be avoided if you declare all the functions
4791 you use that never return as @code{noreturn}. @xref{Function
4794 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4796 @item -Wunknown-pragmas
4797 @opindex Wunknown-pragmas
4798 @opindex Wno-unknown-pragmas
4799 @cindex warning for unknown pragmas
4800 @cindex unknown pragmas, warning
4801 @cindex pragmas, warning of unknown
4802 Warn when a @code{#pragma} directive is encountered that is not understood by
4803 GCC@. If this command-line option is used, warnings are even issued
4804 for unknown pragmas in system header files. This is not the case if
4805 the warnings are only enabled by the @option{-Wall} command-line option.
4808 @opindex Wno-pragmas
4810 Do not warn about misuses of pragmas, such as incorrect parameters,
4811 invalid syntax, or conflicts between pragmas. See also
4812 @option{-Wunknown-pragmas}.
4814 @item -Wstrict-aliasing
4815 @opindex Wstrict-aliasing
4816 @opindex Wno-strict-aliasing
4817 This option is only active when @option{-fstrict-aliasing} is active.
4818 It warns about code that might break the strict aliasing rules that the
4819 compiler is using for optimization. The warning does not catch all
4820 cases, but does attempt to catch the more common pitfalls. It is
4821 included in @option{-Wall}.
4822 It is equivalent to @option{-Wstrict-aliasing=3}
4824 @item -Wstrict-aliasing=n
4825 @opindex Wstrict-aliasing=n
4826 This option is only active when @option{-fstrict-aliasing} is active.
4827 It warns about code that might break the strict aliasing rules that the
4828 compiler is using for optimization.
4829 Higher levels correspond to higher accuracy (fewer false positives).
4830 Higher levels also correspond to more effort, similar to the way @option{-O}
4832 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4834 Level 1: Most aggressive, quick, least accurate.
4835 Possibly useful when higher levels
4836 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4837 false negatives. However, it has many false positives.
4838 Warns for all pointer conversions between possibly incompatible types,
4839 even if never dereferenced. Runs in the front end only.
4841 Level 2: Aggressive, quick, not too precise.
4842 May still have many false positives (not as many as level 1 though),
4843 and few false negatives (but possibly more than level 1).
4844 Unlike level 1, it only warns when an address is taken. Warns about
4845 incomplete types. Runs in the front end only.
4847 Level 3 (default for @option{-Wstrict-aliasing}):
4848 Should have very few false positives and few false
4849 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4850 Takes care of the common pun+dereference pattern in the front end:
4851 @code{*(int*)&some_float}.
4852 If optimization is enabled, it also runs in the back end, where it deals
4853 with multiple statement cases using flow-sensitive points-to information.
4854 Only warns when the converted pointer is dereferenced.
4855 Does not warn about incomplete types.
4857 @item -Wstrict-overflow
4858 @itemx -Wstrict-overflow=@var{n}
4859 @opindex Wstrict-overflow
4860 @opindex Wno-strict-overflow
4861 This option is only active when @option{-fstrict-overflow} is active.
4862 It warns about cases where the compiler optimizes based on the
4863 assumption that signed overflow does not occur. Note that it does not
4864 warn about all cases where the code might overflow: it only warns
4865 about cases where the compiler implements some optimization. Thus
4866 this warning depends on the optimization level.
4868 An optimization that assumes that signed overflow does not occur is
4869 perfectly safe if the values of the variables involved are such that
4870 overflow never does, in fact, occur. Therefore this warning can
4871 easily give a false positive: a warning about code that is not
4872 actually a problem. To help focus on important issues, several
4873 warning levels are defined. No warnings are issued for the use of
4874 undefined signed overflow when estimating how many iterations a loop
4875 requires, in particular when determining whether a loop will be
4879 @item -Wstrict-overflow=1
4880 Warn about cases that are both questionable and easy to avoid. For
4881 example, with @option{-fstrict-overflow}, the compiler simplifies
4882 @code{x + 1 > x} to @code{1}. This level of
4883 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4884 are not, and must be explicitly requested.
4886 @item -Wstrict-overflow=2
4887 Also warn about other cases where a comparison is simplified to a
4888 constant. For example: @code{abs (x) >= 0}. This can only be
4889 simplified when @option{-fstrict-overflow} is in effect, because
4890 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4891 zero. @option{-Wstrict-overflow} (with no level) is the same as
4892 @option{-Wstrict-overflow=2}.
4894 @item -Wstrict-overflow=3
4895 Also warn about other cases where a comparison is simplified. For
4896 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4898 @item -Wstrict-overflow=4
4899 Also warn about other simplifications not covered by the above cases.
4900 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4902 @item -Wstrict-overflow=5
4903 Also warn about cases where the compiler reduces the magnitude of a
4904 constant involved in a comparison. For example: @code{x + 2 > y} is
4905 simplified to @code{x + 1 >= y}. This is reported only at the
4906 highest warning level because this simplification applies to many
4907 comparisons, so this warning level gives a very large number of
4911 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4912 @opindex Wsuggest-attribute=
4913 @opindex Wno-suggest-attribute=
4914 Warn for cases where adding an attribute may be beneficial. The
4915 attributes currently supported are listed below.
4918 @item -Wsuggest-attribute=pure
4919 @itemx -Wsuggest-attribute=const
4920 @itemx -Wsuggest-attribute=noreturn
4921 @opindex Wsuggest-attribute=pure
4922 @opindex Wno-suggest-attribute=pure
4923 @opindex Wsuggest-attribute=const
4924 @opindex Wno-suggest-attribute=const
4925 @opindex Wsuggest-attribute=noreturn
4926 @opindex Wno-suggest-attribute=noreturn
4928 Warn about functions that might be candidates for attributes
4929 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4930 functions visible in other compilation units or (in the case of @code{pure} and
4931 @code{const}) if it cannot prove that the function returns normally. A function
4932 returns normally if it doesn't contain an infinite loop or return abnormally
4933 by throwing, calling @code{abort} or trapping. This analysis requires option
4934 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4935 higher. Higher optimization levels improve the accuracy of the analysis.
4937 @item -Wsuggest-attribute=format
4938 @itemx -Wmissing-format-attribute
4939 @opindex Wsuggest-attribute=format
4940 @opindex Wmissing-format-attribute
4941 @opindex Wno-suggest-attribute=format
4942 @opindex Wno-missing-format-attribute
4946 Warn about function pointers that might be candidates for @code{format}
4947 attributes. Note these are only possible candidates, not absolute ones.
4948 GCC guesses that function pointers with @code{format} attributes that
4949 are used in assignment, initialization, parameter passing or return
4950 statements should have a corresponding @code{format} attribute in the
4951 resulting type. I.e.@: the left-hand side of the assignment or
4952 initialization, the type of the parameter variable, or the return type
4953 of the containing function respectively should also have a @code{format}
4954 attribute to avoid the warning.
4956 GCC also warns about function definitions that might be
4957 candidates for @code{format} attributes. Again, these are only
4958 possible candidates. GCC guesses that @code{format} attributes
4959 might be appropriate for any function that calls a function like
4960 @code{vprintf} or @code{vscanf}, but this might not always be the
4961 case, and some functions for which @code{format} attributes are
4962 appropriate may not be detected.
4965 @item -Wsuggest-final-types
4966 @opindex Wno-suggest-final-types
4967 @opindex Wsuggest-final-types
4968 Warn about types with virtual methods where code quality would be improved
4969 if the type were declared with the C++11 @code{final} specifier,
4971 declared in an anonymous namespace. This allows GCC to more aggressively
4972 devirtualize the polymorphic calls. This warning is more effective with link
4973 time optimization, where the information about the class hierarchy graph is
4976 @item -Wsuggest-final-methods
4977 @opindex Wno-suggest-final-methods
4978 @opindex Wsuggest-final-methods
4979 Warn about virtual methods where code quality would be improved if the method
4980 were declared with the C++11 @code{final} specifier,
4981 or, if possible, its type were
4982 declared in an anonymous namespace or with the @code{final} specifier.
4984 more effective with link time optimization, where the information about the
4985 class hierarchy graph is more complete. It is recommended to first consider
4986 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4989 @item -Wsuggest-override
4990 Warn about overriding virtual functions that are not marked with the override
4996 This option warns on all uses of @code{alloca} in the source.
4998 @item -Walloca-larger-than=@var{n}
4999 This option warns on calls to @code{alloca} that are not bounded by a
5000 controlling predicate limiting its size to @var{n} bytes, or calls to
5001 @code{alloca} where the bound is unknown.
5003 For example, a bounded case of @code{alloca} could be:
5006 void func (size_t n)
5017 In the above example, passing @code{-Walloca=1000} would not issue a
5018 warning because the call to @code{alloca} is known to be at most 1000
5019 bytes. However, if @code{-Walloca=500} was passed, the compiler would
5020 have emitted a warning.
5022 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5023 controlling predicate verifying its size. For example:
5028 void *p = alloca (n);
5033 If @code{-Walloca=500} was passed, the above would trigger a warning,
5034 but this time because of the lack of bounds checking.
5036 Note, that even seemingly correct code involving signed integers could
5040 void func (signed int n)
5050 In the above example, @var{n} could be negative, causing a larger than
5051 expected argument to be implicitly casted into the @code{alloca} call.
5053 This option also warns when @code{alloca} is used in a loop.
5055 This warning is not enabled by @option{-Wall}, and is only active when
5056 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5058 See also @option{-Wvla-larger-than=@var{n}}.
5060 @item -Warray-bounds
5061 @itemx -Warray-bounds=@var{n}
5062 @opindex Wno-array-bounds
5063 @opindex Warray-bounds
5064 This option is only active when @option{-ftree-vrp} is active
5065 (default for @option{-O2} and above). It warns about subscripts to arrays
5066 that are always out of bounds. This warning is enabled by @option{-Wall}.
5069 @item -Warray-bounds=1
5070 This is the warning level of @option{-Warray-bounds} and is enabled
5071 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5073 @item -Warray-bounds=2
5074 This warning level also warns about out of bounds access for
5075 arrays at the end of a struct and for arrays accessed through
5076 pointers. This warning level may give a larger number of
5077 false positives and is deactivated by default.
5080 @item -Wbool-compare
5081 @opindex Wno-bool-compare
5082 @opindex Wbool-compare
5083 Warn about boolean expression compared with an integer value different from
5084 @code{true}/@code{false}. For instance, the following comparison is
5089 if ((n > 1) == 2) @{ @dots{} @}
5091 This warning is enabled by @option{-Wall}.
5093 @item -Wbool-operation
5094 @opindex Wno-bool-operation
5095 @opindex Wbool-operation
5096 Warn about suspicious operations on expressions of a boolean type. For
5097 instance, bitwise negation of a boolean is very likely a bug in the program.
5098 For C, this warning also warns about incrementing or decrementing a boolean,
5099 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5100 Incrementing a boolean is invalid in C++1z, and deprecated otherwise.)
5102 This warning is enabled by @option{-Wall}.
5104 @item -Wduplicated-cond
5105 @opindex Wno-duplicated-cond
5106 @opindex Wduplicated-cond
5107 Warn about duplicated conditions in an if-else-if chain. For instance,
5108 warn for the following code:
5110 if (p->q != NULL) @{ @dots{} @}
5111 else if (p->q != NULL) @{ @dots{} @}
5114 @item -Wframe-address
5115 @opindex Wno-frame-address
5116 @opindex Wframe-address
5117 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5118 is called with an argument greater than 0. Such calls may return indeterminate
5119 values or crash the program. The warning is included in @option{-Wall}.
5121 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5122 @opindex Wno-discarded-qualifiers
5123 @opindex Wdiscarded-qualifiers
5124 Do not warn if type qualifiers on pointers are being discarded.
5125 Typically, the compiler warns if a @code{const char *} variable is
5126 passed to a function that takes a @code{char *} parameter. This option
5127 can be used to suppress such a warning.
5129 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5130 @opindex Wno-discarded-array-qualifiers
5131 @opindex Wdiscarded-array-qualifiers
5132 Do not warn if type qualifiers on arrays which are pointer targets
5133 are being discarded. Typically, the compiler warns if a
5134 @code{const int (*)[]} variable is passed to a function that
5135 takes a @code{int (*)[]} parameter. This option can be used to
5136 suppress such a warning.
5138 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5139 @opindex Wno-incompatible-pointer-types
5140 @opindex Wincompatible-pointer-types
5141 Do not warn when there is a conversion between pointers that have incompatible
5142 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5143 which warns for pointer argument passing or assignment with different
5146 @item -Wno-int-conversion @r{(C and Objective-C only)}
5147 @opindex Wno-int-conversion
5148 @opindex Wint-conversion
5149 Do not warn about incompatible integer to pointer and pointer to integer
5150 conversions. This warning is about implicit conversions; for explicit
5151 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5152 @option{-Wno-pointer-to-int-cast} may be used.
5154 @item -Wno-div-by-zero
5155 @opindex Wno-div-by-zero
5156 @opindex Wdiv-by-zero
5157 Do not warn about compile-time integer division by zero. Floating-point
5158 division by zero is not warned about, as it can be a legitimate way of
5159 obtaining infinities and NaNs.
5161 @item -Wsystem-headers
5162 @opindex Wsystem-headers
5163 @opindex Wno-system-headers
5164 @cindex warnings from system headers
5165 @cindex system headers, warnings from
5166 Print warning messages for constructs found in system header files.
5167 Warnings from system headers are normally suppressed, on the assumption
5168 that they usually do not indicate real problems and would only make the
5169 compiler output harder to read. Using this command-line option tells
5170 GCC to emit warnings from system headers as if they occurred in user
5171 code. However, note that using @option{-Wall} in conjunction with this
5172 option does @emph{not} warn about unknown pragmas in system
5173 headers---for that, @option{-Wunknown-pragmas} must also be used.
5175 @item -Wtautological-compare
5176 @opindex Wtautological-compare
5177 @opindex Wno-tautological-compare
5178 Warn if a self-comparison always evaluates to true or false. This
5179 warning detects various mistakes such as:
5183 if (i > i) @{ @dots{} @}
5185 This warning is enabled by @option{-Wall}.
5188 @opindex Wtrampolines
5189 @opindex Wno-trampolines
5190 Warn about trampolines generated for pointers to nested functions.
5191 A trampoline is a small piece of data or code that is created at run
5192 time on the stack when the address of a nested function is taken, and is
5193 used to call the nested function indirectly. For some targets, it is
5194 made up of data only and thus requires no special treatment. But, for
5195 most targets, it is made up of code and thus requires the stack to be
5196 made executable in order for the program to work properly.
5199 @opindex Wfloat-equal
5200 @opindex Wno-float-equal
5201 Warn if floating-point values are used in equality comparisons.
5203 The idea behind this is that sometimes it is convenient (for the
5204 programmer) to consider floating-point values as approximations to
5205 infinitely precise real numbers. If you are doing this, then you need
5206 to compute (by analyzing the code, or in some other way) the maximum or
5207 likely maximum error that the computation introduces, and allow for it
5208 when performing comparisons (and when producing output, but that's a
5209 different problem). In particular, instead of testing for equality, you
5210 should check to see whether the two values have ranges that overlap; and
5211 this is done with the relational operators, so equality comparisons are
5214 @item -Wtraditional @r{(C and Objective-C only)}
5215 @opindex Wtraditional
5216 @opindex Wno-traditional
5217 Warn about certain constructs that behave differently in traditional and
5218 ISO C@. Also warn about ISO C constructs that have no traditional C
5219 equivalent, and/or problematic constructs that should be avoided.
5223 Macro parameters that appear within string literals in the macro body.
5224 In traditional C macro replacement takes place within string literals,
5225 but in ISO C it does not.
5228 In traditional C, some preprocessor directives did not exist.
5229 Traditional preprocessors only considered a line to be a directive
5230 if the @samp{#} appeared in column 1 on the line. Therefore
5231 @option{-Wtraditional} warns about directives that traditional C
5232 understands but ignores because the @samp{#} does not appear as the
5233 first character on the line. It also suggests you hide directives like
5234 @code{#pragma} not understood by traditional C by indenting them. Some
5235 traditional implementations do not recognize @code{#elif}, so this option
5236 suggests avoiding it altogether.
5239 A function-like macro that appears without arguments.
5242 The unary plus operator.
5245 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5246 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5247 constants.) Note, these suffixes appear in macros defined in the system
5248 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5249 Use of these macros in user code might normally lead to spurious
5250 warnings, however GCC's integrated preprocessor has enough context to
5251 avoid warning in these cases.
5254 A function declared external in one block and then used after the end of
5258 A @code{switch} statement has an operand of type @code{long}.
5261 A non-@code{static} function declaration follows a @code{static} one.
5262 This construct is not accepted by some traditional C compilers.
5265 The ISO type of an integer constant has a different width or
5266 signedness from its traditional type. This warning is only issued if
5267 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5268 typically represent bit patterns, are not warned about.
5271 Usage of ISO string concatenation is detected.
5274 Initialization of automatic aggregates.
5277 Identifier conflicts with labels. Traditional C lacks a separate
5278 namespace for labels.
5281 Initialization of unions. If the initializer is zero, the warning is
5282 omitted. This is done under the assumption that the zero initializer in
5283 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5284 initializer warnings and relies on default initialization to zero in the
5288 Conversions by prototypes between fixed/floating-point values and vice
5289 versa. The absence of these prototypes when compiling with traditional
5290 C causes serious problems. This is a subset of the possible
5291 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5294 Use of ISO C style function definitions. This warning intentionally is
5295 @emph{not} issued for prototype declarations or variadic functions
5296 because these ISO C features appear in your code when using
5297 libiberty's traditional C compatibility macros, @code{PARAMS} and
5298 @code{VPARAMS}. This warning is also bypassed for nested functions
5299 because that feature is already a GCC extension and thus not relevant to
5300 traditional C compatibility.
5303 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5304 @opindex Wtraditional-conversion
5305 @opindex Wno-traditional-conversion
5306 Warn if a prototype causes a type conversion that is different from what
5307 would happen to the same argument in the absence of a prototype. This
5308 includes conversions of fixed point to floating and vice versa, and
5309 conversions changing the width or signedness of a fixed-point argument
5310 except when the same as the default promotion.
5312 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5313 @opindex Wdeclaration-after-statement
5314 @opindex Wno-declaration-after-statement
5315 Warn when a declaration is found after a statement in a block. This
5316 construct, known from C++, was introduced with ISO C99 and is by default
5317 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5322 Warn if an undefined identifier is evaluated in an @code{#if} directive.
5324 @item -Wno-endif-labels
5325 @opindex Wno-endif-labels
5326 @opindex Wendif-labels
5327 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
5332 Warn whenever a local variable or type declaration shadows another
5333 variable, parameter, type, class member (in C++), or instance variable
5334 (in Objective-C) or whenever a built-in function is shadowed. Note
5335 that in C++, the compiler warns if a local variable shadows an
5336 explicit typedef, but not if it shadows a struct/class/enum.
5337 Same as @option{-Wshadow=global}.
5339 @item -Wno-shadow-ivar @r{(Objective-C only)}
5340 @opindex Wno-shadow-ivar
5341 @opindex Wshadow-ivar
5342 Do not warn whenever a local variable shadows an instance variable in an
5345 @item -Wshadow=global
5346 @opindex Wshadow=local
5347 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5349 @item -Wshadow=local
5350 @opindex Wshadow=local
5351 Warn when a local variable shadows another local variable or parameter.
5352 This warning is enabled by @option{-Wshadow=global}.
5354 @item -Wshadow=compatible-local
5355 @opindex Wshadow=compatible-local
5356 Warn when a local variable shadows another local variable or parameter
5357 whose type is compatible with that of the shadowing variable. In C++,
5358 type compatibility here means the type of the shadowing variable can be
5359 converted to that of the shadowed variable. The creation of this flag
5360 (in addition to @option{-Wshadow=local}) is based on the idea that when
5361 a local variable shadows another one of incompatible type, it is most
5362 likely intentional, not a bug or typo, as shown in the following example:
5366 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5368 for (int i = 0; i < N; ++i)
5377 Since the two variable @code{i} in the example above have incompatible types,
5378 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5379 Because their types are incompatible, if a programmer accidentally uses one
5380 in place of the other, type checking will catch that and emit an error or
5381 warning. So not warning (about shadowing) in this case will not lead to
5382 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5383 possibly reduce the number of warnings triggered by intentional shadowing.
5385 This warning is enabled by @option{-Wshadow=local}.
5387 @item -Wlarger-than=@var{len}
5388 @opindex Wlarger-than=@var{len}
5389 @opindex Wlarger-than-@var{len}
5390 Warn whenever an object of larger than @var{len} bytes is defined.
5392 @item -Wframe-larger-than=@var{len}
5393 @opindex Wframe-larger-than
5394 Warn if the size of a function frame is larger than @var{len} bytes.
5395 The computation done to determine the stack frame size is approximate
5396 and not conservative.
5397 The actual requirements may be somewhat greater than @var{len}
5398 even if you do not get a warning. In addition, any space allocated
5399 via @code{alloca}, variable-length arrays, or related constructs
5400 is not included by the compiler when determining
5401 whether or not to issue a warning.
5403 @item -Wno-free-nonheap-object
5404 @opindex Wno-free-nonheap-object
5405 @opindex Wfree-nonheap-object
5406 Do not warn when attempting to free an object that was not allocated
5409 @item -Wstack-usage=@var{len}
5410 @opindex Wstack-usage
5411 Warn if the stack usage of a function might be larger than @var{len} bytes.
5412 The computation done to determine the stack usage is conservative.
5413 Any space allocated via @code{alloca}, variable-length arrays, or related
5414 constructs is included by the compiler when determining whether or not to
5417 The message is in keeping with the output of @option{-fstack-usage}.
5421 If the stack usage is fully static but exceeds the specified amount, it's:
5424 warning: stack usage is 1120 bytes
5427 If the stack usage is (partly) dynamic but bounded, it's:
5430 warning: stack usage might be 1648 bytes
5433 If the stack usage is (partly) dynamic and not bounded, it's:
5436 warning: stack usage might be unbounded
5440 @item -Wunsafe-loop-optimizations
5441 @opindex Wunsafe-loop-optimizations
5442 @opindex Wno-unsafe-loop-optimizations
5443 Warn if the loop cannot be optimized because the compiler cannot
5444 assume anything on the bounds of the loop indices. With
5445 @option{-funsafe-loop-optimizations} warn if the compiler makes
5448 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5449 @opindex Wno-pedantic-ms-format
5450 @opindex Wpedantic-ms-format
5451 When used in combination with @option{-Wformat}
5452 and @option{-pedantic} without GNU extensions, this option
5453 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5454 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5455 which depend on the MS runtime.
5458 @opindex Waligned-new
5459 @opindex Wno-aligned-new
5460 Warn about a new-expression of a type that requires greater alignment
5461 than the @code{alignof(std::max_align_t)} but uses an allocation
5462 function without an explicit alignment parameter. This option is
5463 enabled by @option{-Wall}.
5465 Normally this only warns about global allocation functions, but
5466 @option{-Waligned-new=all} also warns about class member allocation
5469 @item -Wplacement-new
5470 @itemx -Wplacement-new=@var{n}
5471 @opindex Wplacement-new
5472 @opindex Wno-placement-new
5473 Warn about placement new expressions with undefined behavior, such as
5474 constructing an object in a buffer that is smaller than the type of
5475 the object. For example, the placement new expression below is diagnosed
5476 because it attempts to construct an array of 64 integers in a buffer only
5482 This warning is enabled by default.
5485 @item -Wplacement-new=1
5486 This is the default warning level of @option{-Wplacement-new}. At this
5487 level the warning is not issued for some strictly undefined constructs that
5488 GCC allows as extensions for compatibility with legacy code. For example,
5489 the following @code{new} expression is not diagnosed at this level even
5490 though it has undefined behavior according to the C++ standard because
5491 it writes past the end of the one-element array.
5493 struct S @{ int n, a[1]; @};
5494 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5495 new (s->a)int [32]();
5498 @item -Wplacement-new=2
5499 At this level, in addition to diagnosing all the same constructs as at level
5500 1, a diagnostic is also issued for placement new expressions that construct
5501 an object in the last member of structure whose type is an array of a single
5502 element and whose size is less than the size of the object being constructed.
5503 While the previous example would be diagnosed, the following construct makes
5504 use of the flexible member array extension to avoid the warning at level 2.
5506 struct S @{ int n, a[]; @};
5507 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5508 new (s->a)int [32]();
5513 @item -Wpointer-arith
5514 @opindex Wpointer-arith
5515 @opindex Wno-pointer-arith
5516 Warn about anything that depends on the ``size of'' a function type or
5517 of @code{void}. GNU C assigns these types a size of 1, for
5518 convenience in calculations with @code{void *} pointers and pointers
5519 to functions. In C++, warn also when an arithmetic operation involves
5520 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5523 @opindex Wtype-limits
5524 @opindex Wno-type-limits
5525 Warn if a comparison is always true or always false due to the limited
5526 range of the data type, but do not warn for constant expressions. For
5527 example, warn if an unsigned variable is compared against zero with
5528 @code{<} or @code{>=}. This warning is also enabled by
5531 @item -Wbad-function-cast @r{(C and Objective-C only)}
5532 @opindex Wbad-function-cast
5533 @opindex Wno-bad-function-cast
5534 Warn when a function call is cast to a non-matching type.
5535 For example, warn if a call to a function returning an integer type
5536 is cast to a pointer type.
5538 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5539 @opindex Wc90-c99-compat
5540 @opindex Wno-c90-c99-compat
5541 Warn about features not present in ISO C90, but present in ISO C99.
5542 For instance, warn about use of variable length arrays, @code{long long}
5543 type, @code{bool} type, compound literals, designated initializers, and so
5544 on. This option is independent of the standards mode. Warnings are disabled
5545 in the expression that follows @code{__extension__}.
5547 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5548 @opindex Wc99-c11-compat
5549 @opindex Wno-c99-c11-compat
5550 Warn about features not present in ISO C99, but present in ISO C11.
5551 For instance, warn about use of anonymous structures and unions,
5552 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5553 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5554 and so on. This option is independent of the standards mode. Warnings are
5555 disabled in the expression that follows @code{__extension__}.
5557 @item -Wc++-compat @r{(C and Objective-C only)}
5558 @opindex Wc++-compat
5559 Warn about ISO C constructs that are outside of the common subset of
5560 ISO C and ISO C++, e.g.@: request for implicit conversion from
5561 @code{void *} to a pointer to non-@code{void} type.
5563 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5564 @opindex Wc++11-compat
5565 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5566 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5567 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5568 enabled by @option{-Wall}.
5570 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5571 @opindex Wc++14-compat
5572 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5573 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5577 @opindex Wno-cast-qual
5578 Warn whenever a pointer is cast so as to remove a type qualifier from
5579 the target type. For example, warn if a @code{const char *} is cast
5580 to an ordinary @code{char *}.
5582 Also warn when making a cast that introduces a type qualifier in an
5583 unsafe way. For example, casting @code{char **} to @code{const char **}
5584 is unsafe, as in this example:
5587 /* p is char ** value. */
5588 const char **q = (const char **) p;
5589 /* Assignment of readonly string to const char * is OK. */
5591 /* Now char** pointer points to read-only memory. */
5596 @opindex Wcast-align
5597 @opindex Wno-cast-align
5598 Warn whenever a pointer is cast such that the required alignment of the
5599 target is increased. For example, warn if a @code{char *} is cast to
5600 an @code{int *} on machines where integers can only be accessed at
5601 two- or four-byte boundaries.
5603 @item -Wwrite-strings
5604 @opindex Wwrite-strings
5605 @opindex Wno-write-strings
5606 When compiling C, give string constants the type @code{const
5607 char[@var{length}]} so that copying the address of one into a
5608 non-@code{const} @code{char *} pointer produces a warning. These
5609 warnings help you find at compile time code that can try to write
5610 into a string constant, but only if you have been very careful about
5611 using @code{const} in declarations and prototypes. Otherwise, it is
5612 just a nuisance. This is why we did not make @option{-Wall} request
5615 When compiling C++, warn about the deprecated conversion from string
5616 literals to @code{char *}. This warning is enabled by default for C++
5621 @opindex Wno-clobbered
5622 Warn for variables that might be changed by @code{longjmp} or
5623 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5625 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5626 @opindex Wconditionally-supported
5627 @opindex Wno-conditionally-supported
5628 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5631 @opindex Wconversion
5632 @opindex Wno-conversion
5633 Warn for implicit conversions that may alter a value. This includes
5634 conversions between real and integer, like @code{abs (x)} when
5635 @code{x} is @code{double}; conversions between signed and unsigned,
5636 like @code{unsigned ui = -1}; and conversions to smaller types, like
5637 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5638 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5639 changed by the conversion like in @code{abs (2.0)}. Warnings about
5640 conversions between signed and unsigned integers can be disabled by
5641 using @option{-Wno-sign-conversion}.
5643 For C++, also warn for confusing overload resolution for user-defined
5644 conversions; and conversions that never use a type conversion
5645 operator: conversions to @code{void}, the same type, a base class or a
5646 reference to them. Warnings about conversions between signed and
5647 unsigned integers are disabled by default in C++ unless
5648 @option{-Wsign-conversion} is explicitly enabled.
5650 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5651 @opindex Wconversion-null
5652 @opindex Wno-conversion-null
5653 Do not warn for conversions between @code{NULL} and non-pointer
5654 types. @option{-Wconversion-null} is enabled by default.
5656 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5657 @opindex Wzero-as-null-pointer-constant
5658 @opindex Wno-zero-as-null-pointer-constant
5659 Warn when a literal @samp{0} is used as null pointer constant. This can
5660 be useful to facilitate the conversion to @code{nullptr} in C++11.
5662 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5663 @opindex Wsubobject-linkage
5664 @opindex Wno-subobject-linkage
5665 Warn if a class type has a base or a field whose type uses the anonymous
5666 namespace or depends on a type with no linkage. If a type A depends on
5667 a type B with no or internal linkage, defining it in multiple
5668 translation units would be an ODR violation because the meaning of B
5669 is different in each translation unit. If A only appears in a single
5670 translation unit, the best way to silence the warning is to give it
5671 internal linkage by putting it in an anonymous namespace as well. The
5672 compiler doesn't give this warning for types defined in the main .C
5673 file, as those are unlikely to have multiple definitions.
5674 @option{-Wsubobject-linkage} is enabled by default.
5676 @item -Wdangling-else
5677 @opindex Wdangling-else
5678 @opindex Wno-dangling-else
5679 Warn about constructions where there may be confusion to which
5680 @code{if} statement an @code{else} branch belongs. Here is an example of
5695 In C/C++, every @code{else} branch belongs to the innermost possible
5696 @code{if} statement, which in this example is @code{if (b)}. This is
5697 often not what the programmer expected, as illustrated in the above
5698 example by indentation the programmer chose. When there is the
5699 potential for this confusion, GCC issues a warning when this flag
5700 is specified. To eliminate the warning, add explicit braces around
5701 the innermost @code{if} statement so there is no way the @code{else}
5702 can belong to the enclosing @code{if}. The resulting code
5719 This warning is enabled by @option{-Wparentheses}.
5723 @opindex Wno-date-time
5724 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5725 are encountered as they might prevent bit-wise-identical reproducible
5728 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5729 @opindex Wdelete-incomplete
5730 @opindex Wno-delete-incomplete
5731 Warn when deleting a pointer to incomplete type, which may cause
5732 undefined behavior at runtime. This warning is enabled by default.
5734 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5735 @opindex Wuseless-cast
5736 @opindex Wno-useless-cast
5737 Warn when an expression is casted to its own type.
5740 @opindex Wempty-body
5741 @opindex Wno-empty-body
5742 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5743 while} statement. This warning is also enabled by @option{-Wextra}.
5745 @item -Wenum-compare
5746 @opindex Wenum-compare
5747 @opindex Wno-enum-compare
5748 Warn about a comparison between values of different enumerated types.
5749 In C++ enumeral mismatches in conditional expressions are also
5750 diagnosed and the warning is enabled by default. In C this warning is
5751 enabled by @option{-Wall}.
5753 @item -Wjump-misses-init @r{(C, Objective-C only)}
5754 @opindex Wjump-misses-init
5755 @opindex Wno-jump-misses-init
5756 Warn if a @code{goto} statement or a @code{switch} statement jumps
5757 forward across the initialization of a variable, or jumps backward to a
5758 label after the variable has been initialized. This only warns about
5759 variables that are initialized when they are declared. This warning is
5760 only supported for C and Objective-C; in C++ this sort of branch is an
5763 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5764 can be disabled with the @option{-Wno-jump-misses-init} option.
5766 @item -Wsign-compare
5767 @opindex Wsign-compare
5768 @opindex Wno-sign-compare
5769 @cindex warning for comparison of signed and unsigned values
5770 @cindex comparison of signed and unsigned values, warning
5771 @cindex signed and unsigned values, comparison warning
5772 Warn when a comparison between signed and unsigned values could produce
5773 an incorrect result when the signed value is converted to unsigned.
5774 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5775 also enabled by @option{-Wextra}.
5777 @item -Wsign-conversion
5778 @opindex Wsign-conversion
5779 @opindex Wno-sign-conversion
5780 Warn for implicit conversions that may change the sign of an integer
5781 value, like assigning a signed integer expression to an unsigned
5782 integer variable. An explicit cast silences the warning. In C, this
5783 option is enabled also by @option{-Wconversion}.
5785 @item -Wfloat-conversion
5786 @opindex Wfloat-conversion
5787 @opindex Wno-float-conversion
5788 Warn for implicit conversions that reduce the precision of a real value.
5789 This includes conversions from real to integer, and from higher precision
5790 real to lower precision real values. This option is also enabled by
5791 @option{-Wconversion}.
5793 @item -Wno-scalar-storage-order
5794 @opindex -Wno-scalar-storage-order
5795 @opindex -Wscalar-storage-order
5796 Do not warn on suspicious constructs involving reverse scalar storage order.
5798 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5799 @opindex Wsized-deallocation
5800 @opindex Wno-sized-deallocation
5801 Warn about a definition of an unsized deallocation function
5803 void operator delete (void *) noexcept;
5804 void operator delete[] (void *) noexcept;
5806 without a definition of the corresponding sized deallocation function
5808 void operator delete (void *, std::size_t) noexcept;
5809 void operator delete[] (void *, std::size_t) noexcept;
5811 or vice versa. Enabled by @option{-Wextra} along with
5812 @option{-fsized-deallocation}.
5814 @item -Wsizeof-pointer-memaccess
5815 @opindex Wsizeof-pointer-memaccess
5816 @opindex Wno-sizeof-pointer-memaccess
5817 Warn for suspicious length parameters to certain string and memory built-in
5818 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5819 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5820 but a pointer, and suggests a possible fix, or about
5821 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5824 @item -Wsizeof-array-argument
5825 @opindex Wsizeof-array-argument
5826 @opindex Wno-sizeof-array-argument
5827 Warn when the @code{sizeof} operator is applied to a parameter that is
5828 declared as an array in a function definition. This warning is enabled by
5829 default for C and C++ programs.
5831 @item -Wmemset-elt-size
5832 @opindex Wmemset-elt-size
5833 @opindex Wno-memset-elt-size
5834 Warn for suspicious calls to the @code{memset} built-in function, if the
5835 first argument references an array, and the third argument is a number
5836 equal to the number of elements, but not equal to the size of the array
5837 in memory. This indicates that the user has omitted a multiplication by
5838 the element size. This warning is enabled by @option{-Wall}.
5840 @item -Wmemset-transposed-args
5841 @opindex Wmemset-transposed-args
5842 @opindex Wno-memset-transposed-args
5843 Warn for suspicious calls to the @code{memset} built-in function, if the
5844 second argument is not zero and the third argument is zero. This warns e.g.@
5845 about @code{memset (buf, sizeof buf, 0)} where most probably
5846 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5847 is only emitted if the third argument is literal zero. If it is some
5848 expression that is folded to zero, a cast of zero to some type, etc.,
5849 it is far less likely that the user has mistakenly exchanged the arguments
5850 and no warning is emitted. This warning is enabled by @option{-Wall}.
5854 @opindex Wno-address
5855 Warn about suspicious uses of memory addresses. These include using
5856 the address of a function in a conditional expression, such as
5857 @code{void func(void); if (func)}, and comparisons against the memory
5858 address of a string literal, such as @code{if (x == "abc")}. Such
5859 uses typically indicate a programmer error: the address of a function
5860 always evaluates to true, so their use in a conditional usually
5861 indicate that the programmer forgot the parentheses in a function
5862 call; and comparisons against string literals result in unspecified
5863 behavior and are not portable in C, so they usually indicate that the
5864 programmer intended to use @code{strcmp}. This warning is enabled by
5868 @opindex Wlogical-op
5869 @opindex Wno-logical-op
5870 Warn about suspicious uses of logical operators in expressions.
5871 This includes using logical operators in contexts where a
5872 bit-wise operator is likely to be expected. Also warns when
5873 the operands of a logical operator are the same:
5876 if (a < 0 && a < 0) @{ @dots{} @}
5879 @item -Wlogical-not-parentheses
5880 @opindex Wlogical-not-parentheses
5881 @opindex Wno-logical-not-parentheses
5882 Warn about logical not used on the left hand side operand of a comparison.
5883 This option does not warn if the right operand is considered to be a boolean
5884 expression. Its purpose is to detect suspicious code like the following:
5888 if (!a > 1) @{ @dots{} @}
5891 It is possible to suppress the warning by wrapping the LHS into
5894 if ((!a) > 1) @{ @dots{} @}
5897 This warning is enabled by @option{-Wall}.
5899 @item -Waggregate-return
5900 @opindex Waggregate-return
5901 @opindex Wno-aggregate-return
5902 Warn if any functions that return structures or unions are defined or
5903 called. (In languages where you can return an array, this also elicits
5906 @item -Wno-aggressive-loop-optimizations
5907 @opindex Wno-aggressive-loop-optimizations
5908 @opindex Waggressive-loop-optimizations
5909 Warn if in a loop with constant number of iterations the compiler detects
5910 undefined behavior in some statement during one or more of the iterations.
5912 @item -Wno-attributes
5913 @opindex Wno-attributes
5914 @opindex Wattributes
5915 Do not warn if an unexpected @code{__attribute__} is used, such as
5916 unrecognized attributes, function attributes applied to variables,
5917 etc. This does not stop errors for incorrect use of supported
5920 @item -Wno-builtin-macro-redefined
5921 @opindex Wno-builtin-macro-redefined
5922 @opindex Wbuiltin-macro-redefined
5923 Do not warn if certain built-in macros are redefined. This suppresses
5924 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5925 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5927 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5928 @opindex Wstrict-prototypes
5929 @opindex Wno-strict-prototypes
5930 Warn if a function is declared or defined without specifying the
5931 argument types. (An old-style function definition is permitted without
5932 a warning if preceded by a declaration that specifies the argument
5935 @item -Wold-style-declaration @r{(C and Objective-C only)}
5936 @opindex Wold-style-declaration
5937 @opindex Wno-old-style-declaration
5938 Warn for obsolescent usages, according to the C Standard, in a
5939 declaration. For example, warn if storage-class specifiers like
5940 @code{static} are not the first things in a declaration. This warning
5941 is also enabled by @option{-Wextra}.
5943 @item -Wold-style-definition @r{(C and Objective-C only)}
5944 @opindex Wold-style-definition
5945 @opindex Wno-old-style-definition
5946 Warn if an old-style function definition is used. A warning is given
5947 even if there is a previous prototype.
5949 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5950 @opindex Wmissing-parameter-type
5951 @opindex Wno-missing-parameter-type
5952 A function parameter is declared without a type specifier in K&R-style
5959 This warning is also enabled by @option{-Wextra}.
5961 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5962 @opindex Wmissing-prototypes
5963 @opindex Wno-missing-prototypes
5964 Warn if a global function is defined without a previous prototype
5965 declaration. This warning is issued even if the definition itself
5966 provides a prototype. Use this option to detect global functions
5967 that do not have a matching prototype declaration in a header file.
5968 This option is not valid for C++ because all function declarations
5969 provide prototypes and a non-matching declaration declares an
5970 overload rather than conflict with an earlier declaration.
5971 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5973 @item -Wmissing-declarations
5974 @opindex Wmissing-declarations
5975 @opindex Wno-missing-declarations
5976 Warn if a global function is defined without a previous declaration.
5977 Do so even if the definition itself provides a prototype.
5978 Use this option to detect global functions that are not declared in
5979 header files. In C, no warnings are issued for functions with previous
5980 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5981 missing prototypes. In C++, no warnings are issued for function templates,
5982 or for inline functions, or for functions in anonymous namespaces.
5984 @item -Wmissing-field-initializers
5985 @opindex Wmissing-field-initializers
5986 @opindex Wno-missing-field-initializers
5990 Warn if a structure's initializer has some fields missing. For
5991 example, the following code causes such a warning, because
5992 @code{x.h} is implicitly zero:
5995 struct s @{ int f, g, h; @};
5996 struct s x = @{ 3, 4 @};
5999 This option does not warn about designated initializers, so the following
6000 modification does not trigger a warning:
6003 struct s @{ int f, g, h; @};
6004 struct s x = @{ .f = 3, .g = 4 @};
6007 In C++ this option does not warn either about the empty @{ @}
6008 initializer, for example:
6011 struct s @{ int f, g, h; @};
6015 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6016 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6018 @item -Wno-multichar
6019 @opindex Wno-multichar
6021 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6022 Usually they indicate a typo in the user's code, as they have
6023 implementation-defined values, and should not be used in portable code.
6025 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
6026 @opindex Wnormalized=
6027 @opindex Wnormalized
6028 @opindex Wno-normalized
6031 @cindex character set, input normalization
6032 In ISO C and ISO C++, two identifiers are different if they are
6033 different sequences of characters. However, sometimes when characters
6034 outside the basic ASCII character set are used, you can have two
6035 different character sequences that look the same. To avoid confusion,
6036 the ISO 10646 standard sets out some @dfn{normalization rules} which
6037 when applied ensure that two sequences that look the same are turned into
6038 the same sequence. GCC can warn you if you are using identifiers that
6039 have not been normalized; this option controls that warning.
6041 There are four levels of warning supported by GCC@. The default is
6042 @option{-Wnormalized=nfc}, which warns about any identifier that is
6043 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6044 recommended form for most uses. It is equivalent to
6045 @option{-Wnormalized}.
6047 Unfortunately, there are some characters allowed in identifiers by
6048 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6049 identifiers. That is, there's no way to use these symbols in portable
6050 ISO C or C++ and have all your identifiers in NFC@.
6051 @option{-Wnormalized=id} suppresses the warning for these characters.
6052 It is hoped that future versions of the standards involved will correct
6053 this, which is why this option is not the default.
6055 You can switch the warning off for all characters by writing
6056 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6057 only do this if you are using some other normalization scheme (like
6058 ``D''), because otherwise you can easily create bugs that are
6059 literally impossible to see.
6061 Some characters in ISO 10646 have distinct meanings but look identical
6062 in some fonts or display methodologies, especially once formatting has
6063 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6064 LETTER N'', displays just like a regular @code{n} that has been
6065 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6066 normalization scheme to convert all these into a standard form as
6067 well, and GCC warns if your code is not in NFKC if you use
6068 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6069 about every identifier that contains the letter O because it might be
6070 confused with the digit 0, and so is not the default, but may be
6071 useful as a local coding convention if the programming environment
6072 cannot be fixed to display these characters distinctly.
6074 @item -Wno-deprecated
6075 @opindex Wno-deprecated
6076 @opindex Wdeprecated
6077 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6079 @item -Wno-deprecated-declarations
6080 @opindex Wno-deprecated-declarations
6081 @opindex Wdeprecated-declarations
6082 Do not warn about uses of functions (@pxref{Function Attributes}),
6083 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6084 Attributes}) marked as deprecated by using the @code{deprecated}
6088 @opindex Wno-overflow
6090 Do not warn about compile-time overflow in constant expressions.
6095 Warn about One Definition Rule violations during link-time optimization.
6096 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6099 @opindex Wopenm-simd
6100 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6101 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6102 option can be used to relax the cost model.
6104 @item -Woverride-init @r{(C and Objective-C only)}
6105 @opindex Woverride-init
6106 @opindex Wno-override-init
6110 Warn if an initialized field without side effects is overridden when
6111 using designated initializers (@pxref{Designated Inits, , Designated
6114 This warning is included in @option{-Wextra}. To get other
6115 @option{-Wextra} warnings without this one, use @option{-Wextra
6116 -Wno-override-init}.
6118 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6119 @opindex Woverride-init-side-effects
6120 @opindex Wno-override-init-side-effects
6121 Warn if an initialized field with side effects is overridden when
6122 using designated initializers (@pxref{Designated Inits, , Designated
6123 Initializers}). This warning is enabled by default.
6128 Warn if a structure is given the packed attribute, but the packed
6129 attribute has no effect on the layout or size of the structure.
6130 Such structures may be mis-aligned for little benefit. For
6131 instance, in this code, the variable @code{f.x} in @code{struct bar}
6132 is misaligned even though @code{struct bar} does not itself
6133 have the packed attribute:
6140 @} __attribute__((packed));
6148 @item -Wpacked-bitfield-compat
6149 @opindex Wpacked-bitfield-compat
6150 @opindex Wno-packed-bitfield-compat
6151 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6152 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6153 the change can lead to differences in the structure layout. GCC
6154 informs you when the offset of such a field has changed in GCC 4.4.
6155 For example there is no longer a 4-bit padding between field @code{a}
6156 and @code{b} in this structure:
6163 @} __attribute__ ((packed));
6166 This warning is enabled by default. Use
6167 @option{-Wno-packed-bitfield-compat} to disable this warning.
6172 Warn if padding is included in a structure, either to align an element
6173 of the structure or to align the whole structure. Sometimes when this
6174 happens it is possible to rearrange the fields of the structure to
6175 reduce the padding and so make the structure smaller.
6177 @item -Wredundant-decls
6178 @opindex Wredundant-decls
6179 @opindex Wno-redundant-decls
6180 Warn if anything is declared more than once in the same scope, even in
6181 cases where multiple declaration is valid and changes nothing.
6183 @item -Wnested-externs @r{(C and Objective-C only)}
6184 @opindex Wnested-externs
6185 @opindex Wno-nested-externs
6186 Warn if an @code{extern} declaration is encountered within a function.
6188 @item -Wno-inherited-variadic-ctor
6189 @opindex Winherited-variadic-ctor
6190 @opindex Wno-inherited-variadic-ctor
6191 Suppress warnings about use of C++11 inheriting constructors when the
6192 base class inherited from has a C variadic constructor; the warning is
6193 on by default because the ellipsis is not inherited.
6198 Warn if a function that is declared as inline cannot be inlined.
6199 Even with this option, the compiler does not warn about failures to
6200 inline functions declared in system headers.
6202 The compiler uses a variety of heuristics to determine whether or not
6203 to inline a function. For example, the compiler takes into account
6204 the size of the function being inlined and the amount of inlining
6205 that has already been done in the current function. Therefore,
6206 seemingly insignificant changes in the source program can cause the
6207 warnings produced by @option{-Winline} to appear or disappear.
6209 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6210 @opindex Wno-invalid-offsetof
6211 @opindex Winvalid-offsetof
6212 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6213 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6214 to a non-standard-layout type is undefined. In existing C++ implementations,
6215 however, @code{offsetof} typically gives meaningful results.
6216 This flag is for users who are aware that they are
6217 writing nonportable code and who have deliberately chosen to ignore the
6220 The restrictions on @code{offsetof} may be relaxed in a future version
6221 of the C++ standard.
6223 @item -Wint-in-bool-context
6224 @opindex Wint-in-bool-context
6225 @opindex Wno-int-in-bool-context
6226 Warn for suspicious use of integer values where boolean values are expected,
6227 such as conditional expressions (?:) using non-boolean integer constants in
6228 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6229 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6230 for all kinds of multiplications regardless of the data type.
6231 This warning is enabled by @option{-Wall}.
6233 @item -Wno-int-to-pointer-cast
6234 @opindex Wno-int-to-pointer-cast
6235 @opindex Wint-to-pointer-cast
6236 Suppress warnings from casts to pointer type of an integer of a
6237 different size. In C++, casting to a pointer type of smaller size is
6238 an error. @option{Wint-to-pointer-cast} is enabled by default.
6241 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6242 @opindex Wno-pointer-to-int-cast
6243 @opindex Wpointer-to-int-cast
6244 Suppress warnings from casts from a pointer to an integer type of a
6248 @opindex Winvalid-pch
6249 @opindex Wno-invalid-pch
6250 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6251 the search path but can't be used.
6255 @opindex Wno-long-long
6256 Warn if @code{long long} type is used. This is enabled by either
6257 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6258 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6260 @item -Wvariadic-macros
6261 @opindex Wvariadic-macros
6262 @opindex Wno-variadic-macros
6263 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6264 alternate syntax is used in ISO C99 mode. This is enabled by either
6265 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6266 messages, use @option{-Wno-variadic-macros}.
6270 @opindex Wno-varargs
6271 Warn upon questionable usage of the macros used to handle variable
6272 arguments like @code{va_start}. This is default. To inhibit the
6273 warning messages, use @option{-Wno-varargs}.
6275 @item -Wvector-operation-performance
6276 @opindex Wvector-operation-performance
6277 @opindex Wno-vector-operation-performance
6278 Warn if vector operation is not implemented via SIMD capabilities of the
6279 architecture. Mainly useful for the performance tuning.
6280 Vector operation can be implemented @code{piecewise}, which means that the
6281 scalar operation is performed on every vector element;
6282 @code{in parallel}, which means that the vector operation is implemented
6283 using scalars of wider type, which normally is more performance efficient;
6284 and @code{as a single scalar}, which means that vector fits into a
6287 @item -Wno-virtual-move-assign
6288 @opindex Wvirtual-move-assign
6289 @opindex Wno-virtual-move-assign
6290 Suppress warnings about inheriting from a virtual base with a
6291 non-trivial C++11 move assignment operator. This is dangerous because
6292 if the virtual base is reachable along more than one path, it is
6293 moved multiple times, which can mean both objects end up in the
6294 moved-from state. If the move assignment operator is written to avoid
6295 moving from a moved-from object, this warning can be disabled.
6300 Warn if a variable-length array is used in the code.
6301 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6302 the variable-length array.
6304 @item -Wvla-larger-than=@var{n}
6305 If this option is used, the compiler will warn on uses of
6306 variable-length arrays where the size is either unbounded, or bounded
6307 by an argument that can be larger than @var{n} bytes. This is similar
6308 to how @option{-Walloca-larger-than=@var{n}} works, but with
6309 variable-length arrays.
6311 Note that GCC may optimize small variable-length arrays of a known
6312 value into plain arrays, so this warning may not get triggered for
6315 This warning is not enabled by @option{-Wall}, and is only active when
6316 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6318 See also @option{-Walloca-larger-than=@var{n}}.
6320 @item -Wvolatile-register-var
6321 @opindex Wvolatile-register-var
6322 @opindex Wno-volatile-register-var
6323 Warn if a register variable is declared volatile. The volatile
6324 modifier does not inhibit all optimizations that may eliminate reads
6325 and/or writes to register variables. This warning is enabled by
6328 @item -Wdisabled-optimization
6329 @opindex Wdisabled-optimization
6330 @opindex Wno-disabled-optimization
6331 Warn if a requested optimization pass is disabled. This warning does
6332 not generally indicate that there is anything wrong with your code; it
6333 merely indicates that GCC's optimizers are unable to handle the code
6334 effectively. Often, the problem is that your code is too big or too
6335 complex; GCC refuses to optimize programs when the optimization
6336 itself is likely to take inordinate amounts of time.
6338 @item -Wpointer-sign @r{(C and Objective-C only)}
6339 @opindex Wpointer-sign
6340 @opindex Wno-pointer-sign
6341 Warn for pointer argument passing or assignment with different signedness.
6342 This option is only supported for C and Objective-C@. It is implied by
6343 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6344 @option{-Wno-pointer-sign}.
6346 @item -Wstack-protector
6347 @opindex Wstack-protector
6348 @opindex Wno-stack-protector
6349 This option is only active when @option{-fstack-protector} is active. It
6350 warns about functions that are not protected against stack smashing.
6352 @item -Woverlength-strings
6353 @opindex Woverlength-strings
6354 @opindex Wno-overlength-strings
6355 Warn about string constants that are longer than the ``minimum
6356 maximum'' length specified in the C standard. Modern compilers
6357 generally allow string constants that are much longer than the
6358 standard's minimum limit, but very portable programs should avoid
6359 using longer strings.
6361 The limit applies @emph{after} string constant concatenation, and does
6362 not count the trailing NUL@. In C90, the limit was 509 characters; in
6363 C99, it was raised to 4095. C++98 does not specify a normative
6364 minimum maximum, so we do not diagnose overlength strings in C++@.
6366 This option is implied by @option{-Wpedantic}, and can be disabled with
6367 @option{-Wno-overlength-strings}.
6369 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6370 @opindex Wunsuffixed-float-constants
6372 Issue a warning for any floating constant that does not have
6373 a suffix. When used together with @option{-Wsystem-headers} it
6374 warns about such constants in system header files. This can be useful
6375 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6376 from the decimal floating-point extension to C99.
6378 @item -Wno-designated-init @r{(C and Objective-C only)}
6379 Suppress warnings when a positional initializer is used to initialize
6380 a structure that has been marked with the @code{designated_init}
6384 Issue a warning when HSAIL cannot be emitted for the compiled function or
6389 @node Debugging Options
6390 @section Options for Debugging Your Program
6391 @cindex options, debugging
6392 @cindex debugging information options
6394 To tell GCC to emit extra information for use by a debugger, in almost
6395 all cases you need only to add @option{-g} to your other options.
6397 GCC allows you to use @option{-g} with
6398 @option{-O}. The shortcuts taken by optimized code may occasionally
6399 be surprising: some variables you declared may not exist
6400 at all; flow of control may briefly move where you did not expect it;
6401 some statements may not be executed because they compute constant
6402 results or their values are already at hand; some statements may
6403 execute in different places because they have been moved out of loops.
6404 Nevertheless it is possible to debug optimized output. This makes
6405 it reasonable to use the optimizer for programs that might have bugs.
6407 If you are not using some other optimization option, consider
6408 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6409 With no @option{-O} option at all, some compiler passes that collect
6410 information useful for debugging do not run at all, so that
6411 @option{-Og} may result in a better debugging experience.
6416 Produce debugging information in the operating system's native format
6417 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6420 On most systems that use stabs format, @option{-g} enables use of extra
6421 debugging information that only GDB can use; this extra information
6422 makes debugging work better in GDB but probably makes other debuggers
6424 refuse to read the program. If you want to control for certain whether
6425 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6426 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6430 Produce debugging information for use by GDB@. This means to use the
6431 most expressive format available (DWARF, stabs, or the native format
6432 if neither of those are supported), including GDB extensions if at all
6436 @itemx -gdwarf-@var{version}
6438 Produce debugging information in DWARF format (if that is supported).
6439 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6440 for most targets is 4. DWARF Version 5 is only experimental.
6442 Note that with DWARF Version 2, some ports require and always
6443 use some non-conflicting DWARF 3 extensions in the unwind tables.
6445 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6446 for maximum benefit.
6448 GCC no longer supports DWARF Version 1, which is substantially
6449 different than Version 2 and later. For historical reasons, some
6450 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
6451 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6452 in their names, but apply to all currently-supported versions of DWARF.
6456 Produce debugging information in stabs format (if that is supported),
6457 without GDB extensions. This is the format used by DBX on most BSD
6458 systems. On MIPS, Alpha and System V Release 4 systems this option
6459 produces stabs debugging output that is not understood by DBX or SDB@.
6460 On System V Release 4 systems this option requires the GNU assembler.
6464 Produce debugging information in stabs format (if that is supported),
6465 using GNU extensions understood only by the GNU debugger (GDB)@. The
6466 use of these extensions is likely to make other debuggers crash or
6467 refuse to read the program.
6471 Produce debugging information in COFF format (if that is supported).
6472 This is the format used by SDB on most System V systems prior to
6477 Produce debugging information in XCOFF format (if that is supported).
6478 This is the format used by the DBX debugger on IBM RS/6000 systems.
6482 Produce debugging information in XCOFF format (if that is supported),
6483 using GNU extensions understood only by the GNU debugger (GDB)@. The
6484 use of these extensions is likely to make other debuggers crash or
6485 refuse to read the program, and may cause assemblers other than the GNU
6486 assembler (GAS) to fail with an error.
6490 Produce debugging information in Alpha/VMS debug format (if that is
6491 supported). This is the format used by DEBUG on Alpha/VMS systems.
6494 @itemx -ggdb@var{level}
6495 @itemx -gstabs@var{level}
6496 @itemx -gcoff@var{level}
6497 @itemx -gxcoff@var{level}
6498 @itemx -gvms@var{level}
6499 Request debugging information and also use @var{level} to specify how
6500 much information. The default level is 2.
6502 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6505 Level 1 produces minimal information, enough for making backtraces in
6506 parts of the program that you don't plan to debug. This includes
6507 descriptions of functions and external variables, and line number
6508 tables, but no information about local variables.
6510 Level 3 includes extra information, such as all the macro definitions
6511 present in the program. Some debuggers support macro expansion when
6512 you use @option{-g3}.
6514 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6515 confusion with @option{-gdwarf-@var{level}}.
6516 Instead use an additional @option{-g@var{level}} option to change the
6517 debug level for DWARF.
6519 @item -feliminate-unused-debug-symbols
6520 @opindex feliminate-unused-debug-symbols
6521 Produce debugging information in stabs format (if that is supported),
6522 for only symbols that are actually used.
6524 @item -femit-class-debug-always
6525 @opindex femit-class-debug-always
6526 Instead of emitting debugging information for a C++ class in only one
6527 object file, emit it in all object files using the class. This option
6528 should be used only with debuggers that are unable to handle the way GCC
6529 normally emits debugging information for classes because using this
6530 option increases the size of debugging information by as much as a
6533 @item -fno-merge-debug-strings
6534 @opindex fmerge-debug-strings
6535 @opindex fno-merge-debug-strings
6536 Direct the linker to not merge together strings in the debugging
6537 information that are identical in different object files. Merging is
6538 not supported by all assemblers or linkers. Merging decreases the size
6539 of the debug information in the output file at the cost of increasing
6540 link processing time. Merging is enabled by default.
6542 @item -fdebug-prefix-map=@var{old}=@var{new}
6543 @opindex fdebug-prefix-map
6544 When compiling files in directory @file{@var{old}}, record debugging
6545 information describing them as in @file{@var{new}} instead.
6547 @item -fvar-tracking
6548 @opindex fvar-tracking
6549 Run variable tracking pass. It computes where variables are stored at each
6550 position in code. Better debugging information is then generated
6551 (if the debugging information format supports this information).
6553 It is enabled by default when compiling with optimization (@option{-Os},
6554 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6555 the debug info format supports it.
6557 @item -fvar-tracking-assignments
6558 @opindex fvar-tracking-assignments
6559 @opindex fno-var-tracking-assignments
6560 Annotate assignments to user variables early in the compilation and
6561 attempt to carry the annotations over throughout the compilation all the
6562 way to the end, in an attempt to improve debug information while
6563 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6565 It can be enabled even if var-tracking is disabled, in which case
6566 annotations are created and maintained, but discarded at the end.
6567 By default, this flag is enabled together with @option{-fvar-tracking},
6568 except when selective scheduling is enabled.
6571 @opindex gsplit-dwarf
6572 Separate as much DWARF debugging information as possible into a
6573 separate output file with the extension @file{.dwo}. This option allows
6574 the build system to avoid linking files with debug information. To
6575 be useful, this option requires a debugger capable of reading @file{.dwo}
6580 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6582 @item -ggnu-pubnames
6583 @opindex ggnu-pubnames
6584 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6585 suitable for conversion into a GDB@ index. This option is only useful
6586 with a linker that can produce GDB@ index version 7.
6588 @item -fdebug-types-section
6589 @opindex fdebug-types-section
6590 @opindex fno-debug-types-section
6591 When using DWARF Version 4 or higher, type DIEs can be put into
6592 their own @code{.debug_types} section instead of making them part of the
6593 @code{.debug_info} section. It is more efficient to put them in a separate
6594 comdat sections since the linker can then remove duplicates.
6595 But not all DWARF consumers support @code{.debug_types} sections yet
6596 and on some objects @code{.debug_types} produces larger instead of smaller
6597 debugging information.
6599 @item -grecord-gcc-switches
6600 @item -gno-record-gcc-switches
6601 @opindex grecord-gcc-switches
6602 @opindex gno-record-gcc-switches
6603 This switch causes the command-line options used to invoke the
6604 compiler that may affect code generation to be appended to the
6605 DW_AT_producer attribute in DWARF debugging information. The options
6606 are concatenated with spaces separating them from each other and from
6607 the compiler version.
6608 It is enabled by default.
6609 See also @option{-frecord-gcc-switches} for another
6610 way of storing compiler options into the object file.
6612 @item -gstrict-dwarf
6613 @opindex gstrict-dwarf
6614 Disallow using extensions of later DWARF standard version than selected
6615 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6616 DWARF extensions from later standard versions is allowed.
6618 @item -gno-strict-dwarf
6619 @opindex gno-strict-dwarf
6620 Allow using extensions of later DWARF standard version than selected with
6621 @option{-gdwarf-@var{version}}.
6623 @item -gz@r{[}=@var{type}@r{]}
6625 Produce compressed debug sections in DWARF format, if that is supported.
6626 If @var{type} is not given, the default type depends on the capabilities
6627 of the assembler and linker used. @var{type} may be one of
6628 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6629 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6630 compression in traditional GNU format). If the linker doesn't support
6631 writing compressed debug sections, the option is rejected. Otherwise,
6632 if the assembler does not support them, @option{-gz} is silently ignored
6633 when producing object files.
6635 @item -feliminate-dwarf2-dups
6636 @opindex feliminate-dwarf2-dups
6637 Compress DWARF debugging information by eliminating duplicated
6638 information about each symbol. This option only makes sense when
6639 generating DWARF debugging information.
6641 @item -femit-struct-debug-baseonly
6642 @opindex femit-struct-debug-baseonly
6643 Emit debug information for struct-like types
6644 only when the base name of the compilation source file
6645 matches the base name of file in which the struct is defined.
6647 This option substantially reduces the size of debugging information,
6648 but at significant potential loss in type information to the debugger.
6649 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6650 See @option{-femit-struct-debug-detailed} for more detailed control.
6652 This option works only with DWARF debug output.
6654 @item -femit-struct-debug-reduced
6655 @opindex femit-struct-debug-reduced
6656 Emit debug information for struct-like types
6657 only when the base name of the compilation source file
6658 matches the base name of file in which the type is defined,
6659 unless the struct is a template or defined in a system header.
6661 This option significantly reduces the size of debugging information,
6662 with some potential loss in type information to the debugger.
6663 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6664 See @option{-femit-struct-debug-detailed} for more detailed control.
6666 This option works only with DWARF debug output.
6668 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6669 @opindex femit-struct-debug-detailed
6670 Specify the struct-like types
6671 for which the compiler generates debug information.
6672 The intent is to reduce duplicate struct debug information
6673 between different object files within the same program.
6675 This option is a detailed version of
6676 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6677 which serves for most needs.
6679 A specification has the syntax@*
6680 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6682 The optional first word limits the specification to
6683 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6684 A struct type is used directly when it is the type of a variable, member.
6685 Indirect uses arise through pointers to structs.
6686 That is, when use of an incomplete struct is valid, the use is indirect.
6688 @samp{struct one direct; struct two * indirect;}.
6690 The optional second word limits the specification to
6691 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6692 Generic structs are a bit complicated to explain.
6693 For C++, these are non-explicit specializations of template classes,
6694 or non-template classes within the above.
6695 Other programming languages have generics,
6696 but @option{-femit-struct-debug-detailed} does not yet implement them.
6698 The third word specifies the source files for those
6699 structs for which the compiler should emit debug information.
6700 The values @samp{none} and @samp{any} have the normal meaning.
6701 The value @samp{base} means that
6702 the base of name of the file in which the type declaration appears
6703 must match the base of the name of the main compilation file.
6704 In practice, this means that when compiling @file{foo.c}, debug information
6705 is generated for types declared in that file and @file{foo.h},
6706 but not other header files.
6707 The value @samp{sys} means those types satisfying @samp{base}
6708 or declared in system or compiler headers.
6710 You may need to experiment to determine the best settings for your application.
6712 The default is @option{-femit-struct-debug-detailed=all}.
6714 This option works only with DWARF debug output.
6716 @item -fno-dwarf2-cfi-asm
6717 @opindex fdwarf2-cfi-asm
6718 @opindex fno-dwarf2-cfi-asm
6719 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6720 instead of using GAS @code{.cfi_*} directives.
6722 @item -fno-eliminate-unused-debug-types
6723 @opindex feliminate-unused-debug-types
6724 @opindex fno-eliminate-unused-debug-types
6725 Normally, when producing DWARF output, GCC avoids producing debug symbol
6726 output for types that are nowhere used in the source file being compiled.
6727 Sometimes it is useful to have GCC emit debugging
6728 information for all types declared in a compilation
6729 unit, regardless of whether or not they are actually used
6730 in that compilation unit, for example
6731 if, in the debugger, you want to cast a value to a type that is
6732 not actually used in your program (but is declared). More often,
6733 however, this results in a significant amount of wasted space.
6736 @node Optimize Options
6737 @section Options That Control Optimization
6738 @cindex optimize options
6739 @cindex options, optimization
6741 These options control various sorts of optimizations.
6743 Without any optimization option, the compiler's goal is to reduce the
6744 cost of compilation and to make debugging produce the expected
6745 results. Statements are independent: if you stop the program with a
6746 breakpoint between statements, you can then assign a new value to any
6747 variable or change the program counter to any other statement in the
6748 function and get exactly the results you expect from the source
6751 Turning on optimization flags makes the compiler attempt to improve
6752 the performance and/or code size at the expense of compilation time
6753 and possibly the ability to debug the program.
6755 The compiler performs optimization based on the knowledge it has of the
6756 program. Compiling multiple files at once to a single output file mode allows
6757 the compiler to use information gained from all of the files when compiling
6760 Not all optimizations are controlled directly by a flag. Only
6761 optimizations that have a flag are listed in this section.
6763 Most optimizations are only enabled if an @option{-O} level is set on
6764 the command line. Otherwise they are disabled, even if individual
6765 optimization flags are specified.
6767 Depending on the target and how GCC was configured, a slightly different
6768 set of optimizations may be enabled at each @option{-O} level than
6769 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6770 to find out the exact set of optimizations that are enabled at each level.
6771 @xref{Overall Options}, for examples.
6778 Optimize. Optimizing compilation takes somewhat more time, and a lot
6779 more memory for a large function.
6781 With @option{-O}, the compiler tries to reduce code size and execution
6782 time, without performing any optimizations that take a great deal of
6785 @option{-O} turns on the following optimization flags:
6788 -fbranch-count-reg @gol
6789 -fcombine-stack-adjustments @gol
6791 -fcprop-registers @gol
6794 -fdelayed-branch @gol
6796 -fforward-propagate @gol
6797 -fguess-branch-probability @gol
6798 -fif-conversion2 @gol
6799 -fif-conversion @gol
6800 -finline-functions-called-once @gol
6801 -fipa-pure-const @gol
6803 -fipa-reference @gol
6804 -fmerge-constants @gol
6805 -fmove-loop-invariants @gol
6806 -freorder-blocks @gol
6808 -fshrink-wrap-separate @gol
6809 -fsplit-wide-types @gol
6812 -fstore-merging @gol
6816 -ftree-coalesce-vars @gol
6817 -ftree-copy-prop @gol
6819 -ftree-dominator-opts @gol
6821 -ftree-forwprop @gol
6831 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6832 where doing so does not interfere with debugging.
6836 Optimize even more. GCC performs nearly all supported optimizations
6837 that do not involve a space-speed tradeoff.
6838 As compared to @option{-O}, this option increases both compilation time
6839 and the performance of the generated code.
6841 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6842 also turns on the following optimization flags:
6843 @gccoptlist{-fthread-jumps @gol
6844 -falign-functions -falign-jumps @gol
6845 -falign-loops -falign-labels @gol
6848 -fcse-follow-jumps -fcse-skip-blocks @gol
6849 -fdelete-null-pointer-checks @gol
6850 -fdevirtualize -fdevirtualize-speculatively @gol
6851 -fexpensive-optimizations @gol
6852 -fgcse -fgcse-lm @gol
6853 -fhoist-adjacent-loads @gol
6854 -finline-small-functions @gol
6855 -findirect-inlining @gol
6857 -fipa-cp-alignment @gol
6861 -fisolate-erroneous-paths-dereference @gol
6863 -foptimize-sibling-calls @gol
6864 -foptimize-strlen @gol
6865 -fpartial-inlining @gol
6867 -freorder-blocks-algorithm=stc @gol
6868 -freorder-blocks-and-partition -freorder-functions @gol
6869 -frerun-cse-after-loop @gol
6870 -fsched-interblock -fsched-spec @gol
6871 -fschedule-insns -fschedule-insns2 @gol
6872 -fstrict-aliasing -fstrict-overflow @gol
6873 -ftree-builtin-call-dce @gol
6874 -ftree-switch-conversion -ftree-tail-merge @gol
6875 -fcode-hoisting @gol
6880 Please note the warning under @option{-fgcse} about
6881 invoking @option{-O2} on programs that use computed gotos.
6885 Optimize yet more. @option{-O3} turns on all optimizations specified
6886 by @option{-O2} and also turns on the @option{-finline-functions},
6887 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6888 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6889 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
6890 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6891 @option{-ftree-partial-pre}, @option{-fpeel-loops}
6892 and @option{-fipa-cp-clone} options.
6896 Reduce compilation time and make debugging produce the expected
6897 results. This is the default.
6901 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6902 do not typically increase code size. It also performs further
6903 optimizations designed to reduce code size.
6905 @option{-Os} disables the following optimization flags:
6906 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6907 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
6908 -freorder-blocks-and-partition -fprefetch-loop-arrays}
6912 Disregard strict standards compliance. @option{-Ofast} enables all
6913 @option{-O3} optimizations. It also enables optimizations that are not
6914 valid for all standard-compliant programs.
6915 It turns on @option{-ffast-math} and the Fortran-specific
6916 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6920 Optimize debugging experience. @option{-Og} enables optimizations
6921 that do not interfere with debugging. It should be the optimization
6922 level of choice for the standard edit-compile-debug cycle, offering
6923 a reasonable level of optimization while maintaining fast compilation
6924 and a good debugging experience.
6927 If you use multiple @option{-O} options, with or without level numbers,
6928 the last such option is the one that is effective.
6930 Options of the form @option{-f@var{flag}} specify machine-independent
6931 flags. Most flags have both positive and negative forms; the negative
6932 form of @option{-ffoo} is @option{-fno-foo}. In the table
6933 below, only one of the forms is listed---the one you typically
6934 use. You can figure out the other form by either removing @samp{no-}
6937 The following options control specific optimizations. They are either
6938 activated by @option{-O} options or are related to ones that are. You
6939 can use the following flags in the rare cases when ``fine-tuning'' of
6940 optimizations to be performed is desired.
6943 @item -fno-defer-pop
6944 @opindex fno-defer-pop
6945 Always pop the arguments to each function call as soon as that function
6946 returns. For machines that must pop arguments after a function call,
6947 the compiler normally lets arguments accumulate on the stack for several
6948 function calls and pops them all at once.
6950 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6952 @item -fforward-propagate
6953 @opindex fforward-propagate
6954 Perform a forward propagation pass on RTL@. The pass tries to combine two
6955 instructions and checks if the result can be simplified. If loop unrolling
6956 is active, two passes are performed and the second is scheduled after
6959 This option is enabled by default at optimization levels @option{-O},
6960 @option{-O2}, @option{-O3}, @option{-Os}.
6962 @item -ffp-contract=@var{style}
6963 @opindex ffp-contract
6964 @option{-ffp-contract=off} disables floating-point expression contraction.
6965 @option{-ffp-contract=fast} enables floating-point expression contraction
6966 such as forming of fused multiply-add operations if the target has
6967 native support for them.
6968 @option{-ffp-contract=on} enables floating-point expression contraction
6969 if allowed by the language standard. This is currently not implemented
6970 and treated equal to @option{-ffp-contract=off}.
6972 The default is @option{-ffp-contract=fast}.
6974 @item -fomit-frame-pointer
6975 @opindex fomit-frame-pointer
6976 Don't keep the frame pointer in a register for functions that
6977 don't need one. This avoids the instructions to save, set up and
6978 restore frame pointers; it also makes an extra register available
6979 in many functions. @strong{It also makes debugging impossible on
6982 On some machines, such as the VAX, this flag has no effect, because
6983 the standard calling sequence automatically handles the frame pointer
6984 and nothing is saved by pretending it doesn't exist. The
6985 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6986 whether a target machine supports this flag. @xref{Registers,,Register
6987 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6989 The default setting (when not optimizing for
6990 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
6991 @option{-fomit-frame-pointer}. You can configure GCC with the
6992 @option{--enable-frame-pointer} configure option to change the default.
6994 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6996 @item -foptimize-sibling-calls
6997 @opindex foptimize-sibling-calls
6998 Optimize sibling and tail recursive calls.
7000 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7002 @item -foptimize-strlen
7003 @opindex foptimize-strlen
7004 Optimize various standard C string functions (e.g. @code{strlen},
7005 @code{strchr} or @code{strcpy}) and
7006 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7008 Enabled at levels @option{-O2}, @option{-O3}.
7012 Do not expand any functions inline apart from those marked with
7013 the @code{always_inline} attribute. This is the default when not
7016 Single functions can be exempted from inlining by marking them
7017 with the @code{noinline} attribute.
7019 @item -finline-small-functions
7020 @opindex finline-small-functions
7021 Integrate functions into their callers when their body is smaller than expected
7022 function call code (so overall size of program gets smaller). The compiler
7023 heuristically decides which functions are simple enough to be worth integrating
7024 in this way. This inlining applies to all functions, even those not declared
7027 Enabled at level @option{-O2}.
7029 @item -findirect-inlining
7030 @opindex findirect-inlining
7031 Inline also indirect calls that are discovered to be known at compile
7032 time thanks to previous inlining. This option has any effect only
7033 when inlining itself is turned on by the @option{-finline-functions}
7034 or @option{-finline-small-functions} options.
7036 Enabled at level @option{-O2}.
7038 @item -finline-functions
7039 @opindex finline-functions
7040 Consider all functions for inlining, even if they are not declared inline.
7041 The compiler heuristically decides which functions are worth integrating
7044 If all calls to a given function are integrated, and the function is
7045 declared @code{static}, then the function is normally not output as
7046 assembler code in its own right.
7048 Enabled at level @option{-O3}.
7050 @item -finline-functions-called-once
7051 @opindex finline-functions-called-once
7052 Consider all @code{static} functions called once for inlining into their
7053 caller even if they are not marked @code{inline}. If a call to a given
7054 function is integrated, then the function is not output as assembler code
7057 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7059 @item -fearly-inlining
7060 @opindex fearly-inlining
7061 Inline functions marked by @code{always_inline} and functions whose body seems
7062 smaller than the function call overhead early before doing
7063 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7064 makes profiling significantly cheaper and usually inlining faster on programs
7065 having large chains of nested wrapper functions.
7071 Perform interprocedural scalar replacement of aggregates, removal of
7072 unused parameters and replacement of parameters passed by reference
7073 by parameters passed by value.
7075 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7077 @item -finline-limit=@var{n}
7078 @opindex finline-limit
7079 By default, GCC limits the size of functions that can be inlined. This flag
7080 allows coarse control of this limit. @var{n} is the size of functions that
7081 can be inlined in number of pseudo instructions.
7083 Inlining is actually controlled by a number of parameters, which may be
7084 specified individually by using @option{--param @var{name}=@var{value}}.
7085 The @option{-finline-limit=@var{n}} option sets some of these parameters
7089 @item max-inline-insns-single
7090 is set to @var{n}/2.
7091 @item max-inline-insns-auto
7092 is set to @var{n}/2.
7095 See below for a documentation of the individual
7096 parameters controlling inlining and for the defaults of these parameters.
7098 @emph{Note:} there may be no value to @option{-finline-limit} that results
7099 in default behavior.
7101 @emph{Note:} pseudo instruction represents, in this particular context, an
7102 abstract measurement of function's size. In no way does it represent a count
7103 of assembly instructions and as such its exact meaning might change from one
7104 release to an another.
7106 @item -fno-keep-inline-dllexport
7107 @opindex fno-keep-inline-dllexport
7108 This is a more fine-grained version of @option{-fkeep-inline-functions},
7109 which applies only to functions that are declared using the @code{dllexport}
7110 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7113 @item -fkeep-inline-functions
7114 @opindex fkeep-inline-functions
7115 In C, emit @code{static} functions that are declared @code{inline}
7116 into the object file, even if the function has been inlined into all
7117 of its callers. This switch does not affect functions using the
7118 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7119 inline functions into the object file.
7121 @item -fkeep-static-functions
7122 @opindex fkeep-static-functions
7123 Emit @code{static} functions into the object file, even if the function
7126 @item -fkeep-static-consts
7127 @opindex fkeep-static-consts
7128 Emit variables declared @code{static const} when optimization isn't turned
7129 on, even if the variables aren't referenced.
7131 GCC enables this option by default. If you want to force the compiler to
7132 check if a variable is referenced, regardless of whether or not
7133 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7135 @item -fmerge-constants
7136 @opindex fmerge-constants
7137 Attempt to merge identical constants (string constants and floating-point
7138 constants) across compilation units.
7140 This option is the default for optimized compilation if the assembler and
7141 linker support it. Use @option{-fno-merge-constants} to inhibit this
7144 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7146 @item -fmerge-all-constants
7147 @opindex fmerge-all-constants
7148 Attempt to merge identical constants and identical variables.
7150 This option implies @option{-fmerge-constants}. In addition to
7151 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7152 arrays or initialized constant variables with integral or floating-point
7153 types. Languages like C or C++ require each variable, including multiple
7154 instances of the same variable in recursive calls, to have distinct locations,
7155 so using this option results in non-conforming
7158 @item -fmodulo-sched
7159 @opindex fmodulo-sched
7160 Perform swing modulo scheduling immediately before the first scheduling
7161 pass. This pass looks at innermost loops and reorders their
7162 instructions by overlapping different iterations.
7164 @item -fmodulo-sched-allow-regmoves
7165 @opindex fmodulo-sched-allow-regmoves
7166 Perform more aggressive SMS-based modulo scheduling with register moves
7167 allowed. By setting this flag certain anti-dependences edges are
7168 deleted, which triggers the generation of reg-moves based on the
7169 life-range analysis. This option is effective only with
7170 @option{-fmodulo-sched} enabled.
7172 @item -fno-branch-count-reg
7173 @opindex fno-branch-count-reg
7174 Avoid running a pass scanning for opportunities to use ``decrement and
7175 branch'' instructions on a count register instead of generating sequences
7176 of instructions that decrement a register, compare it against zero, and
7177 then branch based upon the result. This option is only meaningful on
7178 architectures that support such instructions, which include x86, PowerPC,
7179 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7180 doesn't remove the decrement and branch instructions from the generated
7181 instruction stream introduced by other optimization passes.
7183 Enabled by default at @option{-O1} and higher.
7185 The default is @option{-fbranch-count-reg}.
7187 @item -fno-function-cse
7188 @opindex fno-function-cse
7189 Do not put function addresses in registers; make each instruction that
7190 calls a constant function contain the function's address explicitly.
7192 This option results in less efficient code, but some strange hacks
7193 that alter the assembler output may be confused by the optimizations
7194 performed when this option is not used.
7196 The default is @option{-ffunction-cse}
7198 @item -fno-zero-initialized-in-bss
7199 @opindex fno-zero-initialized-in-bss
7200 If the target supports a BSS section, GCC by default puts variables that
7201 are initialized to zero into BSS@. This can save space in the resulting
7204 This option turns off this behavior because some programs explicitly
7205 rely on variables going to the data section---e.g., so that the
7206 resulting executable can find the beginning of that section and/or make
7207 assumptions based on that.
7209 The default is @option{-fzero-initialized-in-bss}.
7211 @item -fthread-jumps
7212 @opindex fthread-jumps
7213 Perform optimizations that check to see if a jump branches to a
7214 location where another comparison subsumed by the first is found. If
7215 so, the first branch is redirected to either the destination of the
7216 second branch or a point immediately following it, depending on whether
7217 the condition is known to be true or false.
7219 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7221 @item -fsplit-wide-types
7222 @opindex fsplit-wide-types
7223 When using a type that occupies multiple registers, such as @code{long
7224 long} on a 32-bit system, split the registers apart and allocate them
7225 independently. This normally generates better code for those types,
7226 but may make debugging more difficult.
7228 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7231 @item -fcse-follow-jumps
7232 @opindex fcse-follow-jumps
7233 In common subexpression elimination (CSE), scan through jump instructions
7234 when the target of the jump is not reached by any other path. For
7235 example, when CSE encounters an @code{if} statement with an
7236 @code{else} clause, CSE follows the jump when the condition
7239 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7241 @item -fcse-skip-blocks
7242 @opindex fcse-skip-blocks
7243 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7244 follow jumps that conditionally skip over blocks. When CSE
7245 encounters a simple @code{if} statement with no else clause,
7246 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7247 body of the @code{if}.
7249 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7251 @item -frerun-cse-after-loop
7252 @opindex frerun-cse-after-loop
7253 Re-run common subexpression elimination after loop optimizations are
7256 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7260 Perform a global common subexpression elimination pass.
7261 This pass also performs global constant and copy propagation.
7263 @emph{Note:} When compiling a program using computed gotos, a GCC
7264 extension, you may get better run-time performance if you disable
7265 the global common subexpression elimination pass by adding
7266 @option{-fno-gcse} to the command line.
7268 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7272 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7273 attempts to move loads that are only killed by stores into themselves. This
7274 allows a loop containing a load/store sequence to be changed to a load outside
7275 the loop, and a copy/store within the loop.
7277 Enabled by default when @option{-fgcse} is enabled.
7281 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7282 global common subexpression elimination. This pass attempts to move
7283 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7284 loops containing a load/store sequence can be changed to a load before
7285 the loop and a store after the loop.
7287 Not enabled at any optimization level.
7291 When @option{-fgcse-las} is enabled, the global common subexpression
7292 elimination pass eliminates redundant loads that come after stores to the
7293 same memory location (both partial and full redundancies).
7295 Not enabled at any optimization level.
7297 @item -fgcse-after-reload
7298 @opindex fgcse-after-reload
7299 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7300 pass is performed after reload. The purpose of this pass is to clean up
7303 @item -faggressive-loop-optimizations
7304 @opindex faggressive-loop-optimizations
7305 This option tells the loop optimizer to use language constraints to
7306 derive bounds for the number of iterations of a loop. This assumes that
7307 loop code does not invoke undefined behavior by for example causing signed
7308 integer overflows or out-of-bound array accesses. The bounds for the
7309 number of iterations of a loop are used to guide loop unrolling and peeling
7310 and loop exit test optimizations.
7311 This option is enabled by default.
7313 @item -funconstrained-commons
7314 @opindex funconstrained-commons
7315 This option tells the compiler that variables declared in common blocks
7316 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7317 prevents certain optimizations that depend on knowing the array bounds.
7319 @item -fcrossjumping
7320 @opindex fcrossjumping
7321 Perform cross-jumping transformation.
7322 This transformation unifies equivalent code and saves code size. The
7323 resulting code may or may not perform better than without cross-jumping.
7325 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7327 @item -fauto-inc-dec
7328 @opindex fauto-inc-dec
7329 Combine increments or decrements of addresses with memory accesses.
7330 This pass is always skipped on architectures that do not have
7331 instructions to support this. Enabled by default at @option{-O} and
7332 higher on architectures that support this.
7336 Perform dead code elimination (DCE) on RTL@.
7337 Enabled by default at @option{-O} and higher.
7341 Perform dead store elimination (DSE) on RTL@.
7342 Enabled by default at @option{-O} and higher.
7344 @item -fif-conversion
7345 @opindex fif-conversion
7346 Attempt to transform conditional jumps into branch-less equivalents. This
7347 includes use of conditional moves, min, max, set flags and abs instructions, and
7348 some tricks doable by standard arithmetics. The use of conditional execution
7349 on chips where it is available is controlled by @option{-fif-conversion2}.
7351 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7353 @item -fif-conversion2
7354 @opindex fif-conversion2
7355 Use conditional execution (where available) to transform conditional jumps into
7356 branch-less equivalents.
7358 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7360 @item -fdeclone-ctor-dtor
7361 @opindex fdeclone-ctor-dtor
7362 The C++ ABI requires multiple entry points for constructors and
7363 destructors: one for a base subobject, one for a complete object, and
7364 one for a virtual destructor that calls operator delete afterwards.
7365 For a hierarchy with virtual bases, the base and complete variants are
7366 clones, which means two copies of the function. With this option, the
7367 base and complete variants are changed to be thunks that call a common
7370 Enabled by @option{-Os}.
7372 @item -fdelete-null-pointer-checks
7373 @opindex fdelete-null-pointer-checks
7374 Assume that programs cannot safely dereference null pointers, and that
7375 no code or data element resides at address zero.
7376 This option enables simple constant
7377 folding optimizations at all optimization levels. In addition, other
7378 optimization passes in GCC use this flag to control global dataflow
7379 analyses that eliminate useless checks for null pointers; these assume
7380 that a memory access to address zero always results in a trap, so
7381 that if a pointer is checked after it has already been dereferenced,
7384 Note however that in some environments this assumption is not true.
7385 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7386 for programs that depend on that behavior.
7388 This option is enabled by default on most targets. On Nios II ELF, it
7389 defaults to off. On AVR and CR16, this option is completely disabled.
7391 Passes that use the dataflow information
7392 are enabled independently at different optimization levels.
7394 @item -fdevirtualize
7395 @opindex fdevirtualize
7396 Attempt to convert calls to virtual functions to direct calls. This
7397 is done both within a procedure and interprocedurally as part of
7398 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7399 propagation (@option{-fipa-cp}).
7400 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7402 @item -fdevirtualize-speculatively
7403 @opindex fdevirtualize-speculatively
7404 Attempt to convert calls to virtual functions to speculative direct calls.
7405 Based on the analysis of the type inheritance graph, determine for a given call
7406 the set of likely targets. If the set is small, preferably of size 1, change
7407 the call into a conditional deciding between direct and indirect calls. The
7408 speculative calls enable more optimizations, such as inlining. When they seem
7409 useless after further optimization, they are converted back into original form.
7411 @item -fdevirtualize-at-ltrans
7412 @opindex fdevirtualize-at-ltrans
7413 Stream extra information needed for aggressive devirtualization when running
7414 the link-time optimizer in local transformation mode.
7415 This option enables more devirtualization but
7416 significantly increases the size of streamed data. For this reason it is
7417 disabled by default.
7419 @item -fexpensive-optimizations
7420 @opindex fexpensive-optimizations
7421 Perform a number of minor optimizations that are relatively expensive.
7423 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7427 Attempt to remove redundant extension instructions. This is especially
7428 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7429 registers after writing to their lower 32-bit half.
7431 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7432 @option{-O3}, @option{-Os}.
7434 @item -fno-lifetime-dse
7435 @opindex fno-lifetime-dse
7436 In C++ the value of an object is only affected by changes within its
7437 lifetime: when the constructor begins, the object has an indeterminate
7438 value, and any changes during the lifetime of the object are dead when
7439 the object is destroyed. Normally dead store elimination will take
7440 advantage of this; if your code relies on the value of the object
7441 storage persisting beyond the lifetime of the object, you can use this
7442 flag to disable this optimization. To preserve stores before the
7443 constructor starts (e.g. because your operator new clears the object
7444 storage) but still treat the object as dead after the destructor you,
7445 can use @option{-flifetime-dse=1}. The default behavior can be
7446 explicitly selected with @option{-flifetime-dse=2}.
7447 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7449 @item -flive-range-shrinkage
7450 @opindex flive-range-shrinkage
7451 Attempt to decrease register pressure through register live range
7452 shrinkage. This is helpful for fast processors with small or moderate
7455 @item -fira-algorithm=@var{algorithm}
7456 @opindex fira-algorithm
7457 Use the specified coloring algorithm for the integrated register
7458 allocator. The @var{algorithm} argument can be @samp{priority}, which
7459 specifies Chow's priority coloring, or @samp{CB}, which specifies
7460 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7461 for all architectures, but for those targets that do support it, it is
7462 the default because it generates better code.
7464 @item -fira-region=@var{region}
7465 @opindex fira-region
7466 Use specified regions for the integrated register allocator. The
7467 @var{region} argument should be one of the following:
7472 Use all loops as register allocation regions.
7473 This can give the best results for machines with a small and/or
7474 irregular register set.
7477 Use all loops except for loops with small register pressure
7478 as the regions. This value usually gives
7479 the best results in most cases and for most architectures,
7480 and is enabled by default when compiling with optimization for speed
7481 (@option{-O}, @option{-O2}, @dots{}).
7484 Use all functions as a single region.
7485 This typically results in the smallest code size, and is enabled by default for
7486 @option{-Os} or @option{-O0}.
7490 @item -fira-hoist-pressure
7491 @opindex fira-hoist-pressure
7492 Use IRA to evaluate register pressure in the code hoisting pass for
7493 decisions to hoist expressions. This option usually results in smaller
7494 code, but it can slow the compiler down.
7496 This option is enabled at level @option{-Os} for all targets.
7498 @item -fira-loop-pressure
7499 @opindex fira-loop-pressure
7500 Use IRA to evaluate register pressure in loops for decisions to move
7501 loop invariants. This option usually results in generation
7502 of faster and smaller code on machines with large register files (>= 32
7503 registers), but it can slow the compiler down.
7505 This option is enabled at level @option{-O3} for some targets.
7507 @item -fno-ira-share-save-slots
7508 @opindex fno-ira-share-save-slots
7509 Disable sharing of stack slots used for saving call-used hard
7510 registers living through a call. Each hard register gets a
7511 separate stack slot, and as a result function stack frames are
7514 @item -fno-ira-share-spill-slots
7515 @opindex fno-ira-share-spill-slots
7516 Disable sharing of stack slots allocated for pseudo-registers. Each
7517 pseudo-register that does not get a hard register gets a separate
7518 stack slot, and as a result function stack frames are larger.
7522 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7523 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7524 values if it is profitable.
7526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7528 @item -fdelayed-branch
7529 @opindex fdelayed-branch
7530 If supported for the target machine, attempt to reorder instructions
7531 to exploit instruction slots available after delayed branch
7534 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7536 @item -fschedule-insns
7537 @opindex fschedule-insns
7538 If supported for the target machine, attempt to reorder instructions to
7539 eliminate execution stalls due to required data being unavailable. This
7540 helps machines that have slow floating point or memory load instructions
7541 by allowing other instructions to be issued until the result of the load
7542 or floating-point instruction is required.
7544 Enabled at levels @option{-O2}, @option{-O3}.
7546 @item -fschedule-insns2
7547 @opindex fschedule-insns2
7548 Similar to @option{-fschedule-insns}, but requests an additional pass of
7549 instruction scheduling after register allocation has been done. This is
7550 especially useful on machines with a relatively small number of
7551 registers and where memory load instructions take more than one cycle.
7553 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7555 @item -fno-sched-interblock
7556 @opindex fno-sched-interblock
7557 Don't schedule instructions across basic blocks. This is normally
7558 enabled by default when scheduling before register allocation, i.e.@:
7559 with @option{-fschedule-insns} or at @option{-O2} or higher.
7561 @item -fno-sched-spec
7562 @opindex fno-sched-spec
7563 Don't allow speculative motion of non-load instructions. This is normally
7564 enabled by default when scheduling before register allocation, i.e.@:
7565 with @option{-fschedule-insns} or at @option{-O2} or higher.
7567 @item -fsched-pressure
7568 @opindex fsched-pressure
7569 Enable register pressure sensitive insn scheduling before register
7570 allocation. This only makes sense when scheduling before register
7571 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7572 @option{-O2} or higher. Usage of this option can improve the
7573 generated code and decrease its size by preventing register pressure
7574 increase above the number of available hard registers and subsequent
7575 spills in register allocation.
7577 @item -fsched-spec-load
7578 @opindex fsched-spec-load
7579 Allow speculative motion of some load instructions. This only makes
7580 sense when scheduling before register allocation, i.e.@: with
7581 @option{-fschedule-insns} or at @option{-O2} or higher.
7583 @item -fsched-spec-load-dangerous
7584 @opindex fsched-spec-load-dangerous
7585 Allow speculative motion of more load instructions. This only makes
7586 sense when scheduling before register allocation, i.e.@: with
7587 @option{-fschedule-insns} or at @option{-O2} or higher.
7589 @item -fsched-stalled-insns
7590 @itemx -fsched-stalled-insns=@var{n}
7591 @opindex fsched-stalled-insns
7592 Define how many insns (if any) can be moved prematurely from the queue
7593 of stalled insns into the ready list during the second scheduling pass.
7594 @option{-fno-sched-stalled-insns} means that no insns are moved
7595 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7596 on how many queued insns can be moved prematurely.
7597 @option{-fsched-stalled-insns} without a value is equivalent to
7598 @option{-fsched-stalled-insns=1}.
7600 @item -fsched-stalled-insns-dep
7601 @itemx -fsched-stalled-insns-dep=@var{n}
7602 @opindex fsched-stalled-insns-dep
7603 Define how many insn groups (cycles) are examined for a dependency
7604 on a stalled insn that is a candidate for premature removal from the queue
7605 of stalled insns. This has an effect only during the second scheduling pass,
7606 and only if @option{-fsched-stalled-insns} is used.
7607 @option{-fno-sched-stalled-insns-dep} is equivalent to
7608 @option{-fsched-stalled-insns-dep=0}.
7609 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7610 @option{-fsched-stalled-insns-dep=1}.
7612 @item -fsched2-use-superblocks
7613 @opindex fsched2-use-superblocks
7614 When scheduling after register allocation, use superblock scheduling.
7615 This allows motion across basic block boundaries,
7616 resulting in faster schedules. This option is experimental, as not all machine
7617 descriptions used by GCC model the CPU closely enough to avoid unreliable
7618 results from the algorithm.
7620 This only makes sense when scheduling after register allocation, i.e.@: with
7621 @option{-fschedule-insns2} or at @option{-O2} or higher.
7623 @item -fsched-group-heuristic
7624 @opindex fsched-group-heuristic
7625 Enable the group heuristic in the scheduler. This heuristic favors
7626 the instruction that belongs to a schedule group. This is enabled
7627 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7628 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7630 @item -fsched-critical-path-heuristic
7631 @opindex fsched-critical-path-heuristic
7632 Enable the critical-path heuristic in the scheduler. This heuristic favors
7633 instructions on the critical path. This is enabled by default when
7634 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7635 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7637 @item -fsched-spec-insn-heuristic
7638 @opindex fsched-spec-insn-heuristic
7639 Enable the speculative instruction heuristic in the scheduler. This
7640 heuristic favors speculative instructions with greater dependency weakness.
7641 This is enabled by default when scheduling is enabled, i.e.@:
7642 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7643 or at @option{-O2} or higher.
7645 @item -fsched-rank-heuristic
7646 @opindex fsched-rank-heuristic
7647 Enable the rank heuristic in the scheduler. This heuristic favors
7648 the instruction belonging to a basic block with greater size or frequency.
7649 This is enabled by default when scheduling is enabled, i.e.@:
7650 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7651 at @option{-O2} or higher.
7653 @item -fsched-last-insn-heuristic
7654 @opindex fsched-last-insn-heuristic
7655 Enable the last-instruction heuristic in the scheduler. This heuristic
7656 favors the instruction that is less dependent on the last instruction
7657 scheduled. This is enabled by default when scheduling is enabled,
7658 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7659 at @option{-O2} or higher.
7661 @item -fsched-dep-count-heuristic
7662 @opindex fsched-dep-count-heuristic
7663 Enable the dependent-count heuristic in the scheduler. This heuristic
7664 favors the instruction that has more instructions depending on it.
7665 This is enabled by default when scheduling is enabled, i.e.@:
7666 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7667 at @option{-O2} or higher.
7669 @item -freschedule-modulo-scheduled-loops
7670 @opindex freschedule-modulo-scheduled-loops
7671 Modulo scheduling is performed before traditional scheduling. If a loop
7672 is modulo scheduled, later scheduling passes may change its schedule.
7673 Use this option to control that behavior.
7675 @item -fselective-scheduling
7676 @opindex fselective-scheduling
7677 Schedule instructions using selective scheduling algorithm. Selective
7678 scheduling runs instead of the first scheduler pass.
7680 @item -fselective-scheduling2
7681 @opindex fselective-scheduling2
7682 Schedule instructions using selective scheduling algorithm. Selective
7683 scheduling runs instead of the second scheduler pass.
7685 @item -fsel-sched-pipelining
7686 @opindex fsel-sched-pipelining
7687 Enable software pipelining of innermost loops during selective scheduling.
7688 This option has no effect unless one of @option{-fselective-scheduling} or
7689 @option{-fselective-scheduling2} is turned on.
7691 @item -fsel-sched-pipelining-outer-loops
7692 @opindex fsel-sched-pipelining-outer-loops
7693 When pipelining loops during selective scheduling, also pipeline outer loops.
7694 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7696 @item -fsemantic-interposition
7697 @opindex fsemantic-interposition
7698 Some object formats, like ELF, allow interposing of symbols by the
7700 This means that for symbols exported from the DSO, the compiler cannot perform
7701 interprocedural propagation, inlining and other optimizations in anticipation
7702 that the function or variable in question may change. While this feature is
7703 useful, for example, to rewrite memory allocation functions by a debugging
7704 implementation, it is expensive in the terms of code quality.
7705 With @option{-fno-semantic-interposition} the compiler assumes that
7706 if interposition happens for functions the overwriting function will have
7707 precisely the same semantics (and side effects).
7708 Similarly if interposition happens
7709 for variables, the constructor of the variable will be the same. The flag
7710 has no effect for functions explicitly declared inline
7711 (where it is never allowed for interposition to change semantics)
7712 and for symbols explicitly declared weak.
7715 @opindex fshrink-wrap
7716 Emit function prologues only before parts of the function that need it,
7717 rather than at the top of the function. This flag is enabled by default at
7718 @option{-O} and higher.
7720 @item -fshrink-wrap-separate
7721 @opindex fshrink-wrap-separate
7722 Shrink-wrap separate parts of the prologue and epilogue separately, so that
7723 those parts are only executed when needed.
7724 This option is on by default, but has no effect unless @option{-fshrink-wrap}
7725 is also turned on and the target supports this.
7727 @item -fcaller-saves
7728 @opindex fcaller-saves
7729 Enable allocation of values to registers that are clobbered by
7730 function calls, by emitting extra instructions to save and restore the
7731 registers around such calls. Such allocation is done only when it
7732 seems to result in better code.
7734 This option is always enabled by default on certain machines, usually
7735 those which have no call-preserved registers to use instead.
7737 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7739 @item -fcombine-stack-adjustments
7740 @opindex fcombine-stack-adjustments
7741 Tracks stack adjustments (pushes and pops) and stack memory references
7742 and then tries to find ways to combine them.
7744 Enabled by default at @option{-O1} and higher.
7748 Use caller save registers for allocation if those registers are not used by
7749 any called function. In that case it is not necessary to save and restore
7750 them around calls. This is only possible if called functions are part of
7751 same compilation unit as current function and they are compiled before it.
7753 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
7754 is disabled if generated code will be instrumented for profiling
7755 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
7756 exactly (this happens on targets that do not expose prologues
7757 and epilogues in RTL).
7759 @item -fconserve-stack
7760 @opindex fconserve-stack
7761 Attempt to minimize stack usage. The compiler attempts to use less
7762 stack space, even if that makes the program slower. This option
7763 implies setting the @option{large-stack-frame} parameter to 100
7764 and the @option{large-stack-frame-growth} parameter to 400.
7766 @item -ftree-reassoc
7767 @opindex ftree-reassoc
7768 Perform reassociation on trees. This flag is enabled by default
7769 at @option{-O} and higher.
7771 @item -fcode-hoisting
7772 @opindex fcode-hoisting
7773 Perform code hoisting. Code hoisting tries to move the
7774 evaluation of expressions executed on all paths to the function exit
7775 as early as possible. This is especially useful as a code size
7776 optimization, but it often helps for code speed as well.
7777 This flag is enabled by default at @option{-O2} and higher.
7781 Perform partial redundancy elimination (PRE) on trees. This flag is
7782 enabled by default at @option{-O2} and @option{-O3}.
7784 @item -ftree-partial-pre
7785 @opindex ftree-partial-pre
7786 Make partial redundancy elimination (PRE) more aggressive. This flag is
7787 enabled by default at @option{-O3}.
7789 @item -ftree-forwprop
7790 @opindex ftree-forwprop
7791 Perform forward propagation on trees. This flag is enabled by default
7792 at @option{-O} and higher.
7796 Perform full redundancy elimination (FRE) on trees. The difference
7797 between FRE and PRE is that FRE only considers expressions
7798 that are computed on all paths leading to the redundant computation.
7799 This analysis is faster than PRE, though it exposes fewer redundancies.
7800 This flag is enabled by default at @option{-O} and higher.
7802 @item -ftree-phiprop
7803 @opindex ftree-phiprop
7804 Perform hoisting of loads from conditional pointers on trees. This
7805 pass is enabled by default at @option{-O} and higher.
7807 @item -fhoist-adjacent-loads
7808 @opindex fhoist-adjacent-loads
7809 Speculatively hoist loads from both branches of an if-then-else if the
7810 loads are from adjacent locations in the same structure and the target
7811 architecture has a conditional move instruction. This flag is enabled
7812 by default at @option{-O2} and higher.
7814 @item -ftree-copy-prop
7815 @opindex ftree-copy-prop
7816 Perform copy propagation on trees. This pass eliminates unnecessary
7817 copy operations. This flag is enabled by default at @option{-O} and
7820 @item -fipa-pure-const
7821 @opindex fipa-pure-const
7822 Discover which functions are pure or constant.
7823 Enabled by default at @option{-O} and higher.
7825 @item -fipa-reference
7826 @opindex fipa-reference
7827 Discover which static variables do not escape the
7829 Enabled by default at @option{-O} and higher.
7833 Perform interprocedural pointer analysis and interprocedural modification
7834 and reference analysis. This option can cause excessive memory and
7835 compile-time usage on large compilation units. It is not enabled by
7836 default at any optimization level.
7839 @opindex fipa-profile
7840 Perform interprocedural profile propagation. The functions called only from
7841 cold functions are marked as cold. Also functions executed once (such as
7842 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7843 functions and loop less parts of functions executed once are then optimized for
7845 Enabled by default at @option{-O} and higher.
7849 Perform interprocedural constant propagation.
7850 This optimization analyzes the program to determine when values passed
7851 to functions are constants and then optimizes accordingly.
7852 This optimization can substantially increase performance
7853 if the application has constants passed to functions.
7854 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7856 @item -fipa-cp-clone
7857 @opindex fipa-cp-clone
7858 Perform function cloning to make interprocedural constant propagation stronger.
7859 When enabled, interprocedural constant propagation performs function cloning
7860 when externally visible function can be called with constant arguments.
7861 Because this optimization can create multiple copies of functions,
7862 it may significantly increase code size
7863 (see @option{--param ipcp-unit-growth=@var{value}}).
7864 This flag is enabled by default at @option{-O3}.
7866 @item -fipa-cp-alignment
7867 @opindex -fipa-cp-alignment
7868 When enabled, this optimization propagates alignment of function
7869 parameters to support better vectorization and string operations.
7871 This flag is enabled by default at @option{-O2} and @option{-Os}. It
7872 requires that @option{-fipa-cp} is enabled.
7873 @option{-fipa-cp-alignment} is obsolete, use @option{-fipa-bit-cp} instead.
7876 @opindex -fipa-bit-cp
7877 When enabled, perform ipa bitwise constant propagation. This flag is
7878 enabled by default at @option{-O2}. It requires that @option{-fipa-cp}
7883 Perform Identical Code Folding for functions and read-only variables.
7884 The optimization reduces code size and may disturb unwind stacks by replacing
7885 a function by equivalent one with a different name. The optimization works
7886 more effectively with link time optimization enabled.
7888 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
7889 works on different levels and thus the optimizations are not same - there are
7890 equivalences that are found only by GCC and equivalences found only by Gold.
7892 This flag is enabled by default at @option{-O2} and @option{-Os}.
7894 @item -fisolate-erroneous-paths-dereference
7895 @opindex fisolate-erroneous-paths-dereference
7896 Detect paths that trigger erroneous or undefined behavior due to
7897 dereferencing a null pointer. Isolate those paths from the main control
7898 flow and turn the statement with erroneous or undefined behavior into a trap.
7899 This flag is enabled by default at @option{-O2} and higher and depends on
7900 @option{-fdelete-null-pointer-checks} also being enabled.
7902 @item -fisolate-erroneous-paths-attribute
7903 @opindex fisolate-erroneous-paths-attribute
7904 Detect paths that trigger erroneous or undefined behavior due a null value
7905 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
7906 attribute. Isolate those paths from the main control flow and turn the
7907 statement with erroneous or undefined behavior into a trap. This is not
7908 currently enabled, but may be enabled by @option{-O2} in the future.
7912 Perform forward store motion on trees. This flag is
7913 enabled by default at @option{-O} and higher.
7915 @item -ftree-bit-ccp
7916 @opindex ftree-bit-ccp
7917 Perform sparse conditional bit constant propagation on trees and propagate
7918 pointer alignment information.
7919 This pass only operates on local scalar variables and is enabled by default
7920 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7924 Perform sparse conditional constant propagation (CCP) on trees. This
7925 pass only operates on local scalar variables and is enabled by default
7926 at @option{-O} and higher.
7928 @item -fssa-backprop
7929 @opindex fssa-backprop
7930 Propagate information about uses of a value up the definition chain
7931 in order to simplify the definitions. For example, this pass strips
7932 sign operations if the sign of a value never matters. The flag is
7933 enabled by default at @option{-O} and higher.
7936 @opindex fssa-phiopt
7937 Perform pattern matching on SSA PHI nodes to optimize conditional
7938 code. This pass is enabled by default at @option{-O} and higher.
7940 @item -ftree-switch-conversion
7941 @opindex ftree-switch-conversion
7942 Perform conversion of simple initializations in a switch to
7943 initializations from a scalar array. This flag is enabled by default
7944 at @option{-O2} and higher.
7946 @item -ftree-tail-merge
7947 @opindex ftree-tail-merge
7948 Look for identical code sequences. When found, replace one with a jump to the
7949 other. This optimization is known as tail merging or cross jumping. This flag
7950 is enabled by default at @option{-O2} and higher. The compilation time
7952 be limited using @option{max-tail-merge-comparisons} parameter and
7953 @option{max-tail-merge-iterations} parameter.
7957 Perform dead code elimination (DCE) on trees. This flag is enabled by
7958 default at @option{-O} and higher.
7960 @item -ftree-builtin-call-dce
7961 @opindex ftree-builtin-call-dce
7962 Perform conditional dead code elimination (DCE) for calls to built-in functions
7963 that may set @code{errno} but are otherwise side-effect free. This flag is
7964 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7967 @item -ftree-dominator-opts
7968 @opindex ftree-dominator-opts
7969 Perform a variety of simple scalar cleanups (constant/copy
7970 propagation, redundancy elimination, range propagation and expression
7971 simplification) based on a dominator tree traversal. This also
7972 performs jump threading (to reduce jumps to jumps). This flag is
7973 enabled by default at @option{-O} and higher.
7977 Perform dead store elimination (DSE) on trees. A dead store is a store into
7978 a memory location that is later overwritten by another store without
7979 any intervening loads. In this case the earlier store can be deleted. This
7980 flag is enabled by default at @option{-O} and higher.
7984 Perform loop header copying on trees. This is beneficial since it increases
7985 effectiveness of code motion optimizations. It also saves one jump. This flag
7986 is enabled by default at @option{-O} and higher. It is not enabled
7987 for @option{-Os}, since it usually increases code size.
7989 @item -ftree-loop-optimize
7990 @opindex ftree-loop-optimize
7991 Perform loop optimizations on trees. This flag is enabled by default
7992 at @option{-O} and higher.
7994 @item -ftree-loop-linear
7995 @itemx -floop-interchange
7996 @itemx -floop-strip-mine
7998 @itemx -floop-unroll-and-jam
7999 @opindex ftree-loop-linear
8000 @opindex floop-interchange
8001 @opindex floop-strip-mine
8002 @opindex floop-block
8003 @opindex floop-unroll-and-jam
8004 Perform loop nest optimizations. Same as
8005 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8006 to be configured with @option{--with-isl} to enable the Graphite loop
8007 transformation infrastructure.
8009 @item -fgraphite-identity
8010 @opindex fgraphite-identity
8011 Enable the identity transformation for graphite. For every SCoP we generate
8012 the polyhedral representation and transform it back to gimple. Using
8013 @option{-fgraphite-identity} we can check the costs or benefits of the
8014 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8015 are also performed by the code generator isl, like index splitting and
8016 dead code elimination in loops.
8018 @item -floop-nest-optimize
8019 @opindex floop-nest-optimize
8020 Enable the isl based loop nest optimizer. This is a generic loop nest
8021 optimizer based on the Pluto optimization algorithms. It calculates a loop
8022 structure optimized for data-locality and parallelism. This option
8025 @item -floop-parallelize-all
8026 @opindex floop-parallelize-all
8027 Use the Graphite data dependence analysis to identify loops that can
8028 be parallelized. Parallelize all the loops that can be analyzed to
8029 not contain loop carried dependences without checking that it is
8030 profitable to parallelize the loops.
8032 @item -ftree-coalesce-vars
8033 @opindex ftree-coalesce-vars
8034 While transforming the program out of the SSA representation, attempt to
8035 reduce copying by coalescing versions of different user-defined
8036 variables, instead of just compiler temporaries. This may severely
8037 limit the ability to debug an optimized program compiled with
8038 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8039 prevents SSA coalescing of user variables. This option is enabled by
8040 default if optimization is enabled, and it does very little otherwise.
8042 @item -ftree-loop-if-convert
8043 @opindex ftree-loop-if-convert
8044 Attempt to transform conditional jumps in the innermost loops to
8045 branch-less equivalents. The intent is to remove control-flow from
8046 the innermost loops in order to improve the ability of the
8047 vectorization pass to handle these loops. This is enabled by default
8048 if vectorization is enabled.
8050 @item -ftree-loop-if-convert-stores
8051 @opindex ftree-loop-if-convert-stores
8052 Attempt to also if-convert conditional jumps containing memory writes.
8053 This transformation can be unsafe for multi-threaded programs as it
8054 transforms conditional memory writes into unconditional memory writes.
8057 for (i = 0; i < N; i++)
8063 for (i = 0; i < N; i++)
8064 A[i] = cond ? expr : A[i];
8066 potentially producing data races.
8068 @item -ftree-loop-distribution
8069 @opindex ftree-loop-distribution
8070 Perform loop distribution. This flag can improve cache performance on
8071 big loop bodies and allow further loop optimizations, like
8072 parallelization or vectorization, to take place. For example, the loop
8089 @item -ftree-loop-distribute-patterns
8090 @opindex ftree-loop-distribute-patterns
8091 Perform loop distribution of patterns that can be code generated with
8092 calls to a library. This flag is enabled by default at @option{-O3}.
8094 This pass distributes the initialization loops and generates a call to
8095 memset zero. For example, the loop
8111 and the initialization loop is transformed into a call to memset zero.
8113 @item -ftree-loop-im
8114 @opindex ftree-loop-im
8115 Perform loop invariant motion on trees. This pass moves only invariants that
8116 are hard to handle at RTL level (function calls, operations that expand to
8117 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8118 operands of conditions that are invariant out of the loop, so that we can use
8119 just trivial invariantness analysis in loop unswitching. The pass also includes
8122 @item -ftree-loop-ivcanon
8123 @opindex ftree-loop-ivcanon
8124 Create a canonical counter for number of iterations in loops for which
8125 determining number of iterations requires complicated analysis. Later
8126 optimizations then may determine the number easily. Useful especially
8127 in connection with unrolling.
8131 Perform induction variable optimizations (strength reduction, induction
8132 variable merging and induction variable elimination) on trees.
8134 @item -ftree-parallelize-loops=n
8135 @opindex ftree-parallelize-loops
8136 Parallelize loops, i.e., split their iteration space to run in n threads.
8137 This is only possible for loops whose iterations are independent
8138 and can be arbitrarily reordered. The optimization is only
8139 profitable on multiprocessor machines, for loops that are CPU-intensive,
8140 rather than constrained e.g.@: by memory bandwidth. This option
8141 implies @option{-pthread}, and thus is only supported on targets
8142 that have support for @option{-pthread}.
8146 Perform function-local points-to analysis on trees. This flag is
8147 enabled by default at @option{-O} and higher.
8151 Perform scalar replacement of aggregates. This pass replaces structure
8152 references with scalars to prevent committing structures to memory too
8153 early. This flag is enabled by default at @option{-O} and higher.
8155 @item -fstore-merging
8156 @opindex fstore-merging
8157 Perform merging of narrow stores to consecutive memory addresses. This pass
8158 merges contiguous stores of immediate values narrower than a word into fewer
8159 wider stores to reduce the number of instructions. This is enabled by default
8160 at @option{-O} and higher.
8164 Perform temporary expression replacement during the SSA->normal phase. Single
8165 use/single def temporaries are replaced at their use location with their
8166 defining expression. This results in non-GIMPLE code, but gives the expanders
8167 much more complex trees to work on resulting in better RTL generation. This is
8168 enabled by default at @option{-O} and higher.
8172 Perform straight-line strength reduction on trees. This recognizes related
8173 expressions involving multiplications and replaces them by less expensive
8174 calculations when possible. This is enabled by default at @option{-O} and
8177 @item -ftree-vectorize
8178 @opindex ftree-vectorize
8179 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8180 and @option{-ftree-slp-vectorize} if not explicitly specified.
8182 @item -ftree-loop-vectorize
8183 @opindex ftree-loop-vectorize
8184 Perform loop vectorization on trees. This flag is enabled by default at
8185 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8187 @item -ftree-slp-vectorize
8188 @opindex ftree-slp-vectorize
8189 Perform basic block vectorization on trees. This flag is enabled by default at
8190 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8192 @item -fvect-cost-model=@var{model}
8193 @opindex fvect-cost-model
8194 Alter the cost model used for vectorization. The @var{model} argument
8195 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8196 With the @samp{unlimited} model the vectorized code-path is assumed
8197 to be profitable while with the @samp{dynamic} model a runtime check
8198 guards the vectorized code-path to enable it only for iteration
8199 counts that will likely execute faster than when executing the original
8200 scalar loop. The @samp{cheap} model disables vectorization of
8201 loops where doing so would be cost prohibitive for example due to
8202 required runtime checks for data dependence or alignment but otherwise
8203 is equal to the @samp{dynamic} model.
8204 The default cost model depends on other optimization flags and is
8205 either @samp{dynamic} or @samp{cheap}.
8207 @item -fsimd-cost-model=@var{model}
8208 @opindex fsimd-cost-model
8209 Alter the cost model used for vectorization of loops marked with the OpenMP
8210 or Cilk Plus simd directive. The @var{model} argument should be one of
8211 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8212 have the same meaning as described in @option{-fvect-cost-model} and by
8213 default a cost model defined with @option{-fvect-cost-model} is used.
8217 Perform Value Range Propagation on trees. This is similar to the
8218 constant propagation pass, but instead of values, ranges of values are
8219 propagated. This allows the optimizers to remove unnecessary range
8220 checks like array bound checks and null pointer checks. This is
8221 enabled by default at @option{-O2} and higher. Null pointer check
8222 elimination is only done if @option{-fdelete-null-pointer-checks} is
8226 @opindex fsplit-paths
8227 Split paths leading to loop backedges. This can improve dead code
8228 elimination and common subexpression elimination. This is enabled by
8229 default at @option{-O2} and above.
8231 @item -fsplit-ivs-in-unroller
8232 @opindex fsplit-ivs-in-unroller
8233 Enables expression of values of induction variables in later iterations
8234 of the unrolled loop using the value in the first iteration. This breaks
8235 long dependency chains, thus improving efficiency of the scheduling passes.
8237 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8238 same effect. However, that is not reliable in cases where the loop body
8239 is more complicated than a single basic block. It also does not work at all
8240 on some architectures due to restrictions in the CSE pass.
8242 This optimization is enabled by default.
8244 @item -fvariable-expansion-in-unroller
8245 @opindex fvariable-expansion-in-unroller
8246 With this option, the compiler creates multiple copies of some
8247 local variables when unrolling a loop, which can result in superior code.
8249 @item -fpartial-inlining
8250 @opindex fpartial-inlining
8251 Inline parts of functions. This option has any effect only
8252 when inlining itself is turned on by the @option{-finline-functions}
8253 or @option{-finline-small-functions} options.
8255 Enabled at level @option{-O2}.
8257 @item -fpredictive-commoning
8258 @opindex fpredictive-commoning
8259 Perform predictive commoning optimization, i.e., reusing computations
8260 (especially memory loads and stores) performed in previous
8261 iterations of loops.
8263 This option is enabled at level @option{-O3}.
8265 @item -fprefetch-loop-arrays
8266 @opindex fprefetch-loop-arrays
8267 If supported by the target machine, generate instructions to prefetch
8268 memory to improve the performance of loops that access large arrays.
8270 This option may generate better or worse code; results are highly
8271 dependent on the structure of loops within the source code.
8273 Disabled at level @option{-Os}.
8275 @item -fprintf-return-value
8276 @opindex fprintf-return-value
8277 Substitute constants for known return value of formatted output functions
8278 such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and @code{vsnprintf}
8279 (but not @code{printf} of @code{fprintf}). This transformation allows GCC
8280 to optimize or even eliminate branches based on the known return value of
8281 these functions called with arguments that are either constant, or whose
8282 values are known to be in a range that makes determining the exact return
8283 value possible. For example, both the branch and the body of the @code{if}
8284 statement (but not the call to @code{snprint}) can be optimized away when
8285 @code{i} is a 32-bit or smaller integer because the return value is guaranteed
8290 if (snprintf (buf, "%08x", i) >= sizeof buf)
8294 The @option{-fprintf-return-value} option relies on other optimizations
8295 and yields best results with @option{-O2}. It works in tandem with the
8296 @option{-Wformat-length} option. The @option{-fprintf-return-value}
8297 option is disabled by default.
8300 @itemx -fno-peephole2
8301 @opindex fno-peephole
8302 @opindex fno-peephole2
8303 Disable any machine-specific peephole optimizations. The difference
8304 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8305 are implemented in the compiler; some targets use one, some use the
8306 other, a few use both.
8308 @option{-fpeephole} is enabled by default.
8309 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8311 @item -fno-guess-branch-probability
8312 @opindex fno-guess-branch-probability
8313 Do not guess branch probabilities using heuristics.
8315 GCC uses heuristics to guess branch probabilities if they are
8316 not provided by profiling feedback (@option{-fprofile-arcs}). These
8317 heuristics are based on the control flow graph. If some branch probabilities
8318 are specified by @code{__builtin_expect}, then the heuristics are
8319 used to guess branch probabilities for the rest of the control flow graph,
8320 taking the @code{__builtin_expect} info into account. The interactions
8321 between the heuristics and @code{__builtin_expect} can be complex, and in
8322 some cases, it may be useful to disable the heuristics so that the effects
8323 of @code{__builtin_expect} are easier to understand.
8325 The default is @option{-fguess-branch-probability} at levels
8326 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8328 @item -freorder-blocks
8329 @opindex freorder-blocks
8330 Reorder basic blocks in the compiled function in order to reduce number of
8331 taken branches and improve code locality.
8333 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8335 @item -freorder-blocks-algorithm=@var{algorithm}
8336 @opindex freorder-blocks-algorithm
8337 Use the specified algorithm for basic block reordering. The
8338 @var{algorithm} argument can be @samp{simple}, which does not increase
8339 code size (except sometimes due to secondary effects like alignment),
8340 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8341 put all often executed code together, minimizing the number of branches
8342 executed by making extra copies of code.
8344 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8345 @samp{stc} at levels @option{-O2}, @option{-O3}.
8347 @item -freorder-blocks-and-partition
8348 @opindex freorder-blocks-and-partition
8349 In addition to reordering basic blocks in the compiled function, in order
8350 to reduce number of taken branches, partitions hot and cold basic blocks
8351 into separate sections of the assembly and @file{.o} files, to improve
8352 paging and cache locality performance.
8354 This optimization is automatically turned off in the presence of
8355 exception handling, for linkonce sections, for functions with a user-defined
8356 section attribute and on any architecture that does not support named
8359 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8361 @item -freorder-functions
8362 @opindex freorder-functions
8363 Reorder functions in the object file in order to
8364 improve code locality. This is implemented by using special
8365 subsections @code{.text.hot} for most frequently executed functions and
8366 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8367 the linker so object file format must support named sections and linker must
8368 place them in a reasonable way.
8370 Also profile feedback must be available to make this option effective. See
8371 @option{-fprofile-arcs} for details.
8373 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8375 @item -fstrict-aliasing
8376 @opindex fstrict-aliasing
8377 Allow the compiler to assume the strictest aliasing rules applicable to
8378 the language being compiled. For C (and C++), this activates
8379 optimizations based on the type of expressions. In particular, an
8380 object of one type is assumed never to reside at the same address as an
8381 object of a different type, unless the types are almost the same. For
8382 example, an @code{unsigned int} can alias an @code{int}, but not a
8383 @code{void*} or a @code{double}. A character type may alias any other
8386 @anchor{Type-punning}Pay special attention to code like this:
8399 The practice of reading from a different union member than the one most
8400 recently written to (called ``type-punning'') is common. Even with
8401 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8402 is accessed through the union type. So, the code above works as
8403 expected. @xref{Structures unions enumerations and bit-fields
8404 implementation}. However, this code might not:
8415 Similarly, access by taking the address, casting the resulting pointer
8416 and dereferencing the result has undefined behavior, even if the cast
8417 uses a union type, e.g.:
8421 return ((union a_union *) &d)->i;
8425 The @option{-fstrict-aliasing} option is enabled at levels
8426 @option{-O2}, @option{-O3}, @option{-Os}.
8428 @item -fstrict-overflow
8429 @opindex fstrict-overflow
8430 Allow the compiler to assume strict signed overflow rules, depending
8431 on the language being compiled. For C (and C++) this means that
8432 overflow when doing arithmetic with signed numbers is undefined, which
8433 means that the compiler may assume that it does not happen. This
8434 permits various optimizations. For example, the compiler assumes
8435 that an expression like @code{i + 10 > i} is always true for
8436 signed @code{i}. This assumption is only valid if signed overflow is
8437 undefined, as the expression is false if @code{i + 10} overflows when
8438 using twos complement arithmetic. When this option is in effect any
8439 attempt to determine whether an operation on signed numbers
8440 overflows must be written carefully to not actually involve overflow.
8442 This option also allows the compiler to assume strict pointer
8443 semantics: given a pointer to an object, if adding an offset to that
8444 pointer does not produce a pointer to the same object, the addition is
8445 undefined. This permits the compiler to conclude that @code{p + u >
8446 p} is always true for a pointer @code{p} and unsigned integer
8447 @code{u}. This assumption is only valid because pointer wraparound is
8448 undefined, as the expression is false if @code{p + u} overflows using
8449 twos complement arithmetic.
8451 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8452 that integer signed overflow is fully defined: it wraps. When
8453 @option{-fwrapv} is used, there is no difference between
8454 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8455 integers. With @option{-fwrapv} certain types of overflow are
8456 permitted. For example, if the compiler gets an overflow when doing
8457 arithmetic on constants, the overflowed value can still be used with
8458 @option{-fwrapv}, but not otherwise.
8460 The @option{-fstrict-overflow} option is enabled at levels
8461 @option{-O2}, @option{-O3}, @option{-Os}.
8463 @item -falign-functions
8464 @itemx -falign-functions=@var{n}
8465 @opindex falign-functions
8466 Align the start of functions to the next power-of-two greater than
8467 @var{n}, skipping up to @var{n} bytes. For instance,
8468 @option{-falign-functions=32} aligns functions to the next 32-byte
8469 boundary, but @option{-falign-functions=24} aligns to the next
8470 32-byte boundary only if this can be done by skipping 23 bytes or less.
8472 @option{-fno-align-functions} and @option{-falign-functions=1} are
8473 equivalent and mean that functions are not aligned.
8475 Some assemblers only support this flag when @var{n} is a power of two;
8476 in that case, it is rounded up.
8478 If @var{n} is not specified or is zero, use a machine-dependent default.
8480 Enabled at levels @option{-O2}, @option{-O3}.
8482 @item -falign-labels
8483 @itemx -falign-labels=@var{n}
8484 @opindex falign-labels
8485 Align all branch targets to a power-of-two boundary, skipping up to
8486 @var{n} bytes like @option{-falign-functions}. This option can easily
8487 make code slower, because it must insert dummy operations for when the
8488 branch target is reached in the usual flow of the code.
8490 @option{-fno-align-labels} and @option{-falign-labels=1} are
8491 equivalent and mean that labels are not aligned.
8493 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8494 are greater than this value, then their values are used instead.
8496 If @var{n} is not specified or is zero, use a machine-dependent default
8497 which is very likely to be @samp{1}, meaning no alignment.
8499 Enabled at levels @option{-O2}, @option{-O3}.
8502 @itemx -falign-loops=@var{n}
8503 @opindex falign-loops
8504 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8505 like @option{-falign-functions}. If the loops are
8506 executed many times, this makes up for any execution of the dummy
8509 @option{-fno-align-loops} and @option{-falign-loops=1} are
8510 equivalent and mean that loops are not aligned.
8512 If @var{n} is not specified or is zero, use a machine-dependent default.
8514 Enabled at levels @option{-O2}, @option{-O3}.
8517 @itemx -falign-jumps=@var{n}
8518 @opindex falign-jumps
8519 Align branch targets to a power-of-two boundary, for branch targets
8520 where the targets can only be reached by jumping, skipping up to @var{n}
8521 bytes like @option{-falign-functions}. In this case, no dummy operations
8524 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8525 equivalent and mean that loops are not aligned.
8527 If @var{n} is not specified or is zero, use a machine-dependent default.
8529 Enabled at levels @option{-O2}, @option{-O3}.
8531 @item -funit-at-a-time
8532 @opindex funit-at-a-time
8533 This option is left for compatibility reasons. @option{-funit-at-a-time}
8534 has no effect, while @option{-fno-unit-at-a-time} implies
8535 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8539 @item -fno-toplevel-reorder
8540 @opindex fno-toplevel-reorder
8541 Do not reorder top-level functions, variables, and @code{asm}
8542 statements. Output them in the same order that they appear in the
8543 input file. When this option is used, unreferenced static variables
8544 are not removed. This option is intended to support existing code
8545 that relies on a particular ordering. For new code, it is better to
8546 use attributes when possible.
8548 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8549 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8554 Constructs webs as commonly used for register allocation purposes and assign
8555 each web individual pseudo register. This allows the register allocation pass
8556 to operate on pseudos directly, but also strengthens several other optimization
8557 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8558 however, make debugging impossible, since variables no longer stay in a
8561 Enabled by default with @option{-funroll-loops}.
8563 @item -fwhole-program
8564 @opindex fwhole-program
8565 Assume that the current compilation unit represents the whole program being
8566 compiled. All public functions and variables with the exception of @code{main}
8567 and those merged by attribute @code{externally_visible} become static functions
8568 and in effect are optimized more aggressively by interprocedural optimizers.
8570 This option should not be used in combination with @option{-flto}.
8571 Instead relying on a linker plugin should provide safer and more precise
8574 @item -flto[=@var{n}]
8576 This option runs the standard link-time optimizer. When invoked
8577 with source code, it generates GIMPLE (one of GCC's internal
8578 representations) and writes it to special ELF sections in the object
8579 file. When the object files are linked together, all the function
8580 bodies are read from these ELF sections and instantiated as if they
8581 had been part of the same translation unit.
8583 To use the link-time optimizer, @option{-flto} and optimization
8584 options should be specified at compile time and during the final link.
8585 It is recommended that you compile all the files participating in the
8586 same link with the same options and also specify those options at
8591 gcc -c -O2 -flto foo.c
8592 gcc -c -O2 -flto bar.c
8593 gcc -o myprog -flto -O2 foo.o bar.o
8596 The first two invocations to GCC save a bytecode representation
8597 of GIMPLE into special ELF sections inside @file{foo.o} and
8598 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8599 @file{foo.o} and @file{bar.o}, merges the two files into a single
8600 internal image, and compiles the result as usual. Since both
8601 @file{foo.o} and @file{bar.o} are merged into a single image, this
8602 causes all the interprocedural analyses and optimizations in GCC to
8603 work across the two files as if they were a single one. This means,
8604 for example, that the inliner is able to inline functions in
8605 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8607 Another (simpler) way to enable link-time optimization is:
8610 gcc -o myprog -flto -O2 foo.c bar.c
8613 The above generates bytecode for @file{foo.c} and @file{bar.c},
8614 merges them together into a single GIMPLE representation and optimizes
8615 them as usual to produce @file{myprog}.
8617 The only important thing to keep in mind is that to enable link-time
8618 optimizations you need to use the GCC driver to perform the link step.
8619 GCC then automatically performs link-time optimization if any of the
8620 objects involved were compiled with the @option{-flto} command-line option.
8622 should specify the optimization options to be used for link-time
8623 optimization though GCC tries to be clever at guessing an
8624 optimization level to use from the options used at compile time
8625 if you fail to specify one at link time. You can always override
8626 the automatic decision to do link-time optimization at link time
8627 by passing @option{-fno-lto} to the link command.
8629 To make whole program optimization effective, it is necessary to make
8630 certain whole program assumptions. The compiler needs to know
8631 what functions and variables can be accessed by libraries and runtime
8632 outside of the link-time optimized unit. When supported by the linker,
8633 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8634 to the compiler about used and externally visible symbols. When
8635 the linker plugin is not available, @option{-fwhole-program} should be
8636 used to allow the compiler to make these assumptions, which leads
8637 to more aggressive optimization decisions.
8639 When @option{-fuse-linker-plugin} is not enabled, when a file is
8640 compiled with @option{-flto}, the generated object file is larger than
8641 a regular object file because it contains GIMPLE bytecodes and the usual
8642 final code (see @option{-ffat-lto-objects}. This means that
8643 object files with LTO information can be linked as normal object
8644 files; if @option{-fno-lto} is passed to the linker, no
8645 interprocedural optimizations are applied. Note that when
8646 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8647 but you cannot perform a regular, non-LTO link on them.
8649 Additionally, the optimization flags used to compile individual files
8650 are not necessarily related to those used at link time. For instance,
8653 gcc -c -O0 -ffat-lto-objects -flto foo.c
8654 gcc -c -O0 -ffat-lto-objects -flto bar.c
8655 gcc -o myprog -O3 foo.o bar.o
8658 This produces individual object files with unoptimized assembler
8659 code, but the resulting binary @file{myprog} is optimized at
8660 @option{-O3}. If, instead, the final binary is generated with
8661 @option{-fno-lto}, then @file{myprog} is not optimized.
8663 When producing the final binary, GCC only
8664 applies link-time optimizations to those files that contain bytecode.
8665 Therefore, you can mix and match object files and libraries with
8666 GIMPLE bytecodes and final object code. GCC automatically selects
8667 which files to optimize in LTO mode and which files to link without
8670 There are some code generation flags preserved by GCC when
8671 generating bytecodes, as they need to be used during the final link
8672 stage. Generally options specified at link time override those
8673 specified at compile time.
8675 If you do not specify an optimization level option @option{-O} at
8676 link time, then GCC uses the highest optimization level
8677 used when compiling the object files.
8679 Currently, the following options and their settings are taken from
8680 the first object file that explicitly specifies them:
8681 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8682 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8683 and all the @option{-m} target flags.
8685 Certain ABI-changing flags are required to match in all compilation units,
8686 and trying to override this at link time with a conflicting value
8687 is ignored. This includes options such as @option{-freg-struct-return}
8688 and @option{-fpcc-struct-return}.
8690 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8691 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8692 are passed through to the link stage and merged conservatively for
8693 conflicting translation units. Specifically
8694 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8695 precedence; and for example @option{-ffp-contract=off} takes precedence
8696 over @option{-ffp-contract=fast}. You can override them at link time.
8698 If LTO encounters objects with C linkage declared with incompatible
8699 types in separate translation units to be linked together (undefined
8700 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8701 issued. The behavior is still undefined at run time. Similar
8702 diagnostics may be raised for other languages.
8704 Another feature of LTO is that it is possible to apply interprocedural
8705 optimizations on files written in different languages:
8710 gfortran -c -flto baz.f90
8711 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8714 Notice that the final link is done with @command{g++} to get the C++
8715 runtime libraries and @option{-lgfortran} is added to get the Fortran
8716 runtime libraries. In general, when mixing languages in LTO mode, you
8717 should use the same link command options as when mixing languages in a
8718 regular (non-LTO) compilation.
8720 If object files containing GIMPLE bytecode are stored in a library archive, say
8721 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8722 are using a linker with plugin support. To create static libraries suitable
8723 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8724 and @command{ranlib};
8725 to show the symbols of object files with GIMPLE bytecode, use
8726 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8727 and @command{nm} have been compiled with plugin support. At link time, use the the
8728 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8729 the LTO optimization process:
8732 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8735 With the linker plugin enabled, the linker extracts the needed
8736 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8737 to make them part of the aggregated GIMPLE image to be optimized.
8739 If you are not using a linker with plugin support and/or do not
8740 enable the linker plugin, then the objects inside @file{libfoo.a}
8741 are extracted and linked as usual, but they do not participate
8742 in the LTO optimization process. In order to make a static library suitable
8743 for both LTO optimization and usual linkage, compile its object files with
8744 @option{-flto} @option{-ffat-lto-objects}.
8746 Link-time optimizations do not require the presence of the whole program to
8747 operate. If the program does not require any symbols to be exported, it is
8748 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8749 the interprocedural optimizers to use more aggressive assumptions which may
8750 lead to improved optimization opportunities.
8751 Use of @option{-fwhole-program} is not needed when linker plugin is
8752 active (see @option{-fuse-linker-plugin}).
8754 The current implementation of LTO makes no
8755 attempt to generate bytecode that is portable between different
8756 types of hosts. The bytecode files are versioned and there is a
8757 strict version check, so bytecode files generated in one version of
8758 GCC do not work with an older or newer version of GCC.
8760 Link-time optimization does not work well with generation of debugging
8761 information. Combining @option{-flto} with
8762 @option{-g} is currently experimental and expected to produce unexpected
8765 If you specify the optional @var{n}, the optimization and code
8766 generation done at link time is executed in parallel using @var{n}
8767 parallel jobs by utilizing an installed @command{make} program. The
8768 environment variable @env{MAKE} may be used to override the program
8769 used. The default value for @var{n} is 1.
8771 You can also specify @option{-flto=jobserver} to use GNU make's
8772 job server mode to determine the number of parallel jobs. This
8773 is useful when the Makefile calling GCC is already executing in parallel.
8774 You must prepend a @samp{+} to the command recipe in the parent Makefile
8775 for this to work. This option likely only works if @env{MAKE} is
8778 @item -flto-partition=@var{alg}
8779 @opindex flto-partition
8780 Specify the partitioning algorithm used by the link-time optimizer.
8781 The value is either @samp{1to1} to specify a partitioning mirroring
8782 the original source files or @samp{balanced} to specify partitioning
8783 into equally sized chunks (whenever possible) or @samp{max} to create
8784 new partition for every symbol where possible. Specifying @samp{none}
8785 as an algorithm disables partitioning and streaming completely.
8786 The default value is @samp{balanced}. While @samp{1to1} can be used
8787 as an workaround for various code ordering issues, the @samp{max}
8788 partitioning is intended for internal testing only.
8789 The value @samp{one} specifies that exactly one partition should be
8790 used while the value @samp{none} bypasses partitioning and executes
8791 the link-time optimization step directly from the WPA phase.
8793 @item -flto-odr-type-merging
8794 @opindex flto-odr-type-merging
8795 Enable streaming of mangled types names of C++ types and their unification
8796 at link time. This increases size of LTO object files, but enables
8797 diagnostics about One Definition Rule violations.
8799 @item -flto-compression-level=@var{n}
8800 @opindex flto-compression-level
8801 This option specifies the level of compression used for intermediate
8802 language written to LTO object files, and is only meaningful in
8803 conjunction with LTO mode (@option{-flto}). Valid
8804 values are 0 (no compression) to 9 (maximum compression). Values
8805 outside this range are clamped to either 0 or 9. If the option is not
8806 given, a default balanced compression setting is used.
8808 @item -fuse-linker-plugin
8809 @opindex fuse-linker-plugin
8810 Enables the use of a linker plugin during link-time optimization. This
8811 option relies on plugin support in the linker, which is available in gold
8812 or in GNU ld 2.21 or newer.
8814 This option enables the extraction of object files with GIMPLE bytecode out
8815 of library archives. This improves the quality of optimization by exposing
8816 more code to the link-time optimizer. This information specifies what
8817 symbols can be accessed externally (by non-LTO object or during dynamic
8818 linking). Resulting code quality improvements on binaries (and shared
8819 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8820 See @option{-flto} for a description of the effect of this flag and how to
8823 This option is enabled by default when LTO support in GCC is enabled
8824 and GCC was configured for use with
8825 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8827 @item -ffat-lto-objects
8828 @opindex ffat-lto-objects
8829 Fat LTO objects are object files that contain both the intermediate language
8830 and the object code. This makes them usable for both LTO linking and normal
8831 linking. This option is effective only when compiling with @option{-flto}
8832 and is ignored at link time.
8834 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8835 requires the complete toolchain to be aware of LTO. It requires a linker with
8836 linker plugin support for basic functionality. Additionally,
8837 @command{nm}, @command{ar} and @command{ranlib}
8838 need to support linker plugins to allow a full-featured build environment
8839 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8840 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8841 to these tools. With non fat LTO makefiles need to be modified to use them.
8843 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8846 @item -fcompare-elim
8847 @opindex fcompare-elim
8848 After register allocation and post-register allocation instruction splitting,
8849 identify arithmetic instructions that compute processor flags similar to a
8850 comparison operation based on that arithmetic. If possible, eliminate the
8851 explicit comparison operation.
8853 This pass only applies to certain targets that cannot explicitly represent
8854 the comparison operation before register allocation is complete.
8856 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8858 @item -fcprop-registers
8859 @opindex fcprop-registers
8860 After register allocation and post-register allocation instruction splitting,
8861 perform a copy-propagation pass to try to reduce scheduling dependencies
8862 and occasionally eliminate the copy.
8864 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8866 @item -fprofile-correction
8867 @opindex fprofile-correction
8868 Profiles collected using an instrumented binary for multi-threaded programs may
8869 be inconsistent due to missed counter updates. When this option is specified,
8870 GCC uses heuristics to correct or smooth out such inconsistencies. By
8871 default, GCC emits an error message when an inconsistent profile is detected.
8874 @itemx -fprofile-use=@var{path}
8875 @opindex fprofile-use
8876 Enable profile feedback-directed optimizations,
8877 and the following optimizations
8878 which are generally profitable only with profile feedback available:
8879 @option{-fbranch-probabilities}, @option{-fvpt},
8880 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8881 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
8883 Before you can use this option, you must first generate profiling information.
8884 @xref{Optimize Options}, for information about the @option{-fprofile-generate}
8887 By default, GCC emits an error message if the feedback profiles do not
8888 match the source code. This error can be turned into a warning by using
8889 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8892 If @var{path} is specified, GCC looks at the @var{path} to find
8893 the profile feedback data files. See @option{-fprofile-dir}.
8895 @item -fauto-profile
8896 @itemx -fauto-profile=@var{path}
8897 @opindex fauto-profile
8898 Enable sampling-based feedback-directed optimizations,
8899 and the following optimizations
8900 which are generally profitable only with profile feedback available:
8901 @option{-fbranch-probabilities}, @option{-fvpt},
8902 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8903 @option{-ftree-vectorize},
8904 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
8905 @option{-fpredictive-commoning}, @option{-funswitch-loops},
8906 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
8908 @var{path} is the name of a file containing AutoFDO profile information.
8909 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
8911 Producing an AutoFDO profile data file requires running your program
8912 with the @command{perf} utility on a supported GNU/Linux target system.
8913 For more information, see @uref{https://perf.wiki.kernel.org/}.
8917 perf record -e br_inst_retired:near_taken -b -o perf.data \
8921 Then use the @command{create_gcov} tool to convert the raw profile data
8922 to a format that can be used by GCC.@ You must also supply the
8923 unstripped binary for your program to this tool.
8924 See @uref{https://github.com/google/autofdo}.
8928 create_gcov --binary=your_program.unstripped --profile=perf.data \
8933 The following options control compiler behavior regarding floating-point
8934 arithmetic. These options trade off between speed and
8935 correctness. All must be specifically enabled.
8939 @opindex ffloat-store
8940 Do not store floating-point variables in registers, and inhibit other
8941 options that might change whether a floating-point value is taken from a
8944 @cindex floating-point precision
8945 This option prevents undesirable excess precision on machines such as
8946 the 68000 where the floating registers (of the 68881) keep more
8947 precision than a @code{double} is supposed to have. Similarly for the
8948 x86 architecture. For most programs, the excess precision does only
8949 good, but a few programs rely on the precise definition of IEEE floating
8950 point. Use @option{-ffloat-store} for such programs, after modifying
8951 them to store all pertinent intermediate computations into variables.
8953 @item -fexcess-precision=@var{style}
8954 @opindex fexcess-precision
8955 This option allows further control over excess precision on machines
8956 where floating-point registers have more precision than the IEEE
8957 @code{float} and @code{double} types and the processor does not
8958 support operations rounding to those types. By default,
8959 @option{-fexcess-precision=fast} is in effect; this means that
8960 operations are carried out in the precision of the registers and that
8961 it is unpredictable when rounding to the types specified in the source
8962 code takes place. When compiling C, if
8963 @option{-fexcess-precision=standard} is specified then excess
8964 precision follows the rules specified in ISO C99; in particular,
8965 both casts and assignments cause values to be rounded to their
8966 semantic types (whereas @option{-ffloat-store} only affects
8967 assignments). This option is enabled by default for C if a strict
8968 conformance option such as @option{-std=c99} is used.
8971 @option{-fexcess-precision=standard} is not implemented for languages
8972 other than C, and has no effect if
8973 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8974 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8975 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8976 semantics apply without excess precision, and in the latter, rounding
8981 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8982 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8983 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8985 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8987 This option is not turned on by any @option{-O} option besides
8988 @option{-Ofast} since it can result in incorrect output for programs
8989 that depend on an exact implementation of IEEE or ISO rules/specifications
8990 for math functions. It may, however, yield faster code for programs
8991 that do not require the guarantees of these specifications.
8993 @item -fno-math-errno
8994 @opindex fno-math-errno
8995 Do not set @code{errno} after calling math functions that are executed
8996 with a single instruction, e.g., @code{sqrt}. A program that relies on
8997 IEEE exceptions for math error handling may want to use this flag
8998 for speed while maintaining IEEE arithmetic compatibility.
9000 This option is not turned on by any @option{-O} option since
9001 it can result in incorrect output for programs that depend on
9002 an exact implementation of IEEE or ISO rules/specifications for
9003 math functions. It may, however, yield faster code for programs
9004 that do not require the guarantees of these specifications.
9006 The default is @option{-fmath-errno}.
9008 On Darwin systems, the math library never sets @code{errno}. There is
9009 therefore no reason for the compiler to consider the possibility that
9010 it might, and @option{-fno-math-errno} is the default.
9012 @item -funsafe-math-optimizations
9013 @opindex funsafe-math-optimizations
9015 Allow optimizations for floating-point arithmetic that (a) assume
9016 that arguments and results are valid and (b) may violate IEEE or
9017 ANSI standards. When used at link time, it may include libraries
9018 or startup files that change the default FPU control word or other
9019 similar optimizations.
9021 This option is not turned on by any @option{-O} option since
9022 it can result in incorrect output for programs that depend on
9023 an exact implementation of IEEE or ISO rules/specifications for
9024 math functions. It may, however, yield faster code for programs
9025 that do not require the guarantees of these specifications.
9026 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9027 @option{-fassociative-math} and @option{-freciprocal-math}.
9029 The default is @option{-fno-unsafe-math-optimizations}.
9031 @item -fassociative-math
9032 @opindex fassociative-math
9034 Allow re-association of operands in series of floating-point operations.
9035 This violates the ISO C and C++ language standard by possibly changing
9036 computation result. NOTE: re-ordering may change the sign of zero as
9037 well as ignore NaNs and inhibit or create underflow or overflow (and
9038 thus cannot be used on code that relies on rounding behavior like
9039 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9040 and thus may not be used when ordered comparisons are required.
9041 This option requires that both @option{-fno-signed-zeros} and
9042 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9043 much sense with @option{-frounding-math}. For Fortran the option
9044 is automatically enabled when both @option{-fno-signed-zeros} and
9045 @option{-fno-trapping-math} are in effect.
9047 The default is @option{-fno-associative-math}.
9049 @item -freciprocal-math
9050 @opindex freciprocal-math
9052 Allow the reciprocal of a value to be used instead of dividing by
9053 the value if this enables optimizations. For example @code{x / y}
9054 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9055 is subject to common subexpression elimination. Note that this loses
9056 precision and increases the number of flops operating on the value.
9058 The default is @option{-fno-reciprocal-math}.
9060 @item -ffinite-math-only
9061 @opindex ffinite-math-only
9062 Allow optimizations for floating-point arithmetic that assume
9063 that arguments and results are not NaNs or +-Infs.
9065 This option is not turned on by any @option{-O} option since
9066 it can result in incorrect output for programs that depend on
9067 an exact implementation of IEEE or ISO rules/specifications for
9068 math functions. It may, however, yield faster code for programs
9069 that do not require the guarantees of these specifications.
9071 The default is @option{-fno-finite-math-only}.
9073 @item -fno-signed-zeros
9074 @opindex fno-signed-zeros
9075 Allow optimizations for floating-point arithmetic that ignore the
9076 signedness of zero. IEEE arithmetic specifies the behavior of
9077 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9078 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9079 This option implies that the sign of a zero result isn't significant.
9081 The default is @option{-fsigned-zeros}.
9083 @item -fno-trapping-math
9084 @opindex fno-trapping-math
9085 Compile code assuming that floating-point operations cannot generate
9086 user-visible traps. These traps include division by zero, overflow,
9087 underflow, inexact result and invalid operation. This option requires
9088 that @option{-fno-signaling-nans} be in effect. Setting this option may
9089 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9091 This option should never be turned on by any @option{-O} option since
9092 it can result in incorrect output for programs that depend on
9093 an exact implementation of IEEE or ISO rules/specifications for
9096 The default is @option{-ftrapping-math}.
9098 @item -frounding-math
9099 @opindex frounding-math
9100 Disable transformations and optimizations that assume default floating-point
9101 rounding behavior. This is round-to-zero for all floating point
9102 to integer conversions, and round-to-nearest for all other arithmetic
9103 truncations. This option should be specified for programs that change
9104 the FP rounding mode dynamically, or that may be executed with a
9105 non-default rounding mode. This option disables constant folding of
9106 floating-point expressions at compile time (which may be affected by
9107 rounding mode) and arithmetic transformations that are unsafe in the
9108 presence of sign-dependent rounding modes.
9110 The default is @option{-fno-rounding-math}.
9112 This option is experimental and does not currently guarantee to
9113 disable all GCC optimizations that are affected by rounding mode.
9114 Future versions of GCC may provide finer control of this setting
9115 using C99's @code{FENV_ACCESS} pragma. This command-line option
9116 will be used to specify the default state for @code{FENV_ACCESS}.
9118 @item -fsignaling-nans
9119 @opindex fsignaling-nans
9120 Compile code assuming that IEEE signaling NaNs may generate user-visible
9121 traps during floating-point operations. Setting this option disables
9122 optimizations that may change the number of exceptions visible with
9123 signaling NaNs. This option implies @option{-ftrapping-math}.
9125 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9128 The default is @option{-fno-signaling-nans}.
9130 This option is experimental and does not currently guarantee to
9131 disable all GCC optimizations that affect signaling NaN behavior.
9133 @item -fno-fp-int-builtin-inexact
9134 @opindex fno-fp-int-builtin-inexact
9135 Do not allow the built-in functions @code{ceil}, @code{floor},
9136 @code{round} and @code{trunc}, and their @code{float} and @code{long
9137 double} variants, to generate code that raises the ``inexact''
9138 floating-point exception for noninteger arguments. ISO C99 and C11
9139 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9140 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9143 The default is @option{-ffp-int-builtin-inexact}, allowing the
9144 exception to be raised. This option does nothing unless
9145 @option{-ftrapping-math} is in effect.
9147 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9148 generate a call to a library function then the ``inexact'' exception
9149 may be raised if the library implementation does not follow TS 18661.
9151 @item -fsingle-precision-constant
9152 @opindex fsingle-precision-constant
9153 Treat floating-point constants as single precision instead of
9154 implicitly converting them to double-precision constants.
9156 @item -fcx-limited-range
9157 @opindex fcx-limited-range
9158 When enabled, this option states that a range reduction step is not
9159 needed when performing complex division. Also, there is no checking
9160 whether the result of a complex multiplication or division is @code{NaN
9161 + I*NaN}, with an attempt to rescue the situation in that case. The
9162 default is @option{-fno-cx-limited-range}, but is enabled by
9163 @option{-ffast-math}.
9165 This option controls the default setting of the ISO C99
9166 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9169 @item -fcx-fortran-rules
9170 @opindex fcx-fortran-rules
9171 Complex multiplication and division follow Fortran rules. Range
9172 reduction is done as part of complex division, but there is no checking
9173 whether the result of a complex multiplication or division is @code{NaN
9174 + I*NaN}, with an attempt to rescue the situation in that case.
9176 The default is @option{-fno-cx-fortran-rules}.
9180 The following options control optimizations that may improve
9181 performance, but are not enabled by any @option{-O} options. This
9182 section includes experimental options that may produce broken code.
9185 @item -fbranch-probabilities
9186 @opindex fbranch-probabilities
9187 After running a program compiled with @option{-fprofile-arcs}
9188 (@pxref{Instrumentation Options}),
9189 you can compile it a second time using
9190 @option{-fbranch-probabilities}, to improve optimizations based on
9191 the number of times each branch was taken. When a program
9192 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9193 counts to a file called @file{@var{sourcename}.gcda} for each source
9194 file. The information in this data file is very dependent on the
9195 structure of the generated code, so you must use the same source code
9196 and the same optimization options for both compilations.
9198 With @option{-fbranch-probabilities}, GCC puts a
9199 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9200 These can be used to improve optimization. Currently, they are only
9201 used in one place: in @file{reorg.c}, instead of guessing which path a
9202 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9203 exactly determine which path is taken more often.
9205 @item -fprofile-values
9206 @opindex fprofile-values
9207 If combined with @option{-fprofile-arcs}, it adds code so that some
9208 data about values of expressions in the program is gathered.
9210 With @option{-fbranch-probabilities}, it reads back the data gathered
9211 from profiling values of expressions for usage in optimizations.
9213 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9215 @item -fprofile-reorder-functions
9216 @opindex fprofile-reorder-functions
9217 Function reordering based on profile instrumentation collects
9218 first time of execution of a function and orders these functions
9221 Enabled with @option{-fprofile-use}.
9225 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9226 to add code to gather information about values of expressions.
9228 With @option{-fbranch-probabilities}, it reads back the data gathered
9229 and actually performs the optimizations based on them.
9230 Currently the optimizations include specialization of division operations
9231 using the knowledge about the value of the denominator.
9233 @item -frename-registers
9234 @opindex frename-registers
9235 Attempt to avoid false dependencies in scheduled code by making use
9236 of registers left over after register allocation. This optimization
9237 most benefits processors with lots of registers. Depending on the
9238 debug information format adopted by the target, however, it can
9239 make debugging impossible, since variables no longer stay in
9240 a ``home register''.
9242 Enabled by default with @option{-funroll-loops}.
9244 @item -fschedule-fusion
9245 @opindex fschedule-fusion
9246 Performs a target dependent pass over the instruction stream to schedule
9247 instructions of same type together because target machine can execute them
9248 more efficiently if they are adjacent to each other in the instruction flow.
9250 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9254 Perform tail duplication to enlarge superblock size. This transformation
9255 simplifies the control flow of the function allowing other optimizations to do
9258 Enabled with @option{-fprofile-use}.
9260 @item -funroll-loops
9261 @opindex funroll-loops
9262 Unroll loops whose number of iterations can be determined at compile time or
9263 upon entry to the loop. @option{-funroll-loops} implies
9264 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9265 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9266 a small constant number of iterations). This option makes code larger, and may
9267 or may not make it run faster.
9269 Enabled with @option{-fprofile-use}.
9271 @item -funroll-all-loops
9272 @opindex funroll-all-loops
9273 Unroll all loops, even if their number of iterations is uncertain when
9274 the loop is entered. This usually makes programs run more slowly.
9275 @option{-funroll-all-loops} implies the same options as
9276 @option{-funroll-loops}.
9279 @opindex fpeel-loops
9280 Peels loops for which there is enough information that they do not
9281 roll much (from profile feedback or static analysis). It also turns on
9282 complete loop peeling (i.e.@: complete removal of loops with small constant
9283 number of iterations).
9285 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9287 @item -fmove-loop-invariants
9288 @opindex fmove-loop-invariants
9289 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9290 at level @option{-O1}
9293 @opindex fsplit-loops
9294 Split a loop into two if it contains a condition that's always true
9295 for one side of the iteration space and false for the other.
9297 @item -funswitch-loops
9298 @opindex funswitch-loops
9299 Move branches with loop invariant conditions out of the loop, with duplicates
9300 of the loop on both branches (modified according to result of the condition).
9302 @item -ffunction-sections
9303 @itemx -fdata-sections
9304 @opindex ffunction-sections
9305 @opindex fdata-sections
9306 Place each function or data item into its own section in the output
9307 file if the target supports arbitrary sections. The name of the
9308 function or the name of the data item determines the section's name
9311 Use these options on systems where the linker can perform optimizations
9312 to improve locality of reference in the instruction space. Most systems
9313 using the ELF object format and SPARC processors running Solaris 2 have
9314 linkers with such optimizations. AIX may have these optimizations in
9317 Only use these options when there are significant benefits from doing
9318 so. When you specify these options, the assembler and linker
9319 create larger object and executable files and are also slower.
9320 You cannot use @command{gprof} on all systems if you
9321 specify this option, and you may have problems with debugging if
9322 you specify both this option and @option{-g}.
9324 @item -fbranch-target-load-optimize
9325 @opindex fbranch-target-load-optimize
9326 Perform branch target register load optimization before prologue / epilogue
9328 The use of target registers can typically be exposed only during reload,
9329 thus hoisting loads out of loops and doing inter-block scheduling needs
9330 a separate optimization pass.
9332 @item -fbranch-target-load-optimize2
9333 @opindex fbranch-target-load-optimize2
9334 Perform branch target register load optimization after prologue / epilogue
9337 @item -fbtr-bb-exclusive
9338 @opindex fbtr-bb-exclusive
9339 When performing branch target register load optimization, don't reuse
9340 branch target registers within any basic block.
9343 @opindex fstdarg-opt
9344 Optimize the prologue of variadic argument functions with respect to usage of
9347 @item -fsection-anchors
9348 @opindex fsection-anchors
9349 Try to reduce the number of symbolic address calculations by using
9350 shared ``anchor'' symbols to address nearby objects. This transformation
9351 can help to reduce the number of GOT entries and GOT accesses on some
9354 For example, the implementation of the following function @code{foo}:
9358 int foo (void) @{ return a + b + c; @}
9362 usually calculates the addresses of all three variables, but if you
9363 compile it with @option{-fsection-anchors}, it accesses the variables
9364 from a common anchor point instead. The effect is similar to the
9365 following pseudocode (which isn't valid C):
9370 register int *xr = &x;
9371 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9375 Not all targets support this option.
9377 @item --param @var{name}=@var{value}
9379 In some places, GCC uses various constants to control the amount of
9380 optimization that is done. For example, GCC does not inline functions
9381 that contain more than a certain number of instructions. You can
9382 control some of these constants on the command line using the
9383 @option{--param} option.
9385 The names of specific parameters, and the meaning of the values, are
9386 tied to the internals of the compiler, and are subject to change
9387 without notice in future releases.
9389 In each case, the @var{value} is an integer. The allowable choices for
9393 @item predictable-branch-outcome
9394 When branch is predicted to be taken with probability lower than this threshold
9395 (in percent), then it is considered well predictable. The default is 10.
9397 @item max-rtl-if-conversion-insns
9398 RTL if-conversion tries to remove conditional branches around a block and
9399 replace them with conditionally executed instructions. This parameter
9400 gives the maximum number of instructions in a block which should be
9401 considered for if-conversion. The default is 10, though the compiler will
9402 also use other heuristics to decide whether if-conversion is likely to be
9405 @item max-rtl-if-conversion-predictable-cost
9406 @item max-rtl-if-conversion-unpredictable-cost
9407 RTL if-conversion will try to remove conditional branches around a block
9408 and replace them with conditionally executed instructions. These parameters
9409 give the maximum permissible cost for the sequence that would be generated
9410 by if-conversion depending on whether the branch is statically determined
9411 to be predictable or not. The units for this parameter are the same as
9412 those for the GCC internal seq_cost metric. The compiler will try to
9413 provide a reasonable default for this parameter using the BRANCH_COST
9416 @item max-crossjump-edges
9417 The maximum number of incoming edges to consider for cross-jumping.
9418 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9419 the number of edges incoming to each block. Increasing values mean
9420 more aggressive optimization, making the compilation time increase with
9421 probably small improvement in executable size.
9423 @item min-crossjump-insns
9424 The minimum number of instructions that must be matched at the end
9425 of two blocks before cross-jumping is performed on them. This
9426 value is ignored in the case where all instructions in the block being
9427 cross-jumped from are matched. The default value is 5.
9429 @item max-grow-copy-bb-insns
9430 The maximum code size expansion factor when copying basic blocks
9431 instead of jumping. The expansion is relative to a jump instruction.
9432 The default value is 8.
9434 @item max-goto-duplication-insns
9435 The maximum number of instructions to duplicate to a block that jumps
9436 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9437 passes, GCC factors computed gotos early in the compilation process,
9438 and unfactors them as late as possible. Only computed jumps at the
9439 end of a basic blocks with no more than max-goto-duplication-insns are
9440 unfactored. The default value is 8.
9442 @item max-delay-slot-insn-search
9443 The maximum number of instructions to consider when looking for an
9444 instruction to fill a delay slot. If more than this arbitrary number of
9445 instructions are searched, the time savings from filling the delay slot
9446 are minimal, so stop searching. Increasing values mean more
9447 aggressive optimization, making the compilation time increase with probably
9448 small improvement in execution time.
9450 @item max-delay-slot-live-search
9451 When trying to fill delay slots, the maximum number of instructions to
9452 consider when searching for a block with valid live register
9453 information. Increasing this arbitrarily chosen value means more
9454 aggressive optimization, increasing the compilation time. This parameter
9455 should be removed when the delay slot code is rewritten to maintain the
9458 @item max-gcse-memory
9459 The approximate maximum amount of memory that can be allocated in
9460 order to perform the global common subexpression elimination
9461 optimization. If more memory than specified is required, the
9462 optimization is not done.
9464 @item max-gcse-insertion-ratio
9465 If the ratio of expression insertions to deletions is larger than this value
9466 for any expression, then RTL PRE inserts or removes the expression and thus
9467 leaves partially redundant computations in the instruction stream. The default value is 20.
9469 @item max-pending-list-length
9470 The maximum number of pending dependencies scheduling allows
9471 before flushing the current state and starting over. Large functions
9472 with few branches or calls can create excessively large lists which
9473 needlessly consume memory and resources.
9475 @item max-modulo-backtrack-attempts
9476 The maximum number of backtrack attempts the scheduler should make
9477 when modulo scheduling a loop. Larger values can exponentially increase
9480 @item max-inline-insns-single
9481 Several parameters control the tree inliner used in GCC@.
9482 This number sets the maximum number of instructions (counted in GCC's
9483 internal representation) in a single function that the tree inliner
9484 considers for inlining. This only affects functions declared
9485 inline and methods implemented in a class declaration (C++).
9486 The default value is 400.
9488 @item max-inline-insns-auto
9489 When you use @option{-finline-functions} (included in @option{-O3}),
9490 a lot of functions that would otherwise not be considered for inlining
9491 by the compiler are investigated. To those functions, a different
9492 (more restrictive) limit compared to functions declared inline can
9494 The default value is 40.
9496 @item inline-min-speedup
9497 When estimated performance improvement of caller + callee runtime exceeds this
9498 threshold (in precent), the function can be inlined regardless the limit on
9499 @option{--param max-inline-insns-single} and @option{--param
9500 max-inline-insns-auto}.
9502 @item large-function-insns
9503 The limit specifying really large functions. For functions larger than this
9504 limit after inlining, inlining is constrained by
9505 @option{--param large-function-growth}. This parameter is useful primarily
9506 to avoid extreme compilation time caused by non-linear algorithms used by the
9508 The default value is 2700.
9510 @item large-function-growth
9511 Specifies maximal growth of large function caused by inlining in percents.
9512 The default value is 100 which limits large function growth to 2.0 times
9515 @item large-unit-insns
9516 The limit specifying large translation unit. Growth caused by inlining of
9517 units larger than this limit is limited by @option{--param inline-unit-growth}.
9518 For small units this might be too tight.
9519 For example, consider a unit consisting of function A
9520 that is inline and B that just calls A three times. If B is small relative to
9521 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9522 large units consisting of small inlineable functions, however, the overall unit
9523 growth limit is needed to avoid exponential explosion of code size. Thus for
9524 smaller units, the size is increased to @option{--param large-unit-insns}
9525 before applying @option{--param inline-unit-growth}. The default is 10000.
9527 @item inline-unit-growth
9528 Specifies maximal overall growth of the compilation unit caused by inlining.
9529 The default value is 20 which limits unit growth to 1.2 times the original
9530 size. Cold functions (either marked cold via an attribute or by profile
9531 feedback) are not accounted into the unit size.
9533 @item ipcp-unit-growth
9534 Specifies maximal overall growth of the compilation unit caused by
9535 interprocedural constant propagation. The default value is 10 which limits
9536 unit growth to 1.1 times the original size.
9538 @item large-stack-frame
9539 The limit specifying large stack frames. While inlining the algorithm is trying
9540 to not grow past this limit too much. The default value is 256 bytes.
9542 @item large-stack-frame-growth
9543 Specifies maximal growth of large stack frames caused by inlining in percents.
9544 The default value is 1000 which limits large stack frame growth to 11 times
9547 @item max-inline-insns-recursive
9548 @itemx max-inline-insns-recursive-auto
9549 Specifies the maximum number of instructions an out-of-line copy of a
9550 self-recursive inline
9551 function can grow into by performing recursive inlining.
9553 @option{--param max-inline-insns-recursive} applies to functions
9555 For functions not declared inline, recursive inlining
9556 happens only when @option{-finline-functions} (included in @option{-O3}) is
9557 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9558 default value is 450.
9560 @item max-inline-recursive-depth
9561 @itemx max-inline-recursive-depth-auto
9562 Specifies the maximum recursion depth used for recursive inlining.
9564 @option{--param max-inline-recursive-depth} applies to functions
9565 declared inline. For functions not declared inline, recursive inlining
9566 happens only when @option{-finline-functions} (included in @option{-O3}) is
9567 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9570 @item min-inline-recursive-probability
9571 Recursive inlining is profitable only for function having deep recursion
9572 in average and can hurt for function having little recursion depth by
9573 increasing the prologue size or complexity of function body to other
9576 When profile feedback is available (see @option{-fprofile-generate}) the actual
9577 recursion depth can be guessed from the probability that function recurses
9578 via a given call expression. This parameter limits inlining only to call
9579 expressions whose probability exceeds the given threshold (in percents).
9580 The default value is 10.
9582 @item early-inlining-insns
9583 Specify growth that the early inliner can make. In effect it increases
9584 the amount of inlining for code having a large abstraction penalty.
9585 The default value is 14.
9587 @item max-early-inliner-iterations
9588 Limit of iterations of the early inliner. This basically bounds
9589 the number of nested indirect calls the early inliner can resolve.
9590 Deeper chains are still handled by late inlining.
9592 @item comdat-sharing-probability
9593 Probability (in percent) that C++ inline function with comdat visibility
9594 are shared across multiple compilation units. The default value is 20.
9596 @item profile-func-internal-id
9597 A parameter to control whether to use function internal id in profile
9598 database lookup. If the value is 0, the compiler uses an id that
9599 is based on function assembler name and filename, which makes old profile
9600 data more tolerant to source changes such as function reordering etc.
9601 The default value is 0.
9603 @item min-vect-loop-bound
9604 The minimum number of iterations under which loops are not vectorized
9605 when @option{-ftree-vectorize} is used. The number of iterations after
9606 vectorization needs to be greater than the value specified by this option
9607 to allow vectorization. The default value is 0.
9609 @item gcse-cost-distance-ratio
9610 Scaling factor in calculation of maximum distance an expression
9611 can be moved by GCSE optimizations. This is currently supported only in the
9612 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9613 is with simple expressions, i.e., the expressions that have cost
9614 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9615 hoisting of simple expressions. The default value is 10.
9617 @item gcse-unrestricted-cost
9618 Cost, roughly measured as the cost of a single typical machine
9619 instruction, at which GCSE optimizations do not constrain
9620 the distance an expression can travel. This is currently
9621 supported only in the code hoisting pass. The lesser the cost,
9622 the more aggressive code hoisting is. Specifying 0
9623 allows all expressions to travel unrestricted distances.
9624 The default value is 3.
9626 @item max-hoist-depth
9627 The depth of search in the dominator tree for expressions to hoist.
9628 This is used to avoid quadratic behavior in hoisting algorithm.
9629 The value of 0 does not limit on the search, but may slow down compilation
9630 of huge functions. The default value is 30.
9632 @item max-tail-merge-comparisons
9633 The maximum amount of similar bbs to compare a bb with. This is used to
9634 avoid quadratic behavior in tree tail merging. The default value is 10.
9636 @item max-tail-merge-iterations
9637 The maximum amount of iterations of the pass over the function. This is used to
9638 limit compilation time in tree tail merging. The default value is 2.
9640 @item store-merging-allow-unaligned
9641 Allow the store merging pass to introduce unaligned stores if it is legal to
9642 do so. The default value is 1.
9644 @item max-stores-to-merge
9645 The maximum number of stores to attempt to merge into wider stores in the store
9646 merging pass. The minimum value is 2 and the default is 64.
9648 @item max-unrolled-insns
9649 The maximum number of instructions that a loop may have to be unrolled.
9650 If a loop is unrolled, this parameter also determines how many times
9651 the loop code is unrolled.
9653 @item max-average-unrolled-insns
9654 The maximum number of instructions biased by probabilities of their execution
9655 that a loop may have to be unrolled. If a loop is unrolled,
9656 this parameter also determines how many times the loop code is unrolled.
9658 @item max-unroll-times
9659 The maximum number of unrollings of a single loop.
9661 @item max-peeled-insns
9662 The maximum number of instructions that a loop may have to be peeled.
9663 If a loop is peeled, this parameter also determines how many times
9664 the loop code is peeled.
9666 @item max-peel-times
9667 The maximum number of peelings of a single loop.
9669 @item max-peel-branches
9670 The maximum number of branches on the hot path through the peeled sequence.
9672 @item max-completely-peeled-insns
9673 The maximum number of insns of a completely peeled loop.
9675 @item max-completely-peel-times
9676 The maximum number of iterations of a loop to be suitable for complete peeling.
9678 @item max-completely-peel-loop-nest-depth
9679 The maximum depth of a loop nest suitable for complete peeling.
9681 @item max-unswitch-insns
9682 The maximum number of insns of an unswitched loop.
9684 @item max-unswitch-level
9685 The maximum number of branches unswitched in a single loop.
9687 @item max-loop-headers-insns
9688 The maximum number of insns in loop header duplicated by he copy loop headers
9692 The minimum cost of an expensive expression in the loop invariant motion.
9694 @item iv-consider-all-candidates-bound
9695 Bound on number of candidates for induction variables, below which
9696 all candidates are considered for each use in induction variable
9697 optimizations. If there are more candidates than this,
9698 only the most relevant ones are considered to avoid quadratic time complexity.
9700 @item iv-max-considered-uses
9701 The induction variable optimizations give up on loops that contain more
9702 induction variable uses.
9704 @item iv-always-prune-cand-set-bound
9705 If the number of candidates in the set is smaller than this value,
9706 always try to remove unnecessary ivs from the set
9707 when adding a new one.
9709 @item avg-loop-niter
9710 Average number of iterations of a loop.
9712 @item scev-max-expr-size
9713 Bound on size of expressions used in the scalar evolutions analyzer.
9714 Large expressions slow the analyzer.
9716 @item scev-max-expr-complexity
9717 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9718 Complex expressions slow the analyzer.
9720 @item max-tree-if-conversion-phi-args
9721 Maximum number of arguments in a PHI supported by TREE if conversion
9722 unless the loop is marked with simd pragma.
9724 @item vect-max-version-for-alignment-checks
9725 The maximum number of run-time checks that can be performed when
9726 doing loop versioning for alignment in the vectorizer.
9728 @item vect-max-version-for-alias-checks
9729 The maximum number of run-time checks that can be performed when
9730 doing loop versioning for alias in the vectorizer.
9732 @item vect-max-peeling-for-alignment
9733 The maximum number of loop peels to enhance access alignment
9734 for vectorizer. Value -1 means no limit.
9736 @item max-iterations-to-track
9737 The maximum number of iterations of a loop the brute-force algorithm
9738 for analysis of the number of iterations of the loop tries to evaluate.
9740 @item hot-bb-count-ws-permille
9741 A basic block profile count is considered hot if it contributes to
9742 the given permillage (i.e. 0...1000) of the entire profiled execution.
9744 @item hot-bb-frequency-fraction
9745 Select fraction of the entry block frequency of executions of basic block in
9746 function given basic block needs to have to be considered hot.
9748 @item max-predicted-iterations
9749 The maximum number of loop iterations we predict statically. This is useful
9750 in cases where a function contains a single loop with known bound and
9751 another loop with unknown bound.
9752 The known number of iterations is predicted correctly, while
9753 the unknown number of iterations average to roughly 10. This means that the
9754 loop without bounds appears artificially cold relative to the other one.
9756 @item builtin-expect-probability
9757 Control the probability of the expression having the specified value. This
9758 parameter takes a percentage (i.e. 0 ... 100) as input.
9759 The default probability of 90 is obtained empirically.
9761 @item align-threshold
9763 Select fraction of the maximal frequency of executions of a basic block in
9764 a function to align the basic block.
9766 @item align-loop-iterations
9768 A loop expected to iterate at least the selected number of iterations is
9771 @item tracer-dynamic-coverage
9772 @itemx tracer-dynamic-coverage-feedback
9774 This value is used to limit superblock formation once the given percentage of
9775 executed instructions is covered. This limits unnecessary code size
9778 The @option{tracer-dynamic-coverage-feedback} parameter
9779 is used only when profile
9780 feedback is available. The real profiles (as opposed to statically estimated
9781 ones) are much less balanced allowing the threshold to be larger value.
9783 @item tracer-max-code-growth
9784 Stop tail duplication once code growth has reached given percentage. This is
9785 a rather artificial limit, as most of the duplicates are eliminated later in
9786 cross jumping, so it may be set to much higher values than is the desired code
9789 @item tracer-min-branch-ratio
9791 Stop reverse growth when the reverse probability of best edge is less than this
9792 threshold (in percent).
9794 @item tracer-min-branch-probability
9795 @itemx tracer-min-branch-probability-feedback
9797 Stop forward growth if the best edge has probability lower than this
9800 Similarly to @option{tracer-dynamic-coverage} two parameters are
9801 provided. @option{tracer-min-branch-probability-feedback} is used for
9802 compilation with profile feedback and @option{tracer-min-branch-probability}
9803 compilation without. The value for compilation with profile feedback
9804 needs to be more conservative (higher) in order to make tracer
9807 @item max-cse-path-length
9809 The maximum number of basic blocks on path that CSE considers.
9813 The maximum number of instructions CSE processes before flushing.
9814 The default is 1000.
9816 @item ggc-min-expand
9818 GCC uses a garbage collector to manage its own memory allocation. This
9819 parameter specifies the minimum percentage by which the garbage
9820 collector's heap should be allowed to expand between collections.
9821 Tuning this may improve compilation speed; it has no effect on code
9824 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9825 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9826 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9827 GCC is not able to calculate RAM on a particular platform, the lower
9828 bound of 30% is used. Setting this parameter and
9829 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9830 every opportunity. This is extremely slow, but can be useful for
9833 @item ggc-min-heapsize
9835 Minimum size of the garbage collector's heap before it begins bothering
9836 to collect garbage. The first collection occurs after the heap expands
9837 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9838 tuning this may improve compilation speed, and has no effect on code
9841 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9842 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9843 with a lower bound of 4096 (four megabytes) and an upper bound of
9844 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9845 particular platform, the lower bound is used. Setting this parameter
9846 very large effectively disables garbage collection. Setting this
9847 parameter and @option{ggc-min-expand} to zero causes a full collection
9848 to occur at every opportunity.
9850 @item max-reload-search-insns
9851 The maximum number of instruction reload should look backward for equivalent
9852 register. Increasing values mean more aggressive optimization, making the
9853 compilation time increase with probably slightly better performance.
9854 The default value is 100.
9856 @item max-cselib-memory-locations
9857 The maximum number of memory locations cselib should take into account.
9858 Increasing values mean more aggressive optimization, making the compilation time
9859 increase with probably slightly better performance. The default value is 500.
9861 @item max-sched-ready-insns
9862 The maximum number of instructions ready to be issued the scheduler should
9863 consider at any given time during the first scheduling pass. Increasing
9864 values mean more thorough searches, making the compilation time increase
9865 with probably little benefit. The default value is 100.
9867 @item max-sched-region-blocks
9868 The maximum number of blocks in a region to be considered for
9869 interblock scheduling. The default value is 10.
9871 @item max-pipeline-region-blocks
9872 The maximum number of blocks in a region to be considered for
9873 pipelining in the selective scheduler. The default value is 15.
9875 @item max-sched-region-insns
9876 The maximum number of insns in a region to be considered for
9877 interblock scheduling. The default value is 100.
9879 @item max-pipeline-region-insns
9880 The maximum number of insns in a region to be considered for
9881 pipelining in the selective scheduler. The default value is 200.
9884 The minimum probability (in percents) of reaching a source block
9885 for interblock speculative scheduling. The default value is 40.
9887 @item max-sched-extend-regions-iters
9888 The maximum number of iterations through CFG to extend regions.
9889 A value of 0 (the default) disables region extensions.
9891 @item max-sched-insn-conflict-delay
9892 The maximum conflict delay for an insn to be considered for speculative motion.
9893 The default value is 3.
9895 @item sched-spec-prob-cutoff
9896 The minimal probability of speculation success (in percents), so that
9897 speculative insns are scheduled.
9898 The default value is 40.
9900 @item sched-state-edge-prob-cutoff
9901 The minimum probability an edge must have for the scheduler to save its
9903 The default value is 10.
9905 @item sched-mem-true-dep-cost
9906 Minimal distance (in CPU cycles) between store and load targeting same
9907 memory locations. The default value is 1.
9909 @item selsched-max-lookahead
9910 The maximum size of the lookahead window of selective scheduling. It is a
9911 depth of search for available instructions.
9912 The default value is 50.
9914 @item selsched-max-sched-times
9915 The maximum number of times that an instruction is scheduled during
9916 selective scheduling. This is the limit on the number of iterations
9917 through which the instruction may be pipelined. The default value is 2.
9919 @item selsched-insns-to-rename
9920 The maximum number of best instructions in the ready list that are considered
9921 for renaming in the selective scheduler. The default value is 2.
9924 The minimum value of stage count that swing modulo scheduler
9925 generates. The default value is 2.
9927 @item max-last-value-rtl
9928 The maximum size measured as number of RTLs that can be recorded in an expression
9929 in combiner for a pseudo register as last known value of that register. The default
9932 @item max-combine-insns
9933 The maximum number of instructions the RTL combiner tries to combine.
9934 The default value is 2 at @option{-Og} and 4 otherwise.
9936 @item integer-share-limit
9937 Small integer constants can use a shared data structure, reducing the
9938 compiler's memory usage and increasing its speed. This sets the maximum
9939 value of a shared integer constant. The default value is 256.
9941 @item ssp-buffer-size
9942 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9943 protection when @option{-fstack-protection} is used.
9945 @item min-size-for-stack-sharing
9946 The minimum size of variables taking part in stack slot sharing when not
9947 optimizing. The default value is 32.
9949 @item max-jump-thread-duplication-stmts
9950 Maximum number of statements allowed in a block that needs to be
9951 duplicated when threading jumps.
9953 @item max-fields-for-field-sensitive
9954 Maximum number of fields in a structure treated in
9955 a field sensitive manner during pointer analysis. The default is zero
9956 for @option{-O0} and @option{-O1},
9957 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9959 @item prefetch-latency
9960 Estimate on average number of instructions that are executed before
9961 prefetch finishes. The distance prefetched ahead is proportional
9962 to this constant. Increasing this number may also lead to less
9963 streams being prefetched (see @option{simultaneous-prefetches}).
9965 @item simultaneous-prefetches
9966 Maximum number of prefetches that can run at the same time.
9968 @item l1-cache-line-size
9969 The size of cache line in L1 cache, in bytes.
9972 The size of L1 cache, in kilobytes.
9975 The size of L2 cache, in kilobytes.
9977 @item min-insn-to-prefetch-ratio
9978 The minimum ratio between the number of instructions and the
9979 number of prefetches to enable prefetching in a loop.
9981 @item prefetch-min-insn-to-mem-ratio
9982 The minimum ratio between the number of instructions and the
9983 number of memory references to enable prefetching in a loop.
9985 @item use-canonical-types
9986 Whether the compiler should use the ``canonical'' type system. By
9987 default, this should always be 1, which uses a more efficient internal
9988 mechanism for comparing types in C++ and Objective-C++. However, if
9989 bugs in the canonical type system are causing compilation failures,
9990 set this value to 0 to disable canonical types.
9992 @item switch-conversion-max-branch-ratio
9993 Switch initialization conversion refuses to create arrays that are
9994 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9995 branches in the switch.
9997 @item max-partial-antic-length
9998 Maximum length of the partial antic set computed during the tree
9999 partial redundancy elimination optimization (@option{-ftree-pre}) when
10000 optimizing at @option{-O3} and above. For some sorts of source code
10001 the enhanced partial redundancy elimination optimization can run away,
10002 consuming all of the memory available on the host machine. This
10003 parameter sets a limit on the length of the sets that are computed,
10004 which prevents the runaway behavior. Setting a value of 0 for
10005 this parameter allows an unlimited set length.
10007 @item sccvn-max-scc-size
10008 Maximum size of a strongly connected component (SCC) during SCCVN
10009 processing. If this limit is hit, SCCVN processing for the whole
10010 function is not done and optimizations depending on it are
10011 disabled. The default maximum SCC size is 10000.
10013 @item sccvn-max-alias-queries-per-access
10014 Maximum number of alias-oracle queries we perform when looking for
10015 redundancies for loads and stores. If this limit is hit the search
10016 is aborted and the load or store is not considered redundant. The
10017 number of queries is algorithmically limited to the number of
10018 stores on all paths from the load to the function entry.
10019 The default maximum number of queries is 1000.
10021 @item ira-max-loops-num
10022 IRA uses regional register allocation by default. If a function
10023 contains more loops than the number given by this parameter, only at most
10024 the given number of the most frequently-executed loops form regions
10025 for regional register allocation. The default value of the
10028 @item ira-max-conflict-table-size
10029 Although IRA uses a sophisticated algorithm to compress the conflict
10030 table, the table can still require excessive amounts of memory for
10031 huge functions. If the conflict table for a function could be more
10032 than the size in MB given by this parameter, the register allocator
10033 instead uses a faster, simpler, and lower-quality
10034 algorithm that does not require building a pseudo-register conflict table.
10035 The default value of the parameter is 2000.
10037 @item ira-loop-reserved-regs
10038 IRA can be used to evaluate more accurate register pressure in loops
10039 for decisions to move loop invariants (see @option{-O3}). The number
10040 of available registers reserved for some other purposes is given
10041 by this parameter. The default value of the parameter is 2, which is
10042 the minimal number of registers needed by typical instructions.
10043 This value is the best found from numerous experiments.
10045 @item lra-inheritance-ebb-probability-cutoff
10046 LRA tries to reuse values reloaded in registers in subsequent insns.
10047 This optimization is called inheritance. EBB is used as a region to
10048 do this optimization. The parameter defines a minimal fall-through
10049 edge probability in percentage used to add BB to inheritance EBB in
10050 LRA. The default value of the parameter is 40. The value was chosen
10051 from numerous runs of SPEC2000 on x86-64.
10053 @item loop-invariant-max-bbs-in-loop
10054 Loop invariant motion can be very expensive, both in compilation time and
10055 in amount of needed compile-time memory, with very large loops. Loops
10056 with more basic blocks than this parameter won't have loop invariant
10057 motion optimization performed on them. The default value of the
10058 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10060 @item loop-max-datarefs-for-datadeps
10061 Building data dependencies is expensive for very large loops. This
10062 parameter limits the number of data references in loops that are
10063 considered for data dependence analysis. These large loops are no
10064 handled by the optimizations using loop data dependencies.
10065 The default value is 1000.
10067 @item max-vartrack-size
10068 Sets a maximum number of hash table slots to use during variable
10069 tracking dataflow analysis of any function. If this limit is exceeded
10070 with variable tracking at assignments enabled, analysis for that
10071 function is retried without it, after removing all debug insns from
10072 the function. If the limit is exceeded even without debug insns, var
10073 tracking analysis is completely disabled for the function. Setting
10074 the parameter to zero makes it unlimited.
10076 @item max-vartrack-expr-depth
10077 Sets a maximum number of recursion levels when attempting to map
10078 variable names or debug temporaries to value expressions. This trades
10079 compilation time for more complete debug information. If this is set too
10080 low, value expressions that are available and could be represented in
10081 debug information may end up not being used; setting this higher may
10082 enable the compiler to find more complex debug expressions, but compile
10083 time and memory use may grow. The default is 12.
10085 @item min-nondebug-insn-uid
10086 Use uids starting at this parameter for nondebug insns. The range below
10087 the parameter is reserved exclusively for debug insns created by
10088 @option{-fvar-tracking-assignments}, but debug insns may get
10089 (non-overlapping) uids above it if the reserved range is exhausted.
10091 @item ipa-sra-ptr-growth-factor
10092 IPA-SRA replaces a pointer to an aggregate with one or more new
10093 parameters only when their cumulative size is less or equal to
10094 @option{ipa-sra-ptr-growth-factor} times the size of the original
10097 @item sra-max-scalarization-size-Ospeed
10098 @item sra-max-scalarization-size-Osize
10099 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10100 replace scalar parts of aggregates with uses of independent scalar
10101 variables. These parameters control the maximum size, in storage units,
10102 of aggregate which is considered for replacement when compiling for
10104 (@option{sra-max-scalarization-size-Ospeed}) or size
10105 (@option{sra-max-scalarization-size-Osize}) respectively.
10107 @item tm-max-aggregate-size
10108 When making copies of thread-local variables in a transaction, this
10109 parameter specifies the size in bytes after which variables are
10110 saved with the logging functions as opposed to save/restore code
10111 sequence pairs. This option only applies when using
10114 @item graphite-max-nb-scop-params
10115 To avoid exponential effects in the Graphite loop transforms, the
10116 number of parameters in a Static Control Part (SCoP) is bounded. The
10117 default value is 10 parameters. A variable whose value is unknown at
10118 compilation time and defined outside a SCoP is a parameter of the SCoP.
10120 @item graphite-max-bbs-per-function
10121 To avoid exponential effects in the detection of SCoPs, the size of
10122 the functions analyzed by Graphite is bounded. The default value is
10125 @item loop-block-tile-size
10126 Loop blocking or strip mining transforms, enabled with
10127 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10128 loop in the loop nest by a given number of iterations. The strip
10129 length can be changed using the @option{loop-block-tile-size}
10130 parameter. The default value is 51 iterations.
10132 @item loop-unroll-jam-size
10133 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10134 default value is 4.
10136 @item loop-unroll-jam-depth
10137 Specify the dimension to be unrolled (counting from the most inner loop)
10138 for the @option{-floop-unroll-and-jam}. The default value is 2.
10140 @item ipa-cp-value-list-size
10141 IPA-CP attempts to track all possible values and types passed to a function's
10142 parameter in order to propagate them and perform devirtualization.
10143 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10144 stores per one formal parameter of a function.
10146 @item ipa-cp-eval-threshold
10147 IPA-CP calculates its own score of cloning profitability heuristics
10148 and performs those cloning opportunities with scores that exceed
10149 @option{ipa-cp-eval-threshold}.
10151 @item ipa-cp-recursion-penalty
10152 Percentage penalty the recursive functions will receive when they
10153 are evaluated for cloning.
10155 @item ipa-cp-single-call-penalty
10156 Percentage penalty functions containg a single call to another
10157 function will receive when they are evaluated for cloning.
10160 @item ipa-max-agg-items
10161 IPA-CP is also capable to propagate a number of scalar values passed
10162 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10163 number of such values per one parameter.
10165 @item ipa-cp-loop-hint-bonus
10166 When IPA-CP determines that a cloning candidate would make the number
10167 of iterations of a loop known, it adds a bonus of
10168 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10171 @item ipa-cp-array-index-hint-bonus
10172 When IPA-CP determines that a cloning candidate would make the index of
10173 an array access known, it adds a bonus of
10174 @option{ipa-cp-array-index-hint-bonus} to the profitability
10175 score of the candidate.
10177 @item ipa-max-aa-steps
10178 During its analysis of function bodies, IPA-CP employs alias analysis
10179 in order to track values pointed to by function parameters. In order
10180 not spend too much time analyzing huge functions, it gives up and
10181 consider all memory clobbered after examining
10182 @option{ipa-max-aa-steps} statements modifying memory.
10184 @item lto-partitions
10185 Specify desired number of partitions produced during WHOPR compilation.
10186 The number of partitions should exceed the number of CPUs used for compilation.
10187 The default value is 32.
10189 @item lto-min-partition
10190 Size of minimal partition for WHOPR (in estimated instructions).
10191 This prevents expenses of splitting very small programs into too many
10194 @item lto-max-partition
10195 Size of max partition for WHOPR (in estimated instructions).
10196 to provide an upper bound for individual size of partition.
10197 Meant to be used only with balanced partitioning.
10199 @item cxx-max-namespaces-for-diagnostic-help
10200 The maximum number of namespaces to consult for suggestions when C++
10201 name lookup fails for an identifier. The default is 1000.
10203 @item sink-frequency-threshold
10204 The maximum relative execution frequency (in percents) of the target block
10205 relative to a statement's original block to allow statement sinking of a
10206 statement. Larger numbers result in more aggressive statement sinking.
10207 The default value is 75. A small positive adjustment is applied for
10208 statements with memory operands as those are even more profitable so sink.
10210 @item max-stores-to-sink
10211 The maximum number of conditional store pairs that can be sunk. Set to 0
10212 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10213 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10215 @item allow-store-data-races
10216 Allow optimizers to introduce new data races on stores.
10217 Set to 1 to allow, otherwise to 0. This option is enabled by default
10218 at optimization level @option{-Ofast}.
10220 @item case-values-threshold
10221 The smallest number of different values for which it is best to use a
10222 jump-table instead of a tree of conditional branches. If the value is
10223 0, use the default for the machine. The default is 0.
10225 @item tree-reassoc-width
10226 Set the maximum number of instructions executed in parallel in
10227 reassociated tree. This parameter overrides target dependent
10228 heuristics used by default if has non zero value.
10230 @item sched-pressure-algorithm
10231 Choose between the two available implementations of
10232 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10233 and is the more likely to prevent instructions from being reordered.
10234 Algorithm 2 was designed to be a compromise between the relatively
10235 conservative approach taken by algorithm 1 and the rather aggressive
10236 approach taken by the default scheduler. It relies more heavily on
10237 having a regular register file and accurate register pressure classes.
10238 See @file{haifa-sched.c} in the GCC sources for more details.
10240 The default choice depends on the target.
10242 @item max-slsr-cand-scan
10243 Set the maximum number of existing candidates that are considered when
10244 seeking a basis for a new straight-line strength reduction candidate.
10247 Enable buffer overflow detection for global objects. This kind
10248 of protection is enabled by default if you are using
10249 @option{-fsanitize=address} option.
10250 To disable global objects protection use @option{--param asan-globals=0}.
10253 Enable buffer overflow detection for stack objects. This kind of
10254 protection is enabled by default when using @option{-fsanitize=address}.
10255 To disable stack protection use @option{--param asan-stack=0} option.
10257 @item asan-instrument-reads
10258 Enable buffer overflow detection for memory reads. This kind of
10259 protection is enabled by default when using @option{-fsanitize=address}.
10260 To disable memory reads protection use
10261 @option{--param asan-instrument-reads=0}.
10263 @item asan-instrument-writes
10264 Enable buffer overflow detection for memory writes. This kind of
10265 protection is enabled by default when using @option{-fsanitize=address}.
10266 To disable memory writes protection use
10267 @option{--param asan-instrument-writes=0} option.
10269 @item asan-memintrin
10270 Enable detection for built-in functions. This kind of protection
10271 is enabled by default when using @option{-fsanitize=address}.
10272 To disable built-in functions protection use
10273 @option{--param asan-memintrin=0}.
10275 @item asan-use-after-return
10276 Enable detection of use-after-return. This kind of protection
10277 is enabled by default when using @option{-fsanitize=address} option.
10278 To disable use-after-return detection use
10279 @option{--param asan-use-after-return=0}.
10281 Note: The check is disabled by default at runtime. To enable the check,
10282 you should set environment variable @env{ASAN_OPTIONS} to
10283 @code{detect_stack_use_after_return=1}.
10285 @item asan-instrumentation-with-call-threshold
10286 If number of memory accesses in function being instrumented
10287 is greater or equal to this number, use callbacks instead of inline checks.
10288 E.g. to disable inline code use
10289 @option{--param asan-instrumentation-with-call-threshold=0}.
10291 @item chkp-max-ctor-size
10292 Static constructors generated by Pointer Bounds Checker may become very
10293 large and significantly increase compile time at optimization level
10294 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10295 in a single generated constructor. Default value is 5000.
10297 @item max-fsm-thread-path-insns
10298 Maximum number of instructions to copy when duplicating blocks on a
10299 finite state automaton jump thread path. The default is 100.
10301 @item max-fsm-thread-length
10302 Maximum number of basic blocks on a finite state automaton jump thread
10303 path. The default is 10.
10305 @item max-fsm-thread-paths
10306 Maximum number of new jump thread paths to create for a finite state
10307 automaton. The default is 50.
10309 @item parloops-chunk-size
10310 Chunk size of omp schedule for loops parallelized by parloops. The default
10313 @item parloops-schedule
10314 Schedule type of omp schedule for loops parallelized by parloops (static,
10315 dynamic, guided, auto, runtime). The default is static.
10317 @item max-ssa-name-query-depth
10318 Maximum depth of recursion when querying properties of SSA names in things
10319 like fold routines. One level of recursion corresponds to following a
10322 @item hsa-gen-debug-stores
10323 Enable emission of special debug stores within HSA kernels which are
10324 then read and reported by libgomp plugin. Generation of these stores
10325 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10328 @item max-speculative-devirt-maydefs
10329 The maximum number of may-defs we analyze when looking for a must-def
10330 specifying the dynamic type of an object that invokes a virtual call
10331 we may be able to devirtualize speculatively.
10333 @item max-vrp-switch-assertions
10334 The maximum number of assertions to add along the default edge of a switch
10335 statement during VRP. The default is 10.
10339 @node Instrumentation Options
10340 @section Program Instrumentation Options
10341 @cindex instrumentation options
10342 @cindex program instrumentation options
10343 @cindex run-time error checking options
10344 @cindex profiling options
10345 @cindex options, program instrumentation
10346 @cindex options, run-time error checking
10347 @cindex options, profiling
10349 GCC supports a number of command-line options that control adding
10350 run-time instrumentation to the code it normally generates.
10351 For example, one purpose of instrumentation is collect profiling
10352 statistics for use in finding program hot spots, code coverage
10353 analysis, or profile-guided optimizations.
10354 Another class of program instrumentation is adding run-time checking
10355 to detect programming errors like invalid pointer
10356 dereferences or out-of-bounds array accesses, as well as deliberately
10357 hostile attacks such as stack smashing or C++ vtable hijacking.
10358 There is also a general hook which can be used to implement other
10359 forms of tracing or function-level instrumentation for debug or
10360 program analysis purposes.
10363 @cindex @command{prof}
10366 Generate extra code to write profile information suitable for the
10367 analysis program @command{prof}. You must use this option when compiling
10368 the source files you want data about, and you must also use it when
10371 @cindex @command{gprof}
10374 Generate extra code to write profile information suitable for the
10375 analysis program @command{gprof}. You must use this option when compiling
10376 the source files you want data about, and you must also use it when
10379 @item -fprofile-arcs
10380 @opindex fprofile-arcs
10381 Add code so that program flow @dfn{arcs} are instrumented. During
10382 execution the program records how many times each branch and call is
10383 executed and how many times it is taken or returns. When the compiled
10384 program exits it saves this data to a file called
10385 @file{@var{auxname}.gcda} for each source file. The data may be used for
10386 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10387 test coverage analysis (@option{-ftest-coverage}). Each object file's
10388 @var{auxname} is generated from the name of the output file, if
10389 explicitly specified and it is not the final executable, otherwise it is
10390 the basename of the source file. In both cases any suffix is removed
10391 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10392 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10393 @xref{Cross-profiling}.
10395 @cindex @command{gcov}
10399 This option is used to compile and link code instrumented for coverage
10400 analysis. The option is a synonym for @option{-fprofile-arcs}
10401 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10402 linking). See the documentation for those options for more details.
10407 Compile the source files with @option{-fprofile-arcs} plus optimization
10408 and code generation options. For test coverage analysis, use the
10409 additional @option{-ftest-coverage} option. You do not need to profile
10410 every source file in a program.
10413 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10414 (the latter implies the former).
10417 Run the program on a representative workload to generate the arc profile
10418 information. This may be repeated any number of times. You can run
10419 concurrent instances of your program, and provided that the file system
10420 supports locking, the data files will be correctly updated. Also
10421 @code{fork} calls are detected and correctly handled (double counting
10425 For profile-directed optimizations, compile the source files again with
10426 the same optimization and code generation options plus
10427 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10428 Control Optimization}).
10431 For test coverage analysis, use @command{gcov} to produce human readable
10432 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10433 @command{gcov} documentation for further information.
10437 With @option{-fprofile-arcs}, for each function of your program GCC
10438 creates a program flow graph, then finds a spanning tree for the graph.
10439 Only arcs that are not on the spanning tree have to be instrumented: the
10440 compiler adds code to count the number of times that these arcs are
10441 executed. When an arc is the only exit or only entrance to a block, the
10442 instrumentation code can be added to the block; otherwise, a new basic
10443 block must be created to hold the instrumentation code.
10446 @item -ftest-coverage
10447 @opindex ftest-coverage
10448 Produce a notes file that the @command{gcov} code-coverage utility
10449 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10450 show program coverage. Each source file's note file is called
10451 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10452 above for a description of @var{auxname} and instructions on how to
10453 generate test coverage data. Coverage data matches the source files
10454 more closely if you do not optimize.
10456 @item -fprofile-dir=@var{path}
10457 @opindex fprofile-dir
10459 Set the directory to search for the profile data files in to @var{path}.
10460 This option affects only the profile data generated by
10461 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10462 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10463 and its related options. Both absolute and relative paths can be used.
10464 By default, GCC uses the current directory as @var{path}, thus the
10465 profile data file appears in the same directory as the object file.
10467 @item -fprofile-generate
10468 @itemx -fprofile-generate=@var{path}
10469 @opindex fprofile-generate
10471 Enable options usually used for instrumenting application to produce
10472 profile useful for later recompilation with profile feedback based
10473 optimization. You must use @option{-fprofile-generate} both when
10474 compiling and when linking your program.
10476 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10478 If @var{path} is specified, GCC looks at the @var{path} to find
10479 the profile feedback data files. See @option{-fprofile-dir}.
10481 To optimize the program based on the collected profile information, use
10482 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10484 @item -fprofile-update=@var{method}
10485 @opindex fprofile-update
10487 Alter the update method for an application instrumented for profile
10488 feedback based optimization. The @var{method} argument should be one of
10489 @samp{single} or @samp{atomic}. The first one is useful for single-threaded
10490 applications, while the second one prevents profile corruption by emitting
10493 @strong{Warning:} When an application does not properly join all threads
10494 (or creates an detached thread), a profile file can be still corrupted.
10496 @item -fsanitize=address
10497 @opindex fsanitize=address
10498 Enable AddressSanitizer, a fast memory error detector.
10499 Memory access instructions are instrumented to detect
10500 out-of-bounds and use-after-free bugs.
10501 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10502 more details. The run-time behavior can be influenced using the
10503 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10504 the available options are shown at startup of the instrumented program. See
10505 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10506 for a list of supported options.
10507 The option can't be combined with @option{-fsanitize=thread}.
10509 @item -fsanitize=kernel-address
10510 @opindex fsanitize=kernel-address
10511 Enable AddressSanitizer for Linux kernel.
10512 See @uref{https://github.com/google/kasan/wiki} for more details.
10514 @item -fsanitize=thread
10515 @opindex fsanitize=thread
10516 Enable ThreadSanitizer, a fast data race detector.
10517 Memory access instructions are instrumented to detect
10518 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10519 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10520 environment variable; see
10521 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10523 The option can't be combined with @option{-fsanitize=address}
10524 and/or @option{-fsanitize=leak}.
10526 @item -fsanitize=leak
10527 @opindex fsanitize=leak
10528 Enable LeakSanitizer, a memory leak detector.
10529 This option only matters for linking of executables and
10530 the executable is linked against a library that overrides @code{malloc}
10531 and other allocator functions. See
10532 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
10533 details. The run-time behavior can be influenced using the
10534 @env{LSAN_OPTIONS} environment variable.
10535 The option can't be combined with @option{-fsanitize=thread}.
10537 @item -fsanitize=undefined
10538 @opindex fsanitize=undefined
10539 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
10540 Various computations are instrumented to detect undefined behavior
10541 at runtime. Current suboptions are:
10545 @item -fsanitize=shift
10546 @opindex fsanitize=shift
10547 This option enables checking that the result of a shift operation is
10548 not undefined. Note that what exactly is considered undefined differs
10549 slightly between C and C++, as well as between ISO C90 and C99, etc.
10551 @item -fsanitize=integer-divide-by-zero
10552 @opindex fsanitize=integer-divide-by-zero
10553 Detect integer division by zero as well as @code{INT_MIN / -1} division.
10555 @item -fsanitize=unreachable
10556 @opindex fsanitize=unreachable
10557 With this option, the compiler turns the @code{__builtin_unreachable}
10558 call into a diagnostics message call instead. When reaching the
10559 @code{__builtin_unreachable} call, the behavior is undefined.
10561 @item -fsanitize=vla-bound
10562 @opindex fsanitize=vla-bound
10563 This option instructs the compiler to check that the size of a variable
10564 length array is positive.
10566 @item -fsanitize=null
10567 @opindex fsanitize=null
10568 This option enables pointer checking. Particularly, the application
10569 built with this option turned on will issue an error message when it
10570 tries to dereference a NULL pointer, or if a reference (possibly an
10571 rvalue reference) is bound to a NULL pointer, or if a method is invoked
10572 on an object pointed by a NULL pointer.
10574 @item -fsanitize=return
10575 @opindex fsanitize=return
10576 This option enables return statement checking. Programs
10577 built with this option turned on will issue an error message
10578 when the end of a non-void function is reached without actually
10579 returning a value. This option works in C++ only.
10581 @item -fsanitize=signed-integer-overflow
10582 @opindex fsanitize=signed-integer-overflow
10583 This option enables signed integer overflow checking. We check that
10584 the result of @code{+}, @code{*}, and both unary and binary @code{-}
10585 does not overflow in the signed arithmetics. Note, integer promotion
10586 rules must be taken into account. That is, the following is not an
10589 signed char a = SCHAR_MAX;
10593 @item -fsanitize=bounds
10594 @opindex fsanitize=bounds
10595 This option enables instrumentation of array bounds. Various out of bounds
10596 accesses are detected. Flexible array members, flexible array member-like
10597 arrays, and initializers of variables with static storage are not instrumented.
10599 @item -fsanitize=bounds-strict
10600 @opindex fsanitize=bounds-strict
10601 This option enables strict instrumentation of array bounds. Most out of bounds
10602 accesses are detected, including flexible array members and flexible array
10603 member-like arrays. Initializers of variables with static storage are not
10606 @item -fsanitize=alignment
10607 @opindex fsanitize=alignment
10609 This option enables checking of alignment of pointers when they are
10610 dereferenced, or when a reference is bound to insufficiently aligned target,
10611 or when a method or constructor is invoked on insufficiently aligned object.
10613 @item -fsanitize=object-size
10614 @opindex fsanitize=object-size
10615 This option enables instrumentation of memory references using the
10616 @code{__builtin_object_size} function. Various out of bounds pointer
10617 accesses are detected.
10619 @item -fsanitize=float-divide-by-zero
10620 @opindex fsanitize=float-divide-by-zero
10621 Detect floating-point division by zero. Unlike other similar options,
10622 @option{-fsanitize=float-divide-by-zero} is not enabled by
10623 @option{-fsanitize=undefined}, since floating-point division by zero can
10624 be a legitimate way of obtaining infinities and NaNs.
10626 @item -fsanitize=float-cast-overflow
10627 @opindex fsanitize=float-cast-overflow
10628 This option enables floating-point type to integer conversion checking.
10629 We check that the result of the conversion does not overflow.
10630 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10631 not enabled by @option{-fsanitize=undefined}.
10632 This option does not work well with @code{FE_INVALID} exceptions enabled.
10634 @item -fsanitize=nonnull-attribute
10635 @opindex fsanitize=nonnull-attribute
10637 This option enables instrumentation of calls, checking whether null values
10638 are not passed to arguments marked as requiring a non-null value by the
10639 @code{nonnull} function attribute.
10641 @item -fsanitize=returns-nonnull-attribute
10642 @opindex fsanitize=returns-nonnull-attribute
10644 This option enables instrumentation of return statements in functions
10645 marked with @code{returns_nonnull} function attribute, to detect returning
10646 of null values from such functions.
10648 @item -fsanitize=bool
10649 @opindex fsanitize=bool
10651 This option enables instrumentation of loads from bool. If a value other
10652 than 0/1 is loaded, a run-time error is issued.
10654 @item -fsanitize=enum
10655 @opindex fsanitize=enum
10657 This option enables instrumentation of loads from an enum type. If
10658 a value outside the range of values for the enum type is loaded,
10659 a run-time error is issued.
10661 @item -fsanitize=vptr
10662 @opindex fsanitize=vptr
10664 This option enables instrumentation of C++ member function calls, member
10665 accesses and some conversions between pointers to base and derived classes,
10666 to verify the referenced object has the correct dynamic type.
10670 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10671 @option{-fsanitize=undefined} gives a diagnostic message.
10672 This currently works only for the C family of languages.
10674 @item -fno-sanitize=all
10675 @opindex fno-sanitize=all
10677 This option disables all previously enabled sanitizers.
10678 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10681 @item -fasan-shadow-offset=@var{number}
10682 @opindex fasan-shadow-offset
10683 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10684 It is useful for experimenting with different shadow memory layouts in
10685 Kernel AddressSanitizer.
10687 @item -fsanitize-sections=@var{s1},@var{s2},...
10688 @opindex fsanitize-sections
10689 Sanitize global variables in selected user-defined sections. @var{si} may
10692 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10693 @opindex fsanitize-recover
10694 @opindex fno-sanitize-recover
10695 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10696 mentioned in comma-separated list of @var{opts}. Enabling this option
10697 for a sanitizer component causes it to attempt to continue
10698 running the program as if no error happened. This means multiple
10699 runtime errors can be reported in a single program run, and the exit
10700 code of the program may indicate success even when errors
10701 have been reported. The @option{-fno-sanitize-recover=} option
10702 can be used to alter
10703 this behavior: only the first detected error is reported
10704 and program then exits with a non-zero exit code.
10706 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10707 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10708 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10709 @option{-fsanitize=bounds-strict},
10710 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10711 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
10712 for which this feature is experimental.
10713 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10714 accepted, the former enables recovery for all sanitizers that support it,
10715 the latter disables recovery for all sanitizers that support it.
10717 Even if a recovery mode is turned on the compiler side, it needs to be also
10718 enabled on the runtime library side, otherwise the failures are still fatal.
10719 The runtime library defaults to @code{halt_on_error=0} for
10720 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10721 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10722 setting the @code{halt_on_error} flag in the corresponding environment variable.
10724 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
10726 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10729 Similarly @option{-fno-sanitize-recover} is equivalent to
10731 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10734 @item -fsanitize-undefined-trap-on-error
10735 @opindex fsanitize-undefined-trap-on-error
10736 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10737 report undefined behavior using @code{__builtin_trap} rather than
10738 a @code{libubsan} library routine. The advantage of this is that the
10739 @code{libubsan} library is not needed and is not linked in, so this
10740 is usable even in freestanding environments.
10742 @item -fsanitize-coverage=trace-pc
10743 @opindex fsanitize-coverage=trace-pc
10744 Enable coverage-guided fuzzing code instrumentation.
10745 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10747 @item -fbounds-check
10748 @opindex fbounds-check
10749 For front ends that support it, generate additional code to check that
10750 indices used to access arrays are within the declared range. This is
10751 currently only supported by the Java and Fortran front ends, where
10752 this option defaults to true and false respectively.
10754 @item -fcheck-pointer-bounds
10755 @opindex fcheck-pointer-bounds
10756 @opindex fno-check-pointer-bounds
10757 @cindex Pointer Bounds Checker options
10758 Enable Pointer Bounds Checker instrumentation. Each memory reference
10759 is instrumented with checks of the pointer used for memory access against
10760 bounds associated with that pointer.
10763 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10764 and @option{-mmpx} are required to enable this feature.
10765 MPX-based instrumentation requires
10766 a runtime library to enable MPX in hardware and handle bounds
10767 violation signals. By default when @option{-fcheck-pointer-bounds}
10768 and @option{-mmpx} options are used to link a program, the GCC driver
10769 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10770 Bounds checking on calls to dynamic libraries requires a linker
10771 with @option{-z bndplt} support; if GCC was configured with a linker
10772 without support for this option (including the Gold linker and older
10773 versions of ld), a warning is given if you link with @option{-mmpx}
10774 without also specifying @option{-static}, since the overall effectiveness
10775 of the bounds checking protection is reduced.
10776 See also @option{-static-libmpxwrappers}.
10778 MPX-based instrumentation
10779 may be used for debugging and also may be included in production code
10780 to increase program security. Depending on usage, you may
10781 have different requirements for the runtime library. The current version
10782 of the MPX runtime library is more oriented for use as a debugging
10783 tool. MPX runtime library usage implies @option{-lpthread}. See
10784 also @option{-static-libmpx}. The runtime library behavior can be
10785 influenced using various @env{CHKP_RT_*} environment variables. See
10786 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10789 Generated instrumentation may be controlled by various
10790 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10791 structure field attribute (@pxref{Type Attributes}) and
10792 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10793 (@pxref{Function Attributes}). GCC also provides a number of built-in
10794 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10795 Bounds Checker builtins}, for more information.
10797 @item -fchkp-check-incomplete-type
10798 @opindex fchkp-check-incomplete-type
10799 @opindex fno-chkp-check-incomplete-type
10800 Generate pointer bounds checks for variables with incomplete type.
10801 Enabled by default.
10803 @item -fchkp-narrow-bounds
10804 @opindex fchkp-narrow-bounds
10805 @opindex fno-chkp-narrow-bounds
10806 Controls bounds used by Pointer Bounds Checker for pointers to object
10807 fields. If narrowing is enabled then field bounds are used. Otherwise
10808 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10809 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10811 @item -fchkp-first-field-has-own-bounds
10812 @opindex fchkp-first-field-has-own-bounds
10813 @opindex fno-chkp-first-field-has-own-bounds
10814 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10815 first field in the structure. By default a pointer to the first field has
10816 the same bounds as a pointer to the whole structure.
10818 @item -fchkp-narrow-to-innermost-array
10819 @opindex fchkp-narrow-to-innermost-array
10820 @opindex fno-chkp-narrow-to-innermost-array
10821 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10822 case of nested static array access. By default this option is disabled and
10823 bounds of the outermost array are used.
10825 @item -fchkp-optimize
10826 @opindex fchkp-optimize
10827 @opindex fno-chkp-optimize
10828 Enables Pointer Bounds Checker optimizations. Enabled by default at
10829 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
10831 @item -fchkp-use-fast-string-functions
10832 @opindex fchkp-use-fast-string-functions
10833 @opindex fno-chkp-use-fast-string-functions
10834 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
10835 by Pointer Bounds Checker. Disabled by default.
10837 @item -fchkp-use-nochk-string-functions
10838 @opindex fchkp-use-nochk-string-functions
10839 @opindex fno-chkp-use-nochk-string-functions
10840 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
10841 by Pointer Bounds Checker. Disabled by default.
10843 @item -fchkp-use-static-bounds
10844 @opindex fchkp-use-static-bounds
10845 @opindex fno-chkp-use-static-bounds
10846 Allow Pointer Bounds Checker to generate static bounds holding
10847 bounds of static variables. Enabled by default.
10849 @item -fchkp-use-static-const-bounds
10850 @opindex fchkp-use-static-const-bounds
10851 @opindex fno-chkp-use-static-const-bounds
10852 Use statically-initialized bounds for constant bounds instead of
10853 generating them each time they are required. By default enabled when
10854 @option{-fchkp-use-static-bounds} is enabled.
10856 @item -fchkp-treat-zero-dynamic-size-as-infinite
10857 @opindex fchkp-treat-zero-dynamic-size-as-infinite
10858 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
10859 With this option, objects with incomplete type whose
10860 dynamically-obtained size is zero are treated as having infinite size
10861 instead by Pointer Bounds
10862 Checker. This option may be helpful if a program is linked with a library
10863 missing size information for some symbols. Disabled by default.
10865 @item -fchkp-check-read
10866 @opindex fchkp-check-read
10867 @opindex fno-chkp-check-read
10868 Instructs Pointer Bounds Checker to generate checks for all read
10869 accesses to memory. Enabled by default.
10871 @item -fchkp-check-write
10872 @opindex fchkp-check-write
10873 @opindex fno-chkp-check-write
10874 Instructs Pointer Bounds Checker to generate checks for all write
10875 accesses to memory. Enabled by default.
10877 @item -fchkp-store-bounds
10878 @opindex fchkp-store-bounds
10879 @opindex fno-chkp-store-bounds
10880 Instructs Pointer Bounds Checker to generate bounds stores for
10881 pointer writes. Enabled by default.
10883 @item -fchkp-instrument-calls
10884 @opindex fchkp-instrument-calls
10885 @opindex fno-chkp-instrument-calls
10886 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
10887 Enabled by default.
10889 @item -fchkp-instrument-marked-only
10890 @opindex fchkp-instrument-marked-only
10891 @opindex fno-chkp-instrument-marked-only
10892 Instructs Pointer Bounds Checker to instrument only functions
10893 marked with the @code{bnd_instrument} attribute
10894 (@pxref{Function Attributes}). Disabled by default.
10896 @item -fchkp-use-wrappers
10897 @opindex fchkp-use-wrappers
10898 @opindex fno-chkp-use-wrappers
10899 Allows Pointer Bounds Checker to replace calls to built-in functions
10900 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
10901 is used to link a program, the GCC driver automatically links
10902 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
10903 Enabled by default.
10905 @item -fstack-protector
10906 @opindex fstack-protector
10907 Emit extra code to check for buffer overflows, such as stack smashing
10908 attacks. This is done by adding a guard variable to functions with
10909 vulnerable objects. This includes functions that call @code{alloca}, and
10910 functions with buffers larger than 8 bytes. The guards are initialized
10911 when a function is entered and then checked when the function exits.
10912 If a guard check fails, an error message is printed and the program exits.
10914 @item -fstack-protector-all
10915 @opindex fstack-protector-all
10916 Like @option{-fstack-protector} except that all functions are protected.
10918 @item -fstack-protector-strong
10919 @opindex fstack-protector-strong
10920 Like @option{-fstack-protector} but includes additional functions to
10921 be protected --- those that have local array definitions, or have
10922 references to local frame addresses.
10924 @item -fstack-protector-explicit
10925 @opindex fstack-protector-explicit
10926 Like @option{-fstack-protector} but only protects those functions which
10927 have the @code{stack_protect} attribute.
10929 @item -fstack-check
10930 @opindex fstack-check
10931 Generate code to verify that you do not go beyond the boundary of the
10932 stack. You should specify this flag if you are running in an
10933 environment with multiple threads, but you only rarely need to specify it in
10934 a single-threaded environment since stack overflow is automatically
10935 detected on nearly all systems if there is only one stack.
10937 Note that this switch does not actually cause checking to be done; the
10938 operating system or the language runtime must do that. The switch causes
10939 generation of code to ensure that they see the stack being extended.
10941 You can additionally specify a string parameter: @samp{no} means no
10942 checking, @samp{generic} means force the use of old-style checking,
10943 @samp{specific} means use the best checking method and is equivalent
10944 to bare @option{-fstack-check}.
10946 Old-style checking is a generic mechanism that requires no specific
10947 target support in the compiler but comes with the following drawbacks:
10951 Modified allocation strategy for large objects: they are always
10952 allocated dynamically if their size exceeds a fixed threshold.
10955 Fixed limit on the size of the static frame of functions: when it is
10956 topped by a particular function, stack checking is not reliable and
10957 a warning is issued by the compiler.
10960 Inefficiency: because of both the modified allocation strategy and the
10961 generic implementation, code performance is hampered.
10964 Note that old-style stack checking is also the fallback method for
10965 @samp{specific} if no target support has been added in the compiler.
10967 @item -fstack-limit-register=@var{reg}
10968 @itemx -fstack-limit-symbol=@var{sym}
10969 @itemx -fno-stack-limit
10970 @opindex fstack-limit-register
10971 @opindex fstack-limit-symbol
10972 @opindex fno-stack-limit
10973 Generate code to ensure that the stack does not grow beyond a certain value,
10974 either the value of a register or the address of a symbol. If a larger
10975 stack is required, a signal is raised at run time. For most targets,
10976 the signal is raised before the stack overruns the boundary, so
10977 it is possible to catch the signal without taking special precautions.
10979 For instance, if the stack starts at absolute address @samp{0x80000000}
10980 and grows downwards, you can use the flags
10981 @option{-fstack-limit-symbol=__stack_limit} and
10982 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
10983 of 128KB@. Note that this may only work with the GNU linker.
10985 You can locally override stack limit checking by using the
10986 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
10988 @item -fsplit-stack
10989 @opindex fsplit-stack
10990 Generate code to automatically split the stack before it overflows.
10991 The resulting program has a discontiguous stack which can only
10992 overflow if the program is unable to allocate any more memory. This
10993 is most useful when running threaded programs, as it is no longer
10994 necessary to calculate a good stack size to use for each thread. This
10995 is currently only implemented for the x86 targets running
10998 When code compiled with @option{-fsplit-stack} calls code compiled
10999 without @option{-fsplit-stack}, there may not be much stack space
11000 available for the latter code to run. If compiling all code,
11001 including library code, with @option{-fsplit-stack} is not an option,
11002 then the linker can fix up these calls so that the code compiled
11003 without @option{-fsplit-stack} always has a large stack. Support for
11004 this is implemented in the gold linker in GNU binutils release 2.21
11007 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11008 @opindex fvtable-verify
11009 This option is only available when compiling C++ code.
11010 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11011 feature that verifies at run time, for every virtual call, that
11012 the vtable pointer through which the call is made is valid for the type of
11013 the object, and has not been corrupted or overwritten. If an invalid vtable
11014 pointer is detected at run time, an error is reported and execution of the
11015 program is immediately halted.
11017 This option causes run-time data structures to be built at program startup,
11018 which are used for verifying the vtable pointers.
11019 The options @samp{std} and @samp{preinit}
11020 control the timing of when these data structures are built. In both cases the
11021 data structures are built before execution reaches @code{main}. Using
11022 @option{-fvtable-verify=std} causes the data structures to be built after
11023 shared libraries have been loaded and initialized.
11024 @option{-fvtable-verify=preinit} causes them to be built before shared
11025 libraries have been loaded and initialized.
11027 If this option appears multiple times in the command line with different
11028 values specified, @samp{none} takes highest priority over both @samp{std} and
11029 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11032 @opindex fvtv-debug
11033 When used in conjunction with @option{-fvtable-verify=std} or
11034 @option{-fvtable-verify=preinit}, causes debug versions of the
11035 runtime functions for the vtable verification feature to be called.
11036 This flag also causes the compiler to log information about which
11037 vtable pointers it finds for each class.
11038 This information is written to a file named @file{vtv_set_ptr_data.log}
11039 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11040 if that is defined or the current working directory otherwise.
11042 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11043 file, be sure to delete any existing one.
11046 @opindex fvtv-counts
11047 This is a debugging flag. When used in conjunction with
11048 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11049 causes the compiler to keep track of the total number of virtual calls
11050 it encounters and the number of verifications it inserts. It also
11051 counts the number of calls to certain run-time library functions
11052 that it inserts and logs this information for each compilation unit.
11053 The compiler writes this information to a file named
11054 @file{vtv_count_data.log} in the directory named by the environment
11055 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11056 directory otherwise. It also counts the size of the vtable pointer sets
11057 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11058 in the same directory.
11060 Note: This feature @emph{appends} data to the log files. To get fresh log
11061 files, be sure to delete any existing ones.
11063 @item -finstrument-functions
11064 @opindex finstrument-functions
11065 Generate instrumentation calls for entry and exit to functions. Just
11066 after function entry and just before function exit, the following
11067 profiling functions are called with the address of the current
11068 function and its call site. (On some platforms,
11069 @code{__builtin_return_address} does not work beyond the current
11070 function, so the call site information may not be available to the
11071 profiling functions otherwise.)
11074 void __cyg_profile_func_enter (void *this_fn,
11076 void __cyg_profile_func_exit (void *this_fn,
11080 The first argument is the address of the start of the current function,
11081 which may be looked up exactly in the symbol table.
11083 This instrumentation is also done for functions expanded inline in other
11084 functions. The profiling calls indicate where, conceptually, the
11085 inline function is entered and exited. This means that addressable
11086 versions of such functions must be available. If all your uses of a
11087 function are expanded inline, this may mean an additional expansion of
11088 code size. If you use @code{extern inline} in your C code, an
11089 addressable version of such functions must be provided. (This is
11090 normally the case anyway, but if you get lucky and the optimizer always
11091 expands the functions inline, you might have gotten away without
11092 providing static copies.)
11094 A function may be given the attribute @code{no_instrument_function}, in
11095 which case this instrumentation is not done. This can be used, for
11096 example, for the profiling functions listed above, high-priority
11097 interrupt routines, and any functions from which the profiling functions
11098 cannot safely be called (perhaps signal handlers, if the profiling
11099 routines generate output or allocate memory).
11101 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11102 @opindex finstrument-functions-exclude-file-list
11104 Set the list of functions that are excluded from instrumentation (see
11105 the description of @option{-finstrument-functions}). If the file that
11106 contains a function definition matches with one of @var{file}, then
11107 that function is not instrumented. The match is done on substrings:
11108 if the @var{file} parameter is a substring of the file name, it is
11109 considered to be a match.
11114 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11118 excludes any inline function defined in files whose pathnames
11119 contain @file{/bits/stl} or @file{include/sys}.
11121 If, for some reason, you want to include letter @samp{,} in one of
11122 @var{sym}, write @samp{\,}. For example,
11123 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11124 (note the single quote surrounding the option).
11126 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11127 @opindex finstrument-functions-exclude-function-list
11129 This is similar to @option{-finstrument-functions-exclude-file-list},
11130 but this option sets the list of function names to be excluded from
11131 instrumentation. The function name to be matched is its user-visible
11132 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11133 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11134 match is done on substrings: if the @var{sym} parameter is a substring
11135 of the function name, it is considered to be a match. For C99 and C++
11136 extended identifiers, the function name must be given in UTF-8, not
11137 using universal character names.
11142 @node Preprocessor Options
11143 @section Options Controlling the Preprocessor
11144 @cindex preprocessor options
11145 @cindex options, preprocessor
11147 These options control the C preprocessor, which is run on each C source
11148 file before actual compilation.
11150 If you use the @option{-E} option, nothing is done except preprocessing.
11151 Some of these options make sense only together with @option{-E} because
11152 they cause the preprocessor output to be unsuitable for actual
11156 @item -Wp,@var{option}
11158 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11159 and pass @var{option} directly through to the preprocessor. If
11160 @var{option} contains commas, it is split into multiple options at the
11161 commas. However, many options are modified, translated or interpreted
11162 by the compiler driver before being passed to the preprocessor, and
11163 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11164 interface is undocumented and subject to change, so whenever possible
11165 you should avoid using @option{-Wp} and let the driver handle the
11168 @item -Xpreprocessor @var{option}
11169 @opindex Xpreprocessor
11170 Pass @var{option} as an option to the preprocessor. You can use this to
11171 supply system-specific preprocessor options that GCC does not
11174 If you want to pass an option that takes an argument, you must use
11175 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11177 @item -no-integrated-cpp
11178 @opindex no-integrated-cpp
11179 Perform preprocessing as a separate pass before compilation.
11180 By default, GCC performs preprocessing as an integrated part of
11181 input tokenization and parsing.
11182 If this option is provided, the appropriate language front end
11183 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11184 and Objective-C, respectively) is instead invoked twice,
11185 once for preprocessing only and once for actual compilation
11186 of the preprocessed input.
11187 This option may be useful in conjunction with the @option{-B} or
11188 @option{-wrapper} options to specify an alternate preprocessor or
11189 perform additional processing of the program source between
11190 normal preprocessing and compilation.
11193 @include cppopts.texi
11195 @node Assembler Options
11196 @section Passing Options to the Assembler
11198 @c prevent bad page break with this line
11199 You can pass options to the assembler.
11202 @item -Wa,@var{option}
11204 Pass @var{option} as an option to the assembler. If @var{option}
11205 contains commas, it is split into multiple options at the commas.
11207 @item -Xassembler @var{option}
11208 @opindex Xassembler
11209 Pass @var{option} as an option to the assembler. You can use this to
11210 supply system-specific assembler options that GCC does not
11213 If you want to pass an option that takes an argument, you must use
11214 @option{-Xassembler} twice, once for the option and once for the argument.
11219 @section Options for Linking
11220 @cindex link options
11221 @cindex options, linking
11223 These options come into play when the compiler links object files into
11224 an executable output file. They are meaningless if the compiler is
11225 not doing a link step.
11229 @item @var{object-file-name}
11230 A file name that does not end in a special recognized suffix is
11231 considered to name an object file or library. (Object files are
11232 distinguished from libraries by the linker according to the file
11233 contents.) If linking is done, these object files are used as input
11242 If any of these options is used, then the linker is not run, and
11243 object file names should not be used as arguments. @xref{Overall
11247 @opindex fuse-ld=bfd
11248 Use the @command{bfd} linker instead of the default linker.
11250 @item -fuse-ld=gold
11251 @opindex fuse-ld=gold
11252 Use the @command{gold} linker instead of the default linker.
11255 @item -l@var{library}
11256 @itemx -l @var{library}
11258 Search the library named @var{library} when linking. (The second
11259 alternative with the library as a separate argument is only for
11260 POSIX compliance and is not recommended.)
11262 It makes a difference where in the command you write this option; the
11263 linker searches and processes libraries and object files in the order they
11264 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11265 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11266 to functions in @samp{z}, those functions may not be loaded.
11268 The linker searches a standard list of directories for the library,
11269 which is actually a file named @file{lib@var{library}.a}. The linker
11270 then uses this file as if it had been specified precisely by name.
11272 The directories searched include several standard system directories
11273 plus any that you specify with @option{-L}.
11275 Normally the files found this way are library files---archive files
11276 whose members are object files. The linker handles an archive file by
11277 scanning through it for members which define symbols that have so far
11278 been referenced but not defined. But if the file that is found is an
11279 ordinary object file, it is linked in the usual fashion. The only
11280 difference between using an @option{-l} option and specifying a file name
11281 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11282 and searches several directories.
11286 You need this special case of the @option{-l} option in order to
11287 link an Objective-C or Objective-C++ program.
11289 @item -nostartfiles
11290 @opindex nostartfiles
11291 Do not use the standard system startup files when linking.
11292 The standard system libraries are used normally, unless @option{-nostdlib}
11293 or @option{-nodefaultlibs} is used.
11295 @item -nodefaultlibs
11296 @opindex nodefaultlibs
11297 Do not use the standard system libraries when linking.
11298 Only the libraries you specify are passed to the linker, and options
11299 specifying linkage of the system libraries, such as @option{-static-libgcc}
11300 or @option{-shared-libgcc}, are ignored.
11301 The standard startup files are used normally, unless @option{-nostartfiles}
11304 The compiler may generate calls to @code{memcmp},
11305 @code{memset}, @code{memcpy} and @code{memmove}.
11306 These entries are usually resolved by entries in
11307 libc. These entry points should be supplied through some other
11308 mechanism when this option is specified.
11312 Do not use the standard system startup files or libraries when linking.
11313 No startup files and only the libraries you specify are passed to
11314 the linker, and options specifying linkage of the system libraries, such as
11315 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11317 The compiler may generate calls to @code{memcmp}, @code{memset},
11318 @code{memcpy} and @code{memmove}.
11319 These entries are usually resolved by entries in
11320 libc. These entry points should be supplied through some other
11321 mechanism when this option is specified.
11323 @cindex @option{-lgcc}, use with @option{-nostdlib}
11324 @cindex @option{-nostdlib} and unresolved references
11325 @cindex unresolved references and @option{-nostdlib}
11326 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11327 @cindex @option{-nodefaultlibs} and unresolved references
11328 @cindex unresolved references and @option{-nodefaultlibs}
11329 One of the standard libraries bypassed by @option{-nostdlib} and
11330 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11331 which GCC uses to overcome shortcomings of particular machines, or special
11332 needs for some languages.
11333 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11334 Collection (GCC) Internals},
11335 for more discussion of @file{libgcc.a}.)
11336 In most cases, you need @file{libgcc.a} even when you want to avoid
11337 other standard libraries. In other words, when you specify @option{-nostdlib}
11338 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11339 This ensures that you have no unresolved references to internal GCC
11340 library subroutines.
11341 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11342 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11343 GNU Compiler Collection (GCC) Internals}.)
11347 Produce a position independent executable on targets that support it.
11348 For predictable results, you must also specify the same set of options
11349 used for compilation (@option{-fpie}, @option{-fPIE},
11350 or model suboptions) when you specify this linker option.
11354 Don't produce a position independent executable.
11358 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11359 that support it. This instructs the linker to add all symbols, not
11360 only used ones, to the dynamic symbol table. This option is needed
11361 for some uses of @code{dlopen} or to allow obtaining backtraces
11362 from within a program.
11366 Remove all symbol table and relocation information from the executable.
11370 On systems that support dynamic linking, this prevents linking with the shared
11371 libraries. On other systems, this option has no effect.
11375 Produce a shared object which can then be linked with other objects to
11376 form an executable. Not all systems support this option. For predictable
11377 results, you must also specify the same set of options used for compilation
11378 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11379 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11380 needs to build supplementary stub code for constructors to work. On
11381 multi-libbed systems, @samp{gcc -shared} must select the correct support
11382 libraries to link against. Failing to supply the correct flags may lead
11383 to subtle defects. Supplying them in cases where they are not necessary
11386 @item -shared-libgcc
11387 @itemx -static-libgcc
11388 @opindex shared-libgcc
11389 @opindex static-libgcc
11390 On systems that provide @file{libgcc} as a shared library, these options
11391 force the use of either the shared or static version, respectively.
11392 If no shared version of @file{libgcc} was built when the compiler was
11393 configured, these options have no effect.
11395 There are several situations in which an application should use the
11396 shared @file{libgcc} instead of the static version. The most common
11397 of these is when the application wishes to throw and catch exceptions
11398 across different shared libraries. In that case, each of the libraries
11399 as well as the application itself should use the shared @file{libgcc}.
11401 Therefore, the G++ and GCJ drivers automatically add
11402 @option{-shared-libgcc} whenever you build a shared library or a main
11403 executable, because C++ and Java programs typically use exceptions, so
11404 this is the right thing to do.
11406 If, instead, you use the GCC driver to create shared libraries, you may
11407 find that they are not always linked with the shared @file{libgcc}.
11408 If GCC finds, at its configuration time, that you have a non-GNU linker
11409 or a GNU linker that does not support option @option{--eh-frame-hdr},
11410 it links the shared version of @file{libgcc} into shared libraries
11411 by default. Otherwise, it takes advantage of the linker and optimizes
11412 away the linking with the shared version of @file{libgcc}, linking with
11413 the static version of libgcc by default. This allows exceptions to
11414 propagate through such shared libraries, without incurring relocation
11415 costs at library load time.
11417 However, if a library or main executable is supposed to throw or catch
11418 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11419 for the languages used in the program, or using the option
11420 @option{-shared-libgcc}, such that it is linked with the shared
11423 @item -static-libasan
11424 @opindex static-libasan
11425 When the @option{-fsanitize=address} option is used to link a program,
11426 the GCC driver automatically links against @option{libasan}. If
11427 @file{libasan} is available as a shared library, and the @option{-static}
11428 option is not used, then this links against the shared version of
11429 @file{libasan}. The @option{-static-libasan} option directs the GCC
11430 driver to link @file{libasan} statically, without necessarily linking
11431 other libraries statically.
11433 @item -static-libtsan
11434 @opindex static-libtsan
11435 When the @option{-fsanitize=thread} option is used to link a program,
11436 the GCC driver automatically links against @option{libtsan}. If
11437 @file{libtsan} is available as a shared library, and the @option{-static}
11438 option is not used, then this links against the shared version of
11439 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11440 driver to link @file{libtsan} statically, without necessarily linking
11441 other libraries statically.
11443 @item -static-liblsan
11444 @opindex static-liblsan
11445 When the @option{-fsanitize=leak} option is used to link a program,
11446 the GCC driver automatically links against @option{liblsan}. If
11447 @file{liblsan} is available as a shared library, and the @option{-static}
11448 option is not used, then this links against the shared version of
11449 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11450 driver to link @file{liblsan} statically, without necessarily linking
11451 other libraries statically.
11453 @item -static-libubsan
11454 @opindex static-libubsan
11455 When the @option{-fsanitize=undefined} option is used to link a program,
11456 the GCC driver automatically links against @option{libubsan}. If
11457 @file{libubsan} is available as a shared library, and the @option{-static}
11458 option is not used, then this links against the shared version of
11459 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11460 driver to link @file{libubsan} statically, without necessarily linking
11461 other libraries statically.
11463 @item -static-libmpx
11464 @opindex static-libmpx
11465 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11466 used to link a program, the GCC driver automatically links against
11467 @file{libmpx}. If @file{libmpx} is available as a shared library,
11468 and the @option{-static} option is not used, then this links against
11469 the shared version of @file{libmpx}. The @option{-static-libmpx}
11470 option directs the GCC driver to link @file{libmpx} statically,
11471 without necessarily linking other libraries statically.
11473 @item -static-libmpxwrappers
11474 @opindex static-libmpxwrappers
11475 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11476 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11477 GCC driver automatically links against @file{libmpxwrappers}. If
11478 @file{libmpxwrappers} is available as a shared library, and the
11479 @option{-static} option is not used, then this links against the shared
11480 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11481 option directs the GCC driver to link @file{libmpxwrappers} statically,
11482 without necessarily linking other libraries statically.
11484 @item -static-libstdc++
11485 @opindex static-libstdc++
11486 When the @command{g++} program is used to link a C++ program, it
11487 normally automatically links against @option{libstdc++}. If
11488 @file{libstdc++} is available as a shared library, and the
11489 @option{-static} option is not used, then this links against the
11490 shared version of @file{libstdc++}. That is normally fine. However, it
11491 is sometimes useful to freeze the version of @file{libstdc++} used by
11492 the program without going all the way to a fully static link. The
11493 @option{-static-libstdc++} option directs the @command{g++} driver to
11494 link @file{libstdc++} statically, without necessarily linking other
11495 libraries statically.
11499 Bind references to global symbols when building a shared object. Warn
11500 about any unresolved references (unless overridden by the link editor
11501 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11504 @item -T @var{script}
11506 @cindex linker script
11507 Use @var{script} as the linker script. This option is supported by most
11508 systems using the GNU linker. On some targets, such as bare-board
11509 targets without an operating system, the @option{-T} option may be required
11510 when linking to avoid references to undefined symbols.
11512 @item -Xlinker @var{option}
11514 Pass @var{option} as an option to the linker. You can use this to
11515 supply system-specific linker options that GCC does not recognize.
11517 If you want to pass an option that takes a separate argument, you must use
11518 @option{-Xlinker} twice, once for the option and once for the argument.
11519 For example, to pass @option{-assert definitions}, you must write
11520 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11521 @option{-Xlinker "-assert definitions"}, because this passes the entire
11522 string as a single argument, which is not what the linker expects.
11524 When using the GNU linker, it is usually more convenient to pass
11525 arguments to linker options using the @option{@var{option}=@var{value}}
11526 syntax than as separate arguments. For example, you can specify
11527 @option{-Xlinker -Map=output.map} rather than
11528 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11529 this syntax for command-line options.
11531 @item -Wl,@var{option}
11533 Pass @var{option} as an option to the linker. If @var{option} contains
11534 commas, it is split into multiple options at the commas. You can use this
11535 syntax to pass an argument to the option.
11536 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11537 linker. When using the GNU linker, you can also get the same effect with
11538 @option{-Wl,-Map=output.map}.
11540 @item -u @var{symbol}
11542 Pretend the symbol @var{symbol} is undefined, to force linking of
11543 library modules to define it. You can use @option{-u} multiple times with
11544 different symbols to force loading of additional library modules.
11546 @item -z @var{keyword}
11548 @option{-z} is passed directly on to the linker along with the keyword
11549 @var{keyword}. See the section in the documentation of your linker for
11550 permitted values and their meanings.
11553 @node Directory Options
11554 @section Options for Directory Search
11555 @cindex directory options
11556 @cindex options, directory search
11557 @cindex search path
11559 These options specify directories to search for header files, for
11560 libraries and for parts of the compiler:
11565 Add the directory @var{dir} to the head of the list of directories to be
11566 searched for header files. This can be used to override a system header
11567 file, substituting your own version, since these directories are
11568 searched before the system header file directories. However, you should
11569 not use this option to add directories that contain vendor-supplied
11570 system header files (use @option{-isystem} for that). If you use more than
11571 one @option{-I} option, the directories are scanned in left-to-right
11572 order; the standard system directories come after.
11574 If a standard system include directory, or a directory specified with
11575 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11576 option is ignored. The directory is still searched but as a
11577 system directory at its normal position in the system include chain.
11578 This is to ensure that GCC's procedure to fix buggy system headers and
11579 the ordering for the @code{include_next} directive are not inadvertently changed.
11580 If you really need to change the search order for system directories,
11581 use the @option{-nostdinc} and/or @option{-isystem} options.
11583 @item -iplugindir=@var{dir}
11584 @opindex iplugindir=
11585 Set the directory to search for plugins that are passed
11586 by @option{-fplugin=@var{name}} instead of
11587 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11588 to be used by the user, but only passed by the driver.
11590 @item -iquote@var{dir}
11592 Add the directory @var{dir} to the head of the list of directories to
11593 be searched for header files only for the case of @code{#include
11594 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11595 otherwise just like @option{-I}.
11599 Add directory @var{dir} to the list of directories to be searched
11602 @item -B@var{prefix}
11604 This option specifies where to find the executables, libraries,
11605 include files, and data files of the compiler itself.
11607 The compiler driver program runs one or more of the subprograms
11608 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11609 @var{prefix} as a prefix for each program it tries to run, both with and
11610 without @samp{@var{machine}/@var{version}/} for the corresponding target
11611 machine and compiler version.
11613 For each subprogram to be run, the compiler driver first tries the
11614 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11615 is not specified, the driver tries two standard prefixes,
11616 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11617 those results in a file name that is found, the unmodified program
11618 name is searched for using the directories specified in your
11619 @env{PATH} environment variable.
11621 The compiler checks to see if the path provided by @option{-B}
11622 refers to a directory, and if necessary it adds a directory
11623 separator character at the end of the path.
11625 @option{-B} prefixes that effectively specify directory names also apply
11626 to libraries in the linker, because the compiler translates these
11627 options into @option{-L} options for the linker. They also apply to
11628 include files in the preprocessor, because the compiler translates these
11629 options into @option{-isystem} options for the preprocessor. In this case,
11630 the compiler appends @samp{include} to the prefix.
11632 The runtime support file @file{libgcc.a} can also be searched for using
11633 the @option{-B} prefix, if needed. If it is not found there, the two
11634 standard prefixes above are tried, and that is all. The file is left
11635 out of the link if it is not found by those means.
11637 Another way to specify a prefix much like the @option{-B} prefix is to use
11638 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11641 As a special kludge, if the path provided by @option{-B} is
11642 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11643 9, then it is replaced by @file{[dir/]include}. This is to help
11644 with boot-strapping the compiler.
11646 @item -no-canonical-prefixes
11647 @opindex no-canonical-prefixes
11648 Do not expand any symbolic links, resolve references to @samp{/../}
11649 or @samp{/./}, or make the path absolute when generating a relative
11652 @item --sysroot=@var{dir}
11654 Use @var{dir} as the logical root directory for headers and libraries.
11655 For example, if the compiler normally searches for headers in
11656 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11657 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11659 If you use both this option and the @option{-isysroot} option, then
11660 the @option{--sysroot} option applies to libraries, but the
11661 @option{-isysroot} option applies to header files.
11663 The GNU linker (beginning with version 2.16) has the necessary support
11664 for this option. If your linker does not support this option, the
11665 header file aspect of @option{--sysroot} still works, but the
11666 library aspect does not.
11668 @item --no-sysroot-suffix
11669 @opindex no-sysroot-suffix
11670 For some targets, a suffix is added to the root directory specified
11671 with @option{--sysroot}, depending on the other options used, so that
11672 headers may for example be found in
11673 @file{@var{dir}/@var{suffix}/usr/include} instead of
11674 @file{@var{dir}/usr/include}. This option disables the addition of
11679 This option has been deprecated. Please use @option{-iquote} instead for
11680 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11682 Any directories you specify with @option{-I} options before the @option{-I-}
11683 option are searched only for the case of @code{#include "@var{file}"};
11684 they are not searched for @code{#include <@var{file}>}.
11686 If additional directories are specified with @option{-I} options after
11687 the @option{-I-} option, these directories are searched for all @code{#include}
11688 directives. (Ordinarily @emph{all} @option{-I} directories are used
11691 In addition, the @option{-I-} option inhibits the use of the current
11692 directory (where the current input file came from) as the first search
11693 directory for @code{#include "@var{file}"}. There is no way to
11694 override this effect of @option{-I-}. With @option{-I.} you can specify
11695 searching the directory that is current when the compiler is
11696 invoked. That is not exactly the same as what the preprocessor does
11697 by default, but it is often satisfactory.
11699 @option{-I-} does not inhibit the use of the standard system directories
11700 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11704 @node Code Gen Options
11705 @section Options for Code Generation Conventions
11706 @cindex code generation conventions
11707 @cindex options, code generation
11708 @cindex run-time options
11710 These machine-independent options control the interface conventions
11711 used in code generation.
11713 Most of them have both positive and negative forms; the negative form
11714 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11715 one of the forms is listed---the one that is not the default. You
11716 can figure out the other form by either removing @samp{no-} or adding
11720 @item -fstack-reuse=@var{reuse-level}
11721 @opindex fstack_reuse
11722 This option controls stack space reuse for user declared local/auto variables
11723 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11724 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11725 local variables and temporaries, @samp{named_vars} enables the reuse only for
11726 user defined local variables with names, and @samp{none} disables stack reuse
11727 completely. The default value is @samp{all}. The option is needed when the
11728 program extends the lifetime of a scoped local variable or a compiler generated
11729 temporary beyond the end point defined by the language. When a lifetime of
11730 a variable ends, and if the variable lives in memory, the optimizing compiler
11731 has the freedom to reuse its stack space with other temporaries or scoped
11732 local variables whose live range does not overlap with it. Legacy code extending
11733 local lifetime is likely to break with the stack reuse optimization.
11752 if (*p == 10) // out of scope use of local1
11763 A(int k) : i(k), j(k) @{ @}
11770 void foo(const A& ar)
11777 foo(A(10)); // temp object's lifetime ends when foo returns
11783 ap->i+= 10; // ap references out of scope temp whose space
11784 // is reused with a. What is the value of ap->i?
11789 The lifetime of a compiler generated temporary is well defined by the C++
11790 standard. When a lifetime of a temporary ends, and if the temporary lives
11791 in memory, the optimizing compiler has the freedom to reuse its stack
11792 space with other temporaries or scoped local variables whose live range
11793 does not overlap with it. However some of the legacy code relies on
11794 the behavior of older compilers in which temporaries' stack space is
11795 not reused, the aggressive stack reuse can lead to runtime errors. This
11796 option is used to control the temporary stack reuse optimization.
11800 This option generates traps for signed overflow on addition, subtraction,
11801 multiplication operations.
11802 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11803 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11804 @option{-fwrapv} being effective. Note that only active options override, so
11805 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11806 results in @option{-ftrapv} being effective.
11810 This option instructs the compiler to assume that signed arithmetic
11811 overflow of addition, subtraction and multiplication wraps around
11812 using twos-complement representation. This flag enables some optimizations
11813 and disables others. This option is enabled by default for the Java
11814 front end, as required by the Java language specification.
11815 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11816 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11817 @option{-fwrapv} being effective. Note that only active options override, so
11818 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11819 results in @option{-ftrapv} being effective.
11822 @opindex fexceptions
11823 Enable exception handling. Generates extra code needed to propagate
11824 exceptions. For some targets, this implies GCC generates frame
11825 unwind information for all functions, which can produce significant data
11826 size overhead, although it does not affect execution. If you do not
11827 specify this option, GCC enables it by default for languages like
11828 C++ that normally require exception handling, and disables it for
11829 languages like C that do not normally require it. However, you may need
11830 to enable this option when compiling C code that needs to interoperate
11831 properly with exception handlers written in C++. You may also wish to
11832 disable this option if you are compiling older C++ programs that don't
11833 use exception handling.
11835 @item -fnon-call-exceptions
11836 @opindex fnon-call-exceptions
11837 Generate code that allows trapping instructions to throw exceptions.
11838 Note that this requires platform-specific runtime support that does
11839 not exist everywhere. Moreover, it only allows @emph{trapping}
11840 instructions to throw exceptions, i.e.@: memory references or floating-point
11841 instructions. It does not allow exceptions to be thrown from
11842 arbitrary signal handlers such as @code{SIGALRM}.
11844 @item -fdelete-dead-exceptions
11845 @opindex fdelete-dead-exceptions
11846 Consider that instructions that may throw exceptions but don't otherwise
11847 contribute to the execution of the program can be optimized away.
11848 This option is enabled by default for the Ada front end, as permitted by
11849 the Ada language specification.
11850 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
11852 @item -funwind-tables
11853 @opindex funwind-tables
11854 Similar to @option{-fexceptions}, except that it just generates any needed
11855 static data, but does not affect the generated code in any other way.
11856 You normally do not need to enable this option; instead, a language processor
11857 that needs this handling enables it on your behalf.
11859 @item -fasynchronous-unwind-tables
11860 @opindex fasynchronous-unwind-tables
11861 Generate unwind table in DWARF format, if supported by target machine. The
11862 table is exact at each instruction boundary, so it can be used for stack
11863 unwinding from asynchronous events (such as debugger or garbage collector).
11865 @item -fno-gnu-unique
11866 @opindex fno-gnu-unique
11867 On systems with recent GNU assembler and C library, the C++ compiler
11868 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
11869 of template static data members and static local variables in inline
11870 functions are unique even in the presence of @code{RTLD_LOCAL}; this
11871 is necessary to avoid problems with a library used by two different
11872 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
11873 therefore disagreeing with the other one about the binding of the
11874 symbol. But this causes @code{dlclose} to be ignored for affected
11875 DSOs; if your program relies on reinitialization of a DSO via
11876 @code{dlclose} and @code{dlopen}, you can use
11877 @option{-fno-gnu-unique}.
11879 @item -fpcc-struct-return
11880 @opindex fpcc-struct-return
11881 Return ``short'' @code{struct} and @code{union} values in memory like
11882 longer ones, rather than in registers. This convention is less
11883 efficient, but it has the advantage of allowing intercallability between
11884 GCC-compiled files and files compiled with other compilers, particularly
11885 the Portable C Compiler (pcc).
11887 The precise convention for returning structures in memory depends
11888 on the target configuration macros.
11890 Short structures and unions are those whose size and alignment match
11891 that of some integer type.
11893 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11894 switch is not binary compatible with code compiled with the
11895 @option{-freg-struct-return} switch.
11896 Use it to conform to a non-default application binary interface.
11898 @item -freg-struct-return
11899 @opindex freg-struct-return
11900 Return @code{struct} and @code{union} values in registers when possible.
11901 This is more efficient for small structures than
11902 @option{-fpcc-struct-return}.
11904 If you specify neither @option{-fpcc-struct-return} nor
11905 @option{-freg-struct-return}, GCC defaults to whichever convention is
11906 standard for the target. If there is no standard convention, GCC
11907 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11908 the principal compiler. In those cases, we can choose the standard, and
11909 we chose the more efficient register return alternative.
11911 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11912 switch is not binary compatible with code compiled with the
11913 @option{-fpcc-struct-return} switch.
11914 Use it to conform to a non-default application binary interface.
11916 @item -fshort-enums
11917 @opindex fshort-enums
11918 Allocate to an @code{enum} type only as many bytes as it needs for the
11919 declared range of possible values. Specifically, the @code{enum} type
11920 is equivalent to the smallest integer type that has enough room.
11922 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11923 code that is not binary compatible with code generated without that switch.
11924 Use it to conform to a non-default application binary interface.
11926 @item -fshort-wchar
11927 @opindex fshort-wchar
11928 Override the underlying type for @code{wchar_t} to be @code{short
11929 unsigned int} instead of the default for the target. This option is
11930 useful for building programs to run under WINE@.
11932 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11933 code that is not binary compatible with code generated without that switch.
11934 Use it to conform to a non-default application binary interface.
11937 @opindex fno-common
11938 In C code, controls the placement of uninitialized global variables.
11939 Unix C compilers have traditionally permitted multiple definitions of
11940 such variables in different compilation units by placing the variables
11942 This is the behavior specified by @option{-fcommon}, and is the default
11943 for GCC on most targets.
11944 On the other hand, this behavior is not required by ISO C, and on some
11945 targets may carry a speed or code size penalty on variable references.
11946 The @option{-fno-common} option specifies that the compiler should place
11947 uninitialized global variables in the data section of the object file,
11948 rather than generating them as common blocks.
11949 This has the effect that if the same variable is declared
11950 (without @code{extern}) in two different compilations,
11951 you get a multiple-definition error when you link them.
11952 In this case, you must compile with @option{-fcommon} instead.
11953 Compiling with @option{-fno-common} is useful on targets for which
11954 it provides better performance, or if you wish to verify that the
11955 program will work on other systems that always treat uninitialized
11956 variable declarations this way.
11960 Ignore the @code{#ident} directive.
11962 @item -finhibit-size-directive
11963 @opindex finhibit-size-directive
11964 Don't output a @code{.size} assembler directive, or anything else that
11965 would cause trouble if the function is split in the middle, and the
11966 two halves are placed at locations far apart in memory. This option is
11967 used when compiling @file{crtstuff.c}; you should not need to use it
11970 @item -fverbose-asm
11971 @opindex fverbose-asm
11972 Put extra commentary information in the generated assembly code to
11973 make it more readable. This option is generally only of use to those
11974 who actually need to read the generated assembly code (perhaps while
11975 debugging the compiler itself).
11977 @option{-fno-verbose-asm}, the default, causes the
11978 extra information to be omitted and is useful when comparing two assembler
11981 The added comments include:
11986 information on the compiler version and command-line options,
11989 the source code lines associated with the assembly instructions,
11990 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
11993 hints on which high-level expressions correspond to
11994 the various assembly instruction operands.
11998 For example, given this C source file:
12006 for (i = 0; i < n; i++)
12013 compiling to (x86_64) assembly via @option{-S} and emitting the result
12014 direct to stdout via @option{-o} @option{-}
12017 gcc -S test.c -fverbose-asm -Os -o -
12020 gives output similar to this:
12024 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12031 .type test, @@function
12035 # test.c:4: int total = 0;
12036 xorl %eax, %eax # <retval>
12037 # test.c:6: for (i = 0; i < n; i++)
12038 xorl %edx, %edx # i
12040 # test.c:6: for (i = 0; i < n; i++)
12041 cmpl %edi, %edx # n, i
12043 # test.c:7: total += i * i;
12044 movl %edx, %ecx # i, tmp92
12045 imull %edx, %ecx # i, tmp92
12046 # test.c:6: for (i = 0; i < n; i++)
12048 # test.c:7: total += i * i;
12049 addl %ecx, %eax # tmp92, <retval>
12057 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12058 .section .note.GNU-stack,"",@@progbits
12061 The comments are intended for humans rather than machines and hence the
12062 precise format of the comments is subject to change.
12064 @item -frecord-gcc-switches
12065 @opindex frecord-gcc-switches
12066 This switch causes the command line used to invoke the
12067 compiler to be recorded into the object file that is being created.
12068 This switch is only implemented on some targets and the exact format
12069 of the recording is target and binary file format dependent, but it
12070 usually takes the form of a section containing ASCII text. This
12071 switch is related to the @option{-fverbose-asm} switch, but that
12072 switch only records information in the assembler output file as
12073 comments, so it never reaches the object file.
12074 See also @option{-grecord-gcc-switches} for another
12075 way of storing compiler options into the object file.
12079 @cindex global offset table
12081 Generate position-independent code (PIC) suitable for use in a shared
12082 library, if supported for the target machine. Such code accesses all
12083 constant addresses through a global offset table (GOT)@. The dynamic
12084 loader resolves the GOT entries when the program starts (the dynamic
12085 loader is not part of GCC; it is part of the operating system). If
12086 the GOT size for the linked executable exceeds a machine-specific
12087 maximum size, you get an error message from the linker indicating that
12088 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12089 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12090 on the m68k and RS/6000. The x86 has no such limit.)
12092 Position-independent code requires special support, and therefore works
12093 only on certain machines. For the x86, GCC supports PIC for System V
12094 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12095 position-independent.
12097 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12102 If supported for the target machine, emit position-independent code,
12103 suitable for dynamic linking and avoiding any limit on the size of the
12104 global offset table. This option makes a difference on AArch64, m68k,
12105 PowerPC and SPARC@.
12107 Position-independent code requires special support, and therefore works
12108 only on certain machines.
12110 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12117 These options are similar to @option{-fpic} and @option{-fPIC}, but
12118 generated position independent code can be only linked into executables.
12119 Usually these options are used when @option{-pie} GCC option is
12120 used during linking.
12122 @option{-fpie} and @option{-fPIE} both define the macros
12123 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12124 for @option{-fpie} and 2 for @option{-fPIE}.
12128 Do not use the PLT for external function calls in position-independent code.
12129 Instead, load the callee address at call sites from the GOT and branch to it.
12130 This leads to more efficient code by eliminating PLT stubs and exposing
12131 GOT loads to optimizations. On architectures such as 32-bit x86 where
12132 PLT stubs expect the GOT pointer in a specific register, this gives more
12133 register allocation freedom to the compiler.
12134 Lazy binding requires use of the PLT;
12135 with @option{-fno-plt} all external symbols are resolved at load time.
12137 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12138 through the PLT for specific external functions.
12140 In position-dependent code, a few targets also convert calls to
12141 functions that are marked to not use the PLT to use the GOT instead.
12143 @item -fno-jump-tables
12144 @opindex fno-jump-tables
12145 Do not use jump tables for switch statements even where it would be
12146 more efficient than other code generation strategies. This option is
12147 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12148 building code that forms part of a dynamic linker and cannot
12149 reference the address of a jump table. On some targets, jump tables
12150 do not require a GOT and this option is not needed.
12152 @item -ffixed-@var{reg}
12154 Treat the register named @var{reg} as a fixed register; generated code
12155 should never refer to it (except perhaps as a stack pointer, frame
12156 pointer or in some other fixed role).
12158 @var{reg} must be the name of a register. The register names accepted
12159 are machine-specific and are defined in the @code{REGISTER_NAMES}
12160 macro in the machine description macro file.
12162 This flag does not have a negative form, because it specifies a
12165 @item -fcall-used-@var{reg}
12166 @opindex fcall-used
12167 Treat the register named @var{reg} as an allocable register that is
12168 clobbered by function calls. It may be allocated for temporaries or
12169 variables that do not live across a call. Functions compiled this way
12170 do not save and restore the register @var{reg}.
12172 It is an error to use this flag with the frame pointer or stack pointer.
12173 Use of this flag for other registers that have fixed pervasive roles in
12174 the machine's execution model produces disastrous results.
12176 This flag does not have a negative form, because it specifies a
12179 @item -fcall-saved-@var{reg}
12180 @opindex fcall-saved
12181 Treat the register named @var{reg} as an allocable register saved by
12182 functions. It may be allocated even for temporaries or variables that
12183 live across a call. Functions compiled this way save and restore
12184 the register @var{reg} if they use it.
12186 It is an error to use this flag with the frame pointer or stack pointer.
12187 Use of this flag for other registers that have fixed pervasive roles in
12188 the machine's execution model produces disastrous results.
12190 A different sort of disaster results from the use of this flag for
12191 a register in which function values may be returned.
12193 This flag does not have a negative form, because it specifies a
12196 @item -fpack-struct[=@var{n}]
12197 @opindex fpack-struct
12198 Without a value specified, pack all structure members together without
12199 holes. When a value is specified (which must be a small power of two), pack
12200 structure members according to this value, representing the maximum
12201 alignment (that is, objects with default alignment requirements larger than
12202 this are output potentially unaligned at the next fitting location.
12204 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12205 code that is not binary compatible with code generated without that switch.
12206 Additionally, it makes the code suboptimal.
12207 Use it to conform to a non-default application binary interface.
12209 @item -fleading-underscore
12210 @opindex fleading-underscore
12211 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12212 change the way C symbols are represented in the object file. One use
12213 is to help link with legacy assembly code.
12215 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12216 generate code that is not binary compatible with code generated without that
12217 switch. Use it to conform to a non-default application binary interface.
12218 Not all targets provide complete support for this switch.
12220 @item -ftls-model=@var{model}
12221 @opindex ftls-model
12222 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12223 The @var{model} argument should be one of @samp{global-dynamic},
12224 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12225 Note that the choice is subject to optimization: the compiler may use
12226 a more efficient model for symbols not visible outside of the translation
12227 unit, or if @option{-fpic} is not given on the command line.
12229 The default without @option{-fpic} is @samp{initial-exec}; with
12230 @option{-fpic} the default is @samp{global-dynamic}.
12232 @item -ftrampolines
12233 @opindex ftrampolines
12234 For targets that normally need trampolines for nested functions, always
12235 generate them instead of using descriptors. Otherwise, for targets that
12236 do not need them, like for example HP-PA or IA-64, do nothing.
12238 A trampoline is a small piece of code that is created at run time on the
12239 stack when the address of a nested function is taken, and is used to call
12240 the nested function indirectly. Therefore, it requires the stack to be
12241 made executable in order for the program to work properly.
12243 @option{-fno-trampolines} is enabled by default on a language by language
12244 basis to let the compiler avoid generating them, if it computes that this
12245 is safe, and replace them with descriptors. Descriptors are made up of data
12246 only, but the generated code must be prepared to deal with them. As of this
12247 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12249 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12250 @option{-fno-trampolines} are not binary compatible if nested functions are
12251 present. This option must therefore be used on a program-wide basis and be
12252 manipulated with extreme care.
12254 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12255 @opindex fvisibility
12256 Set the default ELF image symbol visibility to the specified option---all
12257 symbols are marked with this unless overridden within the code.
12258 Using this feature can very substantially improve linking and
12259 load times of shared object libraries, produce more optimized
12260 code, provide near-perfect API export and prevent symbol clashes.
12261 It is @strong{strongly} recommended that you use this in any shared objects
12264 Despite the nomenclature, @samp{default} always means public; i.e.,
12265 available to be linked against from outside the shared object.
12266 @samp{protected} and @samp{internal} are pretty useless in real-world
12267 usage so the only other commonly used option is @samp{hidden}.
12268 The default if @option{-fvisibility} isn't specified is
12269 @samp{default}, i.e., make every symbol public.
12271 A good explanation of the benefits offered by ensuring ELF
12272 symbols have the correct visibility is given by ``How To Write
12273 Shared Libraries'' by Ulrich Drepper (which can be found at
12274 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
12275 solution made possible by this option to marking things hidden when
12276 the default is public is to make the default hidden and mark things
12277 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12278 and @code{__attribute__ ((visibility("default")))} instead of
12279 @code{__declspec(dllexport)} you get almost identical semantics with
12280 identical syntax. This is a great boon to those working with
12281 cross-platform projects.
12283 For those adding visibility support to existing code, you may find
12284 @code{#pragma GCC visibility} of use. This works by you enclosing
12285 the declarations you wish to set visibility for with (for example)
12286 @code{#pragma GCC visibility push(hidden)} and
12287 @code{#pragma GCC visibility pop}.
12288 Bear in mind that symbol visibility should be viewed @strong{as
12289 part of the API interface contract} and thus all new code should
12290 always specify visibility when it is not the default; i.e., declarations
12291 only for use within the local DSO should @strong{always} be marked explicitly
12292 as hidden as so to avoid PLT indirection overheads---making this
12293 abundantly clear also aids readability and self-documentation of the code.
12294 Note that due to ISO C++ specification requirements, @code{operator new} and
12295 @code{operator delete} must always be of default visibility.
12297 Be aware that headers from outside your project, in particular system
12298 headers and headers from any other library you use, may not be
12299 expecting to be compiled with visibility other than the default. You
12300 may need to explicitly say @code{#pragma GCC visibility push(default)}
12301 before including any such headers.
12303 @code{extern} declarations are not affected by @option{-fvisibility}, so
12304 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12305 no modifications. However, this means that calls to @code{extern}
12306 functions with no explicit visibility use the PLT, so it is more
12307 effective to use @code{__attribute ((visibility))} and/or
12308 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12309 declarations should be treated as hidden.
12311 Note that @option{-fvisibility} does affect C++ vague linkage
12312 entities. This means that, for instance, an exception class that is
12313 be thrown between DSOs must be explicitly marked with default
12314 visibility so that the @samp{type_info} nodes are unified between
12317 An overview of these techniques, their benefits and how to use them
12318 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12320 @item -fstrict-volatile-bitfields
12321 @opindex fstrict-volatile-bitfields
12322 This option should be used if accesses to volatile bit-fields (or other
12323 structure fields, although the compiler usually honors those types
12324 anyway) should use a single access of the width of the
12325 field's type, aligned to a natural alignment if possible. For
12326 example, targets with memory-mapped peripheral registers might require
12327 all such accesses to be 16 bits wide; with this flag you can
12328 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12329 is 16 bits on these targets) to force GCC to use 16-bit accesses
12330 instead of, perhaps, a more efficient 32-bit access.
12332 If this option is disabled, the compiler uses the most efficient
12333 instruction. In the previous example, that might be a 32-bit load
12334 instruction, even though that accesses bytes that do not contain
12335 any portion of the bit-field, or memory-mapped registers unrelated to
12336 the one being updated.
12338 In some cases, such as when the @code{packed} attribute is applied to a
12339 structure field, it may not be possible to access the field with a single
12340 read or write that is correctly aligned for the target machine. In this
12341 case GCC falls back to generating multiple accesses rather than code that
12342 will fault or truncate the result at run time.
12344 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12345 not allowed to touch non bit-field members. It is therefore recommended
12346 to define all bits of the field's type as bit-field members.
12348 The default value of this option is determined by the application binary
12349 interface for the target processor.
12351 @item -fsync-libcalls
12352 @opindex fsync-libcalls
12353 This option controls whether any out-of-line instance of the @code{__sync}
12354 family of functions may be used to implement the C++11 @code{__atomic}
12355 family of functions.
12357 The default value of this option is enabled, thus the only useful form
12358 of the option is @option{-fno-sync-libcalls}. This option is used in
12359 the implementation of the @file{libatomic} runtime library.
12363 @node Developer Options
12364 @section GCC Developer Options
12365 @cindex developer options
12366 @cindex debugging GCC
12367 @cindex debug dump options
12368 @cindex dump options
12369 @cindex compilation statistics
12371 This section describes command-line options that are primarily of
12372 interest to GCC developers, including options to support compiler
12373 testing and investigation of compiler bugs and compile-time
12374 performance problems. This includes options that produce debug dumps
12375 at various points in the compilation; that print statistics such as
12376 memory use and execution time; and that print information about GCC's
12377 configuration, such as where it searches for libraries. You should
12378 rarely need to use any of these options for ordinary compilation and
12383 @item -d@var{letters}
12384 @itemx -fdump-rtl-@var{pass}
12385 @itemx -fdump-rtl-@var{pass}=@var{filename}
12387 @opindex fdump-rtl-@var{pass}
12388 Says to make debugging dumps during compilation at times specified by
12389 @var{letters}. This is used for debugging the RTL-based passes of the
12390 compiler. The file names for most of the dumps are made by appending
12391 a pass number and a word to the @var{dumpname}, and the files are
12392 created in the directory of the output file. In case of
12393 @option{=@var{filename}} option, the dump is output on the given file
12394 instead of the pass numbered dump files. Note that the pass number is
12395 assigned as passes are registered into the pass manager. Most passes
12396 are registered in the order that they will execute and for these passes
12397 the number corresponds to the pass execution order. However, passes
12398 registered by plugins, passes specific to compilation targets, or
12399 passes that are otherwise registered after all the other passes are
12400 numbered higher than a pass named "final", even if they are executed
12401 earlier. @var{dumpname} is generated from the name of the output
12402 file if explicitly specified and not an executable, otherwise it is
12403 the basename of the source file. These switches may have different
12404 effects when @option{-E} is used for preprocessing.
12406 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12407 @option{-d} option @var{letters}. Here are the possible
12408 letters for use in @var{pass} and @var{letters}, and their meanings:
12412 @item -fdump-rtl-alignments
12413 @opindex fdump-rtl-alignments
12414 Dump after branch alignments have been computed.
12416 @item -fdump-rtl-asmcons
12417 @opindex fdump-rtl-asmcons
12418 Dump after fixing rtl statements that have unsatisfied in/out constraints.
12420 @item -fdump-rtl-auto_inc_dec
12421 @opindex fdump-rtl-auto_inc_dec
12422 Dump after auto-inc-dec discovery. This pass is only run on
12423 architectures that have auto inc or auto dec instructions.
12425 @item -fdump-rtl-barriers
12426 @opindex fdump-rtl-barriers
12427 Dump after cleaning up the barrier instructions.
12429 @item -fdump-rtl-bbpart
12430 @opindex fdump-rtl-bbpart
12431 Dump after partitioning hot and cold basic blocks.
12433 @item -fdump-rtl-bbro
12434 @opindex fdump-rtl-bbro
12435 Dump after block reordering.
12437 @item -fdump-rtl-btl1
12438 @itemx -fdump-rtl-btl2
12439 @opindex fdump-rtl-btl2
12440 @opindex fdump-rtl-btl2
12441 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
12442 after the two branch
12443 target load optimization passes.
12445 @item -fdump-rtl-bypass
12446 @opindex fdump-rtl-bypass
12447 Dump after jump bypassing and control flow optimizations.
12449 @item -fdump-rtl-combine
12450 @opindex fdump-rtl-combine
12451 Dump after the RTL instruction combination pass.
12453 @item -fdump-rtl-compgotos
12454 @opindex fdump-rtl-compgotos
12455 Dump after duplicating the computed gotos.
12457 @item -fdump-rtl-ce1
12458 @itemx -fdump-rtl-ce2
12459 @itemx -fdump-rtl-ce3
12460 @opindex fdump-rtl-ce1
12461 @opindex fdump-rtl-ce2
12462 @opindex fdump-rtl-ce3
12463 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
12464 @option{-fdump-rtl-ce3} enable dumping after the three
12465 if conversion passes.
12467 @item -fdump-rtl-cprop_hardreg
12468 @opindex fdump-rtl-cprop_hardreg
12469 Dump after hard register copy propagation.
12471 @item -fdump-rtl-csa
12472 @opindex fdump-rtl-csa
12473 Dump after combining stack adjustments.
12475 @item -fdump-rtl-cse1
12476 @itemx -fdump-rtl-cse2
12477 @opindex fdump-rtl-cse1
12478 @opindex fdump-rtl-cse2
12479 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
12480 the two common subexpression elimination passes.
12482 @item -fdump-rtl-dce
12483 @opindex fdump-rtl-dce
12484 Dump after the standalone dead code elimination passes.
12486 @item -fdump-rtl-dbr
12487 @opindex fdump-rtl-dbr
12488 Dump after delayed branch scheduling.
12490 @item -fdump-rtl-dce1
12491 @itemx -fdump-rtl-dce2
12492 @opindex fdump-rtl-dce1
12493 @opindex fdump-rtl-dce2
12494 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
12495 the two dead store elimination passes.
12497 @item -fdump-rtl-eh
12498 @opindex fdump-rtl-eh
12499 Dump after finalization of EH handling code.
12501 @item -fdump-rtl-eh_ranges
12502 @opindex fdump-rtl-eh_ranges
12503 Dump after conversion of EH handling range regions.
12505 @item -fdump-rtl-expand
12506 @opindex fdump-rtl-expand
12507 Dump after RTL generation.
12509 @item -fdump-rtl-fwprop1
12510 @itemx -fdump-rtl-fwprop2
12511 @opindex fdump-rtl-fwprop1
12512 @opindex fdump-rtl-fwprop2
12513 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
12514 dumping after the two forward propagation passes.
12516 @item -fdump-rtl-gcse1
12517 @itemx -fdump-rtl-gcse2
12518 @opindex fdump-rtl-gcse1
12519 @opindex fdump-rtl-gcse2
12520 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
12521 after global common subexpression elimination.
12523 @item -fdump-rtl-init-regs
12524 @opindex fdump-rtl-init-regs
12525 Dump after the initialization of the registers.
12527 @item -fdump-rtl-initvals
12528 @opindex fdump-rtl-initvals
12529 Dump after the computation of the initial value sets.
12531 @item -fdump-rtl-into_cfglayout
12532 @opindex fdump-rtl-into_cfglayout
12533 Dump after converting to cfglayout mode.
12535 @item -fdump-rtl-ira
12536 @opindex fdump-rtl-ira
12537 Dump after iterated register allocation.
12539 @item -fdump-rtl-jump
12540 @opindex fdump-rtl-jump
12541 Dump after the second jump optimization.
12543 @item -fdump-rtl-loop2
12544 @opindex fdump-rtl-loop2
12545 @option{-fdump-rtl-loop2} enables dumping after the rtl
12546 loop optimization passes.
12548 @item -fdump-rtl-mach
12549 @opindex fdump-rtl-mach
12550 Dump after performing the machine dependent reorganization pass, if that
12553 @item -fdump-rtl-mode_sw
12554 @opindex fdump-rtl-mode_sw
12555 Dump after removing redundant mode switches.
12557 @item -fdump-rtl-rnreg
12558 @opindex fdump-rtl-rnreg
12559 Dump after register renumbering.
12561 @item -fdump-rtl-outof_cfglayout
12562 @opindex fdump-rtl-outof_cfglayout
12563 Dump after converting from cfglayout mode.
12565 @item -fdump-rtl-peephole2
12566 @opindex fdump-rtl-peephole2
12567 Dump after the peephole pass.
12569 @item -fdump-rtl-postreload
12570 @opindex fdump-rtl-postreload
12571 Dump after post-reload optimizations.
12573 @item -fdump-rtl-pro_and_epilogue
12574 @opindex fdump-rtl-pro_and_epilogue
12575 Dump after generating the function prologues and epilogues.
12577 @item -fdump-rtl-sched1
12578 @itemx -fdump-rtl-sched2
12579 @opindex fdump-rtl-sched1
12580 @opindex fdump-rtl-sched2
12581 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
12582 after the basic block scheduling passes.
12584 @item -fdump-rtl-ree
12585 @opindex fdump-rtl-ree
12586 Dump after sign/zero extension elimination.
12588 @item -fdump-rtl-seqabstr
12589 @opindex fdump-rtl-seqabstr
12590 Dump after common sequence discovery.
12592 @item -fdump-rtl-shorten
12593 @opindex fdump-rtl-shorten
12594 Dump after shortening branches.
12596 @item -fdump-rtl-sibling
12597 @opindex fdump-rtl-sibling
12598 Dump after sibling call optimizations.
12600 @item -fdump-rtl-split1
12601 @itemx -fdump-rtl-split2
12602 @itemx -fdump-rtl-split3
12603 @itemx -fdump-rtl-split4
12604 @itemx -fdump-rtl-split5
12605 @opindex fdump-rtl-split1
12606 @opindex fdump-rtl-split2
12607 @opindex fdump-rtl-split3
12608 @opindex fdump-rtl-split4
12609 @opindex fdump-rtl-split5
12610 These options enable dumping after five rounds of
12611 instruction splitting.
12613 @item -fdump-rtl-sms
12614 @opindex fdump-rtl-sms
12615 Dump after modulo scheduling. This pass is only run on some
12618 @item -fdump-rtl-stack
12619 @opindex fdump-rtl-stack
12620 Dump after conversion from GCC's ``flat register file'' registers to the
12621 x87's stack-like registers. This pass is only run on x86 variants.
12623 @item -fdump-rtl-subreg1
12624 @itemx -fdump-rtl-subreg2
12625 @opindex fdump-rtl-subreg1
12626 @opindex fdump-rtl-subreg2
12627 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
12628 the two subreg expansion passes.
12630 @item -fdump-rtl-unshare
12631 @opindex fdump-rtl-unshare
12632 Dump after all rtl has been unshared.
12634 @item -fdump-rtl-vartrack
12635 @opindex fdump-rtl-vartrack
12636 Dump after variable tracking.
12638 @item -fdump-rtl-vregs
12639 @opindex fdump-rtl-vregs
12640 Dump after converting virtual registers to hard registers.
12642 @item -fdump-rtl-web
12643 @opindex fdump-rtl-web
12644 Dump after live range splitting.
12646 @item -fdump-rtl-regclass
12647 @itemx -fdump-rtl-subregs_of_mode_init
12648 @itemx -fdump-rtl-subregs_of_mode_finish
12649 @itemx -fdump-rtl-dfinit
12650 @itemx -fdump-rtl-dfinish
12651 @opindex fdump-rtl-regclass
12652 @opindex fdump-rtl-subregs_of_mode_init
12653 @opindex fdump-rtl-subregs_of_mode_finish
12654 @opindex fdump-rtl-dfinit
12655 @opindex fdump-rtl-dfinish
12656 These dumps are defined but always produce empty files.
12659 @itemx -fdump-rtl-all
12661 @opindex fdump-rtl-all
12662 Produce all the dumps listed above.
12666 Annotate the assembler output with miscellaneous debugging information.
12670 Dump all macro definitions, at the end of preprocessing, in addition to
12675 Produce a core dump whenever an error occurs.
12679 Annotate the assembler output with a comment indicating which
12680 pattern and alternative is used. The length of each instruction is
12685 Dump the RTL in the assembler output as a comment before each instruction.
12686 Also turns on @option{-dp} annotation.
12690 Just generate RTL for a function instead of compiling it. Usually used
12691 with @option{-fdump-rtl-expand}.
12694 @item -fdump-noaddr
12695 @opindex fdump-noaddr
12696 When doing debugging dumps, suppress address output. This makes it more
12697 feasible to use diff on debugging dumps for compiler invocations with
12698 different compiler binaries and/or different
12699 text / bss / data / heap / stack / dso start locations.
12702 @opindex freport-bug
12703 Collect and dump debug information into a temporary file if an
12704 internal compiler error (ICE) occurs.
12706 @item -fdump-unnumbered
12707 @opindex fdump-unnumbered
12708 When doing debugging dumps, suppress instruction numbers and address output.
12709 This makes it more feasible to use diff on debugging dumps for compiler
12710 invocations with different options, in particular with and without
12713 @item -fdump-unnumbered-links
12714 @opindex fdump-unnumbered-links
12715 When doing debugging dumps (see @option{-d} option above), suppress
12716 instruction numbers for the links to the previous and next instructions
12719 @item -fdump-translation-unit @r{(C++ only)}
12720 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12721 @opindex fdump-translation-unit
12722 Dump a representation of the tree structure for the entire translation
12723 unit to a file. The file name is made by appending @file{.tu} to the
12724 source file name, and the file is created in the same directory as the
12725 output file. If the @samp{-@var{options}} form is used, @var{options}
12726 controls the details of the dump as described for the
12727 @option{-fdump-tree} options.
12729 @item -fdump-class-hierarchy @r{(C++ only)}
12730 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12731 @opindex fdump-class-hierarchy
12732 Dump a representation of each class's hierarchy and virtual function
12733 table layout to a file. The file name is made by appending
12734 @file{.class} to the source file name, and the file is created in the
12735 same directory as the output file. If the @samp{-@var{options}} form
12736 is used, @var{options} controls the details of the dump as described
12737 for the @option{-fdump-tree} options.
12739 @item -fdump-ipa-@var{switch}
12741 Control the dumping at various stages of inter-procedural analysis
12742 language tree to a file. The file name is generated by appending a
12743 switch specific suffix to the source file name, and the file is created
12744 in the same directory as the output file. The following dumps are
12749 Enables all inter-procedural analysis dumps.
12752 Dumps information about call-graph optimization, unused function removal,
12753 and inlining decisions.
12756 Dump after function inlining.
12760 @item -fdump-passes
12761 @opindex fdump-passes
12762 Dump the list of optimization passes that are turned on and off by
12763 the current command-line options.
12765 @item -fdump-statistics-@var{option}
12766 @opindex fdump-statistics
12767 Enable and control dumping of pass statistics in a separate file. The
12768 file name is generated by appending a suffix ending in
12769 @samp{.statistics} to the source file name, and the file is created in
12770 the same directory as the output file. If the @samp{-@var{option}}
12771 form is used, @samp{-stats} causes counters to be summed over the
12772 whole compilation unit while @samp{-details} dumps every event as
12773 the passes generate them. The default with no option is to sum
12774 counters for each function compiled.
12776 @item -fdump-tree-@var{switch}
12777 @itemx -fdump-tree-@var{switch}-@var{options}
12778 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
12779 @opindex fdump-tree
12780 Control the dumping at various stages of processing the intermediate
12781 language tree to a file. The file name is generated by appending a
12782 switch-specific suffix to the source file name, and the file is
12783 created in the same directory as the output file. In case of
12784 @option{=@var{filename}} option, the dump is output on the given file
12785 instead of the auto named dump files. If the @samp{-@var{options}}
12786 form is used, @var{options} is a list of @samp{-} separated options
12787 which control the details of the dump. Not all options are applicable
12788 to all dumps; those that are not meaningful are ignored. The
12789 following options are available
12793 Print the address of each node. Usually this is not meaningful as it
12794 changes according to the environment and source file. Its primary use
12795 is for tying up a dump file with a debug environment.
12797 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
12798 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
12799 use working backward from mangled names in the assembly file.
12801 When dumping front-end intermediate representations, inhibit dumping
12802 of members of a scope or body of a function merely because that scope
12803 has been reached. Only dump such items when they are directly reachable
12804 by some other path.
12806 When dumping pretty-printed trees, this option inhibits dumping the
12807 bodies of control structures.
12809 When dumping RTL, print the RTL in slim (condensed) form instead of
12810 the default LISP-like representation.
12812 Print a raw representation of the tree. By default, trees are
12813 pretty-printed into a C-like representation.
12815 Enable more detailed dumps (not honored by every dump option). Also
12816 include information from the optimization passes.
12818 Enable dumping various statistics about the pass (not honored by every dump
12821 Enable showing basic block boundaries (disabled in raw dumps).
12823 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
12824 dump a representation of the control flow graph suitable for viewing with
12825 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
12826 the file is pretty-printed as a subgraph, so that GraphViz can render them
12827 all in a single plot.
12829 This option currently only works for RTL dumps, and the RTL is always
12830 dumped in slim form.
12832 Enable showing virtual operands for every statement.
12834 Enable showing line numbers for statements.
12836 Enable showing the unique ID (@code{DECL_UID}) for each variable.
12838 Enable showing the tree dump for each statement.
12840 Enable showing the EH region number holding each statement.
12842 Enable showing scalar evolution analysis details.
12844 Enable showing optimization information (only available in certain
12847 Enable showing missed optimization information (only available in certain
12850 Enable other detailed optimization information (only available in
12852 @item =@var{filename}
12853 Instead of an auto named dump file, output into the given file
12854 name. The file names @file{stdout} and @file{stderr} are treated
12855 specially and are considered already open standard streams. For
12859 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
12860 -fdump-tree-pre=stderr file.c
12863 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
12864 output on to @file{stderr}. If two conflicting dump filenames are
12865 given for the same pass, then the latter option overrides the earlier
12869 @opindex fdump-tree-split-paths
12870 Dump each function after splitting paths to loop backedges. The file
12871 name is made by appending @file{.split-paths} to the source file name.
12874 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
12875 and @option{lineno}.
12878 Turn on all optimization options, i.e., @option{optimized},
12879 @option{missed}, and @option{note}.
12882 The following tree dumps are possible:
12886 @opindex fdump-tree-original
12887 Dump before any tree based optimization, to @file{@var{file}.original}.
12890 @opindex fdump-tree-optimized
12891 Dump after all tree based optimization, to @file{@var{file}.optimized}.
12894 @opindex fdump-tree-gimple
12895 Dump each function before and after the gimplification pass to a file. The
12896 file name is made by appending @file{.gimple} to the source file name.
12899 @opindex fdump-tree-cfg
12900 Dump the control flow graph of each function to a file. The file name is
12901 made by appending @file{.cfg} to the source file name.
12904 @opindex fdump-tree-ch
12905 Dump each function after copying loop headers. The file name is made by
12906 appending @file{.ch} to the source file name.
12909 @opindex fdump-tree-ssa
12910 Dump SSA related information to a file. The file name is made by appending
12911 @file{.ssa} to the source file name.
12914 @opindex fdump-tree-alias
12915 Dump aliasing information for each function. The file name is made by
12916 appending @file{.alias} to the source file name.
12919 @opindex fdump-tree-ccp
12920 Dump each function after CCP@. The file name is made by appending
12921 @file{.ccp} to the source file name.
12924 @opindex fdump-tree-storeccp
12925 Dump each function after STORE-CCP@. The file name is made by appending
12926 @file{.storeccp} to the source file name.
12929 @opindex fdump-tree-pre
12930 Dump trees after partial redundancy elimination and/or code hoisting.
12931 The file name is made by appending @file{.pre} to the source file name.
12934 @opindex fdump-tree-fre
12935 Dump trees after full redundancy elimination. The file name is made
12936 by appending @file{.fre} to the source file name.
12939 @opindex fdump-tree-copyprop
12940 Dump trees after copy propagation. The file name is made
12941 by appending @file{.copyprop} to the source file name.
12943 @item store_copyprop
12944 @opindex fdump-tree-store_copyprop
12945 Dump trees after store copy-propagation. The file name is made
12946 by appending @file{.store_copyprop} to the source file name.
12949 @opindex fdump-tree-dce
12950 Dump each function after dead code elimination. The file name is made by
12951 appending @file{.dce} to the source file name.
12954 @opindex fdump-tree-sra
12955 Dump each function after performing scalar replacement of aggregates. The
12956 file name is made by appending @file{.sra} to the source file name.
12959 @opindex fdump-tree-sink
12960 Dump each function after performing code sinking. The file name is made
12961 by appending @file{.sink} to the source file name.
12964 @opindex fdump-tree-dom
12965 Dump each function after applying dominator tree optimizations. The file
12966 name is made by appending @file{.dom} to the source file name.
12969 @opindex fdump-tree-dse
12970 Dump each function after applying dead store elimination. The file
12971 name is made by appending @file{.dse} to the source file name.
12974 @opindex fdump-tree-phiopt
12975 Dump each function after optimizing PHI nodes into straightline code. The file
12976 name is made by appending @file{.phiopt} to the source file name.
12979 @opindex fdump-tree-backprop
12980 Dump each function after back-propagating use information up the definition
12981 chain. The file name is made by appending @file{.backprop} to the
12985 @opindex fdump-tree-forwprop
12986 Dump each function after forward propagating single use variables. The file
12987 name is made by appending @file{.forwprop} to the source file name.
12990 @opindex fdump-tree-nrv
12991 Dump each function after applying the named return value optimization on
12992 generic trees. The file name is made by appending @file{.nrv} to the source
12996 @opindex fdump-tree-vect
12997 Dump each function after applying vectorization of loops. The file name is
12998 made by appending @file{.vect} to the source file name.
13001 @opindex fdump-tree-slp
13002 Dump each function after applying vectorization of basic blocks. The file name
13003 is made by appending @file{.slp} to the source file name.
13006 @opindex fdump-tree-vrp
13007 Dump each function after Value Range Propagation (VRP). The file name
13008 is made by appending @file{.vrp} to the source file name.
13011 @opindex fdump-tree-evrp
13012 Dump each function after Early Value Range Propagation (EVRP). The file name
13013 is made by appending @file{.evrp} to the source file name.
13016 @opindex fdump-tree-oaccdevlow
13017 Dump each function after applying device-specific OpenACC transformations.
13018 The file name is made by appending @file{.oaccdevlow} to the source file name.
13021 @opindex fdump-tree-all
13022 Enable all the available tree dumps with the flags provided in this option.
13026 @itemx -fopt-info-@var{options}
13027 @itemx -fopt-info-@var{options}=@var{filename}
13029 Controls optimization dumps from various optimization passes. If the
13030 @samp{-@var{options}} form is used, @var{options} is a list of
13031 @samp{-} separated option keywords to select the dump details and
13034 The @var{options} can be divided into two groups: options describing the
13035 verbosity of the dump, and options describing which optimizations
13036 should be included. The options from both the groups can be freely
13037 mixed as they are non-overlapping. However, in case of any conflicts,
13038 the later options override the earlier options on the command
13041 The following options control the dump verbosity:
13045 Print information when an optimization is successfully applied. It is
13046 up to a pass to decide which information is relevant. For example, the
13047 vectorizer passes print the source location of loops which are
13048 successfully vectorized.
13050 Print information about missed optimizations. Individual passes
13051 control which information to include in the output.
13053 Print verbose information about optimizations, such as certain
13054 transformations, more detailed messages about decisions etc.
13056 Print detailed optimization information. This includes
13057 @samp{optimized}, @samp{missed}, and @samp{note}.
13060 One or more of the following option keywords can be used to describe a
13061 group of optimizations:
13065 Enable dumps from all interprocedural optimizations.
13067 Enable dumps from all loop optimizations.
13069 Enable dumps from all inlining optimizations.
13071 Enable dumps from all vectorization optimizations.
13073 Enable dumps from all optimizations. This is a superset of
13074 the optimization groups listed above.
13077 If @var{options} is
13078 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13079 info about successful optimizations from all the passes.
13081 If the @var{filename} is provided, then the dumps from all the
13082 applicable optimizations are concatenated into the @var{filename}.
13083 Otherwise the dump is output onto @file{stderr}. Though multiple
13084 @option{-fopt-info} options are accepted, only one of them can include
13085 a @var{filename}. If other filenames are provided then all but the
13086 first such option are ignored.
13088 Note that the output @var{filename} is overwritten
13089 in case of multiple translation units. If a combined output from
13090 multiple translation units is desired, @file{stderr} should be used
13093 In the following example, the optimization info is output to
13102 gcc -O3 -fopt-info-missed=missed.all
13106 outputs missed optimization report from all the passes into
13107 @file{missed.all}, and this one:
13110 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13114 prints information about missed optimization opportunities from
13115 vectorization passes on @file{stderr}.
13116 Note that @option{-fopt-info-vec-missed} is equivalent to
13117 @option{-fopt-info-missed-vec}.
13119 As another example,
13121 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13125 outputs information about missed optimizations as well as
13126 optimized locations from all the inlining passes into
13132 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13136 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13137 in conflict since only one output file is allowed. In this case, only
13138 the first option takes effect and the subsequent options are
13139 ignored. Thus only @file{vec.miss} is produced which contains
13140 dumps from the vectorizer about missed opportunities.
13142 @item -fsched-verbose=@var{n}
13143 @opindex fsched-verbose
13144 On targets that use instruction scheduling, this option controls the
13145 amount of debugging output the scheduler prints to the dump files.
13147 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13148 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13149 For @var{n} greater than one, it also output basic block probabilities,
13150 detailed ready list information and unit/insn info. For @var{n} greater
13151 than two, it includes RTL at abort point, control-flow and regions info.
13152 And for @var{n} over four, @option{-fsched-verbose} also includes
13157 @item -fenable-@var{kind}-@var{pass}
13158 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13162 This is a set of options that are used to explicitly disable/enable
13163 optimization passes. These options are intended for use for debugging GCC.
13164 Compiler users should use regular options for enabling/disabling
13169 @item -fdisable-ipa-@var{pass}
13170 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13171 statically invoked in the compiler multiple times, the pass name should be
13172 appended with a sequential number starting from 1.
13174 @item -fdisable-rtl-@var{pass}
13175 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13176 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13177 statically invoked in the compiler multiple times, the pass name should be
13178 appended with a sequential number starting from 1. @var{range-list} is a
13179 comma-separated list of function ranges or assembler names. Each range is a number
13180 pair separated by a colon. The range is inclusive in both ends. If the range
13181 is trivial, the number pair can be simplified as a single number. If the
13182 function's call graph node's @var{uid} falls within one of the specified ranges,
13183 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13184 function header of a dump file, and the pass names can be dumped by using
13185 option @option{-fdump-passes}.
13187 @item -fdisable-tree-@var{pass}
13188 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13189 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13192 @item -fenable-ipa-@var{pass}
13193 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13194 statically invoked in the compiler multiple times, the pass name should be
13195 appended with a sequential number starting from 1.
13197 @item -fenable-rtl-@var{pass}
13198 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13199 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13200 description and examples.
13202 @item -fenable-tree-@var{pass}
13203 @itemx -fenable-tree-@var{pass}=@var{range-list}
13204 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13205 of option arguments.
13209 Here are some examples showing uses of these options.
13213 # disable ccp1 for all functions
13214 -fdisable-tree-ccp1
13215 # disable complete unroll for function whose cgraph node uid is 1
13216 -fenable-tree-cunroll=1
13217 # disable gcse2 for functions at the following ranges [1,1],
13218 # [300,400], and [400,1000]
13219 # disable gcse2 for functions foo and foo2
13220 -fdisable-rtl-gcse2=foo,foo2
13221 # disable early inlining
13222 -fdisable-tree-einline
13223 # disable ipa inlining
13224 -fdisable-ipa-inline
13225 # enable tree full unroll
13226 -fenable-tree-unroll
13231 @itemx -fchecking=@var{n}
13233 @opindex fno-checking
13234 Enable internal consistency checking. The default depends on
13235 the compiler configuration. @option{-fchecking=2} enables further
13236 internal consistency checking that might affect code generation.
13238 @item -frandom-seed=@var{string}
13239 @opindex frandom-seed
13240 This option provides a seed that GCC uses in place of
13241 random numbers in generating certain symbol names
13242 that have to be different in every compiled file. It is also used to
13243 place unique stamps in coverage data files and the object files that
13244 produce them. You can use the @option{-frandom-seed} option to produce
13245 reproducibly identical object files.
13247 The @var{string} can either be a number (decimal, octal or hex) or an
13248 arbitrary string (in which case it's converted to a number by
13251 The @var{string} should be different for every file you compile.
13254 @itemx -save-temps=cwd
13255 @opindex save-temps
13256 Store the usual ``temporary'' intermediate files permanently; place them
13257 in the current directory and name them based on the source file. Thus,
13258 compiling @file{foo.c} with @option{-c -save-temps} produces files
13259 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13260 preprocessed @file{foo.i} output file even though the compiler now
13261 normally uses an integrated preprocessor.
13263 When used in combination with the @option{-x} command-line option,
13264 @option{-save-temps} is sensible enough to avoid over writing an
13265 input source file with the same extension as an intermediate file.
13266 The corresponding intermediate file may be obtained by renaming the
13267 source file before using @option{-save-temps}.
13269 If you invoke GCC in parallel, compiling several different source
13270 files that share a common base name in different subdirectories or the
13271 same source file compiled for multiple output destinations, it is
13272 likely that the different parallel compilers will interfere with each
13273 other, and overwrite the temporary files. For instance:
13276 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13277 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13280 may result in @file{foo.i} and @file{foo.o} being written to
13281 simultaneously by both compilers.
13283 @item -save-temps=obj
13284 @opindex save-temps=obj
13285 Store the usual ``temporary'' intermediate files permanently. If the
13286 @option{-o} option is used, the temporary files are based on the
13287 object file. If the @option{-o} option is not used, the
13288 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13293 gcc -save-temps=obj -c foo.c
13294 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13295 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13299 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13300 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13301 @file{dir2/yfoobar.o}.
13303 @item -time@r{[}=@var{file}@r{]}
13305 Report the CPU time taken by each subprocess in the compilation
13306 sequence. For C source files, this is the compiler proper and assembler
13307 (plus the linker if linking is done).
13309 Without the specification of an output file, the output looks like this:
13316 The first number on each line is the ``user time'', that is time spent
13317 executing the program itself. The second number is ``system time'',
13318 time spent executing operating system routines on behalf of the program.
13319 Both numbers are in seconds.
13321 With the specification of an output file, the output is appended to the
13322 named file, and it looks like this:
13325 0.12 0.01 cc1 @var{options}
13326 0.00 0.01 as @var{options}
13329 The ``user time'' and the ``system time'' are moved before the program
13330 name, and the options passed to the program are displayed, so that one
13331 can later tell what file was being compiled, and with which options.
13333 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13334 @opindex fdump-final-insns
13335 Dump the final internal representation (RTL) to @var{file}. If the
13336 optional argument is omitted (or if @var{file} is @code{.}), the name
13337 of the dump file is determined by appending @code{.gkd} to the
13338 compilation output file name.
13340 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13341 @opindex fcompare-debug
13342 @opindex fno-compare-debug
13343 If no error occurs during compilation, run the compiler a second time,
13344 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13345 passed to the second compilation. Dump the final internal
13346 representation in both compilations, and print an error if they differ.
13348 If the equal sign is omitted, the default @option{-gtoggle} is used.
13350 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13351 and nonzero, implicitly enables @option{-fcompare-debug}. If
13352 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13353 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13356 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13357 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13358 of the final representation and the second compilation, preventing even
13359 @env{GCC_COMPARE_DEBUG} from taking effect.
13361 To verify full coverage during @option{-fcompare-debug} testing, set
13362 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13363 which GCC rejects as an invalid option in any actual compilation
13364 (rather than preprocessing, assembly or linking). To get just a
13365 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13366 not overridden} will do.
13368 @item -fcompare-debug-second
13369 @opindex fcompare-debug-second
13370 This option is implicitly passed to the compiler for the second
13371 compilation requested by @option{-fcompare-debug}, along with options to
13372 silence warnings, and omitting other options that would cause
13373 side-effect compiler outputs to files or to the standard output. Dump
13374 files and preserved temporary files are renamed so as to contain the
13375 @code{.gk} additional extension during the second compilation, to avoid
13376 overwriting those generated by the first.
13378 When this option is passed to the compiler driver, it causes the
13379 @emph{first} compilation to be skipped, which makes it useful for little
13380 other than debugging the compiler proper.
13384 Turn off generation of debug info, if leaving out this option
13385 generates it, or turn it on at level 2 otherwise. The position of this
13386 argument in the command line does not matter; it takes effect after all
13387 other options are processed, and it does so only once, no matter how
13388 many times it is given. This is mainly intended to be used with
13389 @option{-fcompare-debug}.
13391 @item -fvar-tracking-assignments-toggle
13392 @opindex fvar-tracking-assignments-toggle
13393 @opindex fno-var-tracking-assignments-toggle
13394 Toggle @option{-fvar-tracking-assignments}, in the same way that
13395 @option{-gtoggle} toggles @option{-g}.
13399 Makes the compiler print out each function name as it is compiled, and
13400 print some statistics about each pass when it finishes.
13402 @item -ftime-report
13403 @opindex ftime-report
13404 Makes the compiler print some statistics about the time consumed by each
13405 pass when it finishes.
13407 @item -ftime-report-details
13408 @opindex ftime-report-details
13409 Record the time consumed by infrastructure parts separately for each pass.
13411 @item -fira-verbose=@var{n}
13412 @opindex fira-verbose
13413 Control the verbosity of the dump file for the integrated register allocator.
13414 The default value is 5. If the value @var{n} is greater or equal to 10,
13415 the dump output is sent to stderr using the same format as @var{n} minus 10.
13418 @opindex flto-report
13419 Prints a report with internal details on the workings of the link-time
13420 optimizer. The contents of this report vary from version to version.
13421 It is meant to be useful to GCC developers when processing object
13422 files in LTO mode (via @option{-flto}).
13424 Disabled by default.
13426 @item -flto-report-wpa
13427 @opindex flto-report-wpa
13428 Like @option{-flto-report}, but only print for the WPA phase of Link
13432 @opindex fmem-report
13433 Makes the compiler print some statistics about permanent memory
13434 allocation when it finishes.
13436 @item -fmem-report-wpa
13437 @opindex fmem-report-wpa
13438 Makes the compiler print some statistics about permanent memory
13439 allocation for the WPA phase only.
13441 @item -fpre-ipa-mem-report
13442 @opindex fpre-ipa-mem-report
13443 @item -fpost-ipa-mem-report
13444 @opindex fpost-ipa-mem-report
13445 Makes the compiler print some statistics about permanent memory
13446 allocation before or after interprocedural optimization.
13448 @item -fprofile-report
13449 @opindex fprofile-report
13450 Makes the compiler print some statistics about consistency of the
13451 (estimated) profile and effect of individual passes.
13453 @item -fstack-usage
13454 @opindex fstack-usage
13455 Makes the compiler output stack usage information for the program, on a
13456 per-function basis. The filename for the dump is made by appending
13457 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13458 the output file, if explicitly specified and it is not an executable,
13459 otherwise it is the basename of the source file. An entry is made up
13464 The name of the function.
13468 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13471 The qualifier @code{static} means that the function manipulates the stack
13472 statically: a fixed number of bytes are allocated for the frame on function
13473 entry and released on function exit; no stack adjustments are otherwise made
13474 in the function. The second field is this fixed number of bytes.
13476 The qualifier @code{dynamic} means that the function manipulates the stack
13477 dynamically: in addition to the static allocation described above, stack
13478 adjustments are made in the body of the function, for example to push/pop
13479 arguments around function calls. If the qualifier @code{bounded} is also
13480 present, the amount of these adjustments is bounded at compile time and
13481 the second field is an upper bound of the total amount of stack used by
13482 the function. If it is not present, the amount of these adjustments is
13483 not bounded at compile time and the second field only represents the
13488 Emit statistics about front-end processing at the end of the compilation.
13489 This option is supported only by the C++ front end, and
13490 the information is generally only useful to the G++ development team.
13492 @item -fdbg-cnt-list
13493 @opindex fdbg-cnt-list
13494 Print the name and the counter upper bound for all debug counters.
13497 @item -fdbg-cnt=@var{counter-value-list}
13499 Set the internal debug counter upper bound. @var{counter-value-list}
13500 is a comma-separated list of @var{name}:@var{value} pairs
13501 which sets the upper bound of each debug counter @var{name} to @var{value}.
13502 All debug counters have the initial upper bound of @code{UINT_MAX};
13503 thus @code{dbg_cnt} returns true always unless the upper bound
13504 is set by this option.
13505 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13506 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13508 @item -print-file-name=@var{library}
13509 @opindex print-file-name
13510 Print the full absolute name of the library file @var{library} that
13511 would be used when linking---and don't do anything else. With this
13512 option, GCC does not compile or link anything; it just prints the
13515 @item -print-multi-directory
13516 @opindex print-multi-directory
13517 Print the directory name corresponding to the multilib selected by any
13518 other switches present in the command line. This directory is supposed
13519 to exist in @env{GCC_EXEC_PREFIX}.
13521 @item -print-multi-lib
13522 @opindex print-multi-lib
13523 Print the mapping from multilib directory names to compiler switches
13524 that enable them. The directory name is separated from the switches by
13525 @samp{;}, and each switch starts with an @samp{@@} instead of the
13526 @samp{-}, without spaces between multiple switches. This is supposed to
13527 ease shell processing.
13529 @item -print-multi-os-directory
13530 @opindex print-multi-os-directory
13531 Print the path to OS libraries for the selected
13532 multilib, relative to some @file{lib} subdirectory. If OS libraries are
13533 present in the @file{lib} subdirectory and no multilibs are used, this is
13534 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
13535 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
13536 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
13537 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
13539 @item -print-multiarch
13540 @opindex print-multiarch
13541 Print the path to OS libraries for the selected multiarch,
13542 relative to some @file{lib} subdirectory.
13544 @item -print-prog-name=@var{program}
13545 @opindex print-prog-name
13546 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
13548 @item -print-libgcc-file-name
13549 @opindex print-libgcc-file-name
13550 Same as @option{-print-file-name=libgcc.a}.
13552 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
13553 but you do want to link with @file{libgcc.a}. You can do:
13556 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
13559 @item -print-search-dirs
13560 @opindex print-search-dirs
13561 Print the name of the configured installation directory and a list of
13562 program and library directories @command{gcc} searches---and don't do anything else.
13564 This is useful when @command{gcc} prints the error message
13565 @samp{installation problem, cannot exec cpp0: No such file or directory}.
13566 To resolve this you either need to put @file{cpp0} and the other compiler
13567 components where @command{gcc} expects to find them, or you can set the environment
13568 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
13569 Don't forget the trailing @samp{/}.
13570 @xref{Environment Variables}.
13572 @item -print-sysroot
13573 @opindex print-sysroot
13574 Print the target sysroot directory that is used during
13575 compilation. This is the target sysroot specified either at configure
13576 time or using the @option{--sysroot} option, possibly with an extra
13577 suffix that depends on compilation options. If no target sysroot is
13578 specified, the option prints nothing.
13580 @item -print-sysroot-headers-suffix
13581 @opindex print-sysroot-headers-suffix
13582 Print the suffix added to the target sysroot when searching for
13583 headers, or give an error if the compiler is not configured with such
13584 a suffix---and don't do anything else.
13587 @opindex dumpmachine
13588 Print the compiler's target machine (for example,
13589 @samp{i686-pc-linux-gnu})---and don't do anything else.
13592 @opindex dumpversion
13593 Print the compiler version (for example, @code{3.0})---and don't do
13598 Print the compiler's built-in specs---and don't do anything else. (This
13599 is used when GCC itself is being built.) @xref{Spec Files}.
13602 @node Submodel Options
13603 @section Machine-Dependent Options
13604 @cindex submodel options
13605 @cindex specifying hardware config
13606 @cindex hardware models and configurations, specifying
13607 @cindex target-dependent options
13608 @cindex machine-dependent options
13610 Each target machine supported by GCC can have its own options---for
13611 example, to allow you to compile for a particular processor variant or
13612 ABI, or to control optimizations specific to that machine. By
13613 convention, the names of machine-specific options start with
13616 Some configurations of the compiler also support additional target-specific
13617 options, usually for compatibility with other compilers on the same
13620 @c This list is ordered alphanumerically by subsection name.
13621 @c It should be the same order and spelling as these options are listed
13622 @c in Machine Dependent Options
13625 * AArch64 Options::
13626 * Adapteva Epiphany Options::
13630 * Blackfin Options::
13635 * DEC Alpha Options::
13639 * GNU/Linux Options::
13649 * MicroBlaze Options::
13652 * MN10300 Options::
13656 * Nios II Options::
13657 * Nvidia PTX Options::
13659 * picoChip Options::
13660 * PowerPC Options::
13662 * RS/6000 and PowerPC Options::
13664 * S/390 and zSeries Options::
13667 * Solaris 2 Options::
13670 * System V Options::
13671 * TILE-Gx Options::
13672 * TILEPro Options::
13677 * VxWorks Options::
13679 * x86 Windows Options::
13680 * Xstormy16 Options::
13682 * zSeries Options::
13685 @node AArch64 Options
13686 @subsection AArch64 Options
13687 @cindex AArch64 Options
13689 These options are defined for AArch64 implementations:
13693 @item -mabi=@var{name}
13695 Generate code for the specified data model. Permissible values
13696 are @samp{ilp32} for SysV-like data model where int, long int and pointers
13697 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
13698 but long int and pointers are 64 bits.
13700 The default depends on the specific target configuration. Note that
13701 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
13702 entire program with the same ABI, and link with a compatible set of libraries.
13705 @opindex mbig-endian
13706 Generate big-endian code. This is the default when GCC is configured for an
13707 @samp{aarch64_be-*-*} target.
13709 @item -mgeneral-regs-only
13710 @opindex mgeneral-regs-only
13711 Generate code which uses only the general-purpose registers. This will prevent
13712 the compiler from using floating-point and Advanced SIMD registers but will not
13713 impose any restrictions on the assembler.
13715 @item -mlittle-endian
13716 @opindex mlittle-endian
13717 Generate little-endian code. This is the default when GCC is configured for an
13718 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13720 @item -mcmodel=tiny
13721 @opindex mcmodel=tiny
13722 Generate code for the tiny code model. The program and its statically defined
13723 symbols must be within 1MB of each other. Programs can be statically or
13724 dynamically linked.
13726 @item -mcmodel=small
13727 @opindex mcmodel=small
13728 Generate code for the small code model. The program and its statically defined
13729 symbols must be within 4GB of each other. Programs can be statically or
13730 dynamically linked. This is the default code model.
13732 @item -mcmodel=large
13733 @opindex mcmodel=large
13734 Generate code for the large code model. This makes no assumptions about
13735 addresses and sizes of sections. Programs can be statically linked only.
13737 @item -mstrict-align
13738 @opindex mstrict-align
13739 Avoid generating memory accesses that may not be aligned on a natural object
13740 boundary as described in the architecture specification.
13742 @item -momit-leaf-frame-pointer
13743 @itemx -mno-omit-leaf-frame-pointer
13744 @opindex momit-leaf-frame-pointer
13745 @opindex mno-omit-leaf-frame-pointer
13746 Omit or keep the frame pointer in leaf functions. The former behavior is the
13749 @item -mtls-dialect=desc
13750 @opindex mtls-dialect=desc
13751 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13752 of TLS variables. This is the default.
13754 @item -mtls-dialect=traditional
13755 @opindex mtls-dialect=traditional
13756 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13759 @item -mtls-size=@var{size}
13761 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13762 This option requires binutils 2.26 or newer.
13764 @item -mfix-cortex-a53-835769
13765 @itemx -mno-fix-cortex-a53-835769
13766 @opindex mfix-cortex-a53-835769
13767 @opindex mno-fix-cortex-a53-835769
13768 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13769 This involves inserting a NOP instruction between memory instructions and
13770 64-bit integer multiply-accumulate instructions.
13772 @item -mfix-cortex-a53-843419
13773 @itemx -mno-fix-cortex-a53-843419
13774 @opindex mfix-cortex-a53-843419
13775 @opindex mno-fix-cortex-a53-843419
13776 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13777 This erratum workaround is made at link time and this will only pass the
13778 corresponding flag to the linker.
13780 @item -mlow-precision-recip-sqrt
13781 @item -mno-low-precision-recip-sqrt
13782 @opindex mlow-precision-recip-sqrt
13783 @opindex mno-low-precision-recip-sqrt
13784 Enable or disable the reciprocal square root approximation.
13785 This option only has an effect if @option{-ffast-math} or
13786 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13787 precision of reciprocal square root results to about 16 bits for
13788 single precision and to 32 bits for double precision.
13790 @item -mlow-precision-sqrt
13791 @item -mno-low-precision-sqrt
13792 @opindex -mlow-precision-sqrt
13793 @opindex -mno-low-precision-sqrt
13794 Enable or disable the square root approximation.
13795 This option only has an effect if @option{-ffast-math} or
13796 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13797 precision of square root results to about 16 bits for
13798 single precision and to 32 bits for double precision.
13799 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13801 @item -mlow-precision-div
13802 @item -mno-low-precision-div
13803 @opindex -mlow-precision-div
13804 @opindex -mno-low-precision-div
13805 Enable or disable the division approximation.
13806 This option only has an effect if @option{-ffast-math} or
13807 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13808 precision of division results to about 16 bits for
13809 single precision and to 32 bits for double precision.
13811 @item -march=@var{name}
13813 Specify the name of the target architecture and, optionally, one or
13814 more feature modifiers. This option has the form
13815 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13817 The permissible values for @var{arch} are @samp{armv8-a},
13818 @samp{armv8.1-a}, @samp{armv8.2-a} or @var{native}.
13820 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
13821 support for the ARMv8.2-A architecture extensions.
13823 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
13824 support for the ARMv8.1-A architecture extension. In particular, it
13825 enables the @samp{+crc} and @samp{+lse} features.
13827 The value @samp{native} is available on native AArch64 GNU/Linux and
13828 causes the compiler to pick the architecture of the host system. This
13829 option has no effect if the compiler is unable to recognize the
13830 architecture of the host system,
13832 The permissible values for @var{feature} are listed in the sub-section
13833 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13834 Feature Modifiers}. Where conflicting feature modifiers are
13835 specified, the right-most feature is used.
13837 GCC uses @var{name} to determine what kind of instructions it can emit
13838 when generating assembly code. If @option{-march} is specified
13839 without either of @option{-mtune} or @option{-mcpu} also being
13840 specified, the code is tuned to perform well across a range of target
13841 processors implementing the target architecture.
13843 @item -mtune=@var{name}
13845 Specify the name of the target processor for which GCC should tune the
13846 performance of the code. Permissible values for this option are:
13847 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
13848 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{qdf24xx},
13849 @samp{thunderx}, @samp{xgene1}, @samp{vulcan}, @samp{cortex-a57.cortex-a53},
13850 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
13851 @samp{cortex-a73.cortex-a53}, @samp{native}.
13853 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
13854 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
13855 specify that GCC should tune for a big.LITTLE system.
13857 Additionally on native AArch64 GNU/Linux systems the value
13858 @samp{native} tunes performance to the host system. This option has no effect
13859 if the compiler is unable to recognize the processor of the host system.
13861 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13862 are specified, the code is tuned to perform well across a range
13863 of target processors.
13865 This option cannot be suffixed by feature modifiers.
13867 @item -mcpu=@var{name}
13869 Specify the name of the target processor, optionally suffixed by one
13870 or more feature modifiers. This option has the form
13871 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13872 the permissible values for @var{cpu} are the same as those available
13873 for @option{-mtune}. The permissible values for @var{feature} are
13874 documented in the sub-section on
13875 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13876 Feature Modifiers}. Where conflicting feature modifiers are
13877 specified, the right-most feature is used.
13879 GCC uses @var{name} to determine what kind of instructions it can emit when
13880 generating assembly code (as if by @option{-march}) and to determine
13881 the target processor for which to tune for performance (as if
13882 by @option{-mtune}). Where this option is used in conjunction
13883 with @option{-march} or @option{-mtune}, those options take precedence
13884 over the appropriate part of this option.
13886 @item -moverride=@var{string}
13888 Override tuning decisions made by the back-end in response to a
13889 @option{-mtune=} switch. The syntax, semantics, and accepted values
13890 for @var{string} in this option are not guaranteed to be consistent
13893 This option is only intended to be useful when developing GCC.
13895 @item -mpc-relative-literal-loads
13896 @opindex mpc-relative-literal-loads
13897 Enable PC-relative literal loads. With this option literal pools are
13898 accessed using a single instruction and emitted after each function. This
13899 limits the maximum size of functions to 1MB. This is enabled by default for
13900 @option{-mcmodel=tiny}.
13904 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
13905 @anchor{aarch64-feature-modifiers}
13906 @cindex @option{-march} feature modifiers
13907 @cindex @option{-mcpu} feature modifiers
13908 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
13909 the following and their inverses @option{no@var{feature}}:
13913 Enable CRC extension. This is on by default for
13914 @option{-march=armv8.1-a}.
13916 Enable Crypto extension. This also enables Advanced SIMD and floating-point
13919 Enable floating-point instructions. This is on by default for all possible
13920 values for options @option{-march} and @option{-mcpu}.
13922 Enable Advanced SIMD instructions. This also enables floating-point
13923 instructions. This is on by default for all possible values for options
13924 @option{-march} and @option{-mcpu}.
13926 Enable Large System Extension instructions. This is on by default for
13927 @option{-march=armv8.1-a}.
13929 Enable FP16 extension. This also enables floating-point instructions.
13933 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
13934 Conversely, @option{nofp} implies @option{nosimd}, which implies
13937 @node Adapteva Epiphany Options
13938 @subsection Adapteva Epiphany Options
13940 These @samp{-m} options are defined for Adapteva Epiphany:
13943 @item -mhalf-reg-file
13944 @opindex mhalf-reg-file
13945 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
13946 That allows code to run on hardware variants that lack these registers.
13948 @item -mprefer-short-insn-regs
13949 @opindex mprefer-short-insn-regs
13950 Preferentially allocate registers that allow short instruction generation.
13951 This can result in increased instruction count, so this may either reduce or
13952 increase overall code size.
13954 @item -mbranch-cost=@var{num}
13955 @opindex mbranch-cost
13956 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13957 This cost is only a heuristic and is not guaranteed to produce
13958 consistent results across releases.
13962 Enable the generation of conditional moves.
13964 @item -mnops=@var{num}
13966 Emit @var{num} NOPs before every other generated instruction.
13968 @item -mno-soft-cmpsf
13969 @opindex mno-soft-cmpsf
13970 For single-precision floating-point comparisons, emit an @code{fsub} instruction
13971 and test the flags. This is faster than a software comparison, but can
13972 get incorrect results in the presence of NaNs, or when two different small
13973 numbers are compared such that their difference is calculated as zero.
13974 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
13975 software comparisons.
13977 @item -mstack-offset=@var{num}
13978 @opindex mstack-offset
13979 Set the offset between the top of the stack and the stack pointer.
13980 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
13981 can be used by leaf functions without stack allocation.
13982 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
13983 Note also that this option changes the ABI; compiling a program with a
13984 different stack offset than the libraries have been compiled with
13985 generally does not work.
13986 This option can be useful if you want to evaluate if a different stack
13987 offset would give you better code, but to actually use a different stack
13988 offset to build working programs, it is recommended to configure the
13989 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
13991 @item -mno-round-nearest
13992 @opindex mno-round-nearest
13993 Make the scheduler assume that the rounding mode has been set to
13994 truncating. The default is @option{-mround-nearest}.
13997 @opindex mlong-calls
13998 If not otherwise specified by an attribute, assume all calls might be beyond
13999 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14000 function address into a register before performing a (otherwise direct) call.
14001 This is the default.
14003 @item -mshort-calls
14004 @opindex short-calls
14005 If not otherwise specified by an attribute, assume all direct calls are
14006 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14007 for direct calls. The default is @option{-mlong-calls}.
14011 Assume addresses can be loaded as 16-bit unsigned values. This does not
14012 apply to function addresses for which @option{-mlong-calls} semantics
14015 @item -mfp-mode=@var{mode}
14017 Set the prevailing mode of the floating-point unit.
14018 This determines the floating-point mode that is provided and expected
14019 at function call and return time. Making this mode match the mode you
14020 predominantly need at function start can make your programs smaller and
14021 faster by avoiding unnecessary mode switches.
14023 @var{mode} can be set to one the following values:
14027 Any mode at function entry is valid, and retained or restored when
14028 the function returns, and when it calls other functions.
14029 This mode is useful for compiling libraries or other compilation units
14030 you might want to incorporate into different programs with different
14031 prevailing FPU modes, and the convenience of being able to use a single
14032 object file outweighs the size and speed overhead for any extra
14033 mode switching that might be needed, compared with what would be needed
14034 with a more specific choice of prevailing FPU mode.
14037 This is the mode used for floating-point calculations with
14038 truncating (i.e.@: round towards zero) rounding mode. That includes
14039 conversion from floating point to integer.
14041 @item round-nearest
14042 This is the mode used for floating-point calculations with
14043 round-to-nearest-or-even rounding mode.
14046 This is the mode used to perform integer calculations in the FPU, e.g.@:
14047 integer multiply, or integer multiply-and-accumulate.
14050 The default is @option{-mfp-mode=caller}
14052 @item -mnosplit-lohi
14053 @itemx -mno-postinc
14054 @itemx -mno-postmodify
14055 @opindex mnosplit-lohi
14056 @opindex mno-postinc
14057 @opindex mno-postmodify
14058 Code generation tweaks that disable, respectively, splitting of 32-bit
14059 loads, generation of post-increment addresses, and generation of
14060 post-modify addresses. The defaults are @option{msplit-lohi},
14061 @option{-mpost-inc}, and @option{-mpost-modify}.
14063 @item -mnovect-double
14064 @opindex mno-vect-double
14065 Change the preferred SIMD mode to SImode. The default is
14066 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14068 @item -max-vect-align=@var{num}
14069 @opindex max-vect-align
14070 The maximum alignment for SIMD vector mode types.
14071 @var{num} may be 4 or 8. The default is 8.
14072 Note that this is an ABI change, even though many library function
14073 interfaces are unaffected if they don't use SIMD vector modes
14074 in places that affect size and/or alignment of relevant types.
14076 @item -msplit-vecmove-early
14077 @opindex msplit-vecmove-early
14078 Split vector moves into single word moves before reload. In theory this
14079 can give better register allocation, but so far the reverse seems to be
14080 generally the case.
14082 @item -m1reg-@var{reg}
14084 Specify a register to hold the constant @minus{}1, which makes loading small negative
14085 constants and certain bitmasks faster.
14086 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14087 which specify use of that register as a fixed register,
14088 and @samp{none}, which means that no register is used for this
14089 purpose. The default is @option{-m1reg-none}.
14094 @subsection ARC Options
14095 @cindex ARC options
14097 The following options control the architecture variant for which code
14100 @c architecture variants
14103 @item -mbarrel-shifter
14104 @opindex mbarrel-shifter
14105 Generate instructions supported by barrel shifter. This is the default
14106 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14108 @item -mcpu=@var{cpu}
14110 Set architecture type, register usage, and instruction scheduling
14111 parameters for @var{cpu}. There are also shortcut alias options
14112 available for backward compatibility and convenience. Supported
14113 values for @var{cpu} are
14120 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14125 Compile for ARC601. Alias: @option{-mARC601}.
14131 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14132 This is the default when configured with @option{--with-cpu=arc700}@.
14136 Compile for ARC EM.
14140 Compile for ARC HS.
14145 @itemx -mdpfp-compact
14146 @opindex mdpfp-compact
14147 FPX: Generate Double Precision FPX instructions, tuned for the compact
14151 @opindex mdpfp-fast
14152 FPX: Generate Double Precision FPX instructions, tuned for the fast
14155 @item -mno-dpfp-lrsr
14156 @opindex mno-dpfp-lrsr
14157 Disable LR and SR instructions from using FPX extension aux registers.
14161 Generate Extended arithmetic instructions. Currently only
14162 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14163 supported. This is always enabled for @option{-mcpu=ARC700}.
14167 Do not generate mpy instructions for ARC700.
14171 Generate 32x16 bit multiply and mac instructions.
14175 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
14179 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
14184 @itemx -mspfp-compact
14185 @opindex mspfp-compact
14186 FPX: Generate Single Precision FPX instructions, tuned for the compact
14190 @opindex mspfp-fast
14191 FPX: Generate Single Precision FPX instructions, tuned for the fast
14196 Enable generation of ARC SIMD instructions via target-specific
14197 builtins. Only valid for @option{-mcpu=ARC700}.
14200 @opindex msoft-float
14201 This option ignored; it is provided for compatibility purposes only.
14202 Software floating point code is emitted by default, and this default
14203 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
14204 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
14205 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
14209 Generate swap instructions.
14213 This enables Locked Load/Store Conditional extension to implement
14214 atomic memopry built-in functions. Not available for ARC 6xx or ARC
14219 Enable DIV/REM instructions for ARCv2 cores.
14221 @item -mcode-density
14222 @opindex mcode-density
14223 Enable code density instructions for ARC EM, default on for ARC HS.
14227 Enable double load/store operations for ARC HS cores.
14229 @item -mtp-regno=@var{regno}
14231 Specify thread pointer register number.
14233 @item -mmpy-option=@var{multo}
14234 @opindex mmpy-option
14235 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
14236 the default value. The recognized values for @var{multo} are:
14240 No multiplier available.
14244 The multiply option is set to w: 16x16 multiplier, fully pipelined.
14245 The following instructions are enabled: MPYW, and MPYUW.
14249 The multiply option is set to wlh1: 32x32 multiplier, fully
14250 pipelined (1 stage). The following instructions are additionally
14251 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
14255 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
14256 (2 stages). The following instructions are additionally enabled: MPY,
14257 MPYU, MPYM, MPYMU, and MPY_S.
14261 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
14262 sequential. The following instructions are additionally enabled: MPY,
14263 MPYU, MPYM, MPYMU, and MPY_S.
14267 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
14268 sequential. The following instructions are additionally enabled: MPY,
14269 MPYU, MPYM, MPYMU, and MPY_S.
14273 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
14274 sequential. The following instructions are additionally enabled: MPY,
14275 MPYU, MPYM, MPYMU, and MPY_S.
14279 This option is only available for ARCv2 cores@.
14281 @item -mfpu=@var{fpu}
14283 Enables specific floating-point hardware extension for ARCv2
14284 core. Supported values for @var{fpu} are:
14290 Enables support for single precision floating point hardware
14295 Enables support for double precision floating point hardware
14296 extensions. The single precision floating point extension is also
14297 enabled. Not available for ARC EM@.
14301 Enables support for double precision floating point hardware
14302 extensions using double precision assist instructions. The single
14303 precision floating point extension is also enabled. This option is
14304 only available for ARC EM@.
14308 Enables support for double precision floating point hardware
14309 extensions using double precision assist instructions, and simple
14310 precision square-root and divide hardware extensions. The single
14311 precision floating point extension is also enabled. This option is
14312 only available for ARC EM@.
14316 Enables support for double precision floating point hardware
14317 extensions using double precision assist instructions, and simple
14318 precision fused multiple and add hardware extension. The single
14319 precision floating point extension is also enabled. This option is
14320 only available for ARC EM@.
14324 Enables support for double precision floating point hardware
14325 extensions using double precision assist instructions, and all simple
14326 precision hardware extensions. The single precision floating point
14327 extension is also enabled. This option is only available for ARC EM@.
14331 Enables support for single precision floating point, and single
14332 precision square-root and divide hardware extensions@.
14336 Enables support for double precision floating point, and double
14337 precision square-root and divide hardware extensions. This option
14338 includes option @samp{fpus_div}. Not available for ARC EM@.
14342 Enables support for single precision floating point, and single
14343 precision fused multiple and add hardware extensions@.
14347 Enables support for double precision floating point, and double
14348 precision fused multiple and add hardware extensions. This option
14349 includes option @samp{fpus_fma}. Not available for ARC EM@.
14353 Enables support for all single precision floating point hardware
14358 Enables support for all single and double precision floating point
14359 hardware extensions. Not available for ARC EM@.
14365 The following options are passed through to the assembler, and also
14366 define preprocessor macro symbols.
14368 @c Flags used by the assembler, but for which we define preprocessor
14369 @c macro symbols as well.
14372 @opindex mdsp-packa
14373 Passed down to the assembler to enable the DSP Pack A extensions.
14374 Also sets the preprocessor symbol @code{__Xdsp_packa}.
14378 Passed down to the assembler to enable the dual viterbi butterfly
14379 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
14381 @c ARC700 4.10 extension instruction
14384 Passed down to the assembler to enable the Locked Load/Store
14385 Conditional extension. Also sets the preprocessor symbol
14390 Passed down to the assembler. Also sets the preprocessor symbol
14391 @code{__Xxmac_d16}.
14395 Passed down to the assembler. Also sets the preprocessor symbol
14398 @c ARC700 4.10 extension instruction
14401 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
14402 extension instruction. Also sets the preprocessor symbol
14405 @c ARC700 4.10 extension instruction
14408 Passed down to the assembler to enable the swap byte ordering
14409 extension instruction. Also sets the preprocessor symbol
14413 @opindex mtelephony
14414 Passed down to the assembler to enable dual and single operand
14415 instructions for telephony. Also sets the preprocessor symbol
14416 @code{__Xtelephony}.
14420 Passed down to the assembler to enable the XY Memory extension. Also
14421 sets the preprocessor symbol @code{__Xxy}.
14425 The following options control how the assembly code is annotated:
14427 @c Assembly annotation options
14431 Annotate assembler instructions with estimated addresses.
14433 @item -mannotate-align
14434 @opindex mannotate-align
14435 Explain what alignment considerations lead to the decision to make an
14436 instruction short or long.
14440 The following options are passed through to the linker:
14442 @c options passed through to the linker
14446 Passed through to the linker, to specify use of the @code{arclinux} emulation.
14447 This option is enabled by default in tool chains built for
14448 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
14449 when profiling is not requested.
14451 @item -marclinux_prof
14452 @opindex marclinux_prof
14453 Passed through to the linker, to specify use of the
14454 @code{arclinux_prof} emulation. This option is enabled by default in
14455 tool chains built for @w{@code{arc-linux-uclibc}} and
14456 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
14460 The following options control the semantics of generated code:
14462 @c semantically relevant code generation options
14465 @opindex mlong-calls
14466 Generate call insns as register indirect calls, thus providing access
14467 to the full 32-bit address range.
14469 @item -mmedium-calls
14470 @opindex mmedium-calls
14471 Don't use less than 25 bit addressing range for calls, which is the
14472 offset available for an unconditional branch-and-link
14473 instruction. Conditional execution of function calls is suppressed, to
14474 allow use of the 25-bit range, rather than the 21-bit range with
14475 conditional branch-and-link. This is the default for tool chains built
14476 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
14480 Do not generate sdata references. This is the default for tool chains
14481 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
14485 @opindex mucb-mcount
14486 Instrument with mcount calls as used in UCB code. I.e. do the
14487 counting in the callee, not the caller. By default ARC instrumentation
14488 counts in the caller.
14490 @item -mvolatile-cache
14491 @opindex mvolatile-cache
14492 Use ordinarily cached memory accesses for volatile references. This is the
14495 @item -mno-volatile-cache
14496 @opindex mno-volatile-cache
14497 Enable cache bypass for volatile references.
14501 The following options fine tune code generation:
14502 @c code generation tuning options
14505 @opindex malign-call
14506 Do alignment optimizations for call instructions.
14508 @item -mauto-modify-reg
14509 @opindex mauto-modify-reg
14510 Enable the use of pre/post modify with register displacement.
14512 @item -mbbit-peephole
14513 @opindex mbbit-peephole
14514 Enable bbit peephole2.
14518 This option disables a target-specific pass in @file{arc_reorg} to
14519 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
14520 generation driven by the combiner pass.
14522 @item -mcase-vector-pcrel
14523 @opindex mcase-vector-pcrel
14524 Use pc-relative switch case tables - this enables case table shortening.
14525 This is the default for @option{-Os}.
14527 @item -mcompact-casesi
14528 @opindex mcompact-casesi
14529 Enable compact casesi pattern. This is the default for @option{-Os},
14530 and only available for ARCv1 cores.
14532 @item -mno-cond-exec
14533 @opindex mno-cond-exec
14534 Disable ARCompact specific pass to generate conditional execution instructions.
14535 Due to delay slot scheduling and interactions between operand numbers,
14536 literal sizes, instruction lengths, and the support for conditional execution,
14537 the target-independent pass to generate conditional execution is often lacking,
14538 so the ARC port has kept a special pass around that tries to find more
14539 conditional execution generating opportunities after register allocation,
14540 branch shortening, and delay slot scheduling have been done. This pass
14541 generally, but not always, improves performance and code size, at the cost of
14542 extra compilation time, which is why there is an option to switch it off.
14543 If you have a problem with call instructions exceeding their allowable
14544 offset range because they are conditionalized, you should consider using
14545 @option{-mmedium-calls} instead.
14547 @item -mearly-cbranchsi
14548 @opindex mearly-cbranchsi
14549 Enable pre-reload use of the cbranchsi pattern.
14551 @item -mexpand-adddi
14552 @opindex mexpand-adddi
14553 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
14554 @code{add.f}, @code{adc} etc.
14556 @item -mindexed-loads
14557 @opindex mindexed-loads
14558 Enable the use of indexed loads. This can be problematic because some
14559 optimizers then assume that indexed stores exist, which is not
14563 Enable Local Register Allocation. This is still experimental for ARC,
14564 so by default the compiler uses standard reload
14565 (i.e. @option{-mno-lra}).
14567 @item -mlra-priority-none
14568 @opindex mlra-priority-none
14569 Don't indicate any priority for target registers.
14571 @item -mlra-priority-compact
14572 @opindex mlra-priority-compact
14573 Indicate target register priority for r0..r3 / r12..r15.
14575 @item -mlra-priority-noncompact
14576 @opindex mlra-priority-noncompact
14577 Reduce target register priority for r0..r3 / r12..r15.
14579 @item -mno-millicode
14580 @opindex mno-millicode
14581 When optimizing for size (using @option{-Os}), prologues and epilogues
14582 that have to save or restore a large number of registers are often
14583 shortened by using call to a special function in libgcc; this is
14584 referred to as a @emph{millicode} call. As these calls can pose
14585 performance issues, and/or cause linking issues when linking in a
14586 nonstandard way, this option is provided to turn off millicode call
14590 @opindex mmixed-code
14591 Tweak register allocation to help 16-bit instruction generation.
14592 This generally has the effect of decreasing the average instruction size
14593 while increasing the instruction count.
14597 Enable 'q' instruction alternatives.
14598 This is the default for @option{-Os}.
14602 Enable Rcq constraint handling - most short code generation depends on this.
14603 This is the default.
14607 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
14608 This is the default.
14610 @item -msize-level=@var{level}
14611 @opindex msize-level
14612 Fine-tune size optimization with regards to instruction lengths and alignment.
14613 The recognized values for @var{level} are:
14616 No size optimization. This level is deprecated and treated like @samp{1}.
14619 Short instructions are used opportunistically.
14622 In addition, alignment of loops and of code after barriers are dropped.
14625 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
14629 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
14630 the behavior when this is not set is equivalent to level @samp{1}.
14632 @item -mtune=@var{cpu}
14634 Set instruction scheduling parameters for @var{cpu}, overriding any implied
14635 by @option{-mcpu=}.
14637 Supported values for @var{cpu} are
14641 Tune for ARC600 cpu.
14644 Tune for ARC601 cpu.
14647 Tune for ARC700 cpu with standard multiplier block.
14650 Tune for ARC700 cpu with XMAC block.
14653 Tune for ARC725D cpu.
14656 Tune for ARC750D cpu.
14660 @item -mmultcost=@var{num}
14662 Cost to assume for a multiply instruction, with @samp{4} being equal to a
14663 normal instruction.
14665 @item -munalign-prob-threshold=@var{probability}
14666 @opindex munalign-prob-threshold
14667 Set probability threshold for unaligning branches.
14668 When tuning for @samp{ARC700} and optimizing for speed, branches without
14669 filled delay slot are preferably emitted unaligned and long, unless
14670 profiling indicates that the probability for the branch to be taken
14671 is below @var{probability}. @xref{Cross-profiling}.
14672 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
14676 The following options are maintained for backward compatibility, but
14677 are now deprecated and will be removed in a future release:
14679 @c Deprecated options
14687 @opindex mbig-endian
14690 Compile code for big endian targets. Use of these options is now
14691 deprecated. Users wanting big-endian code, should use the
14692 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
14693 building the tool chain, for which big-endian is the default.
14695 @item -mlittle-endian
14696 @opindex mlittle-endian
14699 Compile code for little endian targets. Use of these options is now
14700 deprecated. Users wanting little-endian code should use the
14701 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
14702 building the tool chain, for which little-endian is the default.
14704 @item -mbarrel_shifter
14705 @opindex mbarrel_shifter
14706 Replaced by @option{-mbarrel-shifter}.
14708 @item -mdpfp_compact
14709 @opindex mdpfp_compact
14710 Replaced by @option{-mdpfp-compact}.
14713 @opindex mdpfp_fast
14714 Replaced by @option{-mdpfp-fast}.
14717 @opindex mdsp_packa
14718 Replaced by @option{-mdsp-packa}.
14722 Replaced by @option{-mea}.
14726 Replaced by @option{-mmac-24}.
14730 Replaced by @option{-mmac-d16}.
14732 @item -mspfp_compact
14733 @opindex mspfp_compact
14734 Replaced by @option{-mspfp-compact}.
14737 @opindex mspfp_fast
14738 Replaced by @option{-mspfp-fast}.
14740 @item -mtune=@var{cpu}
14742 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14743 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14744 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
14746 @item -multcost=@var{num}
14748 Replaced by @option{-mmultcost}.
14753 @subsection ARM Options
14754 @cindex ARM options
14756 These @samp{-m} options are defined for the ARM port:
14759 @item -mabi=@var{name}
14761 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
14762 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
14765 @opindex mapcs-frame
14766 Generate a stack frame that is compliant with the ARM Procedure Call
14767 Standard for all functions, even if this is not strictly necessary for
14768 correct execution of the code. Specifying @option{-fomit-frame-pointer}
14769 with this option causes the stack frames not to be generated for
14770 leaf functions. The default is @option{-mno-apcs-frame}.
14771 This option is deprecated.
14775 This is a synonym for @option{-mapcs-frame} and is deprecated.
14778 @c not currently implemented
14779 @item -mapcs-stack-check
14780 @opindex mapcs-stack-check
14781 Generate code to check the amount of stack space available upon entry to
14782 every function (that actually uses some stack space). If there is
14783 insufficient space available then either the function
14784 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
14785 called, depending upon the amount of stack space required. The runtime
14786 system is required to provide these functions. The default is
14787 @option{-mno-apcs-stack-check}, since this produces smaller code.
14789 @c not currently implemented
14791 @opindex mapcs-float
14792 Pass floating-point arguments using the floating-point registers. This is
14793 one of the variants of the APCS@. This option is recommended if the
14794 target hardware has a floating-point unit or if a lot of floating-point
14795 arithmetic is going to be performed by the code. The default is
14796 @option{-mno-apcs-float}, since the size of integer-only code is
14797 slightly increased if @option{-mapcs-float} is used.
14799 @c not currently implemented
14800 @item -mapcs-reentrant
14801 @opindex mapcs-reentrant
14802 Generate reentrant, position-independent code. The default is
14803 @option{-mno-apcs-reentrant}.
14806 @item -mthumb-interwork
14807 @opindex mthumb-interwork
14808 Generate code that supports calling between the ARM and Thumb
14809 instruction sets. Without this option, on pre-v5 architectures, the
14810 two instruction sets cannot be reliably used inside one program. The
14811 default is @option{-mno-thumb-interwork}, since slightly larger code
14812 is generated when @option{-mthumb-interwork} is specified. In AAPCS
14813 configurations this option is meaningless.
14815 @item -mno-sched-prolog
14816 @opindex mno-sched-prolog
14817 Prevent the reordering of instructions in the function prologue, or the
14818 merging of those instruction with the instructions in the function's
14819 body. This means that all functions start with a recognizable set
14820 of instructions (or in fact one of a choice from a small set of
14821 different function prologues), and this information can be used to
14822 locate the start of functions inside an executable piece of code. The
14823 default is @option{-msched-prolog}.
14825 @item -mfloat-abi=@var{name}
14826 @opindex mfloat-abi
14827 Specifies which floating-point ABI to use. Permissible values
14828 are: @samp{soft}, @samp{softfp} and @samp{hard}.
14830 Specifying @samp{soft} causes GCC to generate output containing
14831 library calls for floating-point operations.
14832 @samp{softfp} allows the generation of code using hardware floating-point
14833 instructions, but still uses the soft-float calling conventions.
14834 @samp{hard} allows generation of floating-point instructions
14835 and uses FPU-specific calling conventions.
14837 The default depends on the specific target configuration. Note that
14838 the hard-float and soft-float ABIs are not link-compatible; you must
14839 compile your entire program with the same ABI, and link with a
14840 compatible set of libraries.
14842 @item -mlittle-endian
14843 @opindex mlittle-endian
14844 Generate code for a processor running in little-endian mode. This is
14845 the default for all standard configurations.
14848 @opindex mbig-endian
14849 Generate code for a processor running in big-endian mode; the default is
14850 to compile code for a little-endian processor.
14852 @item -march=@var{name}
14854 This specifies the name of the target ARM architecture. GCC uses this
14855 name to determine what kind of instructions it can emit when generating
14856 assembly code. This option can be used in conjunction with or instead
14857 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
14858 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
14859 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
14860 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
14861 @samp{armv6kz}, @samp{armv6s-m},
14862 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
14863 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
14864 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
14865 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
14866 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
14868 Architecture revisions older than @samp{armv4t} are deprecated.
14870 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
14871 the (now mandatory) SVC instruction.
14873 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
14876 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
14879 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
14880 architecture together with the optional CRC32 extensions.
14882 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
14883 architecture. This also enables the features provided by
14884 @option{-march=armv8-a+crc}.
14886 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
14887 architecture. This also enables the features provided by
14888 @option{-march=armv8.1-a}.
14890 @option{-march=armv8.2-a+fp16} enables compiler support for the
14891 ARMv8.2-A architecture with the optional FP16 instructions extension.
14892 This also enables the features provided by @option{-march=armv8.1-a}
14893 and implies @option{-mfp16-format=ieee}.
14895 @option{-march=native} causes the compiler to auto-detect the architecture
14896 of the build computer. At present, this feature is only supported on
14897 GNU/Linux, and not all architectures are recognized. If the auto-detect
14898 is unsuccessful the option has no effect.
14900 @item -mtune=@var{name}
14902 This option specifies the name of the target ARM processor for
14903 which GCC should tune the performance of the code.
14904 For some ARM implementations better performance can be obtained by using
14906 Permissible names are: @samp{arm2}, @samp{arm250},
14907 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
14908 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
14909 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
14910 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
14912 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
14913 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
14914 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
14915 @samp{strongarm1110},
14916 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
14917 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
14918 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
14919 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
14920 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
14921 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
14922 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
14923 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
14924 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
14925 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14926 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
14927 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
14933 @samp{cortex-m0plus},
14934 @samp{cortex-m1.small-multiply},
14935 @samp{cortex-m0.small-multiply},
14936 @samp{cortex-m0plus.small-multiply},
14939 @samp{marvell-pj4},
14940 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
14941 @samp{fa526}, @samp{fa626},
14942 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
14945 Additionally, this option can specify that GCC should tune the performance
14946 of the code for a big.LITTLE system. Permissible names are:
14947 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
14948 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14949 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
14951 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
14952 performance for a blend of processors within architecture @var{arch}.
14953 The aim is to generate code that run well on the current most popular
14954 processors, balancing between optimizations that benefit some CPUs in the
14955 range, and avoiding performance pitfalls of other CPUs. The effects of
14956 this option may change in future GCC versions as CPU models come and go.
14958 @option{-mtune=native} causes the compiler to auto-detect the CPU
14959 of the build computer. At present, this feature is only supported on
14960 GNU/Linux, and not all architectures are recognized. If the auto-detect is
14961 unsuccessful the option has no effect.
14963 @item -mcpu=@var{name}
14965 This specifies the name of the target ARM processor. GCC uses this name
14966 to derive the name of the target ARM architecture (as if specified
14967 by @option{-march}) and the ARM processor type for which to tune for
14968 performance (as if specified by @option{-mtune}). Where this option
14969 is used in conjunction with @option{-march} or @option{-mtune},
14970 those options take precedence over the appropriate part of this option.
14972 Permissible names for this option are the same as those for
14975 @option{-mcpu=generic-@var{arch}} is also permissible, and is
14976 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
14977 See @option{-mtune} for more information.
14979 @option{-mcpu=native} causes the compiler to auto-detect the CPU
14980 of the build computer. At present, this feature is only supported on
14981 GNU/Linux, and not all architectures are recognized. If the auto-detect
14982 is unsuccessful the option has no effect.
14984 @item -mfpu=@var{name}
14986 This specifies what floating-point hardware (or hardware emulation) is
14987 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
14988 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
14989 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
14990 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
14991 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
14992 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
14994 If @option{-msoft-float} is specified this specifies the format of
14995 floating-point values.
14997 If the selected floating-point hardware includes the NEON extension
14998 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
14999 operations are not generated by GCC's auto-vectorization pass unless
15000 @option{-funsafe-math-optimizations} is also specified. This is
15001 because NEON hardware does not fully implement the IEEE 754 standard for
15002 floating-point arithmetic (in particular denormal values are treated as
15003 zero), so the use of NEON instructions may lead to a loss of precision.
15005 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}).
15007 @item -mfp16-format=@var{name}
15008 @opindex mfp16-format
15009 Specify the format of the @code{__fp16} half-precision floating-point type.
15010 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15011 the default is @samp{none}, in which case the @code{__fp16} type is not
15012 defined. @xref{Half-Precision}, for more information.
15014 @item -mstructure-size-boundary=@var{n}
15015 @opindex mstructure-size-boundary
15016 The sizes of all structures and unions are rounded up to a multiple
15017 of the number of bits set by this option. Permissible values are 8, 32
15018 and 64. The default value varies for different toolchains. For the COFF
15019 targeted toolchain the default value is 8. A value of 64 is only allowed
15020 if the underlying ABI supports it.
15022 Specifying a larger number can produce faster, more efficient code, but
15023 can also increase the size of the program. Different values are potentially
15024 incompatible. Code compiled with one value cannot necessarily expect to
15025 work with code or libraries compiled with another value, if they exchange
15026 information using structures or unions.
15028 @item -mabort-on-noreturn
15029 @opindex mabort-on-noreturn
15030 Generate a call to the function @code{abort} at the end of a
15031 @code{noreturn} function. It is executed if the function tries to
15035 @itemx -mno-long-calls
15036 @opindex mlong-calls
15037 @opindex mno-long-calls
15038 Tells the compiler to perform function calls by first loading the
15039 address of the function into a register and then performing a subroutine
15040 call on this register. This switch is needed if the target function
15041 lies outside of the 64-megabyte addressing range of the offset-based
15042 version of subroutine call instruction.
15044 Even if this switch is enabled, not all function calls are turned
15045 into long calls. The heuristic is that static functions, functions
15046 that have the @code{short_call} attribute, functions that are inside
15047 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15048 definitions have already been compiled within the current compilation
15049 unit are not turned into long calls. The exceptions to this rule are
15050 that weak function definitions, functions with the @code{long_call}
15051 attribute or the @code{section} attribute, and functions that are within
15052 the scope of a @code{#pragma long_calls} directive are always
15053 turned into long calls.
15055 This feature is not enabled by default. Specifying
15056 @option{-mno-long-calls} restores the default behavior, as does
15057 placing the function calls within the scope of a @code{#pragma
15058 long_calls_off} directive. Note these switches have no effect on how
15059 the compiler generates code to handle function calls via function
15062 @item -msingle-pic-base
15063 @opindex msingle-pic-base
15064 Treat the register used for PIC addressing as read-only, rather than
15065 loading it in the prologue for each function. The runtime system is
15066 responsible for initializing this register with an appropriate value
15067 before execution begins.
15069 @item -mpic-register=@var{reg}
15070 @opindex mpic-register
15071 Specify the register to be used for PIC addressing.
15072 For standard PIC base case, the default is any suitable register
15073 determined by compiler. For single PIC base case, the default is
15074 @samp{R9} if target is EABI based or stack-checking is enabled,
15075 otherwise the default is @samp{R10}.
15077 @item -mpic-data-is-text-relative
15078 @opindex mpic-data-is-text-relative
15079 Assume that the displacement between the text and data segments is fixed
15080 at static link time. This permits using PC-relative addressing
15081 operations to access data known to be in the data segment. For
15082 non-VxWorks RTP targets, this option is enabled by default. When
15083 disabled on such targets, it will enable @option{-msingle-pic-base} by
15086 @item -mpoke-function-name
15087 @opindex mpoke-function-name
15088 Write the name of each function into the text section, directly
15089 preceding the function prologue. The generated code is similar to this:
15093 .ascii "arm_poke_function_name", 0
15096 .word 0xff000000 + (t1 - t0)
15097 arm_poke_function_name
15099 stmfd sp!, @{fp, ip, lr, pc@}
15103 When performing a stack backtrace, code can inspect the value of
15104 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
15105 location @code{pc - 12} and the top 8 bits are set, then we know that
15106 there is a function name embedded immediately preceding this location
15107 and has length @code{((pc[-3]) & 0xff000000)}.
15114 Select between generating code that executes in ARM and Thumb
15115 states. The default for most configurations is to generate code
15116 that executes in ARM state, but the default can be changed by
15117 configuring GCC with the @option{--with-mode=}@var{state}
15120 You can also override the ARM and Thumb mode for each function
15121 by using the @code{target("thumb")} and @code{target("arm")} function attributes
15122 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
15125 @opindex mtpcs-frame
15126 Generate a stack frame that is compliant with the Thumb Procedure Call
15127 Standard for all non-leaf functions. (A leaf function is one that does
15128 not call any other functions.) The default is @option{-mno-tpcs-frame}.
15130 @item -mtpcs-leaf-frame
15131 @opindex mtpcs-leaf-frame
15132 Generate a stack frame that is compliant with the Thumb Procedure Call
15133 Standard for all leaf functions. (A leaf function is one that does
15134 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
15136 @item -mcallee-super-interworking
15137 @opindex mcallee-super-interworking
15138 Gives all externally visible functions in the file being compiled an ARM
15139 instruction set header which switches to Thumb mode before executing the
15140 rest of the function. This allows these functions to be called from
15141 non-interworking code. This option is not valid in AAPCS configurations
15142 because interworking is enabled by default.
15144 @item -mcaller-super-interworking
15145 @opindex mcaller-super-interworking
15146 Allows calls via function pointers (including virtual functions) to
15147 execute correctly regardless of whether the target code has been
15148 compiled for interworking or not. There is a small overhead in the cost
15149 of executing a function pointer if this option is enabled. This option
15150 is not valid in AAPCS configurations because interworking is enabled
15153 @item -mtp=@var{name}
15155 Specify the access model for the thread local storage pointer. The valid
15156 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
15157 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
15158 (supported in the arm6k architecture), and @samp{auto}, which uses the
15159 best available method for the selected processor. The default setting is
15162 @item -mtls-dialect=@var{dialect}
15163 @opindex mtls-dialect
15164 Specify the dialect to use for accessing thread local storage. Two
15165 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
15166 @samp{gnu} dialect selects the original GNU scheme for supporting
15167 local and global dynamic TLS models. The @samp{gnu2} dialect
15168 selects the GNU descriptor scheme, which provides better performance
15169 for shared libraries. The GNU descriptor scheme is compatible with
15170 the original scheme, but does require new assembler, linker and
15171 library support. Initial and local exec TLS models are unaffected by
15172 this option and always use the original scheme.
15174 @item -mword-relocations
15175 @opindex mword-relocations
15176 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
15177 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
15178 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
15181 @item -mfix-cortex-m3-ldrd
15182 @opindex mfix-cortex-m3-ldrd
15183 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
15184 with overlapping destination and base registers are used. This option avoids
15185 generating these instructions. This option is enabled by default when
15186 @option{-mcpu=cortex-m3} is specified.
15188 @item -munaligned-access
15189 @itemx -mno-unaligned-access
15190 @opindex munaligned-access
15191 @opindex mno-unaligned-access
15192 Enables (or disables) reading and writing of 16- and 32- bit values
15193 from addresses that are not 16- or 32- bit aligned. By default
15194 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
15195 ARMv8-M Baseline architectures, and enabled for all other
15196 architectures. If unaligned access is not enabled then words in packed
15197 data structures are accessed a byte at a time.
15199 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
15200 generated object file to either true or false, depending upon the
15201 setting of this option. If unaligned access is enabled then the
15202 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
15205 @item -mneon-for-64bits
15206 @opindex mneon-for-64bits
15207 Enables using Neon to handle scalar 64-bits operations. This is
15208 disabled by default since the cost of moving data from core registers
15211 @item -mslow-flash-data
15212 @opindex mslow-flash-data
15213 Assume loading data from flash is slower than fetching instruction.
15214 Therefore literal load is minimized for better performance.
15215 This option is only supported when compiling for ARMv7 M-profile and
15218 @item -masm-syntax-unified
15219 @opindex masm-syntax-unified
15220 Assume inline assembler is using unified asm syntax. The default is
15221 currently off which implies divided syntax. This option has no impact
15222 on Thumb2. However, this may change in future releases of GCC.
15223 Divided syntax should be considered deprecated.
15225 @item -mrestrict-it
15226 @opindex mrestrict-it
15227 Restricts generation of IT blocks to conform to the rules of ARMv8.
15228 IT blocks can only contain a single 16-bit instruction from a select
15229 set of instructions. This option is on by default for ARMv8 Thumb mode.
15231 @item -mprint-tune-info
15232 @opindex mprint-tune-info
15233 Print CPU tuning information as comment in assembler file. This is
15234 an option used only for regression testing of the compiler and not
15235 intended for ordinary use in compiling code. This option is disabled
15239 @opindex mpure-code
15240 Do not allow constant data to be placed in code sections.
15241 Additionally, when compiling for ELF object format give all text sections the
15242 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
15243 is only available when generating non-pic code for ARMv7-M targets.
15248 @subsection AVR Options
15249 @cindex AVR Options
15251 These options are defined for AVR implementations:
15254 @item -mmcu=@var{mcu}
15256 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
15258 The default for this option is@tie{}@samp{avr2}.
15260 GCC supports the following AVR devices and ISAs:
15262 @include avr-mmcu.texi
15264 @item -maccumulate-args
15265 @opindex maccumulate-args
15266 Accumulate outgoing function arguments and acquire/release the needed
15267 stack space for outgoing function arguments once in function
15268 prologue/epilogue. Without this option, outgoing arguments are pushed
15269 before calling a function and popped afterwards.
15271 Popping the arguments after the function call can be expensive on
15272 AVR so that accumulating the stack space might lead to smaller
15273 executables because arguments need not to be removed from the
15274 stack after such a function call.
15276 This option can lead to reduced code size for functions that perform
15277 several calls to functions that get their arguments on the stack like
15278 calls to printf-like functions.
15280 @item -mbranch-cost=@var{cost}
15281 @opindex mbranch-cost
15282 Set the branch costs for conditional branch instructions to
15283 @var{cost}. Reasonable values for @var{cost} are small, non-negative
15284 integers. The default branch cost is 0.
15286 @item -mcall-prologues
15287 @opindex mcall-prologues
15288 Functions prologues/epilogues are expanded as calls to appropriate
15289 subroutines. Code size is smaller.
15293 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
15294 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
15295 and @code{long long} is 4 bytes. Please note that this option does not
15296 conform to the C standards, but it results in smaller code
15299 @item -mn-flash=@var{num}
15301 Assume that the flash memory has a size of
15302 @var{num} times 64@tie{}KiB.
15304 @item -mno-interrupts
15305 @opindex mno-interrupts
15306 Generated code is not compatible with hardware interrupts.
15307 Code size is smaller.
15311 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
15312 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
15313 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
15314 the assembler's command line and the @option{--relax} option to the
15315 linker's command line.
15317 Jump relaxing is performed by the linker because jump offsets are not
15318 known before code is located. Therefore, the assembler code generated by the
15319 compiler is the same, but the instructions in the executable may
15320 differ from instructions in the assembler code.
15322 Relaxing must be turned on if linker stubs are needed, see the
15323 section on @code{EIND} and linker stubs below.
15327 Assume that the device supports the Read-Modify-Write
15328 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
15332 Treat the stack pointer register as an 8-bit register,
15333 i.e.@: assume the high byte of the stack pointer is zero.
15334 In general, you don't need to set this option by hand.
15336 This option is used internally by the compiler to select and
15337 build multilibs for architectures @code{avr2} and @code{avr25}.
15338 These architectures mix devices with and without @code{SPH}.
15339 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
15340 the compiler driver adds or removes this option from the compiler
15341 proper's command line, because the compiler then knows if the device
15342 or architecture has an 8-bit stack pointer and thus no @code{SPH}
15347 Use address register @code{X} in a way proposed by the hardware. This means
15348 that @code{X} is only used in indirect, post-increment or
15349 pre-decrement addressing.
15351 Without this option, the @code{X} register may be used in the same way
15352 as @code{Y} or @code{Z} which then is emulated by additional
15354 For example, loading a value with @code{X+const} addressing with a
15355 small non-negative @code{const < 64} to a register @var{Rn} is
15359 adiw r26, const ; X += const
15360 ld @var{Rn}, X ; @var{Rn} = *X
15361 sbiw r26, const ; X -= const
15365 @opindex mtiny-stack
15366 Only change the lower 8@tie{}bits of the stack pointer.
15368 @item -mfract-convert-truncate
15369 @opindex mfract-convert-truncate
15370 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
15373 @opindex nodevicelib
15374 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
15376 @item -Waddr-space-convert
15377 @opindex Waddr-space-convert
15378 Warn about conversions between address spaces in the case where the
15379 resulting address space is not contained in the incoming address space.
15381 @item -Wmisspelled-isr
15382 @opindex Wmisspelled-isr
15383 Warn if the ISR is misspelled, i.e. without __vector prefix.
15384 Enabled by default.
15387 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
15388 @cindex @code{EIND}
15389 Pointers in the implementation are 16@tie{}bits wide.
15390 The address of a function or label is represented as word address so
15391 that indirect jumps and calls can target any code address in the
15392 range of 64@tie{}Ki words.
15394 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
15395 bytes of program memory space, there is a special function register called
15396 @code{EIND} that serves as most significant part of the target address
15397 when @code{EICALL} or @code{EIJMP} instructions are used.
15399 Indirect jumps and calls on these devices are handled as follows by
15400 the compiler and are subject to some limitations:
15405 The compiler never sets @code{EIND}.
15408 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
15409 instructions or might read @code{EIND} directly in order to emulate an
15410 indirect call/jump by means of a @code{RET} instruction.
15413 The compiler assumes that @code{EIND} never changes during the startup
15414 code or during the application. In particular, @code{EIND} is not
15415 saved/restored in function or interrupt service routine
15419 For indirect calls to functions and computed goto, the linker
15420 generates @emph{stubs}. Stubs are jump pads sometimes also called
15421 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
15422 The stub contains a direct jump to the desired address.
15425 Linker relaxation must be turned on so that the linker generates
15426 the stubs correctly in all situations. See the compiler option
15427 @option{-mrelax} and the linker option @option{--relax}.
15428 There are corner cases where the linker is supposed to generate stubs
15429 but aborts without relaxation and without a helpful error message.
15432 The default linker script is arranged for code with @code{EIND = 0}.
15433 If code is supposed to work for a setup with @code{EIND != 0}, a custom
15434 linker script has to be used in order to place the sections whose
15435 name start with @code{.trampolines} into the segment where @code{EIND}
15439 The startup code from libgcc never sets @code{EIND}.
15440 Notice that startup code is a blend of code from libgcc and AVR-LibC.
15441 For the impact of AVR-LibC on @code{EIND}, see the
15442 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
15445 It is legitimate for user-specific startup code to set up @code{EIND}
15446 early, for example by means of initialization code located in
15447 section @code{.init3}. Such code runs prior to general startup code
15448 that initializes RAM and calls constructors, but after the bit
15449 of startup code from AVR-LibC that sets @code{EIND} to the segment
15450 where the vector table is located.
15452 #include <avr/io.h>
15455 __attribute__((section(".init3"),naked,used,no_instrument_function))
15456 init3_set_eind (void)
15458 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
15459 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
15464 The @code{__trampolines_start} symbol is defined in the linker script.
15467 Stubs are generated automatically by the linker if
15468 the following two conditions are met:
15471 @item The address of a label is taken by means of the @code{gs} modifier
15472 (short for @emph{generate stubs}) like so:
15474 LDI r24, lo8(gs(@var{func}))
15475 LDI r25, hi8(gs(@var{func}))
15477 @item The final location of that label is in a code segment
15478 @emph{outside} the segment where the stubs are located.
15482 The compiler emits such @code{gs} modifiers for code labels in the
15483 following situations:
15485 @item Taking address of a function or code label.
15486 @item Computed goto.
15487 @item If prologue-save function is used, see @option{-mcall-prologues}
15488 command-line option.
15489 @item Switch/case dispatch tables. If you do not want such dispatch
15490 tables you can specify the @option{-fno-jump-tables} command-line option.
15491 @item C and C++ constructors/destructors called during startup/shutdown.
15492 @item If the tools hit a @code{gs()} modifier explained above.
15496 Jumping to non-symbolic addresses like so is @emph{not} supported:
15501 /* Call function at word address 0x2 */
15502 return ((int(*)(void)) 0x2)();
15506 Instead, a stub has to be set up, i.e.@: the function has to be called
15507 through a symbol (@code{func_4} in the example):
15512 extern int func_4 (void);
15514 /* Call function at byte address 0x4 */
15519 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
15520 Alternatively, @code{func_4} can be defined in the linker script.
15523 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
15524 @cindex @code{RAMPD}
15525 @cindex @code{RAMPX}
15526 @cindex @code{RAMPY}
15527 @cindex @code{RAMPZ}
15528 Some AVR devices support memories larger than the 64@tie{}KiB range
15529 that can be accessed with 16-bit pointers. To access memory locations
15530 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
15531 register is used as high part of the address:
15532 The @code{X}, @code{Y}, @code{Z} address register is concatenated
15533 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
15534 register, respectively, to get a wide address. Similarly,
15535 @code{RAMPD} is used together with direct addressing.
15539 The startup code initializes the @code{RAMP} special function
15540 registers with zero.
15543 If a @ref{AVR Named Address Spaces,named address space} other than
15544 generic or @code{__flash} is used, then @code{RAMPZ} is set
15545 as needed before the operation.
15548 If the device supports RAM larger than 64@tie{}KiB and the compiler
15549 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
15550 is reset to zero after the operation.
15553 If the device comes with a specific @code{RAMP} register, the ISR
15554 prologue/epilogue saves/restores that SFR and initializes it with
15555 zero in case the ISR code might (implicitly) use it.
15558 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
15559 If you use inline assembler to read from locations outside the
15560 16-bit address range and change one of the @code{RAMP} registers,
15561 you must reset it to zero after the access.
15565 @subsubsection AVR Built-in Macros
15567 GCC defines several built-in macros so that the user code can test
15568 for the presence or absence of features. Almost any of the following
15569 built-in macros are deduced from device capabilities and thus
15570 triggered by the @option{-mmcu=} command-line option.
15572 For even more AVR-specific built-in macros see
15573 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
15578 Build-in macro that resolves to a decimal number that identifies the
15579 architecture and depends on the @option{-mmcu=@var{mcu}} option.
15580 Possible values are:
15582 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
15583 @code{4}, @code{5}, @code{51}, @code{6}
15585 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
15586 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
15590 @code{100}, @code{102}, @code{104},
15591 @code{105}, @code{106}, @code{107}
15593 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
15594 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
15595 If @var{mcu} specifies a device, this built-in macro is set
15596 accordingly. For example, with @option{-mmcu=atmega8} the macro is
15597 defined to @code{4}.
15599 @item __AVR_@var{Device}__
15600 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
15601 the device's name. For example, @option{-mmcu=atmega8} defines the
15602 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
15603 @code{__AVR_ATtiny261A__}, etc.
15605 The built-in macros' names follow
15606 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
15607 the device name as from the AVR user manual. The difference between
15608 @var{Device} in the built-in macro and @var{device} in
15609 @option{-mmcu=@var{device}} is that the latter is always lowercase.
15611 If @var{device} is not a device but only a core architecture like
15612 @samp{avr51}, this macro is not defined.
15614 @item __AVR_DEVICE_NAME__
15615 Setting @option{-mmcu=@var{device}} defines this built-in macro to
15616 the device's name. For example, with @option{-mmcu=atmega8} the macro
15617 is defined to @code{atmega8}.
15619 If @var{device} is not a device but only a core architecture like
15620 @samp{avr51}, this macro is not defined.
15622 @item __AVR_XMEGA__
15623 The device / architecture belongs to the XMEGA family of devices.
15625 @item __AVR_HAVE_ELPM__
15626 The device has the @code{ELPM} instruction.
15628 @item __AVR_HAVE_ELPMX__
15629 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
15630 R@var{n},Z+} instructions.
15632 @item __AVR_HAVE_MOVW__
15633 The device has the @code{MOVW} instruction to perform 16-bit
15634 register-register moves.
15636 @item __AVR_HAVE_LPMX__
15637 The device has the @code{LPM R@var{n},Z} and
15638 @code{LPM R@var{n},Z+} instructions.
15640 @item __AVR_HAVE_MUL__
15641 The device has a hardware multiplier.
15643 @item __AVR_HAVE_JMP_CALL__
15644 The device has the @code{JMP} and @code{CALL} instructions.
15645 This is the case for devices with at least 16@tie{}KiB of program
15648 @item __AVR_HAVE_EIJMP_EICALL__
15649 @itemx __AVR_3_BYTE_PC__
15650 The device has the @code{EIJMP} and @code{EICALL} instructions.
15651 This is the case for devices with more than 128@tie{}KiB of program memory.
15652 This also means that the program counter
15653 (PC) is 3@tie{}bytes wide.
15655 @item __AVR_2_BYTE_PC__
15656 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
15657 with up to 128@tie{}KiB of program memory.
15659 @item __AVR_HAVE_8BIT_SP__
15660 @itemx __AVR_HAVE_16BIT_SP__
15661 The stack pointer (SP) register is treated as 8-bit respectively
15662 16-bit register by the compiler.
15663 The definition of these macros is affected by @option{-mtiny-stack}.
15665 @item __AVR_HAVE_SPH__
15667 The device has the SPH (high part of stack pointer) special function
15668 register or has an 8-bit stack pointer, respectively.
15669 The definition of these macros is affected by @option{-mmcu=} and
15670 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
15673 @item __AVR_HAVE_RAMPD__
15674 @itemx __AVR_HAVE_RAMPX__
15675 @itemx __AVR_HAVE_RAMPY__
15676 @itemx __AVR_HAVE_RAMPZ__
15677 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
15678 @code{RAMPZ} special function register, respectively.
15680 @item __NO_INTERRUPTS__
15681 This macro reflects the @option{-mno-interrupts} command-line option.
15683 @item __AVR_ERRATA_SKIP__
15684 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
15685 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
15686 instructions because of a hardware erratum. Skip instructions are
15687 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
15688 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
15691 @item __AVR_ISA_RMW__
15692 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
15694 @item __AVR_SFR_OFFSET__=@var{offset}
15695 Instructions that can address I/O special function registers directly
15696 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
15697 address as if addressed by an instruction to access RAM like @code{LD}
15698 or @code{STS}. This offset depends on the device architecture and has
15699 to be subtracted from the RAM address in order to get the
15700 respective I/O@tie{}address.
15702 @item __WITH_AVRLIBC__
15703 The compiler is configured to be used together with AVR-Libc.
15704 See the @option{--with-avrlibc} configure option.
15708 @node Blackfin Options
15709 @subsection Blackfin Options
15710 @cindex Blackfin Options
15713 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
15715 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
15716 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
15717 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
15718 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
15719 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
15720 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
15721 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
15722 @samp{bf561}, @samp{bf592}.
15724 The optional @var{sirevision} specifies the silicon revision of the target
15725 Blackfin processor. Any workarounds available for the targeted silicon revision
15726 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
15727 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
15728 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
15729 hexadecimal digits representing the major and minor numbers in the silicon
15730 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
15731 is not defined. If @var{sirevision} is @samp{any}, the
15732 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
15733 If this optional @var{sirevision} is not used, GCC assumes the latest known
15734 silicon revision of the targeted Blackfin processor.
15736 GCC defines a preprocessor macro for the specified @var{cpu}.
15737 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
15738 provided by libgloss to be linked in if @option{-msim} is not given.
15740 Without this option, @samp{bf532} is used as the processor by default.
15742 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
15743 only the preprocessor macro is defined.
15747 Specifies that the program will be run on the simulator. This causes
15748 the simulator BSP provided by libgloss to be linked in. This option
15749 has effect only for @samp{bfin-elf} toolchain.
15750 Certain other options, such as @option{-mid-shared-library} and
15751 @option{-mfdpic}, imply @option{-msim}.
15753 @item -momit-leaf-frame-pointer
15754 @opindex momit-leaf-frame-pointer
15755 Don't keep the frame pointer in a register for leaf functions. This
15756 avoids the instructions to save, set up and restore frame pointers and
15757 makes an extra register available in leaf functions. The option
15758 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
15759 which might make debugging harder.
15761 @item -mspecld-anomaly
15762 @opindex mspecld-anomaly
15763 When enabled, the compiler ensures that the generated code does not
15764 contain speculative loads after jump instructions. If this option is used,
15765 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
15767 @item -mno-specld-anomaly
15768 @opindex mno-specld-anomaly
15769 Don't generate extra code to prevent speculative loads from occurring.
15771 @item -mcsync-anomaly
15772 @opindex mcsync-anomaly
15773 When enabled, the compiler ensures that the generated code does not
15774 contain CSYNC or SSYNC instructions too soon after conditional branches.
15775 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
15777 @item -mno-csync-anomaly
15778 @opindex mno-csync-anomaly
15779 Don't generate extra code to prevent CSYNC or SSYNC instructions from
15780 occurring too soon after a conditional branch.
15784 When enabled, the compiler is free to take advantage of the knowledge that
15785 the entire program fits into the low 64k of memory.
15788 @opindex mno-low-64k
15789 Assume that the program is arbitrarily large. This is the default.
15791 @item -mstack-check-l1
15792 @opindex mstack-check-l1
15793 Do stack checking using information placed into L1 scratchpad memory by the
15796 @item -mid-shared-library
15797 @opindex mid-shared-library
15798 Generate code that supports shared libraries via the library ID method.
15799 This allows for execute in place and shared libraries in an environment
15800 without virtual memory management. This option implies @option{-fPIC}.
15801 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15803 @item -mno-id-shared-library
15804 @opindex mno-id-shared-library
15805 Generate code that doesn't assume ID-based shared libraries are being used.
15806 This is the default.
15808 @item -mleaf-id-shared-library
15809 @opindex mleaf-id-shared-library
15810 Generate code that supports shared libraries via the library ID method,
15811 but assumes that this library or executable won't link against any other
15812 ID shared libraries. That allows the compiler to use faster code for jumps
15815 @item -mno-leaf-id-shared-library
15816 @opindex mno-leaf-id-shared-library
15817 Do not assume that the code being compiled won't link against any ID shared
15818 libraries. Slower code is generated for jump and call insns.
15820 @item -mshared-library-id=n
15821 @opindex mshared-library-id
15822 Specifies the identification number of the ID-based shared library being
15823 compiled. Specifying a value of 0 generates more compact code; specifying
15824 other values forces the allocation of that number to the current
15825 library but is no more space- or time-efficient than omitting this option.
15829 Generate code that allows the data segment to be located in a different
15830 area of memory from the text segment. This allows for execute in place in
15831 an environment without virtual memory management by eliminating relocations
15832 against the text section.
15834 @item -mno-sep-data
15835 @opindex mno-sep-data
15836 Generate code that assumes that the data segment follows the text segment.
15837 This is the default.
15840 @itemx -mno-long-calls
15841 @opindex mlong-calls
15842 @opindex mno-long-calls
15843 Tells the compiler to perform function calls by first loading the
15844 address of the function into a register and then performing a subroutine
15845 call on this register. This switch is needed if the target function
15846 lies outside of the 24-bit addressing range of the offset-based
15847 version of subroutine call instruction.
15849 This feature is not enabled by default. Specifying
15850 @option{-mno-long-calls} restores the default behavior. Note these
15851 switches have no effect on how the compiler generates code to handle
15852 function calls via function pointers.
15856 Link with the fast floating-point library. This library relaxes some of
15857 the IEEE floating-point standard's rules for checking inputs against
15858 Not-a-Number (NAN), in the interest of performance.
15861 @opindex minline-plt
15862 Enable inlining of PLT entries in function calls to functions that are
15863 not known to bind locally. It has no effect without @option{-mfdpic}.
15866 @opindex mmulticore
15867 Build a standalone application for multicore Blackfin processors.
15868 This option causes proper start files and link scripts supporting
15869 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
15870 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
15872 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
15873 selects the one-application-per-core programming model. Without
15874 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
15875 programming model is used. In this model, the main function of Core B
15876 should be named as @code{coreb_main}.
15878 If this option is not used, the single-core application programming
15883 Build a standalone application for Core A of BF561 when using
15884 the one-application-per-core programming model. Proper start files
15885 and link scripts are used to support Core A, and the macro
15886 @code{__BFIN_COREA} is defined.
15887 This option can only be used in conjunction with @option{-mmulticore}.
15891 Build a standalone application for Core B of BF561 when using
15892 the one-application-per-core programming model. Proper start files
15893 and link scripts are used to support Core B, and the macro
15894 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
15895 should be used instead of @code{main}.
15896 This option can only be used in conjunction with @option{-mmulticore}.
15900 Build a standalone application for SDRAM. Proper start files and
15901 link scripts are used to put the application into SDRAM, and the macro
15902 @code{__BFIN_SDRAM} is defined.
15903 The loader should initialize SDRAM before loading the application.
15907 Assume that ICPLBs are enabled at run time. This has an effect on certain
15908 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
15909 are enabled; for standalone applications the default is off.
15913 @subsection C6X Options
15914 @cindex C6X Options
15917 @item -march=@var{name}
15919 This specifies the name of the target architecture. GCC uses this
15920 name to determine what kind of instructions it can emit when generating
15921 assembly code. Permissible names are: @samp{c62x},
15922 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
15925 @opindex mbig-endian
15926 Generate code for a big-endian target.
15928 @item -mlittle-endian
15929 @opindex mlittle-endian
15930 Generate code for a little-endian target. This is the default.
15934 Choose startup files and linker script suitable for the simulator.
15936 @item -msdata=default
15937 @opindex msdata=default
15938 Put small global and static data in the @code{.neardata} section,
15939 which is pointed to by register @code{B14}. Put small uninitialized
15940 global and static data in the @code{.bss} section, which is adjacent
15941 to the @code{.neardata} section. Put small read-only data into the
15942 @code{.rodata} section. The corresponding sections used for large
15943 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
15946 @opindex msdata=all
15947 Put all data, not just small objects, into the sections reserved for
15948 small data, and use addressing relative to the @code{B14} register to
15952 @opindex msdata=none
15953 Make no use of the sections reserved for small data, and use absolute
15954 addresses to access all data. Put all initialized global and static
15955 data in the @code{.fardata} section, and all uninitialized data in the
15956 @code{.far} section. Put all constant data into the @code{.const}
15961 @subsection CRIS Options
15962 @cindex CRIS Options
15964 These options are defined specifically for the CRIS ports.
15967 @item -march=@var{architecture-type}
15968 @itemx -mcpu=@var{architecture-type}
15971 Generate code for the specified architecture. The choices for
15972 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
15973 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
15974 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
15977 @item -mtune=@var{architecture-type}
15979 Tune to @var{architecture-type} everything applicable about the generated
15980 code, except for the ABI and the set of available instructions. The
15981 choices for @var{architecture-type} are the same as for
15982 @option{-march=@var{architecture-type}}.
15984 @item -mmax-stack-frame=@var{n}
15985 @opindex mmax-stack-frame
15986 Warn when the stack frame of a function exceeds @var{n} bytes.
15992 The options @option{-metrax4} and @option{-metrax100} are synonyms for
15993 @option{-march=v3} and @option{-march=v8} respectively.
15995 @item -mmul-bug-workaround
15996 @itemx -mno-mul-bug-workaround
15997 @opindex mmul-bug-workaround
15998 @opindex mno-mul-bug-workaround
15999 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16000 models where it applies. This option is active by default.
16004 Enable CRIS-specific verbose debug-related information in the assembly
16005 code. This option also has the effect of turning off the @samp{#NO_APP}
16006 formatted-code indicator to the assembler at the beginning of the
16011 Do not use condition-code results from previous instruction; always emit
16012 compare and test instructions before use of condition codes.
16014 @item -mno-side-effects
16015 @opindex mno-side-effects
16016 Do not emit instructions with side effects in addressing modes other than
16019 @item -mstack-align
16020 @itemx -mno-stack-align
16021 @itemx -mdata-align
16022 @itemx -mno-data-align
16023 @itemx -mconst-align
16024 @itemx -mno-const-align
16025 @opindex mstack-align
16026 @opindex mno-stack-align
16027 @opindex mdata-align
16028 @opindex mno-data-align
16029 @opindex mconst-align
16030 @opindex mno-const-align
16031 These options (@samp{no-} options) arrange (eliminate arrangements) for the
16032 stack frame, individual data and constants to be aligned for the maximum
16033 single data access size for the chosen CPU model. The default is to
16034 arrange for 32-bit alignment. ABI details such as structure layout are
16035 not affected by these options.
16043 Similar to the stack- data- and const-align options above, these options
16044 arrange for stack frame, writable data and constants to all be 32-bit,
16045 16-bit or 8-bit aligned. The default is 32-bit alignment.
16047 @item -mno-prologue-epilogue
16048 @itemx -mprologue-epilogue
16049 @opindex mno-prologue-epilogue
16050 @opindex mprologue-epilogue
16051 With @option{-mno-prologue-epilogue}, the normal function prologue and
16052 epilogue which set up the stack frame are omitted and no return
16053 instructions or return sequences are generated in the code. Use this
16054 option only together with visual inspection of the compiled code: no
16055 warnings or errors are generated when call-saved registers must be saved,
16056 or storage for local variables needs to be allocated.
16060 @opindex mno-gotplt
16062 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
16063 instruction sequences that load addresses for functions from the PLT part
16064 of the GOT rather than (traditional on other architectures) calls to the
16065 PLT@. The default is @option{-mgotplt}.
16069 Legacy no-op option only recognized with the cris-axis-elf and
16070 cris-axis-linux-gnu targets.
16074 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
16078 This option, recognized for the cris-axis-elf, arranges
16079 to link with input-output functions from a simulator library. Code,
16080 initialized data and zero-initialized data are allocated consecutively.
16084 Like @option{-sim}, but pass linker options to locate initialized data at
16085 0x40000000 and zero-initialized data at 0x80000000.
16089 @subsection CR16 Options
16090 @cindex CR16 Options
16092 These options are defined specifically for the CR16 ports.
16098 Enable the use of multiply-accumulate instructions. Disabled by default.
16102 @opindex mcr16cplus
16104 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
16109 Links the library libsim.a which is in compatible with simulator. Applicable
16110 to ELF compiler only.
16114 Choose integer type as 32-bit wide.
16118 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
16120 @item -mdata-model=@var{model}
16121 @opindex mdata-model
16122 Choose a data model. The choices for @var{model} are @samp{near},
16123 @samp{far} or @samp{medium}. @samp{medium} is default.
16124 However, @samp{far} is not valid with @option{-mcr16c}, as the
16125 CR16C architecture does not support the far data model.
16128 @node Darwin Options
16129 @subsection Darwin Options
16130 @cindex Darwin options
16132 These options are defined for all architectures running the Darwin operating
16135 FSF GCC on Darwin does not create ``fat'' object files; it creates
16136 an object file for the single architecture that GCC was built to
16137 target. Apple's GCC on Darwin does create ``fat'' files if multiple
16138 @option{-arch} options are used; it does so by running the compiler or
16139 linker multiple times and joining the results together with
16142 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
16143 @samp{i686}) is determined by the flags that specify the ISA
16144 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
16145 @option{-force_cpusubtype_ALL} option can be used to override this.
16147 The Darwin tools vary in their behavior when presented with an ISA
16148 mismatch. The assembler, @file{as}, only permits instructions to
16149 be used that are valid for the subtype of the file it is generating,
16150 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
16151 The linker for shared libraries, @file{/usr/bin/libtool}, fails
16152 and prints an error if asked to create a shared library with a less
16153 restrictive subtype than its input files (for instance, trying to put
16154 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
16155 for executables, @command{ld}, quietly gives the executable the most
16156 restrictive subtype of any of its input files.
16161 Add the framework directory @var{dir} to the head of the list of
16162 directories to be searched for header files. These directories are
16163 interleaved with those specified by @option{-I} options and are
16164 scanned in a left-to-right order.
16166 A framework directory is a directory with frameworks in it. A
16167 framework is a directory with a @file{Headers} and/or
16168 @file{PrivateHeaders} directory contained directly in it that ends
16169 in @file{.framework}. The name of a framework is the name of this
16170 directory excluding the @file{.framework}. Headers associated with
16171 the framework are found in one of those two directories, with
16172 @file{Headers} being searched first. A subframework is a framework
16173 directory that is in a framework's @file{Frameworks} directory.
16174 Includes of subframework headers can only appear in a header of a
16175 framework that contains the subframework, or in a sibling subframework
16176 header. Two subframeworks are siblings if they occur in the same
16177 framework. A subframework should not have the same name as a
16178 framework; a warning is issued if this is violated. Currently a
16179 subframework cannot have subframeworks; in the future, the mechanism
16180 may be extended to support this. The standard frameworks can be found
16181 in @file{/System/Library/Frameworks} and
16182 @file{/Library/Frameworks}. An example include looks like
16183 @code{#include <Framework/header.h>}, where @file{Framework} denotes
16184 the name of the framework and @file{header.h} is found in the
16185 @file{PrivateHeaders} or @file{Headers} directory.
16187 @item -iframework@var{dir}
16188 @opindex iframework
16189 Like @option{-F} except the directory is a treated as a system
16190 directory. The main difference between this @option{-iframework} and
16191 @option{-F} is that with @option{-iframework} the compiler does not
16192 warn about constructs contained within header files found via
16193 @var{dir}. This option is valid only for the C family of languages.
16197 Emit debugging information for symbols that are used. For stabs
16198 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
16199 This is by default ON@.
16203 Emit debugging information for all symbols and types.
16205 @item -mmacosx-version-min=@var{version}
16206 The earliest version of MacOS X that this executable will run on
16207 is @var{version}. Typical values of @var{version} include @code{10.1},
16208 @code{10.2}, and @code{10.3.9}.
16210 If the compiler was built to use the system's headers by default,
16211 then the default for this option is the system version on which the
16212 compiler is running, otherwise the default is to make choices that
16213 are compatible with as many systems and code bases as possible.
16217 Enable kernel development mode. The @option{-mkernel} option sets
16218 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
16219 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
16220 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
16221 applicable. This mode also sets @option{-mno-altivec},
16222 @option{-msoft-float}, @option{-fno-builtin} and
16223 @option{-mlong-branch} for PowerPC targets.
16225 @item -mone-byte-bool
16226 @opindex mone-byte-bool
16227 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
16228 By default @code{sizeof(bool)} is @code{4} when compiling for
16229 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
16230 option has no effect on x86.
16232 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
16233 to generate code that is not binary compatible with code generated
16234 without that switch. Using this switch may require recompiling all
16235 other modules in a program, including system libraries. Use this
16236 switch to conform to a non-default data model.
16238 @item -mfix-and-continue
16239 @itemx -ffix-and-continue
16240 @itemx -findirect-data
16241 @opindex mfix-and-continue
16242 @opindex ffix-and-continue
16243 @opindex findirect-data
16244 Generate code suitable for fast turnaround development, such as to
16245 allow GDB to dynamically load @file{.o} files into already-running
16246 programs. @option{-findirect-data} and @option{-ffix-and-continue}
16247 are provided for backwards compatibility.
16251 Loads all members of static archive libraries.
16252 See man ld(1) for more information.
16254 @item -arch_errors_fatal
16255 @opindex arch_errors_fatal
16256 Cause the errors having to do with files that have the wrong architecture
16259 @item -bind_at_load
16260 @opindex bind_at_load
16261 Causes the output file to be marked such that the dynamic linker will
16262 bind all undefined references when the file is loaded or launched.
16266 Produce a Mach-o bundle format file.
16267 See man ld(1) for more information.
16269 @item -bundle_loader @var{executable}
16270 @opindex bundle_loader
16271 This option specifies the @var{executable} that will load the build
16272 output file being linked. See man ld(1) for more information.
16275 @opindex dynamiclib
16276 When passed this option, GCC produces a dynamic library instead of
16277 an executable when linking, using the Darwin @file{libtool} command.
16279 @item -force_cpusubtype_ALL
16280 @opindex force_cpusubtype_ALL
16281 This causes GCC's output file to have the @samp{ALL} subtype, instead of
16282 one controlled by the @option{-mcpu} or @option{-march} option.
16284 @item -allowable_client @var{client_name}
16285 @itemx -client_name
16286 @itemx -compatibility_version
16287 @itemx -current_version
16289 @itemx -dependency-file
16291 @itemx -dylinker_install_name
16293 @itemx -exported_symbols_list
16296 @itemx -flat_namespace
16297 @itemx -force_flat_namespace
16298 @itemx -headerpad_max_install_names
16301 @itemx -install_name
16302 @itemx -keep_private_externs
16303 @itemx -multi_module
16304 @itemx -multiply_defined
16305 @itemx -multiply_defined_unused
16308 @itemx -no_dead_strip_inits_and_terms
16309 @itemx -nofixprebinding
16310 @itemx -nomultidefs
16312 @itemx -noseglinkedit
16313 @itemx -pagezero_size
16315 @itemx -prebind_all_twolevel_modules
16316 @itemx -private_bundle
16318 @itemx -read_only_relocs
16320 @itemx -sectobjectsymbols
16324 @itemx -sectobjectsymbols
16327 @itemx -segs_read_only_addr
16329 @itemx -segs_read_write_addr
16330 @itemx -seg_addr_table
16331 @itemx -seg_addr_table_filename
16332 @itemx -seglinkedit
16334 @itemx -segs_read_only_addr
16335 @itemx -segs_read_write_addr
16336 @itemx -single_module
16338 @itemx -sub_library
16340 @itemx -sub_umbrella
16341 @itemx -twolevel_namespace
16344 @itemx -unexported_symbols_list
16345 @itemx -weak_reference_mismatches
16346 @itemx -whatsloaded
16347 @opindex allowable_client
16348 @opindex client_name
16349 @opindex compatibility_version
16350 @opindex current_version
16351 @opindex dead_strip
16352 @opindex dependency-file
16353 @opindex dylib_file
16354 @opindex dylinker_install_name
16356 @opindex exported_symbols_list
16358 @opindex flat_namespace
16359 @opindex force_flat_namespace
16360 @opindex headerpad_max_install_names
16361 @opindex image_base
16363 @opindex install_name
16364 @opindex keep_private_externs
16365 @opindex multi_module
16366 @opindex multiply_defined
16367 @opindex multiply_defined_unused
16368 @opindex noall_load
16369 @opindex no_dead_strip_inits_and_terms
16370 @opindex nofixprebinding
16371 @opindex nomultidefs
16373 @opindex noseglinkedit
16374 @opindex pagezero_size
16376 @opindex prebind_all_twolevel_modules
16377 @opindex private_bundle
16378 @opindex read_only_relocs
16380 @opindex sectobjectsymbols
16383 @opindex sectcreate
16384 @opindex sectobjectsymbols
16387 @opindex segs_read_only_addr
16388 @opindex segs_read_write_addr
16389 @opindex seg_addr_table
16390 @opindex seg_addr_table_filename
16391 @opindex seglinkedit
16393 @opindex segs_read_only_addr
16394 @opindex segs_read_write_addr
16395 @opindex single_module
16397 @opindex sub_library
16398 @opindex sub_umbrella
16399 @opindex twolevel_namespace
16402 @opindex unexported_symbols_list
16403 @opindex weak_reference_mismatches
16404 @opindex whatsloaded
16405 These options are passed to the Darwin linker. The Darwin linker man page
16406 describes them in detail.
16409 @node DEC Alpha Options
16410 @subsection DEC Alpha Options
16412 These @samp{-m} options are defined for the DEC Alpha implementations:
16415 @item -mno-soft-float
16416 @itemx -msoft-float
16417 @opindex mno-soft-float
16418 @opindex msoft-float
16419 Use (do not use) the hardware floating-point instructions for
16420 floating-point operations. When @option{-msoft-float} is specified,
16421 functions in @file{libgcc.a} are used to perform floating-point
16422 operations. Unless they are replaced by routines that emulate the
16423 floating-point operations, or compiled in such a way as to call such
16424 emulations routines, these routines issue floating-point
16425 operations. If you are compiling for an Alpha without floating-point
16426 operations, you must ensure that the library is built so as not to call
16429 Note that Alpha implementations without floating-point operations are
16430 required to have floating-point registers.
16433 @itemx -mno-fp-regs
16435 @opindex mno-fp-regs
16436 Generate code that uses (does not use) the floating-point register set.
16437 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
16438 register set is not used, floating-point operands are passed in integer
16439 registers as if they were integers and floating-point results are passed
16440 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
16441 so any function with a floating-point argument or return value called by code
16442 compiled with @option{-mno-fp-regs} must also be compiled with that
16445 A typical use of this option is building a kernel that does not use,
16446 and hence need not save and restore, any floating-point registers.
16450 The Alpha architecture implements floating-point hardware optimized for
16451 maximum performance. It is mostly compliant with the IEEE floating-point
16452 standard. However, for full compliance, software assistance is
16453 required. This option generates code fully IEEE-compliant code
16454 @emph{except} that the @var{inexact-flag} is not maintained (see below).
16455 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
16456 defined during compilation. The resulting code is less efficient but is
16457 able to correctly support denormalized numbers and exceptional IEEE
16458 values such as not-a-number and plus/minus infinity. Other Alpha
16459 compilers call this option @option{-ieee_with_no_inexact}.
16461 @item -mieee-with-inexact
16462 @opindex mieee-with-inexact
16463 This is like @option{-mieee} except the generated code also maintains
16464 the IEEE @var{inexact-flag}. Turning on this option causes the
16465 generated code to implement fully-compliant IEEE math. In addition to
16466 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
16467 macro. On some Alpha implementations the resulting code may execute
16468 significantly slower than the code generated by default. Since there is
16469 very little code that depends on the @var{inexact-flag}, you should
16470 normally not specify this option. Other Alpha compilers call this
16471 option @option{-ieee_with_inexact}.
16473 @item -mfp-trap-mode=@var{trap-mode}
16474 @opindex mfp-trap-mode
16475 This option controls what floating-point related traps are enabled.
16476 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
16477 The trap mode can be set to one of four values:
16481 This is the default (normal) setting. The only traps that are enabled
16482 are the ones that cannot be disabled in software (e.g., division by zero
16486 In addition to the traps enabled by @samp{n}, underflow traps are enabled
16490 Like @samp{u}, but the instructions are marked to be safe for software
16491 completion (see Alpha architecture manual for details).
16494 Like @samp{su}, but inexact traps are enabled as well.
16497 @item -mfp-rounding-mode=@var{rounding-mode}
16498 @opindex mfp-rounding-mode
16499 Selects the IEEE rounding mode. Other Alpha compilers call this option
16500 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
16505 Normal IEEE rounding mode. Floating-point numbers are rounded towards
16506 the nearest machine number or towards the even machine number in case
16510 Round towards minus infinity.
16513 Chopped rounding mode. Floating-point numbers are rounded towards zero.
16516 Dynamic rounding mode. A field in the floating-point control register
16517 (@var{fpcr}, see Alpha architecture reference manual) controls the
16518 rounding mode in effect. The C library initializes this register for
16519 rounding towards plus infinity. Thus, unless your program modifies the
16520 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
16523 @item -mtrap-precision=@var{trap-precision}
16524 @opindex mtrap-precision
16525 In the Alpha architecture, floating-point traps are imprecise. This
16526 means without software assistance it is impossible to recover from a
16527 floating trap and program execution normally needs to be terminated.
16528 GCC can generate code that can assist operating system trap handlers
16529 in determining the exact location that caused a floating-point trap.
16530 Depending on the requirements of an application, different levels of
16531 precisions can be selected:
16535 Program precision. This option is the default and means a trap handler
16536 can only identify which program caused a floating-point exception.
16539 Function precision. The trap handler can determine the function that
16540 caused a floating-point exception.
16543 Instruction precision. The trap handler can determine the exact
16544 instruction that caused a floating-point exception.
16547 Other Alpha compilers provide the equivalent options called
16548 @option{-scope_safe} and @option{-resumption_safe}.
16550 @item -mieee-conformant
16551 @opindex mieee-conformant
16552 This option marks the generated code as IEEE conformant. You must not
16553 use this option unless you also specify @option{-mtrap-precision=i} and either
16554 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
16555 is to emit the line @samp{.eflag 48} in the function prologue of the
16556 generated assembly file.
16558 @item -mbuild-constants
16559 @opindex mbuild-constants
16560 Normally GCC examines a 32- or 64-bit integer constant to
16561 see if it can construct it from smaller constants in two or three
16562 instructions. If it cannot, it outputs the constant as a literal and
16563 generates code to load it from the data segment at run time.
16565 Use this option to require GCC to construct @emph{all} integer constants
16566 using code, even if it takes more instructions (the maximum is six).
16568 You typically use this option to build a shared library dynamic
16569 loader. Itself a shared library, it must relocate itself in memory
16570 before it can find the variables and constants in its own data segment.
16588 Indicate whether GCC should generate code to use the optional BWX,
16589 CIX, FIX and MAX instruction sets. The default is to use the instruction
16590 sets supported by the CPU type specified via @option{-mcpu=} option or that
16591 of the CPU on which GCC was built if none is specified.
16594 @itemx -mfloat-ieee
16595 @opindex mfloat-vax
16596 @opindex mfloat-ieee
16597 Generate code that uses (does not use) VAX F and G floating-point
16598 arithmetic instead of IEEE single and double precision.
16600 @item -mexplicit-relocs
16601 @itemx -mno-explicit-relocs
16602 @opindex mexplicit-relocs
16603 @opindex mno-explicit-relocs
16604 Older Alpha assemblers provided no way to generate symbol relocations
16605 except via assembler macros. Use of these macros does not allow
16606 optimal instruction scheduling. GNU binutils as of version 2.12
16607 supports a new syntax that allows the compiler to explicitly mark
16608 which relocations should apply to which instructions. This option
16609 is mostly useful for debugging, as GCC detects the capabilities of
16610 the assembler when it is built and sets the default accordingly.
16613 @itemx -mlarge-data
16614 @opindex msmall-data
16615 @opindex mlarge-data
16616 When @option{-mexplicit-relocs} is in effect, static data is
16617 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
16618 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
16619 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
16620 16-bit relocations off of the @code{$gp} register. This limits the
16621 size of the small data area to 64KB, but allows the variables to be
16622 directly accessed via a single instruction.
16624 The default is @option{-mlarge-data}. With this option the data area
16625 is limited to just below 2GB@. Programs that require more than 2GB of
16626 data must use @code{malloc} or @code{mmap} to allocate the data in the
16627 heap instead of in the program's data segment.
16629 When generating code for shared libraries, @option{-fpic} implies
16630 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
16633 @itemx -mlarge-text
16634 @opindex msmall-text
16635 @opindex mlarge-text
16636 When @option{-msmall-text} is used, the compiler assumes that the
16637 code of the entire program (or shared library) fits in 4MB, and is
16638 thus reachable with a branch instruction. When @option{-msmall-data}
16639 is used, the compiler can assume that all local symbols share the
16640 same @code{$gp} value, and thus reduce the number of instructions
16641 required for a function call from 4 to 1.
16643 The default is @option{-mlarge-text}.
16645 @item -mcpu=@var{cpu_type}
16647 Set the instruction set and instruction scheduling parameters for
16648 machine type @var{cpu_type}. You can specify either the @samp{EV}
16649 style name or the corresponding chip number. GCC supports scheduling
16650 parameters for the EV4, EV5 and EV6 family of processors and
16651 chooses the default values for the instruction set from the processor
16652 you specify. If you do not specify a processor type, GCC defaults
16653 to the processor on which the compiler was built.
16655 Supported values for @var{cpu_type} are
16661 Schedules as an EV4 and has no instruction set extensions.
16665 Schedules as an EV5 and has no instruction set extensions.
16669 Schedules as an EV5 and supports the BWX extension.
16674 Schedules as an EV5 and supports the BWX and MAX extensions.
16678 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
16682 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
16685 Native toolchains also support the value @samp{native},
16686 which selects the best architecture option for the host processor.
16687 @option{-mcpu=native} has no effect if GCC does not recognize
16690 @item -mtune=@var{cpu_type}
16692 Set only the instruction scheduling parameters for machine type
16693 @var{cpu_type}. The instruction set is not changed.
16695 Native toolchains also support the value @samp{native},
16696 which selects the best architecture option for the host processor.
16697 @option{-mtune=native} has no effect if GCC does not recognize
16700 @item -mmemory-latency=@var{time}
16701 @opindex mmemory-latency
16702 Sets the latency the scheduler should assume for typical memory
16703 references as seen by the application. This number is highly
16704 dependent on the memory access patterns used by the application
16705 and the size of the external cache on the machine.
16707 Valid options for @var{time} are
16711 A decimal number representing clock cycles.
16717 The compiler contains estimates of the number of clock cycles for
16718 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
16719 (also called Dcache, Scache, and Bcache), as well as to main memory.
16720 Note that L3 is only valid for EV5.
16726 @subsection FR30 Options
16727 @cindex FR30 Options
16729 These options are defined specifically for the FR30 port.
16733 @item -msmall-model
16734 @opindex msmall-model
16735 Use the small address space model. This can produce smaller code, but
16736 it does assume that all symbolic values and addresses fit into a
16741 Assume that runtime support has been provided and so there is no need
16742 to include the simulator library (@file{libsim.a}) on the linker
16748 @subsection FT32 Options
16749 @cindex FT32 Options
16751 These options are defined specifically for the FT32 port.
16757 Specifies that the program will be run on the simulator. This causes
16758 an alternate runtime startup and library to be linked.
16759 You must not use this option when generating programs that will run on
16760 real hardware; you must provide your own runtime library for whatever
16761 I/O functions are needed.
16765 Enable Local Register Allocation. This is still experimental for FT32,
16766 so by default the compiler uses standard reload.
16770 Do not use div and mod instructions.
16775 @subsection FRV Options
16776 @cindex FRV Options
16782 Only use the first 32 general-purpose registers.
16787 Use all 64 general-purpose registers.
16792 Use only the first 32 floating-point registers.
16797 Use all 64 floating-point registers.
16800 @opindex mhard-float
16802 Use hardware instructions for floating-point operations.
16805 @opindex msoft-float
16807 Use library routines for floating-point operations.
16812 Dynamically allocate condition code registers.
16817 Do not try to dynamically allocate condition code registers, only
16818 use @code{icc0} and @code{fcc0}.
16823 Change ABI to use double word insns.
16828 Do not use double word instructions.
16833 Use floating-point double instructions.
16836 @opindex mno-double
16838 Do not use floating-point double instructions.
16843 Use media instructions.
16848 Do not use media instructions.
16853 Use multiply and add/subtract instructions.
16856 @opindex mno-muladd
16858 Do not use multiply and add/subtract instructions.
16863 Select the FDPIC ABI, which uses function descriptors to represent
16864 pointers to functions. Without any PIC/PIE-related options, it
16865 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
16866 assumes GOT entries and small data are within a 12-bit range from the
16867 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
16868 are computed with 32 bits.
16869 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16872 @opindex minline-plt
16874 Enable inlining of PLT entries in function calls to functions that are
16875 not known to bind locally. It has no effect without @option{-mfdpic}.
16876 It's enabled by default if optimizing for speed and compiling for
16877 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
16878 optimization option such as @option{-O3} or above is present in the
16884 Assume a large TLS segment when generating thread-local code.
16889 Do not assume a large TLS segment when generating thread-local code.
16894 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
16895 that is known to be in read-only sections. It's enabled by default,
16896 except for @option{-fpic} or @option{-fpie}: even though it may help
16897 make the global offset table smaller, it trades 1 instruction for 4.
16898 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
16899 one of which may be shared by multiple symbols, and it avoids the need
16900 for a GOT entry for the referenced symbol, so it's more likely to be a
16901 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
16903 @item -multilib-library-pic
16904 @opindex multilib-library-pic
16906 Link with the (library, not FD) pic libraries. It's implied by
16907 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
16908 @option{-fpic} without @option{-mfdpic}. You should never have to use
16912 @opindex mlinked-fp
16914 Follow the EABI requirement of always creating a frame pointer whenever
16915 a stack frame is allocated. This option is enabled by default and can
16916 be disabled with @option{-mno-linked-fp}.
16919 @opindex mlong-calls
16921 Use indirect addressing to call functions outside the current
16922 compilation unit. This allows the functions to be placed anywhere
16923 within the 32-bit address space.
16925 @item -malign-labels
16926 @opindex malign-labels
16928 Try to align labels to an 8-byte boundary by inserting NOPs into the
16929 previous packet. This option only has an effect when VLIW packing
16930 is enabled. It doesn't create new packets; it merely adds NOPs to
16933 @item -mlibrary-pic
16934 @opindex mlibrary-pic
16936 Generate position-independent EABI code.
16941 Use only the first four media accumulator registers.
16946 Use all eight media accumulator registers.
16951 Pack VLIW instructions.
16956 Do not pack VLIW instructions.
16959 @opindex mno-eflags
16961 Do not mark ABI switches in e_flags.
16964 @opindex mcond-move
16966 Enable the use of conditional-move instructions (default).
16968 This switch is mainly for debugging the compiler and will likely be removed
16969 in a future version.
16971 @item -mno-cond-move
16972 @opindex mno-cond-move
16974 Disable the use of conditional-move instructions.
16976 This switch is mainly for debugging the compiler and will likely be removed
16977 in a future version.
16982 Enable the use of conditional set instructions (default).
16984 This switch is mainly for debugging the compiler and will likely be removed
16985 in a future version.
16990 Disable the use of conditional set instructions.
16992 This switch is mainly for debugging the compiler and will likely be removed
16993 in a future version.
16996 @opindex mcond-exec
16998 Enable the use of conditional execution (default).
17000 This switch is mainly for debugging the compiler and will likely be removed
17001 in a future version.
17003 @item -mno-cond-exec
17004 @opindex mno-cond-exec
17006 Disable the use of conditional execution.
17008 This switch is mainly for debugging the compiler and will likely be removed
17009 in a future version.
17011 @item -mvliw-branch
17012 @opindex mvliw-branch
17014 Run a pass to pack branches into VLIW instructions (default).
17016 This switch is mainly for debugging the compiler and will likely be removed
17017 in a future version.
17019 @item -mno-vliw-branch
17020 @opindex mno-vliw-branch
17022 Do not run a pass to pack branches into VLIW instructions.
17024 This switch is mainly for debugging the compiler and will likely be removed
17025 in a future version.
17027 @item -mmulti-cond-exec
17028 @opindex mmulti-cond-exec
17030 Enable optimization of @code{&&} and @code{||} in conditional execution
17033 This switch is mainly for debugging the compiler and will likely be removed
17034 in a future version.
17036 @item -mno-multi-cond-exec
17037 @opindex mno-multi-cond-exec
17039 Disable optimization of @code{&&} and @code{||} in conditional execution.
17041 This switch is mainly for debugging the compiler and will likely be removed
17042 in a future version.
17044 @item -mnested-cond-exec
17045 @opindex mnested-cond-exec
17047 Enable nested conditional execution optimizations (default).
17049 This switch is mainly for debugging the compiler and will likely be removed
17050 in a future version.
17052 @item -mno-nested-cond-exec
17053 @opindex mno-nested-cond-exec
17055 Disable nested conditional execution optimizations.
17057 This switch is mainly for debugging the compiler and will likely be removed
17058 in a future version.
17060 @item -moptimize-membar
17061 @opindex moptimize-membar
17063 This switch removes redundant @code{membar} instructions from the
17064 compiler-generated code. It is enabled by default.
17066 @item -mno-optimize-membar
17067 @opindex mno-optimize-membar
17069 This switch disables the automatic removal of redundant @code{membar}
17070 instructions from the generated code.
17072 @item -mtomcat-stats
17073 @opindex mtomcat-stats
17075 Cause gas to print out tomcat statistics.
17077 @item -mcpu=@var{cpu}
17080 Select the processor type for which to generate code. Possible values are
17081 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
17082 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
17086 @node GNU/Linux Options
17087 @subsection GNU/Linux Options
17089 These @samp{-m} options are defined for GNU/Linux targets:
17094 Use the GNU C library. This is the default except
17095 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
17096 @samp{*-*-linux-*android*} targets.
17100 Use uClibc C library. This is the default on
17101 @samp{*-*-linux-*uclibc*} targets.
17105 Use the musl C library. This is the default on
17106 @samp{*-*-linux-*musl*} targets.
17110 Use Bionic C library. This is the default on
17111 @samp{*-*-linux-*android*} targets.
17115 Compile code compatible with Android platform. This is the default on
17116 @samp{*-*-linux-*android*} targets.
17118 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
17119 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
17120 this option makes the GCC driver pass Android-specific options to the linker.
17121 Finally, this option causes the preprocessor macro @code{__ANDROID__}
17124 @item -tno-android-cc
17125 @opindex tno-android-cc
17126 Disable compilation effects of @option{-mandroid}, i.e., do not enable
17127 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
17128 @option{-fno-rtti} by default.
17130 @item -tno-android-ld
17131 @opindex tno-android-ld
17132 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
17133 linking options to the linker.
17137 @node H8/300 Options
17138 @subsection H8/300 Options
17140 These @samp{-m} options are defined for the H8/300 implementations:
17145 Shorten some address references at link time, when possible; uses the
17146 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
17147 ld, Using ld}, for a fuller description.
17151 Generate code for the H8/300H@.
17155 Generate code for the H8S@.
17159 Generate code for the H8S and H8/300H in the normal mode. This switch
17160 must be used either with @option{-mh} or @option{-ms}.
17164 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
17168 Extended registers are stored on stack before execution of function
17169 with monitor attribute. Default option is @option{-mexr}.
17170 This option is valid only for H8S targets.
17174 Extended registers are not stored on stack before execution of function
17175 with monitor attribute. Default option is @option{-mno-exr}.
17176 This option is valid only for H8S targets.
17180 Make @code{int} data 32 bits by default.
17183 @opindex malign-300
17184 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
17185 The default for the H8/300H and H8S is to align longs and floats on
17187 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
17188 This option has no effect on the H8/300.
17192 @subsection HPPA Options
17193 @cindex HPPA Options
17195 These @samp{-m} options are defined for the HPPA family of computers:
17198 @item -march=@var{architecture-type}
17200 Generate code for the specified architecture. The choices for
17201 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
17202 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
17203 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
17204 architecture option for your machine. Code compiled for lower numbered
17205 architectures runs on higher numbered architectures, but not the
17208 @item -mpa-risc-1-0
17209 @itemx -mpa-risc-1-1
17210 @itemx -mpa-risc-2-0
17211 @opindex mpa-risc-1-0
17212 @opindex mpa-risc-1-1
17213 @opindex mpa-risc-2-0
17214 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
17216 @item -mjump-in-delay
17217 @opindex mjump-in-delay
17218 This option is ignored and provided for compatibility purposes only.
17220 @item -mdisable-fpregs
17221 @opindex mdisable-fpregs
17222 Prevent floating-point registers from being used in any manner. This is
17223 necessary for compiling kernels that perform lazy context switching of
17224 floating-point registers. If you use this option and attempt to perform
17225 floating-point operations, the compiler aborts.
17227 @item -mdisable-indexing
17228 @opindex mdisable-indexing
17229 Prevent the compiler from using indexing address modes. This avoids some
17230 rather obscure problems when compiling MIG generated code under MACH@.
17232 @item -mno-space-regs
17233 @opindex mno-space-regs
17234 Generate code that assumes the target has no space registers. This allows
17235 GCC to generate faster indirect calls and use unscaled index address modes.
17237 Such code is suitable for level 0 PA systems and kernels.
17239 @item -mfast-indirect-calls
17240 @opindex mfast-indirect-calls
17241 Generate code that assumes calls never cross space boundaries. This
17242 allows GCC to emit code that performs faster indirect calls.
17244 This option does not work in the presence of shared libraries or nested
17247 @item -mfixed-range=@var{register-range}
17248 @opindex mfixed-range
17249 Generate code treating the given register range as fixed registers.
17250 A fixed register is one that the register allocator cannot use. This is
17251 useful when compiling kernel code. A register range is specified as
17252 two registers separated by a dash. Multiple register ranges can be
17253 specified separated by a comma.
17255 @item -mlong-load-store
17256 @opindex mlong-load-store
17257 Generate 3-instruction load and store sequences as sometimes required by
17258 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
17261 @item -mportable-runtime
17262 @opindex mportable-runtime
17263 Use the portable calling conventions proposed by HP for ELF systems.
17267 Enable the use of assembler directives only GAS understands.
17269 @item -mschedule=@var{cpu-type}
17271 Schedule code according to the constraints for the machine type
17272 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
17273 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
17274 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
17275 proper scheduling option for your machine. The default scheduling is
17279 @opindex mlinker-opt
17280 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
17281 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
17282 linkers in which they give bogus error messages when linking some programs.
17285 @opindex msoft-float
17286 Generate output containing library calls for floating point.
17287 @strong{Warning:} the requisite libraries are not available for all HPPA
17288 targets. Normally the facilities of the machine's usual C compiler are
17289 used, but this cannot be done directly in cross-compilation. You must make
17290 your own arrangements to provide suitable library functions for
17293 @option{-msoft-float} changes the calling convention in the output file;
17294 therefore, it is only useful if you compile @emph{all} of a program with
17295 this option. In particular, you need to compile @file{libgcc.a}, the
17296 library that comes with GCC, with @option{-msoft-float} in order for
17301 Generate the predefine, @code{_SIO}, for server IO@. The default is
17302 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
17303 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
17304 options are available under HP-UX and HI-UX@.
17308 Use options specific to GNU @command{ld}.
17309 This passes @option{-shared} to @command{ld} when
17310 building a shared library. It is the default when GCC is configured,
17311 explicitly or implicitly, with the GNU linker. This option does not
17312 affect which @command{ld} is called; it only changes what parameters
17313 are passed to that @command{ld}.
17314 The @command{ld} that is called is determined by the
17315 @option{--with-ld} configure option, GCC's program search path, and
17316 finally by the user's @env{PATH}. The linker used by GCC can be printed
17317 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
17318 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17322 Use options specific to HP @command{ld}.
17323 This passes @option{-b} to @command{ld} when building
17324 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
17325 links. It is the default when GCC is configured, explicitly or
17326 implicitly, with the HP linker. This option does not affect
17327 which @command{ld} is called; it only changes what parameters are passed to that
17329 The @command{ld} that is called is determined by the @option{--with-ld}
17330 configure option, GCC's program search path, and finally by the user's
17331 @env{PATH}. The linker used by GCC can be printed using @samp{which
17332 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
17333 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17336 @opindex mno-long-calls
17337 Generate code that uses long call sequences. This ensures that a call
17338 is always able to reach linker generated stubs. The default is to generate
17339 long calls only when the distance from the call site to the beginning
17340 of the function or translation unit, as the case may be, exceeds a
17341 predefined limit set by the branch type being used. The limits for
17342 normal calls are 7,600,000 and 240,000 bytes, respectively for the
17343 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
17346 Distances are measured from the beginning of functions when using the
17347 @option{-ffunction-sections} option, or when using the @option{-mgas}
17348 and @option{-mno-portable-runtime} options together under HP-UX with
17351 It is normally not desirable to use this option as it degrades
17352 performance. However, it may be useful in large applications,
17353 particularly when partial linking is used to build the application.
17355 The types of long calls used depends on the capabilities of the
17356 assembler and linker, and the type of code being generated. The
17357 impact on systems that support long absolute calls, and long pic
17358 symbol-difference or pc-relative calls should be relatively small.
17359 However, an indirect call is used on 32-bit ELF systems in pic code
17360 and it is quite long.
17362 @item -munix=@var{unix-std}
17364 Generate compiler predefines and select a startfile for the specified
17365 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
17366 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
17367 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
17368 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
17369 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
17372 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
17373 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
17374 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
17375 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
17376 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
17377 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
17379 It is @emph{important} to note that this option changes the interfaces
17380 for various library routines. It also affects the operational behavior
17381 of the C library. Thus, @emph{extreme} care is needed in using this
17384 Library code that is intended to operate with more than one UNIX
17385 standard must test, set and restore the variable @code{__xpg4_extended_mask}
17386 as appropriate. Most GNU software doesn't provide this capability.
17390 Suppress the generation of link options to search libdld.sl when the
17391 @option{-static} option is specified on HP-UX 10 and later.
17395 The HP-UX implementation of setlocale in libc has a dependency on
17396 libdld.sl. There isn't an archive version of libdld.sl. Thus,
17397 when the @option{-static} option is specified, special link options
17398 are needed to resolve this dependency.
17400 On HP-UX 10 and later, the GCC driver adds the necessary options to
17401 link with libdld.sl when the @option{-static} option is specified.
17402 This causes the resulting binary to be dynamic. On the 64-bit port,
17403 the linkers generate dynamic binaries by default in any case. The
17404 @option{-nolibdld} option can be used to prevent the GCC driver from
17405 adding these link options.
17409 Add support for multithreading with the @dfn{dce thread} library
17410 under HP-UX@. This option sets flags for both the preprocessor and
17414 @node IA-64 Options
17415 @subsection IA-64 Options
17416 @cindex IA-64 Options
17418 These are the @samp{-m} options defined for the Intel IA-64 architecture.
17422 @opindex mbig-endian
17423 Generate code for a big-endian target. This is the default for HP-UX@.
17425 @item -mlittle-endian
17426 @opindex mlittle-endian
17427 Generate code for a little-endian target. This is the default for AIX5
17433 @opindex mno-gnu-as
17434 Generate (or don't) code for the GNU assembler. This is the default.
17435 @c Also, this is the default if the configure option @option{--with-gnu-as}
17441 @opindex mno-gnu-ld
17442 Generate (or don't) code for the GNU linker. This is the default.
17443 @c Also, this is the default if the configure option @option{--with-gnu-ld}
17448 Generate code that does not use a global pointer register. The result
17449 is not position independent code, and violates the IA-64 ABI@.
17451 @item -mvolatile-asm-stop
17452 @itemx -mno-volatile-asm-stop
17453 @opindex mvolatile-asm-stop
17454 @opindex mno-volatile-asm-stop
17455 Generate (or don't) a stop bit immediately before and after volatile asm
17458 @item -mregister-names
17459 @itemx -mno-register-names
17460 @opindex mregister-names
17461 @opindex mno-register-names
17462 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
17463 the stacked registers. This may make assembler output more readable.
17469 Disable (or enable) optimizations that use the small data section. This may
17470 be useful for working around optimizer bugs.
17472 @item -mconstant-gp
17473 @opindex mconstant-gp
17474 Generate code that uses a single constant global pointer value. This is
17475 useful when compiling kernel code.
17479 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
17480 This is useful when compiling firmware code.
17482 @item -minline-float-divide-min-latency
17483 @opindex minline-float-divide-min-latency
17484 Generate code for inline divides of floating-point values
17485 using the minimum latency algorithm.
17487 @item -minline-float-divide-max-throughput
17488 @opindex minline-float-divide-max-throughput
17489 Generate code for inline divides of floating-point values
17490 using the maximum throughput algorithm.
17492 @item -mno-inline-float-divide
17493 @opindex mno-inline-float-divide
17494 Do not generate inline code for divides of floating-point values.
17496 @item -minline-int-divide-min-latency
17497 @opindex minline-int-divide-min-latency
17498 Generate code for inline divides of integer values
17499 using the minimum latency algorithm.
17501 @item -minline-int-divide-max-throughput
17502 @opindex minline-int-divide-max-throughput
17503 Generate code for inline divides of integer values
17504 using the maximum throughput algorithm.
17506 @item -mno-inline-int-divide
17507 @opindex mno-inline-int-divide
17508 Do not generate inline code for divides of integer values.
17510 @item -minline-sqrt-min-latency
17511 @opindex minline-sqrt-min-latency
17512 Generate code for inline square roots
17513 using the minimum latency algorithm.
17515 @item -minline-sqrt-max-throughput
17516 @opindex minline-sqrt-max-throughput
17517 Generate code for inline square roots
17518 using the maximum throughput algorithm.
17520 @item -mno-inline-sqrt
17521 @opindex mno-inline-sqrt
17522 Do not generate inline code for @code{sqrt}.
17525 @itemx -mno-fused-madd
17526 @opindex mfused-madd
17527 @opindex mno-fused-madd
17528 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
17529 instructions. The default is to use these instructions.
17531 @item -mno-dwarf2-asm
17532 @itemx -mdwarf2-asm
17533 @opindex mno-dwarf2-asm
17534 @opindex mdwarf2-asm
17535 Don't (or do) generate assembler code for the DWARF line number debugging
17536 info. This may be useful when not using the GNU assembler.
17538 @item -mearly-stop-bits
17539 @itemx -mno-early-stop-bits
17540 @opindex mearly-stop-bits
17541 @opindex mno-early-stop-bits
17542 Allow stop bits to be placed earlier than immediately preceding the
17543 instruction that triggered the stop bit. This can improve instruction
17544 scheduling, but does not always do so.
17546 @item -mfixed-range=@var{register-range}
17547 @opindex mfixed-range
17548 Generate code treating the given register range as fixed registers.
17549 A fixed register is one that the register allocator cannot use. This is
17550 useful when compiling kernel code. A register range is specified as
17551 two registers separated by a dash. Multiple register ranges can be
17552 specified separated by a comma.
17554 @item -mtls-size=@var{tls-size}
17556 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
17559 @item -mtune=@var{cpu-type}
17561 Tune the instruction scheduling for a particular CPU, Valid values are
17562 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
17563 and @samp{mckinley}.
17569 Generate code for a 32-bit or 64-bit environment.
17570 The 32-bit environment sets int, long and pointer to 32 bits.
17571 The 64-bit environment sets int to 32 bits and long and pointer
17572 to 64 bits. These are HP-UX specific flags.
17574 @item -mno-sched-br-data-spec
17575 @itemx -msched-br-data-spec
17576 @opindex mno-sched-br-data-spec
17577 @opindex msched-br-data-spec
17578 (Dis/En)able data speculative scheduling before reload.
17579 This results in generation of @code{ld.a} instructions and
17580 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17581 The default setting is disabled.
17583 @item -msched-ar-data-spec
17584 @itemx -mno-sched-ar-data-spec
17585 @opindex msched-ar-data-spec
17586 @opindex mno-sched-ar-data-spec
17587 (En/Dis)able data speculative scheduling after reload.
17588 This results in generation of @code{ld.a} instructions and
17589 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17590 The default setting is enabled.
17592 @item -mno-sched-control-spec
17593 @itemx -msched-control-spec
17594 @opindex mno-sched-control-spec
17595 @opindex msched-control-spec
17596 (Dis/En)able control speculative scheduling. This feature is
17597 available only during region scheduling (i.e.@: before reload).
17598 This results in generation of the @code{ld.s} instructions and
17599 the corresponding check instructions @code{chk.s}.
17600 The default setting is disabled.
17602 @item -msched-br-in-data-spec
17603 @itemx -mno-sched-br-in-data-spec
17604 @opindex msched-br-in-data-spec
17605 @opindex mno-sched-br-in-data-spec
17606 (En/Dis)able speculative scheduling of the instructions that
17607 are dependent on the data speculative loads before reload.
17608 This is effective only with @option{-msched-br-data-spec} enabled.
17609 The default setting is enabled.
17611 @item -msched-ar-in-data-spec
17612 @itemx -mno-sched-ar-in-data-spec
17613 @opindex msched-ar-in-data-spec
17614 @opindex mno-sched-ar-in-data-spec
17615 (En/Dis)able speculative scheduling of the instructions that
17616 are dependent on the data speculative loads after reload.
17617 This is effective only with @option{-msched-ar-data-spec} enabled.
17618 The default setting is enabled.
17620 @item -msched-in-control-spec
17621 @itemx -mno-sched-in-control-spec
17622 @opindex msched-in-control-spec
17623 @opindex mno-sched-in-control-spec
17624 (En/Dis)able speculative scheduling of the instructions that
17625 are dependent on the control speculative loads.
17626 This is effective only with @option{-msched-control-spec} enabled.
17627 The default setting is enabled.
17629 @item -mno-sched-prefer-non-data-spec-insns
17630 @itemx -msched-prefer-non-data-spec-insns
17631 @opindex mno-sched-prefer-non-data-spec-insns
17632 @opindex msched-prefer-non-data-spec-insns
17633 If enabled, data-speculative instructions are chosen for schedule
17634 only if there are no other choices at the moment. This makes
17635 the use of the data speculation much more conservative.
17636 The default setting is disabled.
17638 @item -mno-sched-prefer-non-control-spec-insns
17639 @itemx -msched-prefer-non-control-spec-insns
17640 @opindex mno-sched-prefer-non-control-spec-insns
17641 @opindex msched-prefer-non-control-spec-insns
17642 If enabled, control-speculative instructions are chosen for schedule
17643 only if there are no other choices at the moment. This makes
17644 the use of the control speculation much more conservative.
17645 The default setting is disabled.
17647 @item -mno-sched-count-spec-in-critical-path
17648 @itemx -msched-count-spec-in-critical-path
17649 @opindex mno-sched-count-spec-in-critical-path
17650 @opindex msched-count-spec-in-critical-path
17651 If enabled, speculative dependencies are considered during
17652 computation of the instructions priorities. This makes the use of the
17653 speculation a bit more conservative.
17654 The default setting is disabled.
17656 @item -msched-spec-ldc
17657 @opindex msched-spec-ldc
17658 Use a simple data speculation check. This option is on by default.
17660 @item -msched-control-spec-ldc
17661 @opindex msched-spec-ldc
17662 Use a simple check for control speculation. This option is on by default.
17664 @item -msched-stop-bits-after-every-cycle
17665 @opindex msched-stop-bits-after-every-cycle
17666 Place a stop bit after every cycle when scheduling. This option is on
17669 @item -msched-fp-mem-deps-zero-cost
17670 @opindex msched-fp-mem-deps-zero-cost
17671 Assume that floating-point stores and loads are not likely to cause a conflict
17672 when placed into the same instruction group. This option is disabled by
17675 @item -msel-sched-dont-check-control-spec
17676 @opindex msel-sched-dont-check-control-spec
17677 Generate checks for control speculation in selective scheduling.
17678 This flag is disabled by default.
17680 @item -msched-max-memory-insns=@var{max-insns}
17681 @opindex msched-max-memory-insns
17682 Limit on the number of memory insns per instruction group, giving lower
17683 priority to subsequent memory insns attempting to schedule in the same
17684 instruction group. Frequently useful to prevent cache bank conflicts.
17685 The default value is 1.
17687 @item -msched-max-memory-insns-hard-limit
17688 @opindex msched-max-memory-insns-hard-limit
17689 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
17690 disallowing more than that number in an instruction group.
17691 Otherwise, the limit is ``soft'', meaning that non-memory operations
17692 are preferred when the limit is reached, but memory operations may still
17698 @subsection LM32 Options
17699 @cindex LM32 options
17701 These @option{-m} options are defined for the LatticeMico32 architecture:
17704 @item -mbarrel-shift-enabled
17705 @opindex mbarrel-shift-enabled
17706 Enable barrel-shift instructions.
17708 @item -mdivide-enabled
17709 @opindex mdivide-enabled
17710 Enable divide and modulus instructions.
17712 @item -mmultiply-enabled
17713 @opindex multiply-enabled
17714 Enable multiply instructions.
17716 @item -msign-extend-enabled
17717 @opindex msign-extend-enabled
17718 Enable sign extend instructions.
17720 @item -muser-enabled
17721 @opindex muser-enabled
17722 Enable user-defined instructions.
17727 @subsection M32C Options
17728 @cindex M32C options
17731 @item -mcpu=@var{name}
17733 Select the CPU for which code is generated. @var{name} may be one of
17734 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
17735 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
17736 the M32C/80 series.
17740 Specifies that the program will be run on the simulator. This causes
17741 an alternate runtime library to be linked in which supports, for
17742 example, file I/O@. You must not use this option when generating
17743 programs that will run on real hardware; you must provide your own
17744 runtime library for whatever I/O functions are needed.
17746 @item -memregs=@var{number}
17748 Specifies the number of memory-based pseudo-registers GCC uses
17749 during code generation. These pseudo-registers are used like real
17750 registers, so there is a tradeoff between GCC's ability to fit the
17751 code into available registers, and the performance penalty of using
17752 memory instead of registers. Note that all modules in a program must
17753 be compiled with the same value for this option. Because of that, you
17754 must not use this option with GCC's default runtime libraries.
17758 @node M32R/D Options
17759 @subsection M32R/D Options
17760 @cindex M32R/D options
17762 These @option{-m} options are defined for Renesas M32R/D architectures:
17767 Generate code for the M32R/2@.
17771 Generate code for the M32R/X@.
17775 Generate code for the M32R@. This is the default.
17777 @item -mmodel=small
17778 @opindex mmodel=small
17779 Assume all objects live in the lower 16MB of memory (so that their addresses
17780 can be loaded with the @code{ld24} instruction), and assume all subroutines
17781 are reachable with the @code{bl} instruction.
17782 This is the default.
17784 The addressability of a particular object can be set with the
17785 @code{model} attribute.
17787 @item -mmodel=medium
17788 @opindex mmodel=medium
17789 Assume objects may be anywhere in the 32-bit address space (the compiler
17790 generates @code{seth/add3} instructions to load their addresses), and
17791 assume all subroutines are reachable with the @code{bl} instruction.
17793 @item -mmodel=large
17794 @opindex mmodel=large
17795 Assume objects may be anywhere in the 32-bit address space (the compiler
17796 generates @code{seth/add3} instructions to load their addresses), and
17797 assume subroutines may not be reachable with the @code{bl} instruction
17798 (the compiler generates the much slower @code{seth/add3/jl}
17799 instruction sequence).
17802 @opindex msdata=none
17803 Disable use of the small data area. Variables are put into
17804 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
17805 @code{section} attribute has been specified).
17806 This is the default.
17808 The small data area consists of sections @code{.sdata} and @code{.sbss}.
17809 Objects may be explicitly put in the small data area with the
17810 @code{section} attribute using one of these sections.
17812 @item -msdata=sdata
17813 @opindex msdata=sdata
17814 Put small global and static data in the small data area, but do not
17815 generate special code to reference them.
17818 @opindex msdata=use
17819 Put small global and static data in the small data area, and generate
17820 special instructions to reference them.
17824 @cindex smaller data references
17825 Put global and static objects less than or equal to @var{num} bytes
17826 into the small data or BSS sections instead of the normal data or BSS
17827 sections. The default value of @var{num} is 8.
17828 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
17829 for this option to have any effect.
17831 All modules should be compiled with the same @option{-G @var{num}} value.
17832 Compiling with different values of @var{num} may or may not work; if it
17833 doesn't the linker gives an error message---incorrect code is not
17838 Makes the M32R-specific code in the compiler display some statistics
17839 that might help in debugging programs.
17841 @item -malign-loops
17842 @opindex malign-loops
17843 Align all loops to a 32-byte boundary.
17845 @item -mno-align-loops
17846 @opindex mno-align-loops
17847 Do not enforce a 32-byte alignment for loops. This is the default.
17849 @item -missue-rate=@var{number}
17850 @opindex missue-rate=@var{number}
17851 Issue @var{number} instructions per cycle. @var{number} can only be 1
17854 @item -mbranch-cost=@var{number}
17855 @opindex mbranch-cost=@var{number}
17856 @var{number} can only be 1 or 2. If it is 1 then branches are
17857 preferred over conditional code, if it is 2, then the opposite applies.
17859 @item -mflush-trap=@var{number}
17860 @opindex mflush-trap=@var{number}
17861 Specifies the trap number to use to flush the cache. The default is
17862 12. Valid numbers are between 0 and 15 inclusive.
17864 @item -mno-flush-trap
17865 @opindex mno-flush-trap
17866 Specifies that the cache cannot be flushed by using a trap.
17868 @item -mflush-func=@var{name}
17869 @opindex mflush-func=@var{name}
17870 Specifies the name of the operating system function to call to flush
17871 the cache. The default is @samp{_flush_cache}, but a function call
17872 is only used if a trap is not available.
17874 @item -mno-flush-func
17875 @opindex mno-flush-func
17876 Indicates that there is no OS function for flushing the cache.
17880 @node M680x0 Options
17881 @subsection M680x0 Options
17882 @cindex M680x0 options
17884 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
17885 The default settings depend on which architecture was selected when
17886 the compiler was configured; the defaults for the most common choices
17890 @item -march=@var{arch}
17892 Generate code for a specific M680x0 or ColdFire instruction set
17893 architecture. Permissible values of @var{arch} for M680x0
17894 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
17895 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
17896 architectures are selected according to Freescale's ISA classification
17897 and the permissible values are: @samp{isaa}, @samp{isaaplus},
17898 @samp{isab} and @samp{isac}.
17900 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
17901 code for a ColdFire target. The @var{arch} in this macro is one of the
17902 @option{-march} arguments given above.
17904 When used together, @option{-march} and @option{-mtune} select code
17905 that runs on a family of similar processors but that is optimized
17906 for a particular microarchitecture.
17908 @item -mcpu=@var{cpu}
17910 Generate code for a specific M680x0 or ColdFire processor.
17911 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
17912 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
17913 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
17914 below, which also classifies the CPUs into families:
17916 @multitable @columnfractions 0.20 0.80
17917 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
17918 @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}
17919 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
17920 @item @samp{5206e} @tab @samp{5206e}
17921 @item @samp{5208} @tab @samp{5207} @samp{5208}
17922 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
17923 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
17924 @item @samp{5216} @tab @samp{5214} @samp{5216}
17925 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
17926 @item @samp{5225} @tab @samp{5224} @samp{5225}
17927 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17928 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17929 @item @samp{5249} @tab @samp{5249}
17930 @item @samp{5250} @tab @samp{5250}
17931 @item @samp{5271} @tab @samp{5270} @samp{5271}
17932 @item @samp{5272} @tab @samp{5272}
17933 @item @samp{5275} @tab @samp{5274} @samp{5275}
17934 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17935 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17936 @item @samp{5307} @tab @samp{5307}
17937 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17938 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17939 @item @samp{5407} @tab @samp{5407}
17940 @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}
17943 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17944 @var{arch} is compatible with @var{cpu}. Other combinations of
17945 @option{-mcpu} and @option{-march} are rejected.
17947 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17948 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17949 where the value of @var{family} is given by the table above.
17951 @item -mtune=@var{tune}
17953 Tune the code for a particular microarchitecture within the
17954 constraints set by @option{-march} and @option{-mcpu}.
17955 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17956 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17957 and @samp{cpu32}. The ColdFire microarchitectures
17958 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17960 You can also use @option{-mtune=68020-40} for code that needs
17961 to run relatively well on 68020, 68030 and 68040 targets.
17962 @option{-mtune=68020-60} is similar but includes 68060 targets
17963 as well. These two options select the same tuning decisions as
17964 @option{-m68020-40} and @option{-m68020-60} respectively.
17966 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17967 when tuning for 680x0 architecture @var{arch}. It also defines
17968 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17969 option is used. If GCC is tuning for a range of architectures,
17970 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17971 it defines the macros for every architecture in the range.
17973 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17974 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17975 of the arguments given above.
17981 Generate output for a 68000. This is the default
17982 when the compiler is configured for 68000-based systems.
17983 It is equivalent to @option{-march=68000}.
17985 Use this option for microcontrollers with a 68000 or EC000 core,
17986 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17990 Generate output for a 68010. This is the default
17991 when the compiler is configured for 68010-based systems.
17992 It is equivalent to @option{-march=68010}.
17998 Generate output for a 68020. This is the default
17999 when the compiler is configured for 68020-based systems.
18000 It is equivalent to @option{-march=68020}.
18004 Generate output for a 68030. This is the default when the compiler is
18005 configured for 68030-based systems. It is equivalent to
18006 @option{-march=68030}.
18010 Generate output for a 68040. This is the default when the compiler is
18011 configured for 68040-based systems. It is equivalent to
18012 @option{-march=68040}.
18014 This option inhibits the use of 68881/68882 instructions that have to be
18015 emulated by software on the 68040. Use this option if your 68040 does not
18016 have code to emulate those instructions.
18020 Generate output for a 68060. This is the default when the compiler is
18021 configured for 68060-based systems. It is equivalent to
18022 @option{-march=68060}.
18024 This option inhibits the use of 68020 and 68881/68882 instructions that
18025 have to be emulated by software on the 68060. Use this option if your 68060
18026 does not have code to emulate those instructions.
18030 Generate output for a CPU32. This is the default
18031 when the compiler is configured for CPU32-based systems.
18032 It is equivalent to @option{-march=cpu32}.
18034 Use this option for microcontrollers with a
18035 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
18036 68336, 68340, 68341, 68349 and 68360.
18040 Generate output for a 520X ColdFire CPU@. This is the default
18041 when the compiler is configured for 520X-based systems.
18042 It is equivalent to @option{-mcpu=5206}, and is now deprecated
18043 in favor of that option.
18045 Use this option for microcontroller with a 5200 core, including
18046 the MCF5202, MCF5203, MCF5204 and MCF5206.
18050 Generate output for a 5206e ColdFire CPU@. The option is now
18051 deprecated in favor of the equivalent @option{-mcpu=5206e}.
18055 Generate output for a member of the ColdFire 528X family.
18056 The option is now deprecated in favor of the equivalent
18057 @option{-mcpu=528x}.
18061 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
18062 in favor of the equivalent @option{-mcpu=5307}.
18066 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
18067 in favor of the equivalent @option{-mcpu=5407}.
18071 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
18072 This includes use of hardware floating-point instructions.
18073 The option is equivalent to @option{-mcpu=547x}, and is now
18074 deprecated in favor of that option.
18078 Generate output for a 68040, without using any of the new instructions.
18079 This results in code that can run relatively efficiently on either a
18080 68020/68881 or a 68030 or a 68040. The generated code does use the
18081 68881 instructions that are emulated on the 68040.
18083 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
18087 Generate output for a 68060, without using any of the new instructions.
18088 This results in code that can run relatively efficiently on either a
18089 68020/68881 or a 68030 or a 68040. The generated code does use the
18090 68881 instructions that are emulated on the 68060.
18092 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
18096 @opindex mhard-float
18098 Generate floating-point instructions. This is the default for 68020
18099 and above, and for ColdFire devices that have an FPU@. It defines the
18100 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
18101 on ColdFire targets.
18104 @opindex msoft-float
18105 Do not generate floating-point instructions; use library calls instead.
18106 This is the default for 68000, 68010, and 68832 targets. It is also
18107 the default for ColdFire devices that have no FPU.
18113 Generate (do not generate) ColdFire hardware divide and remainder
18114 instructions. If @option{-march} is used without @option{-mcpu},
18115 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
18116 architectures. Otherwise, the default is taken from the target CPU
18117 (either the default CPU, or the one specified by @option{-mcpu}). For
18118 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
18119 @option{-mcpu=5206e}.
18121 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
18125 Consider type @code{int} to be 16 bits wide, like @code{short int}.
18126 Additionally, parameters passed on the stack are also aligned to a
18127 16-bit boundary even on targets whose API mandates promotion to 32-bit.
18131 Do not consider type @code{int} to be 16 bits wide. This is the default.
18134 @itemx -mno-bitfield
18135 @opindex mnobitfield
18136 @opindex mno-bitfield
18137 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
18138 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
18142 Do use the bit-field instructions. The @option{-m68020} option implies
18143 @option{-mbitfield}. This is the default if you use a configuration
18144 designed for a 68020.
18148 Use a different function-calling convention, in which functions
18149 that take a fixed number of arguments return with the @code{rtd}
18150 instruction, which pops their arguments while returning. This
18151 saves one instruction in the caller since there is no need to pop
18152 the arguments there.
18154 This calling convention is incompatible with the one normally
18155 used on Unix, so you cannot use it if you need to call libraries
18156 compiled with the Unix compiler.
18158 Also, you must provide function prototypes for all functions that
18159 take variable numbers of arguments (including @code{printf});
18160 otherwise incorrect code is generated for calls to those
18163 In addition, seriously incorrect code results if you call a
18164 function with too many arguments. (Normally, extra arguments are
18165 harmlessly ignored.)
18167 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
18168 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
18172 Do not use the calling conventions selected by @option{-mrtd}.
18173 This is the default.
18176 @itemx -mno-align-int
18177 @opindex malign-int
18178 @opindex mno-align-int
18179 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
18180 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
18181 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
18182 Aligning variables on 32-bit boundaries produces code that runs somewhat
18183 faster on processors with 32-bit busses at the expense of more memory.
18185 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
18186 aligns structures containing the above types differently than
18187 most published application binary interface specifications for the m68k.
18191 Use the pc-relative addressing mode of the 68000 directly, instead of
18192 using a global offset table. At present, this option implies @option{-fpic},
18193 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
18194 not presently supported with @option{-mpcrel}, though this could be supported for
18195 68020 and higher processors.
18197 @item -mno-strict-align
18198 @itemx -mstrict-align
18199 @opindex mno-strict-align
18200 @opindex mstrict-align
18201 Do not (do) assume that unaligned memory references are handled by
18205 Generate code that allows the data segment to be located in a different
18206 area of memory from the text segment. This allows for execute-in-place in
18207 an environment without virtual memory management. This option implies
18210 @item -mno-sep-data
18211 Generate code that assumes that the data segment follows the text segment.
18212 This is the default.
18214 @item -mid-shared-library
18215 Generate code that supports shared libraries via the library ID method.
18216 This allows for execute-in-place and shared libraries in an environment
18217 without virtual memory management. This option implies @option{-fPIC}.
18219 @item -mno-id-shared-library
18220 Generate code that doesn't assume ID-based shared libraries are being used.
18221 This is the default.
18223 @item -mshared-library-id=n
18224 Specifies the identification number of the ID-based shared library being
18225 compiled. Specifying a value of 0 generates more compact code; specifying
18226 other values forces the allocation of that number to the current
18227 library, but is no more space- or time-efficient than omitting this option.
18233 When generating position-independent code for ColdFire, generate code
18234 that works if the GOT has more than 8192 entries. This code is
18235 larger and slower than code generated without this option. On M680x0
18236 processors, this option is not needed; @option{-fPIC} suffices.
18238 GCC normally uses a single instruction to load values from the GOT@.
18239 While this is relatively efficient, it only works if the GOT
18240 is smaller than about 64k. Anything larger causes the linker
18241 to report an error such as:
18243 @cindex relocation truncated to fit (ColdFire)
18245 relocation truncated to fit: R_68K_GOT16O foobar
18248 If this happens, you should recompile your code with @option{-mxgot}.
18249 It should then work with very large GOTs. However, code generated with
18250 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
18251 the value of a global symbol.
18253 Note that some linkers, including newer versions of the GNU linker,
18254 can create multiple GOTs and sort GOT entries. If you have such a linker,
18255 you should only need to use @option{-mxgot} when compiling a single
18256 object file that accesses more than 8192 GOT entries. Very few do.
18258 These options have no effect unless GCC is generating
18259 position-independent code.
18263 @node MCore Options
18264 @subsection MCore Options
18265 @cindex MCore options
18267 These are the @samp{-m} options defined for the Motorola M*Core
18273 @itemx -mno-hardlit
18275 @opindex mno-hardlit
18276 Inline constants into the code stream if it can be done in two
18277 instructions or less.
18283 Use the divide instruction. (Enabled by default).
18285 @item -mrelax-immediate
18286 @itemx -mno-relax-immediate
18287 @opindex mrelax-immediate
18288 @opindex mno-relax-immediate
18289 Allow arbitrary-sized immediates in bit operations.
18291 @item -mwide-bitfields
18292 @itemx -mno-wide-bitfields
18293 @opindex mwide-bitfields
18294 @opindex mno-wide-bitfields
18295 Always treat bit-fields as @code{int}-sized.
18297 @item -m4byte-functions
18298 @itemx -mno-4byte-functions
18299 @opindex m4byte-functions
18300 @opindex mno-4byte-functions
18301 Force all functions to be aligned to a 4-byte boundary.
18303 @item -mcallgraph-data
18304 @itemx -mno-callgraph-data
18305 @opindex mcallgraph-data
18306 @opindex mno-callgraph-data
18307 Emit callgraph information.
18310 @itemx -mno-slow-bytes
18311 @opindex mslow-bytes
18312 @opindex mno-slow-bytes
18313 Prefer word access when reading byte quantities.
18315 @item -mlittle-endian
18316 @itemx -mbig-endian
18317 @opindex mlittle-endian
18318 @opindex mbig-endian
18319 Generate code for a little-endian target.
18325 Generate code for the 210 processor.
18329 Assume that runtime support has been provided and so omit the
18330 simulator library (@file{libsim.a)} from the linker command line.
18332 @item -mstack-increment=@var{size}
18333 @opindex mstack-increment
18334 Set the maximum amount for a single stack increment operation. Large
18335 values can increase the speed of programs that contain functions
18336 that need a large amount of stack space, but they can also trigger a
18337 segmentation fault if the stack is extended too much. The default
18343 @subsection MeP Options
18344 @cindex MeP options
18350 Enables the @code{abs} instruction, which is the absolute difference
18351 between two registers.
18355 Enables all the optional instructions---average, multiply, divide, bit
18356 operations, leading zero, absolute difference, min/max, clip, and
18362 Enables the @code{ave} instruction, which computes the average of two
18365 @item -mbased=@var{n}
18367 Variables of size @var{n} bytes or smaller are placed in the
18368 @code{.based} section by default. Based variables use the @code{$tp}
18369 register as a base register, and there is a 128-byte limit to the
18370 @code{.based} section.
18374 Enables the bit operation instructions---bit test (@code{btstm}), set
18375 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
18376 test-and-set (@code{tas}).
18378 @item -mc=@var{name}
18380 Selects which section constant data is placed in. @var{name} may
18381 be @samp{tiny}, @samp{near}, or @samp{far}.
18385 Enables the @code{clip} instruction. Note that @option{-mclip} is not
18386 useful unless you also provide @option{-mminmax}.
18388 @item -mconfig=@var{name}
18390 Selects one of the built-in core configurations. Each MeP chip has
18391 one or more modules in it; each module has a core CPU and a variety of
18392 coprocessors, optional instructions, and peripherals. The
18393 @code{MeP-Integrator} tool, not part of GCC, provides these
18394 configurations through this option; using this option is the same as
18395 using all the corresponding command-line options. The default
18396 configuration is @samp{default}.
18400 Enables the coprocessor instructions. By default, this is a 32-bit
18401 coprocessor. Note that the coprocessor is normally enabled via the
18402 @option{-mconfig=} option.
18406 Enables the 32-bit coprocessor's instructions.
18410 Enables the 64-bit coprocessor's instructions.
18414 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
18418 Causes constant variables to be placed in the @code{.near} section.
18422 Enables the @code{div} and @code{divu} instructions.
18426 Generate big-endian code.
18430 Generate little-endian code.
18432 @item -mio-volatile
18433 @opindex mio-volatile
18434 Tells the compiler that any variable marked with the @code{io}
18435 attribute is to be considered volatile.
18439 Causes variables to be assigned to the @code{.far} section by default.
18443 Enables the @code{leadz} (leading zero) instruction.
18447 Causes variables to be assigned to the @code{.near} section by default.
18451 Enables the @code{min} and @code{max} instructions.
18455 Enables the multiplication and multiply-accumulate instructions.
18459 Disables all the optional instructions enabled by @option{-mall-opts}.
18463 Enables the @code{repeat} and @code{erepeat} instructions, used for
18464 low-overhead looping.
18468 Causes all variables to default to the @code{.tiny} section. Note
18469 that there is a 65536-byte limit to this section. Accesses to these
18470 variables use the @code{%gp} base register.
18474 Enables the saturation instructions. Note that the compiler does not
18475 currently generate these itself, but this option is included for
18476 compatibility with other tools, like @code{as}.
18480 Link the SDRAM-based runtime instead of the default ROM-based runtime.
18484 Link the simulator run-time libraries.
18488 Link the simulator runtime libraries, excluding built-in support
18489 for reset and exception vectors and tables.
18493 Causes all functions to default to the @code{.far} section. Without
18494 this option, functions default to the @code{.near} section.
18496 @item -mtiny=@var{n}
18498 Variables that are @var{n} bytes or smaller are allocated to the
18499 @code{.tiny} section. These variables use the @code{$gp} base
18500 register. The default for this option is 4, but note that there's a
18501 65536-byte limit to the @code{.tiny} section.
18505 @node MicroBlaze Options
18506 @subsection MicroBlaze Options
18507 @cindex MicroBlaze Options
18512 @opindex msoft-float
18513 Use software emulation for floating point (default).
18516 @opindex mhard-float
18517 Use hardware floating-point instructions.
18521 Do not optimize block moves, use @code{memcpy}.
18523 @item -mno-clearbss
18524 @opindex mno-clearbss
18525 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
18527 @item -mcpu=@var{cpu-type}
18529 Use features of, and schedule code for, the given CPU.
18530 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
18531 where @var{X} is a major version, @var{YY} is the minor version, and
18532 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
18533 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
18535 @item -mxl-soft-mul
18536 @opindex mxl-soft-mul
18537 Use software multiply emulation (default).
18539 @item -mxl-soft-div
18540 @opindex mxl-soft-div
18541 Use software emulation for divides (default).
18543 @item -mxl-barrel-shift
18544 @opindex mxl-barrel-shift
18545 Use the hardware barrel shifter.
18547 @item -mxl-pattern-compare
18548 @opindex mxl-pattern-compare
18549 Use pattern compare instructions.
18551 @item -msmall-divides
18552 @opindex msmall-divides
18553 Use table lookup optimization for small signed integer divisions.
18555 @item -mxl-stack-check
18556 @opindex mxl-stack-check
18557 This option is deprecated. Use @option{-fstack-check} instead.
18560 @opindex mxl-gp-opt
18561 Use GP-relative @code{.sdata}/@code{.sbss} sections.
18563 @item -mxl-multiply-high
18564 @opindex mxl-multiply-high
18565 Use multiply high instructions for high part of 32x32 multiply.
18567 @item -mxl-float-convert
18568 @opindex mxl-float-convert
18569 Use hardware floating-point conversion instructions.
18571 @item -mxl-float-sqrt
18572 @opindex mxl-float-sqrt
18573 Use hardware floating-point square root instruction.
18576 @opindex mbig-endian
18577 Generate code for a big-endian target.
18579 @item -mlittle-endian
18580 @opindex mlittle-endian
18581 Generate code for a little-endian target.
18584 @opindex mxl-reorder
18585 Use reorder instructions (swap and byte reversed load/store).
18587 @item -mxl-mode-@var{app-model}
18588 Select application model @var{app-model}. Valid models are
18591 normal executable (default), uses startup code @file{crt0.o}.
18594 for use with Xilinx Microprocessor Debugger (XMD) based
18595 software intrusive debug agent called xmdstub. This uses startup file
18596 @file{crt1.o} and sets the start address of the program to 0x800.
18599 for applications that are loaded using a bootloader.
18600 This model uses startup file @file{crt2.o} which does not contain a processor
18601 reset vector handler. This is suitable for transferring control on a
18602 processor reset to the bootloader rather than the application.
18605 for applications that do not require any of the
18606 MicroBlaze vectors. This option may be useful for applications running
18607 within a monitoring application. This model uses @file{crt3.o} as a startup file.
18610 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
18611 @option{-mxl-mode-@var{app-model}}.
18616 @subsection MIPS Options
18617 @cindex MIPS options
18623 Generate big-endian code.
18627 Generate little-endian code. This is the default for @samp{mips*el-*-*}
18630 @item -march=@var{arch}
18632 Generate code that runs on @var{arch}, which can be the name of a
18633 generic MIPS ISA, or the name of a particular processor.
18635 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
18636 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
18637 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
18638 @samp{mips64r5} and @samp{mips64r6}.
18639 The processor names are:
18640 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
18641 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
18642 @samp{5kc}, @samp{5kf},
18644 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
18645 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
18646 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
18647 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
18648 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
18651 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
18653 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
18654 @samp{m5100}, @samp{m5101},
18655 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
18658 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
18659 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
18660 @samp{rm7000}, @samp{rm9000},
18661 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
18664 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
18665 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
18666 @samp{xlr} and @samp{xlp}.
18667 The special value @samp{from-abi} selects the
18668 most compatible architecture for the selected ABI (that is,
18669 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
18671 The native Linux/GNU toolchain also supports the value @samp{native},
18672 which selects the best architecture option for the host processor.
18673 @option{-march=native} has no effect if GCC does not recognize
18676 In processor names, a final @samp{000} can be abbreviated as @samp{k}
18677 (for example, @option{-march=r2k}). Prefixes are optional, and
18678 @samp{vr} may be written @samp{r}.
18680 Names of the form @samp{@var{n}f2_1} refer to processors with
18681 FPUs clocked at half the rate of the core, names of the form
18682 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
18683 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
18684 processors with FPUs clocked a ratio of 3:2 with respect to the core.
18685 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
18686 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
18687 accepted as synonyms for @samp{@var{n}f1_1}.
18689 GCC defines two macros based on the value of this option. The first
18690 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
18691 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
18692 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
18693 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
18694 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
18696 Note that the @code{_MIPS_ARCH} macro uses the processor names given
18697 above. In other words, it has the full prefix and does not
18698 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
18699 the macro names the resolved architecture (either @code{"mips1"} or
18700 @code{"mips3"}). It names the default architecture when no
18701 @option{-march} option is given.
18703 @item -mtune=@var{arch}
18705 Optimize for @var{arch}. Among other things, this option controls
18706 the way instructions are scheduled, and the perceived cost of arithmetic
18707 operations. The list of @var{arch} values is the same as for
18710 When this option is not used, GCC optimizes for the processor
18711 specified by @option{-march}. By using @option{-march} and
18712 @option{-mtune} together, it is possible to generate code that
18713 runs on a family of processors, but optimize the code for one
18714 particular member of that family.
18716 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
18717 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
18718 @option{-march} ones described above.
18722 Equivalent to @option{-march=mips1}.
18726 Equivalent to @option{-march=mips2}.
18730 Equivalent to @option{-march=mips3}.
18734 Equivalent to @option{-march=mips4}.
18738 Equivalent to @option{-march=mips32}.
18742 Equivalent to @option{-march=mips32r3}.
18746 Equivalent to @option{-march=mips32r5}.
18750 Equivalent to @option{-march=mips32r6}.
18754 Equivalent to @option{-march=mips64}.
18758 Equivalent to @option{-march=mips64r2}.
18762 Equivalent to @option{-march=mips64r3}.
18766 Equivalent to @option{-march=mips64r5}.
18770 Equivalent to @option{-march=mips64r6}.
18775 @opindex mno-mips16
18776 Generate (do not generate) MIPS16 code. If GCC is targeting a
18777 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
18779 MIPS16 code generation can also be controlled on a per-function basis
18780 by means of @code{mips16} and @code{nomips16} attributes.
18781 @xref{Function Attributes}, for more information.
18783 @item -mflip-mips16
18784 @opindex mflip-mips16
18785 Generate MIPS16 code on alternating functions. This option is provided
18786 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
18787 not intended for ordinary use in compiling user code.
18789 @item -minterlink-compressed
18790 @item -mno-interlink-compressed
18791 @opindex minterlink-compressed
18792 @opindex mno-interlink-compressed
18793 Require (do not require) that code using the standard (uncompressed) MIPS ISA
18794 be link-compatible with MIPS16 and microMIPS code, and vice versa.
18796 For example, code using the standard ISA encoding cannot jump directly
18797 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
18798 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
18799 knows that the target of the jump is not compressed.
18801 @item -minterlink-mips16
18802 @itemx -mno-interlink-mips16
18803 @opindex minterlink-mips16
18804 @opindex mno-interlink-mips16
18805 Aliases of @option{-minterlink-compressed} and
18806 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
18807 and are retained for backwards compatibility.
18819 Generate code for the given ABI@.
18821 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
18822 generates 64-bit code when you select a 64-bit architecture, but you
18823 can use @option{-mgp32} to get 32-bit code instead.
18825 For information about the O64 ABI, see
18826 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
18828 GCC supports a variant of the o32 ABI in which floating-point registers
18829 are 64 rather than 32 bits wide. You can select this combination with
18830 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
18831 and @code{mfhc1} instructions and is therefore only supported for
18832 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
18834 The register assignments for arguments and return values remain the
18835 same, but each scalar value is passed in a single 64-bit register
18836 rather than a pair of 32-bit registers. For example, scalar
18837 floating-point values are returned in @samp{$f0} only, not a
18838 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
18839 remains the same in that the even-numbered double-precision registers
18842 Two additional variants of the o32 ABI are supported to enable
18843 a transition from 32-bit to 64-bit registers. These are FPXX
18844 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
18845 The FPXX extension mandates that all code must execute correctly
18846 when run using 32-bit or 64-bit registers. The code can be interlinked
18847 with either FP32 or FP64, but not both.
18848 The FP64A extension is similar to the FP64 extension but forbids the
18849 use of odd-numbered single-precision registers. This can be used
18850 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
18851 processors and allows both FP32 and FP64A code to interlink and
18852 run in the same process without changing FPU modes.
18855 @itemx -mno-abicalls
18857 @opindex mno-abicalls
18858 Generate (do not generate) code that is suitable for SVR4-style
18859 dynamic objects. @option{-mabicalls} is the default for SVR4-based
18864 Generate (do not generate) code that is fully position-independent,
18865 and that can therefore be linked into shared libraries. This option
18866 only affects @option{-mabicalls}.
18868 All @option{-mabicalls} code has traditionally been position-independent,
18869 regardless of options like @option{-fPIC} and @option{-fpic}. However,
18870 as an extension, the GNU toolchain allows executables to use absolute
18871 accesses for locally-binding symbols. It can also use shorter GP
18872 initialization sequences and generate direct calls to locally-defined
18873 functions. This mode is selected by @option{-mno-shared}.
18875 @option{-mno-shared} depends on binutils 2.16 or higher and generates
18876 objects that can only be linked by the GNU linker. However, the option
18877 does not affect the ABI of the final executable; it only affects the ABI
18878 of relocatable objects. Using @option{-mno-shared} generally makes
18879 executables both smaller and quicker.
18881 @option{-mshared} is the default.
18887 Assume (do not assume) that the static and dynamic linkers
18888 support PLTs and copy relocations. This option only affects
18889 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
18890 has no effect without @option{-msym32}.
18892 You can make @option{-mplt} the default by configuring
18893 GCC with @option{--with-mips-plt}. The default is
18894 @option{-mno-plt} otherwise.
18900 Lift (do not lift) the usual restrictions on the size of the global
18903 GCC normally uses a single instruction to load values from the GOT@.
18904 While this is relatively efficient, it only works if the GOT
18905 is smaller than about 64k. Anything larger causes the linker
18906 to report an error such as:
18908 @cindex relocation truncated to fit (MIPS)
18910 relocation truncated to fit: R_MIPS_GOT16 foobar
18913 If this happens, you should recompile your code with @option{-mxgot}.
18914 This works with very large GOTs, although the code is also
18915 less efficient, since it takes three instructions to fetch the
18916 value of a global symbol.
18918 Note that some linkers can create multiple GOTs. If you have such a
18919 linker, you should only need to use @option{-mxgot} when a single object
18920 file accesses more than 64k's worth of GOT entries. Very few do.
18922 These options have no effect unless GCC is generating position
18927 Assume that general-purpose registers are 32 bits wide.
18931 Assume that general-purpose registers are 64 bits wide.
18935 Assume that floating-point registers are 32 bits wide.
18939 Assume that floating-point registers are 64 bits wide.
18943 Do not assume the width of floating-point registers.
18946 @opindex mhard-float
18947 Use floating-point coprocessor instructions.
18950 @opindex msoft-float
18951 Do not use floating-point coprocessor instructions. Implement
18952 floating-point calculations using library calls instead.
18956 Equivalent to @option{-msoft-float}, but additionally asserts that the
18957 program being compiled does not perform any floating-point operations.
18958 This option is presently supported only by some bare-metal MIPS
18959 configurations, where it may select a special set of libraries
18960 that lack all floating-point support (including, for example, the
18961 floating-point @code{printf} formats).
18962 If code compiled with @option{-mno-float} accidentally contains
18963 floating-point operations, it is likely to suffer a link-time
18964 or run-time failure.
18966 @item -msingle-float
18967 @opindex msingle-float
18968 Assume that the floating-point coprocessor only supports single-precision
18971 @item -mdouble-float
18972 @opindex mdouble-float
18973 Assume that the floating-point coprocessor supports double-precision
18974 operations. This is the default.
18977 @itemx -mno-odd-spreg
18978 @opindex modd-spreg
18979 @opindex mno-odd-spreg
18980 Enable the use of odd-numbered single-precision floating-point registers
18981 for the o32 ABI. This is the default for processors that are known to
18982 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18986 @itemx -mabs=legacy
18988 @opindex mabs=legacy
18989 These options control the treatment of the special not-a-number (NaN)
18990 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18991 @code{neg.@i{fmt}} machine instructions.
18993 By default or when @option{-mabs=legacy} is used the legacy
18994 treatment is selected. In this case these instructions are considered
18995 arithmetic and avoided where correct operation is required and the
18996 input operand might be a NaN. A longer sequence of instructions that
18997 manipulate the sign bit of floating-point datum manually is used
18998 instead unless the @option{-ffinite-math-only} option has also been
19001 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
19002 this case these instructions are considered non-arithmetic and therefore
19003 operating correctly in all cases, including in particular where the
19004 input operand is a NaN. These instructions are therefore always used
19005 for the respective operations.
19008 @itemx -mnan=legacy
19010 @opindex mnan=legacy
19011 These options control the encoding of the special not-a-number (NaN)
19012 IEEE 754 floating-point data.
19014 The @option{-mnan=legacy} option selects the legacy encoding. In this
19015 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
19016 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
19017 by the first bit of their trailing significand field being 1.
19019 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
19020 this case qNaNs are denoted by the first bit of their trailing
19021 significand field being 1, whereas sNaNs are denoted by the first bit of
19022 their trailing significand field being 0.
19024 The default is @option{-mnan=legacy} unless GCC has been configured with
19025 @option{--with-nan=2008}.
19031 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
19032 implement atomic memory built-in functions. When neither option is
19033 specified, GCC uses the instructions if the target architecture
19036 @option{-mllsc} is useful if the runtime environment can emulate the
19037 instructions and @option{-mno-llsc} can be useful when compiling for
19038 nonstandard ISAs. You can make either option the default by
19039 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
19040 respectively. @option{--with-llsc} is the default for some
19041 configurations; see the installation documentation for details.
19047 Use (do not use) revision 1 of the MIPS DSP ASE@.
19048 @xref{MIPS DSP Built-in Functions}. This option defines the
19049 preprocessor macro @code{__mips_dsp}. It also defines
19050 @code{__mips_dsp_rev} to 1.
19056 Use (do not use) revision 2 of the MIPS DSP ASE@.
19057 @xref{MIPS DSP Built-in Functions}. This option defines the
19058 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
19059 It also defines @code{__mips_dsp_rev} to 2.
19062 @itemx -mno-smartmips
19063 @opindex msmartmips
19064 @opindex mno-smartmips
19065 Use (do not use) the MIPS SmartMIPS ASE.
19067 @item -mpaired-single
19068 @itemx -mno-paired-single
19069 @opindex mpaired-single
19070 @opindex mno-paired-single
19071 Use (do not use) paired-single floating-point instructions.
19072 @xref{MIPS Paired-Single Support}. This option requires
19073 hardware floating-point support to be enabled.
19079 Use (do not use) MIPS Digital Media Extension instructions.
19080 This option can only be used when generating 64-bit code and requires
19081 hardware floating-point support to be enabled.
19086 @opindex mno-mips3d
19087 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
19088 The option @option{-mips3d} implies @option{-mpaired-single}.
19091 @itemx -mno-micromips
19092 @opindex mmicromips
19093 @opindex mno-mmicromips
19094 Generate (do not generate) microMIPS code.
19096 MicroMIPS code generation can also be controlled on a per-function basis
19097 by means of @code{micromips} and @code{nomicromips} attributes.
19098 @xref{Function Attributes}, for more information.
19104 Use (do not use) MT Multithreading instructions.
19110 Use (do not use) the MIPS MCU ASE instructions.
19116 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
19122 Use (do not use) the MIPS Virtualization Application Specific instructions.
19128 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
19132 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
19133 an explanation of the default and the way that the pointer size is
19138 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
19140 The default size of @code{int}s, @code{long}s and pointers depends on
19141 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
19142 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
19143 32-bit @code{long}s. Pointers are the same size as @code{long}s,
19144 or the same size as integer registers, whichever is smaller.
19150 Assume (do not assume) that all symbols have 32-bit values, regardless
19151 of the selected ABI@. This option is useful in combination with
19152 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
19153 to generate shorter and faster references to symbolic addresses.
19157 Put definitions of externally-visible data in a small data section
19158 if that data is no bigger than @var{num} bytes. GCC can then generate
19159 more efficient accesses to the data; see @option{-mgpopt} for details.
19161 The default @option{-G} option depends on the configuration.
19163 @item -mlocal-sdata
19164 @itemx -mno-local-sdata
19165 @opindex mlocal-sdata
19166 @opindex mno-local-sdata
19167 Extend (do not extend) the @option{-G} behavior to local data too,
19168 such as to static variables in C@. @option{-mlocal-sdata} is the
19169 default for all configurations.
19171 If the linker complains that an application is using too much small data,
19172 you might want to try rebuilding the less performance-critical parts with
19173 @option{-mno-local-sdata}. You might also want to build large
19174 libraries with @option{-mno-local-sdata}, so that the libraries leave
19175 more room for the main program.
19177 @item -mextern-sdata
19178 @itemx -mno-extern-sdata
19179 @opindex mextern-sdata
19180 @opindex mno-extern-sdata
19181 Assume (do not assume) that externally-defined data is in
19182 a small data section if the size of that data is within the @option{-G} limit.
19183 @option{-mextern-sdata} is the default for all configurations.
19185 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
19186 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
19187 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
19188 is placed in a small data section. If @var{Var} is defined by another
19189 module, you must either compile that module with a high-enough
19190 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
19191 definition. If @var{Var} is common, you must link the application
19192 with a high-enough @option{-G} setting.
19194 The easiest way of satisfying these restrictions is to compile
19195 and link every module with the same @option{-G} option. However,
19196 you may wish to build a library that supports several different
19197 small data limits. You can do this by compiling the library with
19198 the highest supported @option{-G} setting and additionally using
19199 @option{-mno-extern-sdata} to stop the library from making assumptions
19200 about externally-defined data.
19206 Use (do not use) GP-relative accesses for symbols that are known to be
19207 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
19208 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
19211 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
19212 might not hold the value of @code{_gp}. For example, if the code is
19213 part of a library that might be used in a boot monitor, programs that
19214 call boot monitor routines pass an unknown value in @code{$gp}.
19215 (In such situations, the boot monitor itself is usually compiled
19216 with @option{-G0}.)
19218 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
19219 @option{-mno-extern-sdata}.
19221 @item -membedded-data
19222 @itemx -mno-embedded-data
19223 @opindex membedded-data
19224 @opindex mno-embedded-data
19225 Allocate variables to the read-only data section first if possible, then
19226 next in the small data section if possible, otherwise in data. This gives
19227 slightly slower code than the default, but reduces the amount of RAM required
19228 when executing, and thus may be preferred for some embedded systems.
19230 @item -muninit-const-in-rodata
19231 @itemx -mno-uninit-const-in-rodata
19232 @opindex muninit-const-in-rodata
19233 @opindex mno-uninit-const-in-rodata
19234 Put uninitialized @code{const} variables in the read-only data section.
19235 This option is only meaningful in conjunction with @option{-membedded-data}.
19237 @item -mcode-readable=@var{setting}
19238 @opindex mcode-readable
19239 Specify whether GCC may generate code that reads from executable sections.
19240 There are three possible settings:
19243 @item -mcode-readable=yes
19244 Instructions may freely access executable sections. This is the
19247 @item -mcode-readable=pcrel
19248 MIPS16 PC-relative load instructions can access executable sections,
19249 but other instructions must not do so. This option is useful on 4KSc
19250 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
19251 It is also useful on processors that can be configured to have a dual
19252 instruction/data SRAM interface and that, like the M4K, automatically
19253 redirect PC-relative loads to the instruction RAM.
19255 @item -mcode-readable=no
19256 Instructions must not access executable sections. This option can be
19257 useful on targets that are configured to have a dual instruction/data
19258 SRAM interface but that (unlike the M4K) do not automatically redirect
19259 PC-relative loads to the instruction RAM.
19262 @item -msplit-addresses
19263 @itemx -mno-split-addresses
19264 @opindex msplit-addresses
19265 @opindex mno-split-addresses
19266 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
19267 relocation operators. This option has been superseded by
19268 @option{-mexplicit-relocs} but is retained for backwards compatibility.
19270 @item -mexplicit-relocs
19271 @itemx -mno-explicit-relocs
19272 @opindex mexplicit-relocs
19273 @opindex mno-explicit-relocs
19274 Use (do not use) assembler relocation operators when dealing with symbolic
19275 addresses. The alternative, selected by @option{-mno-explicit-relocs},
19276 is to use assembler macros instead.
19278 @option{-mexplicit-relocs} is the default if GCC was configured
19279 to use an assembler that supports relocation operators.
19281 @item -mcheck-zero-division
19282 @itemx -mno-check-zero-division
19283 @opindex mcheck-zero-division
19284 @opindex mno-check-zero-division
19285 Trap (do not trap) on integer division by zero.
19287 The default is @option{-mcheck-zero-division}.
19289 @item -mdivide-traps
19290 @itemx -mdivide-breaks
19291 @opindex mdivide-traps
19292 @opindex mdivide-breaks
19293 MIPS systems check for division by zero by generating either a
19294 conditional trap or a break instruction. Using traps results in
19295 smaller code, but is only supported on MIPS II and later. Also, some
19296 versions of the Linux kernel have a bug that prevents trap from
19297 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
19298 allow conditional traps on architectures that support them and
19299 @option{-mdivide-breaks} to force the use of breaks.
19301 The default is usually @option{-mdivide-traps}, but this can be
19302 overridden at configure time using @option{--with-divide=breaks}.
19303 Divide-by-zero checks can be completely disabled using
19304 @option{-mno-check-zero-division}.
19309 @opindex mno-memcpy
19310 Force (do not force) the use of @code{memcpy} for non-trivial block
19311 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
19312 most constant-sized copies.
19315 @itemx -mno-long-calls
19316 @opindex mlong-calls
19317 @opindex mno-long-calls
19318 Disable (do not disable) use of the @code{jal} instruction. Calling
19319 functions using @code{jal} is more efficient but requires the caller
19320 and callee to be in the same 256 megabyte segment.
19322 This option has no effect on abicalls code. The default is
19323 @option{-mno-long-calls}.
19329 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
19330 instructions, as provided by the R4650 ISA@.
19336 Enable (disable) use of the @code{madd} and @code{msub} integer
19337 instructions. The default is @option{-mimadd} on architectures
19338 that support @code{madd} and @code{msub} except for the 74k
19339 architecture where it was found to generate slower code.
19342 @itemx -mno-fused-madd
19343 @opindex mfused-madd
19344 @opindex mno-fused-madd
19345 Enable (disable) use of the floating-point multiply-accumulate
19346 instructions, when they are available. The default is
19347 @option{-mfused-madd}.
19349 On the R8000 CPU when multiply-accumulate instructions are used,
19350 the intermediate product is calculated to infinite precision
19351 and is not subject to the FCSR Flush to Zero bit. This may be
19352 undesirable in some circumstances. On other processors the result
19353 is numerically identical to the equivalent computation using
19354 separate multiply, add, subtract and negate instructions.
19358 Tell the MIPS assembler to not run its preprocessor over user
19359 assembler files (with a @samp{.s} suffix) when assembling them.
19364 @opindex mno-fix-24k
19365 Work around the 24K E48 (lost data on stores during refill) errata.
19366 The workarounds are implemented by the assembler rather than by GCC@.
19369 @itemx -mno-fix-r4000
19370 @opindex mfix-r4000
19371 @opindex mno-fix-r4000
19372 Work around certain R4000 CPU errata:
19375 A double-word or a variable shift may give an incorrect result if executed
19376 immediately after starting an integer division.
19378 A double-word or a variable shift may give an incorrect result if executed
19379 while an integer multiplication is in progress.
19381 An integer division may give an incorrect result if started in a delay slot
19382 of a taken branch or a jump.
19386 @itemx -mno-fix-r4400
19387 @opindex mfix-r4400
19388 @opindex mno-fix-r4400
19389 Work around certain R4400 CPU errata:
19392 A double-word or a variable shift may give an incorrect result if executed
19393 immediately after starting an integer division.
19397 @itemx -mno-fix-r10000
19398 @opindex mfix-r10000
19399 @opindex mno-fix-r10000
19400 Work around certain R10000 errata:
19403 @code{ll}/@code{sc} sequences may not behave atomically on revisions
19404 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
19407 This option can only be used if the target architecture supports
19408 branch-likely instructions. @option{-mfix-r10000} is the default when
19409 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
19413 @itemx -mno-fix-rm7000
19414 @opindex mfix-rm7000
19415 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
19416 workarounds are implemented by the assembler rather than by GCC@.
19419 @itemx -mno-fix-vr4120
19420 @opindex mfix-vr4120
19421 Work around certain VR4120 errata:
19424 @code{dmultu} does not always produce the correct result.
19426 @code{div} and @code{ddiv} do not always produce the correct result if one
19427 of the operands is negative.
19429 The workarounds for the division errata rely on special functions in
19430 @file{libgcc.a}. At present, these functions are only provided by
19431 the @code{mips64vr*-elf} configurations.
19433 Other VR4120 errata require a NOP to be inserted between certain pairs of
19434 instructions. These errata are handled by the assembler, not by GCC itself.
19437 @opindex mfix-vr4130
19438 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
19439 workarounds are implemented by the assembler rather than by GCC,
19440 although GCC avoids using @code{mflo} and @code{mfhi} if the
19441 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
19442 instructions are available instead.
19445 @itemx -mno-fix-sb1
19447 Work around certain SB-1 CPU core errata.
19448 (This flag currently works around the SB-1 revision 2
19449 ``F1'' and ``F2'' floating-point errata.)
19451 @item -mr10k-cache-barrier=@var{setting}
19452 @opindex mr10k-cache-barrier
19453 Specify whether GCC should insert cache barriers to avoid the
19454 side-effects of speculation on R10K processors.
19456 In common with many processors, the R10K tries to predict the outcome
19457 of a conditional branch and speculatively executes instructions from
19458 the ``taken'' branch. It later aborts these instructions if the
19459 predicted outcome is wrong. However, on the R10K, even aborted
19460 instructions can have side effects.
19462 This problem only affects kernel stores and, depending on the system,
19463 kernel loads. As an example, a speculatively-executed store may load
19464 the target memory into cache and mark the cache line as dirty, even if
19465 the store itself is later aborted. If a DMA operation writes to the
19466 same area of memory before the ``dirty'' line is flushed, the cached
19467 data overwrites the DMA-ed data. See the R10K processor manual
19468 for a full description, including other potential problems.
19470 One workaround is to insert cache barrier instructions before every memory
19471 access that might be speculatively executed and that might have side
19472 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
19473 controls GCC's implementation of this workaround. It assumes that
19474 aborted accesses to any byte in the following regions does not have
19479 the memory occupied by the current function's stack frame;
19482 the memory occupied by an incoming stack argument;
19485 the memory occupied by an object with a link-time-constant address.
19488 It is the kernel's responsibility to ensure that speculative
19489 accesses to these regions are indeed safe.
19491 If the input program contains a function declaration such as:
19497 then the implementation of @code{foo} must allow @code{j foo} and
19498 @code{jal foo} to be executed speculatively. GCC honors this
19499 restriction for functions it compiles itself. It expects non-GCC
19500 functions (such as hand-written assembly code) to do the same.
19502 The option has three forms:
19505 @item -mr10k-cache-barrier=load-store
19506 Insert a cache barrier before a load or store that might be
19507 speculatively executed and that might have side effects even
19510 @item -mr10k-cache-barrier=store
19511 Insert a cache barrier before a store that might be speculatively
19512 executed and that might have side effects even if aborted.
19514 @item -mr10k-cache-barrier=none
19515 Disable the insertion of cache barriers. This is the default setting.
19518 @item -mflush-func=@var{func}
19519 @itemx -mno-flush-func
19520 @opindex mflush-func
19521 Specifies the function to call to flush the I and D caches, or to not
19522 call any such function. If called, the function must take the same
19523 arguments as the common @code{_flush_func}, that is, the address of the
19524 memory range for which the cache is being flushed, the size of the
19525 memory range, and the number 3 (to flush both caches). The default
19526 depends on the target GCC was configured for, but commonly is either
19527 @code{_flush_func} or @code{__cpu_flush}.
19529 @item mbranch-cost=@var{num}
19530 @opindex mbranch-cost
19531 Set the cost of branches to roughly @var{num} ``simple'' instructions.
19532 This cost is only a heuristic and is not guaranteed to produce
19533 consistent results across releases. A zero cost redundantly selects
19534 the default, which is based on the @option{-mtune} setting.
19536 @item -mbranch-likely
19537 @itemx -mno-branch-likely
19538 @opindex mbranch-likely
19539 @opindex mno-branch-likely
19540 Enable or disable use of Branch Likely instructions, regardless of the
19541 default for the selected architecture. By default, Branch Likely
19542 instructions may be generated if they are supported by the selected
19543 architecture. An exception is for the MIPS32 and MIPS64 architectures
19544 and processors that implement those architectures; for those, Branch
19545 Likely instructions are not be generated by default because the MIPS32
19546 and MIPS64 architectures specifically deprecate their use.
19548 @item -mcompact-branches=never
19549 @itemx -mcompact-branches=optimal
19550 @itemx -mcompact-branches=always
19551 @opindex mcompact-branches=never
19552 @opindex mcompact-branches=optimal
19553 @opindex mcompact-branches=always
19554 These options control which form of branches will be generated. The
19555 default is @option{-mcompact-branches=optimal}.
19557 The @option{-mcompact-branches=never} option ensures that compact branch
19558 instructions will never be generated.
19560 The @option{-mcompact-branches=always} option ensures that a compact
19561 branch instruction will be generated if available. If a compact branch
19562 instruction is not available, a delay slot form of the branch will be
19565 This option is supported from MIPS Release 6 onwards.
19567 The @option{-mcompact-branches=optimal} option will cause a delay slot
19568 branch to be used if one is available in the current ISA and the delay
19569 slot is successfully filled. If the delay slot is not filled, a compact
19570 branch will be chosen if one is available.
19572 @item -mfp-exceptions
19573 @itemx -mno-fp-exceptions
19574 @opindex mfp-exceptions
19575 Specifies whether FP exceptions are enabled. This affects how
19576 FP instructions are scheduled for some processors.
19577 The default is that FP exceptions are
19580 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
19581 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
19584 @item -mvr4130-align
19585 @itemx -mno-vr4130-align
19586 @opindex mvr4130-align
19587 The VR4130 pipeline is two-way superscalar, but can only issue two
19588 instructions together if the first one is 8-byte aligned. When this
19589 option is enabled, GCC aligns pairs of instructions that it
19590 thinks should execute in parallel.
19592 This option only has an effect when optimizing for the VR4130.
19593 It normally makes code faster, but at the expense of making it bigger.
19594 It is enabled by default at optimization level @option{-O3}.
19599 Enable (disable) generation of @code{synci} instructions on
19600 architectures that support it. The @code{synci} instructions (if
19601 enabled) are generated when @code{__builtin___clear_cache} is
19604 This option defaults to @option{-mno-synci}, but the default can be
19605 overridden by configuring GCC with @option{--with-synci}.
19607 When compiling code for single processor systems, it is generally safe
19608 to use @code{synci}. However, on many multi-core (SMP) systems, it
19609 does not invalidate the instruction caches on all cores and may lead
19610 to undefined behavior.
19612 @item -mrelax-pic-calls
19613 @itemx -mno-relax-pic-calls
19614 @opindex mrelax-pic-calls
19615 Try to turn PIC calls that are normally dispatched via register
19616 @code{$25} into direct calls. This is only possible if the linker can
19617 resolve the destination at link time and if the destination is within
19618 range for a direct call.
19620 @option{-mrelax-pic-calls} is the default if GCC was configured to use
19621 an assembler and a linker that support the @code{.reloc} assembly
19622 directive and @option{-mexplicit-relocs} is in effect. With
19623 @option{-mno-explicit-relocs}, this optimization can be performed by the
19624 assembler and the linker alone without help from the compiler.
19626 @item -mmcount-ra-address
19627 @itemx -mno-mcount-ra-address
19628 @opindex mmcount-ra-address
19629 @opindex mno-mcount-ra-address
19630 Emit (do not emit) code that allows @code{_mcount} to modify the
19631 calling function's return address. When enabled, this option extends
19632 the usual @code{_mcount} interface with a new @var{ra-address}
19633 parameter, which has type @code{intptr_t *} and is passed in register
19634 @code{$12}. @code{_mcount} can then modify the return address by
19635 doing both of the following:
19638 Returning the new address in register @code{$31}.
19640 Storing the new address in @code{*@var{ra-address}},
19641 if @var{ra-address} is nonnull.
19644 The default is @option{-mno-mcount-ra-address}.
19646 @item -mframe-header-opt
19647 @itemx -mno-frame-header-opt
19648 @opindex mframe-header-opt
19649 Enable (disable) frame header optimization in the o32 ABI. When using the
19650 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
19651 function to write out register arguments. When enabled, this optimization
19652 will suppress the allocation of the frame header if it can be determined that
19655 This optimization is off by default at all optimization levels.
19660 @subsection MMIX Options
19661 @cindex MMIX Options
19663 These options are defined for the MMIX:
19667 @itemx -mno-libfuncs
19669 @opindex mno-libfuncs
19670 Specify that intrinsic library functions are being compiled, passing all
19671 values in registers, no matter the size.
19674 @itemx -mno-epsilon
19676 @opindex mno-epsilon
19677 Generate floating-point comparison instructions that compare with respect
19678 to the @code{rE} epsilon register.
19680 @item -mabi=mmixware
19682 @opindex mabi=mmixware
19684 Generate code that passes function parameters and return values that (in
19685 the called function) are seen as registers @code{$0} and up, as opposed to
19686 the GNU ABI which uses global registers @code{$231} and up.
19688 @item -mzero-extend
19689 @itemx -mno-zero-extend
19690 @opindex mzero-extend
19691 @opindex mno-zero-extend
19692 When reading data from memory in sizes shorter than 64 bits, use (do not
19693 use) zero-extending load instructions by default, rather than
19694 sign-extending ones.
19697 @itemx -mno-knuthdiv
19699 @opindex mno-knuthdiv
19700 Make the result of a division yielding a remainder have the same sign as
19701 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
19702 remainder follows the sign of the dividend. Both methods are
19703 arithmetically valid, the latter being almost exclusively used.
19705 @item -mtoplevel-symbols
19706 @itemx -mno-toplevel-symbols
19707 @opindex mtoplevel-symbols
19708 @opindex mno-toplevel-symbols
19709 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
19710 code can be used with the @code{PREFIX} assembly directive.
19714 Generate an executable in the ELF format, rather than the default
19715 @samp{mmo} format used by the @command{mmix} simulator.
19717 @item -mbranch-predict
19718 @itemx -mno-branch-predict
19719 @opindex mbranch-predict
19720 @opindex mno-branch-predict
19721 Use (do not use) the probable-branch instructions, when static branch
19722 prediction indicates a probable branch.
19724 @item -mbase-addresses
19725 @itemx -mno-base-addresses
19726 @opindex mbase-addresses
19727 @opindex mno-base-addresses
19728 Generate (do not generate) code that uses @emph{base addresses}. Using a
19729 base address automatically generates a request (handled by the assembler
19730 and the linker) for a constant to be set up in a global register. The
19731 register is used for one or more base address requests within the range 0
19732 to 255 from the value held in the register. The generally leads to short
19733 and fast code, but the number of different data items that can be
19734 addressed is limited. This means that a program that uses lots of static
19735 data may require @option{-mno-base-addresses}.
19737 @item -msingle-exit
19738 @itemx -mno-single-exit
19739 @opindex msingle-exit
19740 @opindex mno-single-exit
19741 Force (do not force) generated code to have a single exit point in each
19745 @node MN10300 Options
19746 @subsection MN10300 Options
19747 @cindex MN10300 options
19749 These @option{-m} options are defined for Matsushita MN10300 architectures:
19754 Generate code to avoid bugs in the multiply instructions for the MN10300
19755 processors. This is the default.
19757 @item -mno-mult-bug
19758 @opindex mno-mult-bug
19759 Do not generate code to avoid bugs in the multiply instructions for the
19760 MN10300 processors.
19764 Generate code using features specific to the AM33 processor.
19768 Do not generate code using features specific to the AM33 processor. This
19773 Generate code using features specific to the AM33/2.0 processor.
19777 Generate code using features specific to the AM34 processor.
19779 @item -mtune=@var{cpu-type}
19781 Use the timing characteristics of the indicated CPU type when
19782 scheduling instructions. This does not change the targeted processor
19783 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
19784 @samp{am33-2} or @samp{am34}.
19786 @item -mreturn-pointer-on-d0
19787 @opindex mreturn-pointer-on-d0
19788 When generating a function that returns a pointer, return the pointer
19789 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
19790 only in @code{a0}, and attempts to call such functions without a prototype
19791 result in errors. Note that this option is on by default; use
19792 @option{-mno-return-pointer-on-d0} to disable it.
19796 Do not link in the C run-time initialization object file.
19800 Indicate to the linker that it should perform a relaxation optimization pass
19801 to shorten branches, calls and absolute memory addresses. This option only
19802 has an effect when used on the command line for the final link step.
19804 This option makes symbolic debugging impossible.
19808 Allow the compiler to generate @emph{Long Instruction Word}
19809 instructions if the target is the @samp{AM33} or later. This is the
19810 default. This option defines the preprocessor macro @code{__LIW__}.
19814 Do not allow the compiler to generate @emph{Long Instruction Word}
19815 instructions. This option defines the preprocessor macro
19820 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
19821 instructions if the target is the @samp{AM33} or later. This is the
19822 default. This option defines the preprocessor macro @code{__SETLB__}.
19826 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
19827 instructions. This option defines the preprocessor macro
19828 @code{__NO_SETLB__}.
19832 @node Moxie Options
19833 @subsection Moxie Options
19834 @cindex Moxie Options
19840 Generate big-endian code. This is the default for @samp{moxie-*-*}
19845 Generate little-endian code.
19849 Generate mul.x and umul.x instructions. This is the default for
19850 @samp{moxiebox-*-*} configurations.
19854 Do not link in the C run-time initialization object file.
19858 @node MSP430 Options
19859 @subsection MSP430 Options
19860 @cindex MSP430 Options
19862 These options are defined for the MSP430:
19868 Force assembly output to always use hex constants. Normally such
19869 constants are signed decimals, but this option is available for
19870 testsuite and/or aesthetic purposes.
19874 Select the MCU to target. This is used to create a C preprocessor
19875 symbol based upon the MCU name, converted to upper case and pre- and
19876 post-fixed with @samp{__}. This in turn is used by the
19877 @file{msp430.h} header file to select an MCU-specific supplementary
19880 The option also sets the ISA to use. If the MCU name is one that is
19881 known to only support the 430 ISA then that is selected, otherwise the
19882 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
19883 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
19884 name selects the 430X ISA.
19886 In addition an MCU-specific linker script is added to the linker
19887 command line. The script's name is the name of the MCU with
19888 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
19889 command line defines the C preprocessor symbol @code{__XXX__} and
19890 cause the linker to search for a script called @file{xxx.ld}.
19892 This option is also passed on to the assembler.
19895 @itemx -mno-warn-mcu
19897 @opindex mno-warn-mcu
19898 This option enables or disables warnings about conflicts between the
19899 MCU name specified by the @option{-mmcu} option and the ISA set by the
19900 @option{-mcpu} option and/or the hardware multiply support set by the
19901 @option{-mhwmult} option. It also toggles warnings about unrecognized
19902 MCU names. This option is on by default.
19906 Specifies the ISA to use. Accepted values are @samp{msp430},
19907 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
19908 @option{-mmcu=} option should be used to select the ISA.
19912 Link to the simulator runtime libraries and linker script. Overrides
19913 any scripts that would be selected by the @option{-mmcu=} option.
19917 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
19921 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
19925 This option is passed to the assembler and linker, and allows the
19926 linker to perform certain optimizations that cannot be done until
19931 Describes the type of hardware multiply supported by the target.
19932 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19933 for the original 16-bit-only multiply supported by early MCUs.
19934 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19935 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19936 A value of @samp{auto} can also be given. This tells GCC to deduce
19937 the hardware multiply support based upon the MCU name provided by the
19938 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
19939 the MCU name is not recognized then no hardware multiply support is
19940 assumed. @code{auto} is the default setting.
19942 Hardware multiplies are normally performed by calling a library
19943 routine. This saves space in the generated code. When compiling at
19944 @option{-O3} or higher however the hardware multiplier is invoked
19945 inline. This makes for bigger, but faster code.
19947 The hardware multiply routines disable interrupts whilst running and
19948 restore the previous interrupt state when they finish. This makes
19949 them safe to use inside interrupt handlers as well as in normal code.
19953 Enable the use of a minimum runtime environment - no static
19954 initializers or constructors. This is intended for memory-constrained
19955 devices. The compiler includes special symbols in some objects
19956 that tell the linker and runtime which code fragments are required.
19958 @item -mcode-region=
19959 @itemx -mdata-region=
19960 @opindex mcode-region
19961 @opindex mdata-region
19962 These options tell the compiler where to place functions and data that
19963 do not have one of the @code{lower}, @code{upper}, @code{either} or
19964 @code{section} attributes. Possible values are @code{lower},
19965 @code{upper}, @code{either} or @code{any}. The first three behave
19966 like the corresponding attribute. The fourth possible value -
19967 @code{any} - is the default. It leaves placement entirely up to the
19968 linker script and how it assigns the standard sections
19969 (@code{.text}, @code{.data}, etc) to the memory regions.
19971 @item -msilicon-errata=
19972 @opindex msilicon-errata
19973 This option passes on a request to assembler to enable the fixes for
19974 the named silicon errata.
19976 @item -msilicon-errata-warn=
19977 @opindex msilicon-errata-warn
19978 This option passes on a request to the assembler to enable warning
19979 messages when a silicon errata might need to be applied.
19983 @node NDS32 Options
19984 @subsection NDS32 Options
19985 @cindex NDS32 Options
19987 These options are defined for NDS32 implementations:
19992 @opindex mbig-endian
19993 Generate code in big-endian mode.
19995 @item -mlittle-endian
19996 @opindex mlittle-endian
19997 Generate code in little-endian mode.
19999 @item -mreduced-regs
20000 @opindex mreduced-regs
20001 Use reduced-set registers for register allocation.
20004 @opindex mfull-regs
20005 Use full-set registers for register allocation.
20009 Generate conditional move instructions.
20013 Do not generate conditional move instructions.
20017 Generate performance extension instructions.
20019 @item -mno-perf-ext
20020 @opindex mno-perf-ext
20021 Do not generate performance extension instructions.
20025 Generate v3 push25/pop25 instructions.
20028 @opindex mno-v3push
20029 Do not generate v3 push25/pop25 instructions.
20033 Generate 16-bit instructions.
20036 @opindex mno-16-bit
20037 Do not generate 16-bit instructions.
20039 @item -misr-vector-size=@var{num}
20040 @opindex misr-vector-size
20041 Specify the size of each interrupt vector, which must be 4 or 16.
20043 @item -mcache-block-size=@var{num}
20044 @opindex mcache-block-size
20045 Specify the size of each cache block,
20046 which must be a power of 2 between 4 and 512.
20048 @item -march=@var{arch}
20050 Specify the name of the target architecture.
20052 @item -mcmodel=@var{code-model}
20054 Set the code model to one of
20057 All the data and read-only data segments must be within 512KB addressing space.
20058 The text segment must be within 16MB addressing space.
20059 @item @samp{medium}
20060 The data segment must be within 512KB while the read-only data segment can be
20061 within 4GB addressing space. The text segment should be still within 16MB
20064 All the text and data segments can be within 4GB addressing space.
20068 @opindex mctor-dtor
20069 Enable constructor/destructor feature.
20073 Guide linker to relax instructions.
20077 @node Nios II Options
20078 @subsection Nios II Options
20079 @cindex Nios II options
20080 @cindex Altera Nios II options
20082 These are the options defined for the Altera Nios II processor.
20088 @cindex smaller data references
20089 Put global and static objects less than or equal to @var{num} bytes
20090 into the small data or BSS sections instead of the normal data or BSS
20091 sections. The default value of @var{num} is 8.
20093 @item -mgpopt=@var{option}
20098 Generate (do not generate) GP-relative accesses. The following
20099 @var{option} names are recognized:
20104 Do not generate GP-relative accesses.
20107 Generate GP-relative accesses for small data objects that are not
20108 external, weak, or uninitialized common symbols.
20109 Also use GP-relative addressing for objects that
20110 have been explicitly placed in a small data section via a @code{section}
20114 As for @samp{local}, but also generate GP-relative accesses for
20115 small data objects that are external, weak, or common. If you use this option,
20116 you must ensure that all parts of your program (including libraries) are
20117 compiled with the same @option{-G} setting.
20120 Generate GP-relative accesses for all data objects in the program. If you
20121 use this option, the entire data and BSS segments
20122 of your program must fit in 64K of memory and you must use an appropriate
20123 linker script to allocate them within the addressable range of the
20127 Generate GP-relative addresses for function pointers as well as data
20128 pointers. If you use this option, the entire text, data, and BSS segments
20129 of your program must fit in 64K of memory and you must use an appropriate
20130 linker script to allocate them within the addressable range of the
20135 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
20136 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
20138 The default is @option{-mgpopt} except when @option{-fpic} or
20139 @option{-fPIC} is specified to generate position-independent code.
20140 Note that the Nios II ABI does not permit GP-relative accesses from
20143 You may need to specify @option{-mno-gpopt} explicitly when building
20144 programs that include large amounts of small data, including large
20145 GOT data sections. In this case, the 16-bit offset for GP-relative
20146 addressing may not be large enough to allow access to the entire
20147 small data section.
20153 Generate little-endian (default) or big-endian (experimental) code,
20156 @item -march=@var{arch}
20158 This specifies the name of the target Nios II architecture. GCC uses this
20159 name to determine what kind of instructions it can emit when generating
20160 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
20162 The preprocessor macro @code{__nios2_arch__} is available to programs,
20163 with value 1 or 2, indicating the targeted ISA level.
20165 @item -mbypass-cache
20166 @itemx -mno-bypass-cache
20167 @opindex mno-bypass-cache
20168 @opindex mbypass-cache
20169 Force all load and store instructions to always bypass cache by
20170 using I/O variants of the instructions. The default is not to
20173 @item -mno-cache-volatile
20174 @itemx -mcache-volatile
20175 @opindex mcache-volatile
20176 @opindex mno-cache-volatile
20177 Volatile memory access bypass the cache using the I/O variants of
20178 the load and store instructions. The default is not to bypass the cache.
20180 @item -mno-fast-sw-div
20181 @itemx -mfast-sw-div
20182 @opindex mno-fast-sw-div
20183 @opindex mfast-sw-div
20184 Do not use table-based fast divide for small numbers. The default
20185 is to use the fast divide at @option{-O3} and above.
20189 @itemx -mno-hw-mulx
20193 @opindex mno-hw-mul
20195 @opindex mno-hw-mulx
20197 @opindex mno-hw-div
20199 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
20200 instructions by the compiler. The default is to emit @code{mul}
20201 and not emit @code{div} and @code{mulx}.
20207 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
20208 CDX (code density) instructions. Enabling these instructions also
20209 requires @option{-march=r2}. Since these instructions are optional
20210 extensions to the R2 architecture, the default is not to emit them.
20212 @item -mcustom-@var{insn}=@var{N}
20213 @itemx -mno-custom-@var{insn}
20214 @opindex mcustom-@var{insn}
20215 @opindex mno-custom-@var{insn}
20216 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
20217 custom instruction with encoding @var{N} when generating code that uses
20218 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
20219 instruction 253 for single-precision floating-point add operations instead
20220 of the default behavior of using a library call.
20222 The following values of @var{insn} are supported. Except as otherwise
20223 noted, floating-point operations are expected to be implemented with
20224 normal IEEE 754 semantics and correspond directly to the C operators or the
20225 equivalent GCC built-in functions (@pxref{Other Builtins}).
20227 Single-precision floating point:
20230 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
20231 Binary arithmetic operations.
20237 Unary absolute value.
20239 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
20240 Comparison operations.
20242 @item @samp{fmins}, @samp{fmaxs}
20243 Floating-point minimum and maximum. These instructions are only
20244 generated if @option{-ffinite-math-only} is specified.
20246 @item @samp{fsqrts}
20247 Unary square root operation.
20249 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
20250 Floating-point trigonometric and exponential functions. These instructions
20251 are only generated if @option{-funsafe-math-optimizations} is also specified.
20255 Double-precision floating point:
20258 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
20259 Binary arithmetic operations.
20265 Unary absolute value.
20267 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
20268 Comparison operations.
20270 @item @samp{fmind}, @samp{fmaxd}
20271 Double-precision minimum and maximum. These instructions are only
20272 generated if @option{-ffinite-math-only} is specified.
20274 @item @samp{fsqrtd}
20275 Unary square root operation.
20277 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
20278 Double-precision trigonometric and exponential functions. These instructions
20279 are only generated if @option{-funsafe-math-optimizations} is also specified.
20285 @item @samp{fextsd}
20286 Conversion from single precision to double precision.
20288 @item @samp{ftruncds}
20289 Conversion from double precision to single precision.
20291 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
20292 Conversion from floating point to signed or unsigned integer types, with
20293 truncation towards zero.
20296 Conversion from single-precision floating point to signed integer,
20297 rounding to the nearest integer and ties away from zero.
20298 This corresponds to the @code{__builtin_lroundf} function when
20299 @option{-fno-math-errno} is used.
20301 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
20302 Conversion from signed or unsigned integer types to floating-point types.
20306 In addition, all of the following transfer instructions for internal
20307 registers X and Y must be provided to use any of the double-precision
20308 floating-point instructions. Custom instructions taking two
20309 double-precision source operands expect the first operand in the
20310 64-bit register X. The other operand (or only operand of a unary
20311 operation) is given to the custom arithmetic instruction with the
20312 least significant half in source register @var{src1} and the most
20313 significant half in @var{src2}. A custom instruction that returns a
20314 double-precision result returns the most significant 32 bits in the
20315 destination register and the other half in 32-bit register Y.
20316 GCC automatically generates the necessary code sequences to write
20317 register X and/or read register Y when double-precision floating-point
20318 instructions are used.
20323 Write @var{src1} into the least significant half of X and @var{src2} into
20324 the most significant half of X.
20327 Write @var{src1} into Y.
20329 @item @samp{frdxhi}, @samp{frdxlo}
20330 Read the most or least (respectively) significant half of X and store it in
20334 Read the value of Y and store it into @var{dest}.
20337 Note that you can gain more local control over generation of Nios II custom
20338 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
20339 and @code{target("no-custom-@var{insn}")} function attributes
20340 (@pxref{Function Attributes})
20341 or pragmas (@pxref{Function Specific Option Pragmas}).
20343 @item -mcustom-fpu-cfg=@var{name}
20344 @opindex mcustom-fpu-cfg
20346 This option enables a predefined, named set of custom instruction encodings
20347 (see @option{-mcustom-@var{insn}} above).
20348 Currently, the following sets are defined:
20350 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
20351 @gccoptlist{-mcustom-fmuls=252 @gol
20352 -mcustom-fadds=253 @gol
20353 -mcustom-fsubs=254 @gol
20354 -fsingle-precision-constant}
20356 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
20357 @gccoptlist{-mcustom-fmuls=252 @gol
20358 -mcustom-fadds=253 @gol
20359 -mcustom-fsubs=254 @gol
20360 -mcustom-fdivs=255 @gol
20361 -fsingle-precision-constant}
20363 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
20364 @gccoptlist{-mcustom-floatus=243 @gol
20365 -mcustom-fixsi=244 @gol
20366 -mcustom-floatis=245 @gol
20367 -mcustom-fcmpgts=246 @gol
20368 -mcustom-fcmples=249 @gol
20369 -mcustom-fcmpeqs=250 @gol
20370 -mcustom-fcmpnes=251 @gol
20371 -mcustom-fmuls=252 @gol
20372 -mcustom-fadds=253 @gol
20373 -mcustom-fsubs=254 @gol
20374 -mcustom-fdivs=255 @gol
20375 -fsingle-precision-constant}
20377 Custom instruction assignments given by individual
20378 @option{-mcustom-@var{insn}=} options override those given by
20379 @option{-mcustom-fpu-cfg=}, regardless of the
20380 order of the options on the command line.
20382 Note that you can gain more local control over selection of a FPU
20383 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
20384 function attribute (@pxref{Function Attributes})
20385 or pragma (@pxref{Function Specific Option Pragmas}).
20389 These additional @samp{-m} options are available for the Altera Nios II
20390 ELF (bare-metal) target:
20396 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
20397 startup and termination code, and is typically used in conjunction with
20398 @option{-msys-crt0=} to specify the location of the alternate startup code
20399 provided by the HAL BSP.
20403 Link with a limited version of the C library, @option{-lsmallc}, rather than
20406 @item -msys-crt0=@var{startfile}
20408 @var{startfile} is the file name of the startfile (crt0) to use
20409 when linking. This option is only useful in conjunction with @option{-mhal}.
20411 @item -msys-lib=@var{systemlib}
20413 @var{systemlib} is the library name of the library that provides
20414 low-level system calls required by the C library,
20415 e.g. @code{read} and @code{write}.
20416 This option is typically used to link with a library provided by a HAL BSP.
20420 @node Nvidia PTX Options
20421 @subsection Nvidia PTX Options
20422 @cindex Nvidia PTX options
20423 @cindex nvptx options
20425 These options are defined for Nvidia PTX:
20433 Generate code for 32-bit or 64-bit ABI.
20436 @opindex mmainkernel
20437 Link in code for a __main kernel. This is for stand-alone instead of
20438 offloading execution.
20442 Apply partitioned execution optimizations. This is the default when any
20443 level of optimization is selected.
20447 @node PDP-11 Options
20448 @subsection PDP-11 Options
20449 @cindex PDP-11 Options
20451 These options are defined for the PDP-11:
20456 Use hardware FPP floating point. This is the default. (FIS floating
20457 point on the PDP-11/40 is not supported.)
20460 @opindex msoft-float
20461 Do not use hardware floating point.
20465 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
20469 Return floating-point results in memory. This is the default.
20473 Generate code for a PDP-11/40.
20477 Generate code for a PDP-11/45. This is the default.
20481 Generate code for a PDP-11/10.
20483 @item -mbcopy-builtin
20484 @opindex mbcopy-builtin
20485 Use inline @code{movmemhi} patterns for copying memory. This is the
20490 Do not use inline @code{movmemhi} patterns for copying memory.
20496 Use 16-bit @code{int}. This is the default.
20502 Use 32-bit @code{int}.
20505 @itemx -mno-float32
20507 @opindex mno-float32
20508 Use 64-bit @code{float}. This is the default.
20511 @itemx -mno-float64
20513 @opindex mno-float64
20514 Use 32-bit @code{float}.
20518 Use @code{abshi2} pattern. This is the default.
20522 Do not use @code{abshi2} pattern.
20524 @item -mbranch-expensive
20525 @opindex mbranch-expensive
20526 Pretend that branches are expensive. This is for experimenting with
20527 code generation only.
20529 @item -mbranch-cheap
20530 @opindex mbranch-cheap
20531 Do not pretend that branches are expensive. This is the default.
20535 Use Unix assembler syntax. This is the default when configured for
20536 @samp{pdp11-*-bsd}.
20540 Use DEC assembler syntax. This is the default when configured for any
20541 PDP-11 target other than @samp{pdp11-*-bsd}.
20544 @node picoChip Options
20545 @subsection picoChip Options
20546 @cindex picoChip options
20548 These @samp{-m} options are defined for picoChip implementations:
20552 @item -mae=@var{ae_type}
20554 Set the instruction set, register set, and instruction scheduling
20555 parameters for array element type @var{ae_type}. Supported values
20556 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
20558 @option{-mae=ANY} selects a completely generic AE type. Code
20559 generated with this option runs on any of the other AE types. The
20560 code is not as efficient as it would be if compiled for a specific
20561 AE type, and some types of operation (e.g., multiplication) do not
20562 work properly on all types of AE.
20564 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
20565 for compiled code, and is the default.
20567 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
20568 option may suffer from poor performance of byte (char) manipulation,
20569 since the DSP AE does not provide hardware support for byte load/stores.
20571 @item -msymbol-as-address
20572 Enable the compiler to directly use a symbol name as an address in a
20573 load/store instruction, without first loading it into a
20574 register. Typically, the use of this option generates larger
20575 programs, which run faster than when the option isn't used. However, the
20576 results vary from program to program, so it is left as a user option,
20577 rather than being permanently enabled.
20579 @item -mno-inefficient-warnings
20580 Disables warnings about the generation of inefficient code. These
20581 warnings can be generated, for example, when compiling code that
20582 performs byte-level memory operations on the MAC AE type. The MAC AE has
20583 no hardware support for byte-level memory operations, so all byte
20584 load/stores must be synthesized from word load/store operations. This is
20585 inefficient and a warning is generated to indicate
20586 that you should rewrite the code to avoid byte operations, or to target
20587 an AE type that has the necessary hardware support. This option disables
20592 @node PowerPC Options
20593 @subsection PowerPC Options
20594 @cindex PowerPC options
20596 These are listed under @xref{RS/6000 and PowerPC Options}.
20599 @subsection RL78 Options
20600 @cindex RL78 Options
20606 Links in additional target libraries to support operation within a
20615 Specifies the type of hardware multiplication and division support to
20616 be used. The simplest is @code{none}, which uses software for both
20617 multiplication and division. This is the default. The @code{g13}
20618 value is for the hardware multiply/divide peripheral found on the
20619 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
20620 the multiplication and division instructions supported by the RL78/G14
20621 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
20622 the value @code{mg10} is an alias for @code{none}.
20624 In addition a C preprocessor macro is defined, based upon the setting
20625 of this option. Possible values are: @code{__RL78_MUL_NONE__},
20626 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
20633 Specifies the RL78 core to target. The default is the G14 core, also
20634 known as an S3 core or just RL78. The G13 or S2 core does not have
20635 multiply or divide instructions, instead it uses a hardware peripheral
20636 for these operations. The G10 or S1 core does not have register
20637 banks, so it uses a different calling convention.
20639 If this option is set it also selects the type of hardware multiply
20640 support to use, unless this is overridden by an explicit
20641 @option{-mmul=none} option on the command line. Thus specifying
20642 @option{-mcpu=g13} enables the use of the G13 hardware multiply
20643 peripheral and specifying @option{-mcpu=g10} disables the use of
20644 hardware multiplications altogether.
20646 Note, although the RL78/G14 core is the default target, specifying
20647 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
20648 change the behavior of the toolchain since it also enables G14
20649 hardware multiply support. If these options are not specified on the
20650 command line then software multiplication routines will be used even
20651 though the code targets the RL78 core. This is for backwards
20652 compatibility with older toolchains which did not have hardware
20653 multiply and divide support.
20655 In addition a C preprocessor macro is defined, based upon the setting
20656 of this option. Possible values are: @code{__RL78_G10__},
20657 @code{__RL78_G13__} or @code{__RL78_G14__}.
20667 These are aliases for the corresponding @option{-mcpu=} option. They
20668 are provided for backwards compatibility.
20672 Allow the compiler to use all of the available registers. By default
20673 registers @code{r24..r31} are reserved for use in interrupt handlers.
20674 With this option enabled these registers can be used in ordinary
20677 @item -m64bit-doubles
20678 @itemx -m32bit-doubles
20679 @opindex m64bit-doubles
20680 @opindex m32bit-doubles
20681 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20682 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20683 @option{-m32bit-doubles}.
20685 @item -msave-mduc-in-interrupts
20686 @item -mno-save-mduc-in-interrupts
20687 @opindex msave-mduc-in-interrupts
20688 @opindex mno-save-mduc-in-interrupts
20689 Specifies that interrupt handler functions should preserve the
20690 MDUC registers. This is only necessary if normal code might use
20691 the MDUC registers, for example because it performs multiplication
20692 and division operations. The default is to ignore the MDUC registers
20693 as this makes the interrupt handlers faster. The target option -mg13
20694 needs to be passed for this to work as this feature is only available
20695 on the G13 target (S2 core). The MDUC registers will only be saved
20696 if the interrupt handler performs a multiplication or division
20697 operation or it calls another function.
20701 @node RS/6000 and PowerPC Options
20702 @subsection IBM RS/6000 and PowerPC Options
20703 @cindex RS/6000 and PowerPC Options
20704 @cindex IBM RS/6000 and PowerPC Options
20706 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
20708 @item -mpowerpc-gpopt
20709 @itemx -mno-powerpc-gpopt
20710 @itemx -mpowerpc-gfxopt
20711 @itemx -mno-powerpc-gfxopt
20714 @itemx -mno-powerpc64
20718 @itemx -mno-popcntb
20720 @itemx -mno-popcntd
20729 @itemx -mno-hard-dfp
20730 @opindex mpowerpc-gpopt
20731 @opindex mno-powerpc-gpopt
20732 @opindex mpowerpc-gfxopt
20733 @opindex mno-powerpc-gfxopt
20734 @opindex mpowerpc64
20735 @opindex mno-powerpc64
20739 @opindex mno-popcntb
20741 @opindex mno-popcntd
20747 @opindex mno-mfpgpr
20749 @opindex mno-hard-dfp
20750 You use these options to specify which instructions are available on the
20751 processor you are using. The default value of these options is
20752 determined when configuring GCC@. Specifying the
20753 @option{-mcpu=@var{cpu_type}} overrides the specification of these
20754 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
20755 rather than the options listed above.
20757 Specifying @option{-mpowerpc-gpopt} allows
20758 GCC to use the optional PowerPC architecture instructions in the
20759 General Purpose group, including floating-point square root. Specifying
20760 @option{-mpowerpc-gfxopt} allows GCC to
20761 use the optional PowerPC architecture instructions in the Graphics
20762 group, including floating-point select.
20764 The @option{-mmfcrf} option allows GCC to generate the move from
20765 condition register field instruction implemented on the POWER4
20766 processor and other processors that support the PowerPC V2.01
20768 The @option{-mpopcntb} option allows GCC to generate the popcount and
20769 double-precision FP reciprocal estimate instruction implemented on the
20770 POWER5 processor and other processors that support the PowerPC V2.02
20772 The @option{-mpopcntd} option allows GCC to generate the popcount
20773 instruction implemented on the POWER7 processor and other processors
20774 that support the PowerPC V2.06 architecture.
20775 The @option{-mfprnd} option allows GCC to generate the FP round to
20776 integer instructions implemented on the POWER5+ processor and other
20777 processors that support the PowerPC V2.03 architecture.
20778 The @option{-mcmpb} option allows GCC to generate the compare bytes
20779 instruction implemented on the POWER6 processor and other processors
20780 that support the PowerPC V2.05 architecture.
20781 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
20782 general-purpose register instructions implemented on the POWER6X
20783 processor and other processors that support the extended PowerPC V2.05
20785 The @option{-mhard-dfp} option allows GCC to generate the decimal
20786 floating-point instructions implemented on some POWER processors.
20788 The @option{-mpowerpc64} option allows GCC to generate the additional
20789 64-bit instructions that are found in the full PowerPC64 architecture
20790 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
20791 @option{-mno-powerpc64}.
20793 @item -mcpu=@var{cpu_type}
20795 Set architecture type, register usage, and
20796 instruction scheduling parameters for machine type @var{cpu_type}.
20797 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
20798 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
20799 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
20800 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
20801 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
20802 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
20803 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
20804 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
20805 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
20806 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
20807 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
20810 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
20811 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
20812 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
20813 architecture machine types, with an appropriate, generic processor
20814 model assumed for scheduling purposes.
20816 The other options specify a specific processor. Code generated under
20817 those options runs best on that processor, and may not run at all on
20820 The @option{-mcpu} options automatically enable or disable the
20823 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
20824 -mpopcntb -mpopcntd -mpowerpc64 @gol
20825 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
20826 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
20827 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
20828 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
20830 The particular options set for any particular CPU varies between
20831 compiler versions, depending on what setting seems to produce optimal
20832 code for that CPU; it doesn't necessarily reflect the actual hardware's
20833 capabilities. If you wish to set an individual option to a particular
20834 value, you may specify it after the @option{-mcpu} option, like
20835 @option{-mcpu=970 -mno-altivec}.
20837 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
20838 not enabled or disabled by the @option{-mcpu} option at present because
20839 AIX does not have full support for these options. You may still
20840 enable or disable them individually if you're sure it'll work in your
20843 @item -mtune=@var{cpu_type}
20845 Set the instruction scheduling parameters for machine type
20846 @var{cpu_type}, but do not set the architecture type or register usage,
20847 as @option{-mcpu=@var{cpu_type}} does. The same
20848 values for @var{cpu_type} are used for @option{-mtune} as for
20849 @option{-mcpu}. If both are specified, the code generated uses the
20850 architecture and registers set by @option{-mcpu}, but the
20851 scheduling parameters set by @option{-mtune}.
20853 @item -mcmodel=small
20854 @opindex mcmodel=small
20855 Generate PowerPC64 code for the small model: The TOC is limited to
20858 @item -mcmodel=medium
20859 @opindex mcmodel=medium
20860 Generate PowerPC64 code for the medium model: The TOC and other static
20861 data may be up to a total of 4G in size.
20863 @item -mcmodel=large
20864 @opindex mcmodel=large
20865 Generate PowerPC64 code for the large model: The TOC may be up to 4G
20866 in size. Other data and code is only limited by the 64-bit address
20870 @itemx -mno-altivec
20872 @opindex mno-altivec
20873 Generate code that uses (does not use) AltiVec instructions, and also
20874 enable the use of built-in functions that allow more direct access to
20875 the AltiVec instruction set. You may also need to set
20876 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
20879 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
20880 @option{-maltivec=be}, the element order for AltiVec intrinsics such
20881 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
20882 match array element order corresponding to the endianness of the
20883 target. That is, element zero identifies the leftmost element in a
20884 vector register when targeting a big-endian platform, and identifies
20885 the rightmost element in a vector register when targeting a
20886 little-endian platform.
20889 @opindex maltivec=be
20890 Generate AltiVec instructions using big-endian element order,
20891 regardless of whether the target is big- or little-endian. This is
20892 the default 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.
20900 @opindex maltivec=le
20901 Generate AltiVec instructions using little-endian element order,
20902 regardless of whether the target is big- or little-endian. This is
20903 the default when targeting a little-endian platform. This option is
20904 currently ignored when targeting a big-endian platform.
20906 The element order is used to interpret element numbers in AltiVec
20907 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20908 @code{vec_insert}. By default, these match array element order
20909 corresponding to the endianness for the target.
20914 @opindex mno-vrsave
20915 Generate VRSAVE instructions when generating AltiVec code.
20917 @item -mgen-cell-microcode
20918 @opindex mgen-cell-microcode
20919 Generate Cell microcode instructions.
20921 @item -mwarn-cell-microcode
20922 @opindex mwarn-cell-microcode
20923 Warn when a Cell microcode instruction is emitted. An example
20924 of a Cell microcode instruction is a variable shift.
20927 @opindex msecure-plt
20928 Generate code that allows @command{ld} and @command{ld.so}
20929 to build executables and shared
20930 libraries with non-executable @code{.plt} and @code{.got} sections.
20932 32-bit SYSV ABI option.
20936 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
20938 requires @code{.plt} and @code{.got}
20939 sections that are both writable and executable.
20940 This is a PowerPC 32-bit SYSV ABI option.
20946 This switch enables or disables the generation of ISEL instructions.
20948 @item -misel=@var{yes/no}
20949 This switch has been deprecated. Use @option{-misel} and
20950 @option{-mno-isel} instead.
20954 Enable Local Register Allocation. This is still experimental for PowerPC,
20955 so by default the compiler uses standard reload
20956 (i.e. @option{-mno-lra}).
20962 This switch enables or disables the generation of SPE simd
20968 @opindex mno-paired
20969 This switch enables or disables the generation of PAIRED simd
20972 @item -mspe=@var{yes/no}
20973 This option has been deprecated. Use @option{-mspe} and
20974 @option{-mno-spe} instead.
20980 Generate code that uses (does not use) vector/scalar (VSX)
20981 instructions, and also enable the use of built-in functions that allow
20982 more direct access to the VSX instruction set.
20987 @opindex mno-crypto
20988 Enable the use (disable) of the built-in functions that allow direct
20989 access to the cryptographic instructions that were added in version
20990 2.07 of the PowerPC ISA.
20992 @item -mdirect-move
20993 @itemx -mno-direct-move
20994 @opindex mdirect-move
20995 @opindex mno-direct-move
20996 Generate code that uses (does not use) the instructions to move data
20997 between the general purpose registers and the vector/scalar (VSX)
20998 registers that were added in version 2.07 of the PowerPC ISA.
21004 Enable (disable) the use of the built-in functions that allow direct
21005 access to the Hardware Transactional Memory (HTM) instructions that
21006 were added in version 2.07 of the PowerPC ISA.
21008 @item -mpower8-fusion
21009 @itemx -mno-power8-fusion
21010 @opindex mpower8-fusion
21011 @opindex mno-power8-fusion
21012 Generate code that keeps (does not keeps) some integer operations
21013 adjacent so that the instructions can be fused together on power8 and
21016 @item -mpower8-vector
21017 @itemx -mno-power8-vector
21018 @opindex mpower8-vector
21019 @opindex mno-power8-vector
21020 Generate code that uses (does not use) the vector and scalar
21021 instructions that were added in version 2.07 of the PowerPC ISA. Also
21022 enable the use of built-in functions that allow more direct access to
21023 the vector instructions.
21025 @item -mquad-memory
21026 @itemx -mno-quad-memory
21027 @opindex mquad-memory
21028 @opindex mno-quad-memory
21029 Generate code that uses (does not use) the non-atomic quad word memory
21030 instructions. The @option{-mquad-memory} option requires use of
21033 @item -mquad-memory-atomic
21034 @itemx -mno-quad-memory-atomic
21035 @opindex mquad-memory-atomic
21036 @opindex mno-quad-memory-atomic
21037 Generate code that uses (does not use) the atomic quad word memory
21038 instructions. The @option{-mquad-memory-atomic} option requires use of
21041 @item -mupper-regs-di
21042 @itemx -mno-upper-regs-di
21043 @opindex mupper-regs-di
21044 @opindex mno-upper-regs-di
21045 Generate code that uses (does not use) the scalar instructions that
21046 target all 64 registers in the vector/scalar floating point register
21047 set that were added in version 2.06 of the PowerPC ISA when processing
21048 integers. @option{-mupper-regs-di} is turned on by default if you use
21049 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21050 @option{-mcpu=power9}, or @option{-mvsx} options.
21052 @item -mupper-regs-df
21053 @itemx -mno-upper-regs-df
21054 @opindex mupper-regs-df
21055 @opindex mno-upper-regs-df
21056 Generate code that uses (does not use) the scalar double precision
21057 instructions that target all 64 registers in the vector/scalar
21058 floating point register set that were added in version 2.06 of the
21059 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
21060 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21061 @option{-mcpu=power9}, or @option{-mvsx} options.
21063 @item -mupper-regs-sf
21064 @itemx -mno-upper-regs-sf
21065 @opindex mupper-regs-sf
21066 @opindex mno-upper-regs-sf
21067 Generate code that uses (does not use) the scalar single precision
21068 instructions that target all 64 registers in the vector/scalar
21069 floating point register set that were added in version 2.07 of the
21070 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
21071 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
21072 @option{-mcpu=power9} options.
21075 @itemx -mno-upper-regs
21076 @opindex mupper-regs
21077 @opindex mno-upper-regs
21078 Generate code that uses (does not use) the scalar
21079 instructions that target all 64 registers in the vector/scalar
21080 floating point register set, depending on the model of the machine.
21082 If the @option{-mno-upper-regs} option is used, it turns off both
21083 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
21086 @itemx -mno-float128
21088 @opindex mno-float128
21089 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
21090 and use either software emulation for IEEE 128-bit floating point or
21091 hardware instructions.
21093 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
21094 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
21095 option. The @option{-mfloat128} option only works on PowerPC 64-bit
21098 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
21099 @option{-mfloat128} option will also enable the generation of ISA 3.0
21100 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
21101 floating point will be done with software emulation.
21103 @item -mfloat128-hardware
21104 @itemx -mno-float128-hardware
21105 @opindex mfloat128-hardware
21106 @opindex mno-float128-hardware
21107 Enable/disable using ISA 3.0 hardware instructions to support the
21108 @var{__float128} data type.
21110 If you use @option{-mfloat128-hardware}, it will enable the option
21111 @option{-mfloat128} as well.
21113 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
21114 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
21115 the IEEE 128-bit floating point support will not be enabled.
21117 @item -mfloat-gprs=@var{yes/single/double/no}
21118 @itemx -mfloat-gprs
21119 @opindex mfloat-gprs
21120 This switch enables or disables the generation of floating-point
21121 operations on the general-purpose registers for architectures that
21124 The argument @samp{yes} or @samp{single} enables the use of
21125 single-precision floating-point operations.
21127 The argument @samp{double} enables the use of single and
21128 double-precision floating-point operations.
21130 The argument @samp{no} disables floating-point operations on the
21131 general-purpose registers.
21133 This option is currently only available on the MPC854x.
21139 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
21140 targets (including GNU/Linux). The 32-bit environment sets int, long
21141 and pointer to 32 bits and generates code that runs on any PowerPC
21142 variant. The 64-bit environment sets int to 32 bits and long and
21143 pointer to 64 bits, and generates code for PowerPC64, as for
21144 @option{-mpowerpc64}.
21147 @itemx -mno-fp-in-toc
21148 @itemx -mno-sum-in-toc
21149 @itemx -mminimal-toc
21151 @opindex mno-fp-in-toc
21152 @opindex mno-sum-in-toc
21153 @opindex mminimal-toc
21154 Modify generation of the TOC (Table Of Contents), which is created for
21155 every executable file. The @option{-mfull-toc} option is selected by
21156 default. In that case, GCC allocates at least one TOC entry for
21157 each unique non-automatic variable reference in your program. GCC
21158 also places floating-point constants in the TOC@. However, only
21159 16,384 entries are available in the TOC@.
21161 If you receive a linker error message that saying you have overflowed
21162 the available TOC space, you can reduce the amount of TOC space used
21163 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
21164 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
21165 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
21166 generate code to calculate the sum of an address and a constant at
21167 run time instead of putting that sum into the TOC@. You may specify one
21168 or both of these options. Each causes GCC to produce very slightly
21169 slower and larger code at the expense of conserving TOC space.
21171 If you still run out of space in the TOC even when you specify both of
21172 these options, specify @option{-mminimal-toc} instead. This option causes
21173 GCC to make only one TOC entry for every file. When you specify this
21174 option, GCC produces code that is slower and larger but which
21175 uses extremely little TOC space. You may wish to use this option
21176 only on files that contain less frequently-executed code.
21182 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
21183 @code{long} type, and the infrastructure needed to support them.
21184 Specifying @option{-maix64} implies @option{-mpowerpc64},
21185 while @option{-maix32} disables the 64-bit ABI and
21186 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
21189 @itemx -mno-xl-compat
21190 @opindex mxl-compat
21191 @opindex mno-xl-compat
21192 Produce code that conforms more closely to IBM XL compiler semantics
21193 when using AIX-compatible ABI@. Pass floating-point arguments to
21194 prototyped functions beyond the register save area (RSA) on the stack
21195 in addition to argument FPRs. Do not assume that most significant
21196 double in 128-bit long double value is properly rounded when comparing
21197 values and converting to double. Use XL symbol names for long double
21200 The AIX calling convention was extended but not initially documented to
21201 handle an obscure K&R C case of calling a function that takes the
21202 address of its arguments with fewer arguments than declared. IBM XL
21203 compilers access floating-point arguments that do not fit in the
21204 RSA from the stack when a subroutine is compiled without
21205 optimization. Because always storing floating-point arguments on the
21206 stack is inefficient and rarely needed, this option is not enabled by
21207 default and only is necessary when calling subroutines compiled by IBM
21208 XL compilers without optimization.
21212 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
21213 application written to use message passing with special startup code to
21214 enable the application to run. The system must have PE installed in the
21215 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
21216 must be overridden with the @option{-specs=} option to specify the
21217 appropriate directory location. The Parallel Environment does not
21218 support threads, so the @option{-mpe} option and the @option{-pthread}
21219 option are incompatible.
21221 @item -malign-natural
21222 @itemx -malign-power
21223 @opindex malign-natural
21224 @opindex malign-power
21225 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
21226 @option{-malign-natural} overrides the ABI-defined alignment of larger
21227 types, such as floating-point doubles, on their natural size-based boundary.
21228 The option @option{-malign-power} instructs GCC to follow the ABI-specified
21229 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
21231 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
21235 @itemx -mhard-float
21236 @opindex msoft-float
21237 @opindex mhard-float
21238 Generate code that does not use (uses) the floating-point register set.
21239 Software floating-point emulation is provided if you use the
21240 @option{-msoft-float} option, and pass the option to GCC when linking.
21242 @item -msingle-float
21243 @itemx -mdouble-float
21244 @opindex msingle-float
21245 @opindex mdouble-float
21246 Generate code for single- or double-precision floating-point operations.
21247 @option{-mdouble-float} implies @option{-msingle-float}.
21250 @opindex msimple-fpu
21251 Do not generate @code{sqrt} and @code{div} instructions for hardware
21252 floating-point unit.
21254 @item -mfpu=@var{name}
21256 Specify type of floating-point unit. Valid values for @var{name} are
21257 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
21258 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
21259 @samp{sp_full} (equivalent to @option{-msingle-float}),
21260 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
21263 @opindex mxilinx-fpu
21264 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
21267 @itemx -mno-multiple
21269 @opindex mno-multiple
21270 Generate code that uses (does not use) the load multiple word
21271 instructions and the store multiple word instructions. These
21272 instructions are generated by default on POWER systems, and not
21273 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
21274 PowerPC systems, since those instructions do not work when the
21275 processor is in little-endian mode. The exceptions are PPC740 and
21276 PPC750 which permit these instructions in little-endian mode.
21281 @opindex mno-string
21282 Generate code that uses (does not use) the load string instructions
21283 and the store string word instructions to save multiple registers and
21284 do small block moves. These instructions are generated by default on
21285 POWER systems, and not generated on PowerPC systems. Do not use
21286 @option{-mstring} on little-endian PowerPC systems, since those
21287 instructions do not work when the processor is in little-endian mode.
21288 The exceptions are PPC740 and PPC750 which permit these instructions
21289 in little-endian mode.
21294 @opindex mno-update
21295 Generate code that uses (does not use) the load or store instructions
21296 that update the base register to the address of the calculated memory
21297 location. These instructions are generated by default. If you use
21298 @option{-mno-update}, there is a small window between the time that the
21299 stack pointer is updated and the address of the previous frame is
21300 stored, which means code that walks the stack frame across interrupts or
21301 signals may get corrupted data.
21303 @item -mavoid-indexed-addresses
21304 @itemx -mno-avoid-indexed-addresses
21305 @opindex mavoid-indexed-addresses
21306 @opindex mno-avoid-indexed-addresses
21307 Generate code that tries to avoid (not avoid) the use of indexed load
21308 or store instructions. These instructions can incur a performance
21309 penalty on Power6 processors in certain situations, such as when
21310 stepping through large arrays that cross a 16M boundary. This option
21311 is enabled by default when targeting Power6 and disabled otherwise.
21314 @itemx -mno-fused-madd
21315 @opindex mfused-madd
21316 @opindex mno-fused-madd
21317 Generate code that uses (does not use) the floating-point multiply and
21318 accumulate instructions. These instructions are generated by default
21319 if hardware floating point is used. The machine-dependent
21320 @option{-mfused-madd} option is now mapped to the machine-independent
21321 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21322 mapped to @option{-ffp-contract=off}.
21328 Generate code that uses (does not use) the half-word multiply and
21329 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
21330 These instructions are generated by default when targeting those
21337 Generate code that uses (does not use) the string-search @samp{dlmzb}
21338 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
21339 generated by default when targeting those processors.
21341 @item -mno-bit-align
21343 @opindex mno-bit-align
21344 @opindex mbit-align
21345 On System V.4 and embedded PowerPC systems do not (do) force structures
21346 and unions that contain bit-fields to be aligned to the base type of the
21349 For example, by default a structure containing nothing but 8
21350 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
21351 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
21352 the structure is aligned to a 1-byte boundary and is 1 byte in
21355 @item -mno-strict-align
21356 @itemx -mstrict-align
21357 @opindex mno-strict-align
21358 @opindex mstrict-align
21359 On System V.4 and embedded PowerPC systems do not (do) assume that
21360 unaligned memory references are handled by the system.
21362 @item -mrelocatable
21363 @itemx -mno-relocatable
21364 @opindex mrelocatable
21365 @opindex mno-relocatable
21366 Generate code that allows (does not allow) a static executable to be
21367 relocated to a different address at run time. A simple embedded
21368 PowerPC system loader should relocate the entire contents of
21369 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
21370 a table of 32-bit addresses generated by this option. For this to
21371 work, all objects linked together must be compiled with
21372 @option{-mrelocatable} or @option{-mrelocatable-lib}.
21373 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
21375 @item -mrelocatable-lib
21376 @itemx -mno-relocatable-lib
21377 @opindex mrelocatable-lib
21378 @opindex mno-relocatable-lib
21379 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
21380 @code{.fixup} section to allow static executables to be relocated at
21381 run time, but @option{-mrelocatable-lib} does not use the smaller stack
21382 alignment of @option{-mrelocatable}. Objects compiled with
21383 @option{-mrelocatable-lib} may be linked with objects compiled with
21384 any combination of the @option{-mrelocatable} options.
21390 On System V.4 and embedded PowerPC systems do not (do) assume that
21391 register 2 contains a pointer to a global area pointing to the addresses
21392 used in the program.
21395 @itemx -mlittle-endian
21397 @opindex mlittle-endian
21398 On System V.4 and embedded PowerPC systems compile code for the
21399 processor in little-endian mode. The @option{-mlittle-endian} option is
21400 the same as @option{-mlittle}.
21403 @itemx -mbig-endian
21405 @opindex mbig-endian
21406 On System V.4 and embedded PowerPC systems compile code for the
21407 processor in big-endian mode. The @option{-mbig-endian} option is
21408 the same as @option{-mbig}.
21410 @item -mdynamic-no-pic
21411 @opindex mdynamic-no-pic
21412 On Darwin and Mac OS X systems, compile code so that it is not
21413 relocatable, but that its external references are relocatable. The
21414 resulting code is suitable for applications, but not shared
21417 @item -msingle-pic-base
21418 @opindex msingle-pic-base
21419 Treat the register used for PIC addressing as read-only, rather than
21420 loading it in the prologue for each function. The runtime system is
21421 responsible for initializing this register with an appropriate value
21422 before execution begins.
21424 @item -mprioritize-restricted-insns=@var{priority}
21425 @opindex mprioritize-restricted-insns
21426 This option controls the priority that is assigned to
21427 dispatch-slot restricted instructions during the second scheduling
21428 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
21429 or @samp{2} to assign no, highest, or second-highest (respectively)
21430 priority to dispatch-slot restricted
21433 @item -msched-costly-dep=@var{dependence_type}
21434 @opindex msched-costly-dep
21435 This option controls which dependences are considered costly
21436 by the target during instruction scheduling. The argument
21437 @var{dependence_type} takes one of the following values:
21441 No dependence is costly.
21444 All dependences are costly.
21446 @item @samp{true_store_to_load}
21447 A true dependence from store to load is costly.
21449 @item @samp{store_to_load}
21450 Any dependence from store to load is costly.
21453 Any dependence for which the latency is greater than or equal to
21454 @var{number} is costly.
21457 @item -minsert-sched-nops=@var{scheme}
21458 @opindex minsert-sched-nops
21459 This option controls which NOP insertion scheme is used during
21460 the second scheduling pass. The argument @var{scheme} takes one of the
21468 Pad with NOPs any dispatch group that has vacant issue slots,
21469 according to the scheduler's grouping.
21471 @item @samp{regroup_exact}
21472 Insert NOPs to force costly dependent insns into
21473 separate groups. Insert exactly as many NOPs as needed to force an insn
21474 to a new group, according to the estimated processor grouping.
21477 Insert NOPs to force costly dependent insns into
21478 separate groups. Insert @var{number} NOPs to force an insn to a new group.
21482 @opindex mcall-sysv
21483 On System V.4 and embedded PowerPC systems compile code using calling
21484 conventions that adhere to the March 1995 draft of the System V
21485 Application Binary Interface, PowerPC processor supplement. This is the
21486 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
21488 @item -mcall-sysv-eabi
21490 @opindex mcall-sysv-eabi
21491 @opindex mcall-eabi
21492 Specify both @option{-mcall-sysv} and @option{-meabi} options.
21494 @item -mcall-sysv-noeabi
21495 @opindex mcall-sysv-noeabi
21496 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
21498 @item -mcall-aixdesc
21500 On System V.4 and embedded PowerPC systems compile code for the AIX
21504 @opindex mcall-linux
21505 On System V.4 and embedded PowerPC systems compile code for the
21506 Linux-based GNU system.
21508 @item -mcall-freebsd
21509 @opindex mcall-freebsd
21510 On System V.4 and embedded PowerPC systems compile code for the
21511 FreeBSD operating system.
21513 @item -mcall-netbsd
21514 @opindex mcall-netbsd
21515 On System V.4 and embedded PowerPC systems compile code for the
21516 NetBSD operating system.
21518 @item -mcall-openbsd
21519 @opindex mcall-netbsd
21520 On System V.4 and embedded PowerPC systems compile code for the
21521 OpenBSD operating system.
21523 @item -maix-struct-return
21524 @opindex maix-struct-return
21525 Return all structures in memory (as specified by the AIX ABI)@.
21527 @item -msvr4-struct-return
21528 @opindex msvr4-struct-return
21529 Return structures smaller than 8 bytes in registers (as specified by the
21532 @item -mabi=@var{abi-type}
21534 Extend the current ABI with a particular extension, or remove such extension.
21535 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
21536 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
21537 @samp{elfv1}, @samp{elfv2}@.
21541 Extend the current ABI with SPE ABI extensions. This does not change
21542 the default ABI, instead it adds the SPE ABI extensions to the current
21546 @opindex mabi=no-spe
21547 Disable Book-E SPE ABI extensions for the current ABI@.
21549 @item -mabi=ibmlongdouble
21550 @opindex mabi=ibmlongdouble
21551 Change the current ABI to use IBM extended-precision long double.
21552 This is a PowerPC 32-bit SYSV ABI option.
21554 @item -mabi=ieeelongdouble
21555 @opindex mabi=ieeelongdouble
21556 Change the current ABI to use IEEE extended-precision long double.
21557 This is a PowerPC 32-bit Linux ABI option.
21560 @opindex mabi=elfv1
21561 Change the current ABI to use the ELFv1 ABI.
21562 This is the default ABI for big-endian PowerPC 64-bit Linux.
21563 Overriding the default ABI requires special system support and is
21564 likely to fail in spectacular ways.
21567 @opindex mabi=elfv2
21568 Change the current ABI to use the ELFv2 ABI.
21569 This is the default ABI for little-endian PowerPC 64-bit Linux.
21570 Overriding the default ABI requires special system support and is
21571 likely to fail in spectacular ways.
21573 @item -mgnu-attribute
21574 @itemx -mno-gnu-attribute
21575 @opindex mgnu-attribute
21576 @opindex mno-gnu-attribute
21577 Emit .gnu_attribute assembly directives to set tag/value pairs in a
21578 .gnu.attributes section that specify ABI variations in function
21579 parameters or return values.
21582 @itemx -mno-prototype
21583 @opindex mprototype
21584 @opindex mno-prototype
21585 On System V.4 and embedded PowerPC systems assume that all calls to
21586 variable argument functions are properly prototyped. Otherwise, the
21587 compiler must insert an instruction before every non-prototyped call to
21588 set or clear bit 6 of the condition code register (@code{CR}) to
21589 indicate whether floating-point values are passed in the floating-point
21590 registers in case the function takes variable arguments. With
21591 @option{-mprototype}, only calls to prototyped variable argument functions
21592 set or clear the bit.
21596 On embedded PowerPC systems, assume that the startup module is called
21597 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
21598 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
21603 On embedded PowerPC systems, assume that the startup module is called
21604 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
21609 On embedded PowerPC systems, assume that the startup module is called
21610 @file{crt0.o} and the standard C libraries are @file{libads.a} and
21613 @item -myellowknife
21614 @opindex myellowknife
21615 On embedded PowerPC systems, assume that the startup module is called
21616 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
21621 On System V.4 and embedded PowerPC systems, specify that you are
21622 compiling for a VxWorks system.
21626 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
21627 header to indicate that @samp{eabi} extended relocations are used.
21633 On System V.4 and embedded PowerPC systems do (do not) adhere to the
21634 Embedded Applications Binary Interface (EABI), which is a set of
21635 modifications to the System V.4 specifications. Selecting @option{-meabi}
21636 means that the stack is aligned to an 8-byte boundary, a function
21637 @code{__eabi} is called from @code{main} to set up the EABI
21638 environment, and the @option{-msdata} option can use both @code{r2} and
21639 @code{r13} to point to two separate small data areas. Selecting
21640 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
21641 no EABI initialization function is called from @code{main}, and the
21642 @option{-msdata} option only uses @code{r13} to point to a single
21643 small data area. The @option{-meabi} option is on by default if you
21644 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
21647 @opindex msdata=eabi
21648 On System V.4 and embedded PowerPC systems, put small initialized
21649 @code{const} global and static data in the @code{.sdata2} section, which
21650 is pointed to by register @code{r2}. Put small initialized
21651 non-@code{const} global and static data in the @code{.sdata} section,
21652 which is pointed to by register @code{r13}. Put small uninitialized
21653 global and static data in the @code{.sbss} section, which is adjacent to
21654 the @code{.sdata} section. The @option{-msdata=eabi} option is
21655 incompatible with the @option{-mrelocatable} option. The
21656 @option{-msdata=eabi} option also sets the @option{-memb} option.
21659 @opindex msdata=sysv
21660 On System V.4 and embedded PowerPC systems, put small global and static
21661 data in the @code{.sdata} section, which is pointed to by register
21662 @code{r13}. Put small uninitialized global and static data in the
21663 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
21664 The @option{-msdata=sysv} option is incompatible with the
21665 @option{-mrelocatable} option.
21667 @item -msdata=default
21669 @opindex msdata=default
21671 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
21672 compile code the same as @option{-msdata=eabi}, otherwise compile code the
21673 same as @option{-msdata=sysv}.
21676 @opindex msdata=data
21677 On System V.4 and embedded PowerPC systems, put small global
21678 data in the @code{.sdata} section. Put small uninitialized global
21679 data in the @code{.sbss} section. Do not use register @code{r13}
21680 to address small data however. This is the default behavior unless
21681 other @option{-msdata} options are used.
21685 @opindex msdata=none
21687 On embedded PowerPC systems, put all initialized global and static data
21688 in the @code{.data} section, and all uninitialized data in the
21689 @code{.bss} section.
21691 @item -mblock-move-inline-limit=@var{num}
21692 @opindex mblock-move-inline-limit
21693 Inline all block moves (such as calls to @code{memcpy} or structure
21694 copies) less than or equal to @var{num} bytes. The minimum value for
21695 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
21696 targets. The default value is target-specific.
21700 @cindex smaller data references (PowerPC)
21701 @cindex .sdata/.sdata2 references (PowerPC)
21702 On embedded PowerPC systems, put global and static items less than or
21703 equal to @var{num} bytes into the small data or BSS sections instead of
21704 the normal data or BSS section. By default, @var{num} is 8. The
21705 @option{-G @var{num}} switch is also passed to the linker.
21706 All modules should be compiled with the same @option{-G @var{num}} value.
21709 @itemx -mno-regnames
21711 @opindex mno-regnames
21712 On System V.4 and embedded PowerPC systems do (do not) emit register
21713 names in the assembly language output using symbolic forms.
21716 @itemx -mno-longcall
21718 @opindex mno-longcall
21719 By default assume that all calls are far away so that a longer and more
21720 expensive calling sequence is required. This is required for calls
21721 farther than 32 megabytes (33,554,432 bytes) from the current location.
21722 A short call is generated if the compiler knows
21723 the call cannot be that far away. This setting can be overridden by
21724 the @code{shortcall} function attribute, or by @code{#pragma
21727 Some linkers are capable of detecting out-of-range calls and generating
21728 glue code on the fly. On these systems, long calls are unnecessary and
21729 generate slower code. As of this writing, the AIX linker can do this,
21730 as can the GNU linker for PowerPC/64. It is planned to add this feature
21731 to the GNU linker for 32-bit PowerPC systems as well.
21733 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
21734 callee, L42}, plus a @dfn{branch island} (glue code). The two target
21735 addresses represent the callee and the branch island. The
21736 Darwin/PPC linker prefers the first address and generates a @code{bl
21737 callee} if the PPC @code{bl} instruction reaches the callee directly;
21738 otherwise, the linker generates @code{bl L42} to call the branch
21739 island. The branch island is appended to the body of the
21740 calling function; it computes the full 32-bit address of the callee
21743 On Mach-O (Darwin) systems, this option directs the compiler emit to
21744 the glue for every direct call, and the Darwin linker decides whether
21745 to use or discard it.
21747 In the future, GCC may ignore all longcall specifications
21748 when the linker is known to generate glue.
21750 @item -mtls-markers
21751 @itemx -mno-tls-markers
21752 @opindex mtls-markers
21753 @opindex mno-tls-markers
21754 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
21755 specifying the function argument. The relocation allows the linker to
21756 reliably associate function call with argument setup instructions for
21757 TLS optimization, which in turn allows GCC to better schedule the
21762 Adds support for multithreading with the @dfn{pthreads} library.
21763 This option sets flags for both the preprocessor and linker.
21768 This option enables use of the reciprocal estimate and
21769 reciprocal square root estimate instructions with additional
21770 Newton-Raphson steps to increase precision instead of doing a divide or
21771 square root and divide for floating-point arguments. You should use
21772 the @option{-ffast-math} option when using @option{-mrecip} (or at
21773 least @option{-funsafe-math-optimizations},
21774 @option{-ffinite-math-only}, @option{-freciprocal-math} and
21775 @option{-fno-trapping-math}). Note that while the throughput of the
21776 sequence is generally higher than the throughput of the non-reciprocal
21777 instruction, the precision of the sequence can be decreased by up to 2
21778 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
21781 @item -mrecip=@var{opt}
21782 @opindex mrecip=opt
21783 This option controls which reciprocal estimate instructions
21784 may be used. @var{opt} is a comma-separated list of options, which may
21785 be preceded by a @code{!} to invert the option:
21790 Enable all estimate instructions.
21793 Enable the default instructions, equivalent to @option{-mrecip}.
21796 Disable all estimate instructions, equivalent to @option{-mno-recip}.
21799 Enable the reciprocal approximation instructions for both
21800 single and double precision.
21803 Enable the single-precision reciprocal approximation instructions.
21806 Enable the double-precision reciprocal approximation instructions.
21809 Enable the reciprocal square root approximation instructions for both
21810 single and double precision.
21813 Enable the single-precision reciprocal square root approximation instructions.
21816 Enable the double-precision reciprocal square root approximation instructions.
21820 So, for example, @option{-mrecip=all,!rsqrtd} enables
21821 all of the reciprocal estimate instructions, except for the
21822 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
21823 which handle the double-precision reciprocal square root calculations.
21825 @item -mrecip-precision
21826 @itemx -mno-recip-precision
21827 @opindex mrecip-precision
21828 Assume (do not assume) that the reciprocal estimate instructions
21829 provide higher-precision estimates than is mandated by the PowerPC
21830 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
21831 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
21832 The double-precision square root estimate instructions are not generated by
21833 default on low-precision machines, since they do not provide an
21834 estimate that converges after three steps.
21836 @item -mveclibabi=@var{type}
21837 @opindex mveclibabi
21838 Specifies the ABI type to use for vectorizing intrinsics using an
21839 external library. The only type supported at present is @samp{mass},
21840 which specifies to use IBM's Mathematical Acceleration Subsystem
21841 (MASS) libraries for vectorizing intrinsics using external libraries.
21842 GCC currently emits calls to @code{acosd2}, @code{acosf4},
21843 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
21844 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
21845 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
21846 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
21847 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
21848 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
21849 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
21850 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
21851 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
21852 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
21853 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
21854 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
21855 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
21856 for power7. Both @option{-ftree-vectorize} and
21857 @option{-funsafe-math-optimizations} must also be enabled. The MASS
21858 libraries must be specified at link time.
21863 Generate (do not generate) the @code{friz} instruction when the
21864 @option{-funsafe-math-optimizations} option is used to optimize
21865 rounding of floating-point values to 64-bit integer and back to floating
21866 point. The @code{friz} instruction does not return the same value if
21867 the floating-point number is too large to fit in an integer.
21869 @item -mpointers-to-nested-functions
21870 @itemx -mno-pointers-to-nested-functions
21871 @opindex mpointers-to-nested-functions
21872 Generate (do not generate) code to load up the static chain register
21873 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
21874 systems where a function pointer points to a 3-word descriptor giving
21875 the function address, TOC value to be loaded in register @code{r2}, and
21876 static chain value to be loaded in register @code{r11}. The
21877 @option{-mpointers-to-nested-functions} is on by default. You cannot
21878 call through pointers to nested functions or pointers
21879 to functions compiled in other languages that use the static chain if
21880 you use @option{-mno-pointers-to-nested-functions}.
21882 @item -msave-toc-indirect
21883 @itemx -mno-save-toc-indirect
21884 @opindex msave-toc-indirect
21885 Generate (do not generate) code to save the TOC value in the reserved
21886 stack location in the function prologue if the function calls through
21887 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
21888 saved in the prologue, it is saved just before the call through the
21889 pointer. The @option{-mno-save-toc-indirect} option is the default.
21891 @item -mcompat-align-parm
21892 @itemx -mno-compat-align-parm
21893 @opindex mcompat-align-parm
21894 Generate (do not generate) code to pass structure parameters with a
21895 maximum alignment of 64 bits, for compatibility with older versions
21898 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
21899 structure parameter on a 128-bit boundary when that structure contained
21900 a member requiring 128-bit alignment. This is corrected in more
21901 recent versions of GCC. This option may be used to generate code
21902 that is compatible with functions compiled with older versions of
21905 The @option{-mno-compat-align-parm} option is the default.
21909 @subsection RX Options
21912 These command-line options are defined for RX targets:
21915 @item -m64bit-doubles
21916 @itemx -m32bit-doubles
21917 @opindex m64bit-doubles
21918 @opindex m32bit-doubles
21919 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21920 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21921 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
21922 works on 32-bit values, which is why the default is
21923 @option{-m32bit-doubles}.
21929 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
21930 floating-point hardware. The default is enabled for the RX600
21931 series and disabled for the RX200 series.
21933 Floating-point instructions are only generated for 32-bit floating-point
21934 values, however, so the FPU hardware is not used for doubles if the
21935 @option{-m64bit-doubles} option is used.
21937 @emph{Note} If the @option{-fpu} option is enabled then
21938 @option{-funsafe-math-optimizations} is also enabled automatically.
21939 This is because the RX FPU instructions are themselves unsafe.
21941 @item -mcpu=@var{name}
21943 Selects the type of RX CPU to be targeted. Currently three types are
21944 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
21945 the specific @samp{RX610} CPU. The default is @samp{RX600}.
21947 The only difference between @samp{RX600} and @samp{RX610} is that the
21948 @samp{RX610} does not support the @code{MVTIPL} instruction.
21950 The @samp{RX200} series does not have a hardware floating-point unit
21951 and so @option{-nofpu} is enabled by default when this type is
21954 @item -mbig-endian-data
21955 @itemx -mlittle-endian-data
21956 @opindex mbig-endian-data
21957 @opindex mlittle-endian-data
21958 Store data (but not code) in the big-endian format. The default is
21959 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
21962 @item -msmall-data-limit=@var{N}
21963 @opindex msmall-data-limit
21964 Specifies the maximum size in bytes of global and static variables
21965 which can be placed into the small data area. Using the small data
21966 area can lead to smaller and faster code, but the size of area is
21967 limited and it is up to the programmer to ensure that the area does
21968 not overflow. Also when the small data area is used one of the RX's
21969 registers (usually @code{r13}) is reserved for use pointing to this
21970 area, so it is no longer available for use by the compiler. This
21971 could result in slower and/or larger code if variables are pushed onto
21972 the stack instead of being held in this register.
21974 Note, common variables (variables that have not been initialized) and
21975 constants are not placed into the small data area as they are assigned
21976 to other sections in the output executable.
21978 The default value is zero, which disables this feature. Note, this
21979 feature is not enabled by default with higher optimization levels
21980 (@option{-O2} etc) because of the potentially detrimental effects of
21981 reserving a register. It is up to the programmer to experiment and
21982 discover whether this feature is of benefit to their program. See the
21983 description of the @option{-mpid} option for a description of how the
21984 actual register to hold the small data area pointer is chosen.
21990 Use the simulator runtime. The default is to use the libgloss
21991 board-specific runtime.
21993 @item -mas100-syntax
21994 @itemx -mno-as100-syntax
21995 @opindex mas100-syntax
21996 @opindex mno-as100-syntax
21997 When generating assembler output use a syntax that is compatible with
21998 Renesas's AS100 assembler. This syntax can also be handled by the GAS
21999 assembler, but it has some restrictions so it is not generated by default.
22001 @item -mmax-constant-size=@var{N}
22002 @opindex mmax-constant-size
22003 Specifies the maximum size, in bytes, of a constant that can be used as
22004 an operand in a RX instruction. Although the RX instruction set does
22005 allow constants of up to 4 bytes in length to be used in instructions,
22006 a longer value equates to a longer instruction. Thus in some
22007 circumstances it can be beneficial to restrict the size of constants
22008 that are used in instructions. Constants that are too big are instead
22009 placed into a constant pool and referenced via register indirection.
22011 The value @var{N} can be between 0 and 4. A value of 0 (the default)
22012 or 4 means that constants of any size are allowed.
22016 Enable linker relaxation. Linker relaxation is a process whereby the
22017 linker attempts to reduce the size of a program by finding shorter
22018 versions of various instructions. Disabled by default.
22020 @item -mint-register=@var{N}
22021 @opindex mint-register
22022 Specify the number of registers to reserve for fast interrupt handler
22023 functions. The value @var{N} can be between 0 and 4. A value of 1
22024 means that register @code{r13} is reserved for the exclusive use
22025 of fast interrupt handlers. A value of 2 reserves @code{r13} and
22026 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
22027 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
22028 A value of 0, the default, does not reserve any registers.
22030 @item -msave-acc-in-interrupts
22031 @opindex msave-acc-in-interrupts
22032 Specifies that interrupt handler functions should preserve the
22033 accumulator register. This is only necessary if normal code might use
22034 the accumulator register, for example because it performs 64-bit
22035 multiplications. The default is to ignore the accumulator as this
22036 makes the interrupt handlers faster.
22042 Enables the generation of position independent data. When enabled any
22043 access to constant data is done via an offset from a base address
22044 held in a register. This allows the location of constant data to be
22045 determined at run time without requiring the executable to be
22046 relocated, which is a benefit to embedded applications with tight
22047 memory constraints. Data that can be modified is not affected by this
22050 Note, using this feature reserves a register, usually @code{r13}, for
22051 the constant data base address. This can result in slower and/or
22052 larger code, especially in complicated functions.
22054 The actual register chosen to hold the constant data base address
22055 depends upon whether the @option{-msmall-data-limit} and/or the
22056 @option{-mint-register} command-line options are enabled. Starting
22057 with register @code{r13} and proceeding downwards, registers are
22058 allocated first to satisfy the requirements of @option{-mint-register},
22059 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
22060 is possible for the small data area register to be @code{r8} if both
22061 @option{-mint-register=4} and @option{-mpid} are specified on the
22064 By default this feature is not enabled. The default can be restored
22065 via the @option{-mno-pid} command-line option.
22067 @item -mno-warn-multiple-fast-interrupts
22068 @itemx -mwarn-multiple-fast-interrupts
22069 @opindex mno-warn-multiple-fast-interrupts
22070 @opindex mwarn-multiple-fast-interrupts
22071 Prevents GCC from issuing a warning message if it finds more than one
22072 fast interrupt handler when it is compiling a file. The default is to
22073 issue a warning for each extra fast interrupt handler found, as the RX
22074 only supports one such interrupt.
22076 @item -mallow-string-insns
22077 @itemx -mno-allow-string-insns
22078 @opindex mallow-string-insns
22079 @opindex mno-allow-string-insns
22080 Enables or disables the use of the string manipulation instructions
22081 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
22082 @code{SWHILE} and also the @code{RMPA} instruction. These
22083 instructions may prefetch data, which is not safe to do if accessing
22084 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
22085 for more information).
22087 The default is to allow these instructions, but it is not possible for
22088 GCC to reliably detect all circumstances where a string instruction
22089 might be used to access an I/O register, so their use cannot be
22090 disabled automatically. Instead it is reliant upon the programmer to
22091 use the @option{-mno-allow-string-insns} option if their program
22092 accesses I/O space.
22094 When the instructions are enabled GCC defines the C preprocessor
22095 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
22096 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
22102 Use only (or not only) @code{JSR} instructions to access functions.
22103 This option can be used when code size exceeds the range of @code{BSR}
22104 instructions. Note that @option{-mno-jsr} does not mean to not use
22105 @code{JSR} but instead means that any type of branch may be used.
22108 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
22109 has special significance to the RX port when used with the
22110 @code{interrupt} function attribute. This attribute indicates a
22111 function intended to process fast interrupts. GCC ensures
22112 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
22113 and/or @code{r13} and only provided that the normal use of the
22114 corresponding registers have been restricted via the
22115 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
22118 @node S/390 and zSeries Options
22119 @subsection S/390 and zSeries Options
22120 @cindex S/390 and zSeries Options
22122 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
22126 @itemx -msoft-float
22127 @opindex mhard-float
22128 @opindex msoft-float
22129 Use (do not use) the hardware floating-point instructions and registers
22130 for floating-point operations. When @option{-msoft-float} is specified,
22131 functions in @file{libgcc.a} are used to perform floating-point
22132 operations. When @option{-mhard-float} is specified, the compiler
22133 generates IEEE floating-point instructions. This is the default.
22136 @itemx -mno-hard-dfp
22138 @opindex mno-hard-dfp
22139 Use (do not use) the hardware decimal-floating-point instructions for
22140 decimal-floating-point operations. When @option{-mno-hard-dfp} is
22141 specified, functions in @file{libgcc.a} are used to perform
22142 decimal-floating-point operations. When @option{-mhard-dfp} is
22143 specified, the compiler generates decimal-floating-point hardware
22144 instructions. This is the default for @option{-march=z9-ec} or higher.
22146 @item -mlong-double-64
22147 @itemx -mlong-double-128
22148 @opindex mlong-double-64
22149 @opindex mlong-double-128
22150 These switches control the size of @code{long double} type. A size
22151 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22152 type. This is the default.
22155 @itemx -mno-backchain
22156 @opindex mbackchain
22157 @opindex mno-backchain
22158 Store (do not store) the address of the caller's frame as backchain pointer
22159 into the callee's stack frame.
22160 A backchain may be needed to allow debugging using tools that do not understand
22161 DWARF call frame information.
22162 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
22163 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
22164 the backchain is placed into the topmost word of the 96/160 byte register
22167 In general, code compiled with @option{-mbackchain} is call-compatible with
22168 code compiled with @option{-mmo-backchain}; however, use of the backchain
22169 for debugging purposes usually requires that the whole binary is built with
22170 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
22171 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22172 to build a linux kernel use @option{-msoft-float}.
22174 The default is to not maintain the backchain.
22176 @item -mpacked-stack
22177 @itemx -mno-packed-stack
22178 @opindex mpacked-stack
22179 @opindex mno-packed-stack
22180 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
22181 specified, the compiler uses the all fields of the 96/160 byte register save
22182 area only for their default purpose; unused fields still take up stack space.
22183 When @option{-mpacked-stack} is specified, register save slots are densely
22184 packed at the top of the register save area; unused space is reused for other
22185 purposes, allowing for more efficient use of the available stack space.
22186 However, when @option{-mbackchain} is also in effect, the topmost word of
22187 the save area is always used to store the backchain, and the return address
22188 register is always saved two words below the backchain.
22190 As long as the stack frame backchain is not used, code generated with
22191 @option{-mpacked-stack} is call-compatible with code generated with
22192 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
22193 S/390 or zSeries generated code that uses the stack frame backchain at run
22194 time, not just for debugging purposes. Such code is not call-compatible
22195 with code compiled with @option{-mpacked-stack}. Also, note that the
22196 combination of @option{-mbackchain},
22197 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22198 to build a linux kernel use @option{-msoft-float}.
22200 The default is to not use the packed stack layout.
22203 @itemx -mno-small-exec
22204 @opindex msmall-exec
22205 @opindex mno-small-exec
22206 Generate (or do not generate) code using the @code{bras} instruction
22207 to do subroutine calls.
22208 This only works reliably if the total executable size does not
22209 exceed 64k. The default is to use the @code{basr} instruction instead,
22210 which does not have this limitation.
22216 When @option{-m31} is specified, generate code compliant to the
22217 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
22218 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
22219 particular to generate 64-bit instructions. For the @samp{s390}
22220 targets, the default is @option{-m31}, while the @samp{s390x}
22221 targets default to @option{-m64}.
22227 When @option{-mzarch} is specified, generate code using the
22228 instructions available on z/Architecture.
22229 When @option{-mesa} is specified, generate code using the
22230 instructions available on ESA/390. Note that @option{-mesa} is
22231 not possible with @option{-m64}.
22232 When generating code compliant to the GNU/Linux for S/390 ABI,
22233 the default is @option{-mesa}. When generating code compliant
22234 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
22240 The @option{-mhtm} option enables a set of builtins making use of
22241 instructions available with the transactional execution facility
22242 introduced with the IBM zEnterprise EC12 machine generation
22243 @ref{S/390 System z Built-in Functions}.
22244 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
22250 When @option{-mvx} is specified, generate code using the instructions
22251 available with the vector extension facility introduced with the IBM
22252 z13 machine generation.
22253 This option changes the ABI for some vector type values with regard to
22254 alignment and calling conventions. In case vector type values are
22255 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
22256 command will be added to mark the resulting binary with the ABI used.
22257 @option{-mvx} is enabled by default when using @option{-march=z13}.
22260 @itemx -mno-zvector
22262 @opindex mno-zvector
22263 The @option{-mzvector} option enables vector language extensions and
22264 builtins using instructions available with the vector extension
22265 facility introduced with the IBM z13 machine generation.
22266 This option adds support for @samp{vector} to be used as a keyword to
22267 define vector type variables and arguments. @samp{vector} is only
22268 available when GNU extensions are enabled. It will not be expanded
22269 when requesting strict standard compliance e.g. with @option{-std=c99}.
22270 In addition to the GCC low-level builtins @option{-mzvector} enables
22271 a set of builtins added for compatibility with AltiVec-style
22272 implementations like Power and Cell. In order to make use of these
22273 builtins the header file @file{vecintrin.h} needs to be included.
22274 @option{-mzvector} is disabled by default.
22280 Generate (or do not generate) code using the @code{mvcle} instruction
22281 to perform block moves. When @option{-mno-mvcle} is specified,
22282 use a @code{mvc} loop instead. This is the default unless optimizing for
22289 Print (or do not print) additional debug information when compiling.
22290 The default is to not print debug information.
22292 @item -march=@var{cpu-type}
22294 Generate code that runs on @var{cpu-type}, which is the name of a
22295 system representing a certain processor type. Possible values for
22296 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
22297 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
22298 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
22301 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
22302 @samp{g6} are deprecated and will be removed with future releases.
22304 Specifying @samp{native} as cpu type can be used to select the best
22305 architecture option for the host processor.
22306 @option{-march=native} has no effect if GCC does not recognize the
22309 @item -mtune=@var{cpu-type}
22311 Tune to @var{cpu-type} everything applicable about the generated code,
22312 except for the ABI and the set of available instructions.
22313 The list of @var{cpu-type} values is the same as for @option{-march}.
22314 The default is the value used for @option{-march}.
22317 @itemx -mno-tpf-trace
22318 @opindex mtpf-trace
22319 @opindex mno-tpf-trace
22320 Generate code that adds (does not add) in TPF OS specific branches to trace
22321 routines in the operating system. This option is off by default, even
22322 when compiling for the TPF OS@.
22325 @itemx -mno-fused-madd
22326 @opindex mfused-madd
22327 @opindex mno-fused-madd
22328 Generate code that uses (does not use) the floating-point multiply and
22329 accumulate instructions. These instructions are generated by default if
22330 hardware floating point is used.
22332 @item -mwarn-framesize=@var{framesize}
22333 @opindex mwarn-framesize
22334 Emit a warning if the current function exceeds the given frame size. Because
22335 this is a compile-time check it doesn't need to be a real problem when the program
22336 runs. It is intended to identify functions that most probably cause
22337 a stack overflow. It is useful to be used in an environment with limited stack
22338 size e.g.@: the linux kernel.
22340 @item -mwarn-dynamicstack
22341 @opindex mwarn-dynamicstack
22342 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
22343 arrays. This is generally a bad idea with a limited stack size.
22345 @item -mstack-guard=@var{stack-guard}
22346 @itemx -mstack-size=@var{stack-size}
22347 @opindex mstack-guard
22348 @opindex mstack-size
22349 If these options are provided the S/390 back end emits additional instructions in
22350 the function prologue that trigger a trap if the stack size is @var{stack-guard}
22351 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
22352 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
22353 the frame size of the compiled function is chosen.
22354 These options are intended to be used to help debugging stack overflow problems.
22355 The additionally emitted code causes only little overhead and hence can also be
22356 used in production-like systems without greater performance degradation. The given
22357 values have to be exact powers of 2 and @var{stack-size} has to be greater than
22358 @var{stack-guard} without exceeding 64k.
22359 In order to be efficient the extra code makes the assumption that the stack starts
22360 at an address aligned to the value given by @var{stack-size}.
22361 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
22363 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
22365 If the hotpatch option is enabled, a ``hot-patching'' function
22366 prologue is generated for all functions in the compilation unit.
22367 The funtion label is prepended with the given number of two-byte
22368 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
22369 the label, 2 * @var{post-halfwords} bytes are appended, using the
22370 largest NOP like instructions the architecture allows (maximum
22373 If both arguments are zero, hotpatching is disabled.
22375 This option can be overridden for individual functions with the
22376 @code{hotpatch} attribute.
22379 @node Score Options
22380 @subsection Score Options
22381 @cindex Score Options
22383 These options are defined for Score implementations:
22388 Compile code for big-endian mode. This is the default.
22392 Compile code for little-endian mode.
22396 Disable generation of @code{bcnz} instructions.
22400 Enable generation of unaligned load and store instructions.
22404 Enable the use of multiply-accumulate instructions. Disabled by default.
22408 Specify the SCORE5 as the target architecture.
22412 Specify the SCORE5U of the target architecture.
22416 Specify the SCORE7 as the target architecture. This is the default.
22420 Specify the SCORE7D as the target architecture.
22424 @subsection SH Options
22426 These @samp{-m} options are defined for the SH implementations:
22431 Generate code for the SH1.
22435 Generate code for the SH2.
22438 Generate code for the SH2e.
22442 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
22443 that the floating-point unit is not used.
22445 @item -m2a-single-only
22446 @opindex m2a-single-only
22447 Generate code for the SH2a-FPU, in such a way that no double-precision
22448 floating-point operations are used.
22451 @opindex m2a-single
22452 Generate code for the SH2a-FPU assuming the floating-point unit is in
22453 single-precision mode by default.
22457 Generate code for the SH2a-FPU assuming the floating-point unit is in
22458 double-precision mode by default.
22462 Generate code for the SH3.
22466 Generate code for the SH3e.
22470 Generate code for the SH4 without a floating-point unit.
22472 @item -m4-single-only
22473 @opindex m4-single-only
22474 Generate code for the SH4 with a floating-point unit that only
22475 supports single-precision arithmetic.
22479 Generate code for the SH4 assuming the floating-point unit is in
22480 single-precision mode by default.
22484 Generate code for the SH4.
22488 Generate code for SH4-100.
22490 @item -m4-100-nofpu
22491 @opindex m4-100-nofpu
22492 Generate code for SH4-100 in such a way that the
22493 floating-point unit is not used.
22495 @item -m4-100-single
22496 @opindex m4-100-single
22497 Generate code for SH4-100 assuming the floating-point unit is in
22498 single-precision mode by default.
22500 @item -m4-100-single-only
22501 @opindex m4-100-single-only
22502 Generate code for SH4-100 in such a way that no double-precision
22503 floating-point operations are used.
22507 Generate code for SH4-200.
22509 @item -m4-200-nofpu
22510 @opindex m4-200-nofpu
22511 Generate code for SH4-200 without in such a way that the
22512 floating-point unit is not used.
22514 @item -m4-200-single
22515 @opindex m4-200-single
22516 Generate code for SH4-200 assuming the floating-point unit is in
22517 single-precision mode by default.
22519 @item -m4-200-single-only
22520 @opindex m4-200-single-only
22521 Generate code for SH4-200 in such a way that no double-precision
22522 floating-point operations are used.
22526 Generate code for SH4-300.
22528 @item -m4-300-nofpu
22529 @opindex m4-300-nofpu
22530 Generate code for SH4-300 without in such a way that the
22531 floating-point unit is not used.
22533 @item -m4-300-single
22534 @opindex m4-300-single
22535 Generate code for SH4-300 in such a way that no double-precision
22536 floating-point operations are used.
22538 @item -m4-300-single-only
22539 @opindex m4-300-single-only
22540 Generate code for SH4-300 in such a way that no double-precision
22541 floating-point operations are used.
22545 Generate code for SH4-340 (no MMU, no FPU).
22549 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
22554 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
22555 floating-point unit is not used.
22557 @item -m4a-single-only
22558 @opindex m4a-single-only
22559 Generate code for the SH4a, in such a way that no double-precision
22560 floating-point operations are used.
22563 @opindex m4a-single
22564 Generate code for the SH4a assuming the floating-point unit is in
22565 single-precision mode by default.
22569 Generate code for the SH4a.
22573 Same as @option{-m4a-nofpu}, except that it implicitly passes
22574 @option{-dsp} to the assembler. GCC doesn't generate any DSP
22575 instructions at the moment.
22579 Compile code for the processor in big-endian mode.
22583 Compile code for the processor in little-endian mode.
22587 Align doubles at 64-bit boundaries. Note that this changes the calling
22588 conventions, and thus some functions from the standard C library do
22589 not work unless you recompile it first with @option{-mdalign}.
22593 Shorten some address references at link time, when possible; uses the
22594 linker option @option{-relax}.
22598 Use 32-bit offsets in @code{switch} tables. The default is to use
22603 Enable the use of bit manipulation instructions on SH2A.
22607 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
22608 alignment constraints.
22612 Comply with the calling conventions defined by Renesas.
22615 @opindex mno-renesas
22616 Comply with the calling conventions defined for GCC before the Renesas
22617 conventions were available. This option is the default for all
22618 targets of the SH toolchain.
22621 @opindex mnomacsave
22622 Mark the @code{MAC} register as call-clobbered, even if
22623 @option{-mrenesas} is given.
22629 Control the IEEE compliance of floating-point comparisons, which affects the
22630 handling of cases where the result of a comparison is unordered. By default
22631 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
22632 enabled @option{-mno-ieee} is implicitly set, which results in faster
22633 floating-point greater-equal and less-equal comparisons. The implicit settings
22634 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
22636 @item -minline-ic_invalidate
22637 @opindex minline-ic_invalidate
22638 Inline code to invalidate instruction cache entries after setting up
22639 nested function trampolines.
22640 This option has no effect if @option{-musermode} is in effect and the selected
22641 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
22643 If the selected code generation option does not allow the use of the @code{icbi}
22644 instruction, and @option{-musermode} is not in effect, the inlined code
22645 manipulates the instruction cache address array directly with an associative
22646 write. This not only requires privileged mode at run time, but it also
22647 fails if the cache line had been mapped via the TLB and has become unmapped.
22651 Dump instruction size and location in the assembly code.
22654 @opindex mpadstruct
22655 This option is deprecated. It pads structures to multiple of 4 bytes,
22656 which is incompatible with the SH ABI@.
22658 @item -matomic-model=@var{model}
22659 @opindex matomic-model=@var{model}
22660 Sets the model of atomic operations and additional parameters as a comma
22661 separated list. For details on the atomic built-in functions see
22662 @ref{__atomic Builtins}. The following models and parameters are supported:
22667 Disable compiler generated atomic sequences and emit library calls for atomic
22668 operations. This is the default if the target is not @code{sh*-*-linux*}.
22671 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
22672 built-in functions. The generated atomic sequences require additional support
22673 from the interrupt/exception handling code of the system and are only suitable
22674 for SH3* and SH4* single-core systems. This option is enabled by default when
22675 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
22676 this option also partially utilizes the hardware atomic instructions
22677 @code{movli.l} and @code{movco.l} to create more efficient code, unless
22678 @samp{strict} is specified.
22681 Generate software atomic sequences that use a variable in the thread control
22682 block. This is a variation of the gUSA sequences which can also be used on
22683 SH1* and SH2* targets. The generated atomic sequences require additional
22684 support from the interrupt/exception handling code of the system and are only
22685 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
22686 parameter has to be specified as well.
22689 Generate software atomic sequences that temporarily disable interrupts by
22690 setting @code{SR.IMASK = 1111}. This model works only when the program runs
22691 in privileged mode and is only suitable for single-core systems. Additional
22692 support from the interrupt/exception handling code of the system is not
22693 required. This model is enabled by default when the target is
22694 @code{sh*-*-linux*} and SH1* or SH2*.
22697 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
22698 instructions only. This is only available on SH4A and is suitable for
22699 multi-core systems. Since the hardware instructions support only 32 bit atomic
22700 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
22701 Code compiled with this option is also compatible with other software
22702 atomic model interrupt/exception handling systems if executed on an SH4A
22703 system. Additional support from the interrupt/exception handling code of the
22704 system is not required for this model.
22707 This parameter specifies the offset in bytes of the variable in the thread
22708 control block structure that should be used by the generated atomic sequences
22709 when the @samp{soft-tcb} model has been selected. For other models this
22710 parameter is ignored. The specified value must be an integer multiple of four
22711 and in the range 0-1020.
22714 This parameter prevents mixed usage of multiple atomic models, even if they
22715 are compatible, and makes the compiler generate atomic sequences of the
22716 specified model only.
22722 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
22723 Notice that depending on the particular hardware and software configuration
22724 this can degrade overall performance due to the operand cache line flushes
22725 that are implied by the @code{tas.b} instruction. On multi-core SH4A
22726 processors the @code{tas.b} instruction must be used with caution since it
22727 can result in data corruption for certain cache configurations.
22730 @opindex mprefergot
22731 When generating position-independent code, emit function calls using
22732 the Global Offset Table instead of the Procedure Linkage Table.
22735 @itemx -mno-usermode
22737 @opindex mno-usermode
22738 Don't allow (allow) the compiler generating privileged mode code. Specifying
22739 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
22740 inlined code would not work in user mode. @option{-musermode} is the default
22741 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
22742 @option{-musermode} has no effect, since there is no user mode.
22744 @item -multcost=@var{number}
22745 @opindex multcost=@var{number}
22746 Set the cost to assume for a multiply insn.
22748 @item -mdiv=@var{strategy}
22749 @opindex mdiv=@var{strategy}
22750 Set the division strategy to be used for integer division operations.
22751 @var{strategy} can be one of:
22756 Calls a library function that uses the single-step division instruction
22757 @code{div1} to perform the operation. Division by zero calculates an
22758 unspecified result and does not trap. This is the default except for SH4,
22759 SH2A and SHcompact.
22762 Calls a library function that performs the operation in double precision
22763 floating point. Division by zero causes a floating-point exception. This is
22764 the default for SHcompact with FPU. Specifying this for targets that do not
22765 have a double precision FPU defaults to @code{call-div1}.
22768 Calls a library function that uses a lookup table for small divisors and
22769 the @code{div1} instruction with case distinction for larger divisors. Division
22770 by zero calculates an unspecified result and does not trap. This is the default
22771 for SH4. Specifying this for targets that do not have dynamic shift
22772 instructions defaults to @code{call-div1}.
22776 When a division strategy has not been specified the default strategy is
22777 selected based on the current target. For SH2A the default strategy is to
22778 use the @code{divs} and @code{divu} instructions instead of library function
22781 @item -maccumulate-outgoing-args
22782 @opindex maccumulate-outgoing-args
22783 Reserve space once for outgoing arguments in the function prologue rather
22784 than around each call. Generally beneficial for performance and size. Also
22785 needed for unwinding to avoid changing the stack frame around conditional code.
22787 @item -mdivsi3_libfunc=@var{name}
22788 @opindex mdivsi3_libfunc=@var{name}
22789 Set the name of the library function used for 32-bit signed division to
22791 This only affects the name used in the @samp{call} division strategies, and
22792 the compiler still expects the same sets of input/output/clobbered registers as
22793 if this option were not present.
22795 @item -mfixed-range=@var{register-range}
22796 @opindex mfixed-range
22797 Generate code treating the given register range as fixed registers.
22798 A fixed register is one that the register allocator can not use. This is
22799 useful when compiling kernel code. A register range is specified as
22800 two registers separated by a dash. Multiple register ranges can be
22801 specified separated by a comma.
22803 @item -mbranch-cost=@var{num}
22804 @opindex mbranch-cost=@var{num}
22805 Assume @var{num} to be the cost for a branch instruction. Higher numbers
22806 make the compiler try to generate more branch-free code if possible.
22807 If not specified the value is selected depending on the processor type that
22808 is being compiled for.
22811 @itemx -mno-zdcbranch
22812 @opindex mzdcbranch
22813 @opindex mno-zdcbranch
22814 Assume (do not assume) that zero displacement conditional branch instructions
22815 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
22816 compiler prefers zero displacement branch code sequences. This is
22817 enabled by default when generating code for SH4 and SH4A. It can be explicitly
22818 disabled by specifying @option{-mno-zdcbranch}.
22820 @item -mcbranch-force-delay-slot
22821 @opindex mcbranch-force-delay-slot
22822 Force the usage of delay slots for conditional branches, which stuffs the delay
22823 slot with a @code{nop} if a suitable instruction can't be found. By default
22824 this option is disabled. It can be enabled to work around hardware bugs as
22825 found in the original SH7055.
22828 @itemx -mno-fused-madd
22829 @opindex mfused-madd
22830 @opindex mno-fused-madd
22831 Generate code that uses (does not use) the floating-point multiply and
22832 accumulate instructions. These instructions are generated by default
22833 if hardware floating point is used. The machine-dependent
22834 @option{-mfused-madd} option is now mapped to the machine-independent
22835 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22836 mapped to @option{-ffp-contract=off}.
22842 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
22843 and cosine approximations. The option @option{-mfsca} must be used in
22844 combination with @option{-funsafe-math-optimizations}. It is enabled by default
22845 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
22846 approximations even if @option{-funsafe-math-optimizations} is in effect.
22852 Allow or disallow the compiler to emit the @code{fsrra} instruction for
22853 reciprocal square root approximations. The option @option{-mfsrra} must be used
22854 in combination with @option{-funsafe-math-optimizations} and
22855 @option{-ffinite-math-only}. It is enabled by default when generating code for
22856 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
22857 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
22860 @item -mpretend-cmove
22861 @opindex mpretend-cmove
22862 Prefer zero-displacement conditional branches for conditional move instruction
22863 patterns. This can result in faster code on the SH4 processor.
22867 Generate code using the FDPIC ABI.
22871 @node Solaris 2 Options
22872 @subsection Solaris 2 Options
22873 @cindex Solaris 2 options
22875 These @samp{-m} options are supported on Solaris 2:
22878 @item -mclear-hwcap
22879 @opindex mclear-hwcap
22880 @option{-mclear-hwcap} tells the compiler to remove the hardware
22881 capabilities generated by the Solaris assembler. This is only necessary
22882 when object files use ISA extensions not supported by the current
22883 machine, but check at runtime whether or not to use them.
22885 @item -mimpure-text
22886 @opindex mimpure-text
22887 @option{-mimpure-text}, used in addition to @option{-shared}, tells
22888 the compiler to not pass @option{-z text} to the linker when linking a
22889 shared object. Using this option, you can link position-dependent
22890 code into a shared object.
22892 @option{-mimpure-text} suppresses the ``relocations remain against
22893 allocatable but non-writable sections'' linker error message.
22894 However, the necessary relocations trigger copy-on-write, and the
22895 shared object is not actually shared across processes. Instead of
22896 using @option{-mimpure-text}, you should compile all source code with
22897 @option{-fpic} or @option{-fPIC}.
22901 These switches are supported in addition to the above on Solaris 2:
22906 Add support for multithreading using the POSIX threads library. This
22907 option sets flags for both the preprocessor and linker. This option does
22908 not affect the thread safety of object code produced by the compiler or
22909 that of libraries supplied with it.
22913 This is a synonym for @option{-pthreads}.
22916 @node SPARC Options
22917 @subsection SPARC Options
22918 @cindex SPARC options
22920 These @samp{-m} options are supported on the SPARC:
22923 @item -mno-app-regs
22925 @opindex mno-app-regs
22927 Specify @option{-mapp-regs} to generate output using the global registers
22928 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
22929 global register 1, each global register 2 through 4 is then treated as an
22930 allocable register that is clobbered by function calls. This is the default.
22932 To be fully SVR4 ABI-compliant at the cost of some performance loss,
22933 specify @option{-mno-app-regs}. You should compile libraries and system
22934 software with this option.
22940 With @option{-mflat}, the compiler does not generate save/restore instructions
22941 and uses a ``flat'' or single register window model. This model is compatible
22942 with the regular register window model. The local registers and the input
22943 registers (0--5) are still treated as ``call-saved'' registers and are
22944 saved on the stack as needed.
22946 With @option{-mno-flat} (the default), the compiler generates save/restore
22947 instructions (except for leaf functions). This is the normal operating mode.
22950 @itemx -mhard-float
22952 @opindex mhard-float
22953 Generate output containing floating-point instructions. This is the
22957 @itemx -msoft-float
22959 @opindex msoft-float
22960 Generate output containing library calls for floating point.
22961 @strong{Warning:} the requisite libraries are not available for all SPARC
22962 targets. Normally the facilities of the machine's usual C compiler are
22963 used, but this cannot be done directly in cross-compilation. You must make
22964 your own arrangements to provide suitable library functions for
22965 cross-compilation. The embedded targets @samp{sparc-*-aout} and
22966 @samp{sparclite-*-*} do provide software floating-point support.
22968 @option{-msoft-float} changes the calling convention in the output file;
22969 therefore, it is only useful if you compile @emph{all} of a program with
22970 this option. In particular, you need to compile @file{libgcc.a}, the
22971 library that comes with GCC, with @option{-msoft-float} in order for
22974 @item -mhard-quad-float
22975 @opindex mhard-quad-float
22976 Generate output containing quad-word (long double) floating-point
22979 @item -msoft-quad-float
22980 @opindex msoft-quad-float
22981 Generate output containing library calls for quad-word (long double)
22982 floating-point instructions. The functions called are those specified
22983 in the SPARC ABI@. This is the default.
22985 As of this writing, there are no SPARC implementations that have hardware
22986 support for the quad-word floating-point instructions. They all invoke
22987 a trap handler for one of these instructions, and then the trap handler
22988 emulates the effect of the instruction. Because of the trap handler overhead,
22989 this is much slower than calling the ABI library routines. Thus the
22990 @option{-msoft-quad-float} option is the default.
22992 @item -mno-unaligned-doubles
22993 @itemx -munaligned-doubles
22994 @opindex mno-unaligned-doubles
22995 @opindex munaligned-doubles
22996 Assume that doubles have 8-byte alignment. This is the default.
22998 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
22999 alignment only if they are contained in another type, or if they have an
23000 absolute address. Otherwise, it assumes they have 4-byte alignment.
23001 Specifying this option avoids some rare compatibility problems with code
23002 generated by other compilers. It is not the default because it results
23003 in a performance loss, especially for floating-point code.
23006 @itemx -mno-user-mode
23007 @opindex muser-mode
23008 @opindex mno-user-mode
23009 Do not generate code that can only run in supervisor mode. This is relevant
23010 only for the @code{casa} instruction emitted for the LEON3 processor. This
23013 @item -mfaster-structs
23014 @itemx -mno-faster-structs
23015 @opindex mfaster-structs
23016 @opindex mno-faster-structs
23017 With @option{-mfaster-structs}, the compiler assumes that structures
23018 should have 8-byte alignment. This enables the use of pairs of
23019 @code{ldd} and @code{std} instructions for copies in structure
23020 assignment, in place of twice as many @code{ld} and @code{st} pairs.
23021 However, the use of this changed alignment directly violates the SPARC
23022 ABI@. Thus, it's intended only for use on targets where the developer
23023 acknowledges that their resulting code is not directly in line with
23024 the rules of the ABI@.
23026 @item -mstd-struct-return
23027 @itemx -mno-std-struct-return
23028 @opindex mstd-struct-return
23029 @opindex mno-std-struct-return
23030 With @option{-mstd-struct-return}, the compiler generates checking code
23031 in functions returning structures or unions to detect size mismatches
23032 between the two sides of function calls, as per the 32-bit ABI@.
23034 The default is @option{-mno-std-struct-return}. This option has no effect
23037 @item -mcpu=@var{cpu_type}
23039 Set the instruction set, register set, and instruction scheduling parameters
23040 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23041 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
23042 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
23043 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
23044 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
23045 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
23047 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
23048 which selects the best architecture option for the host processor.
23049 @option{-mcpu=native} has no effect if GCC does not recognize
23052 Default instruction scheduling parameters are used for values that select
23053 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
23054 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
23056 Here is a list of each supported architecture and their supported
23064 supersparc, hypersparc, leon, leon3
23067 f930, f934, sparclite86x
23073 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
23076 By default (unless configured otherwise), GCC generates code for the V7
23077 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
23078 additionally optimizes it for the Cypress CY7C602 chip, as used in the
23079 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
23080 SPARCStation 1, 2, IPX etc.
23082 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
23083 architecture. The only difference from V7 code is that the compiler emits
23084 the integer multiply and integer divide instructions which exist in SPARC-V8
23085 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
23086 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
23089 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
23090 the SPARC architecture. This adds the integer multiply, integer divide step
23091 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
23092 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
23093 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
23094 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
23095 MB86934 chip, which is the more recent SPARClite with FPU@.
23097 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
23098 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
23099 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
23100 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
23101 optimizes it for the TEMIC SPARClet chip.
23103 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
23104 architecture. This adds 64-bit integer and floating-point move instructions,
23105 3 additional floating-point condition code registers and conditional move
23106 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
23107 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
23108 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
23109 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
23110 @option{-mcpu=niagara}, the compiler additionally optimizes it for
23111 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
23112 additionally optimizes it for Sun UltraSPARC T2 chips. With
23113 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
23114 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
23115 additionally optimizes it for Sun UltraSPARC T4 chips. With
23116 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
23117 Oracle SPARC M7 chips.
23119 @item -mtune=@var{cpu_type}
23121 Set the instruction scheduling parameters for machine type
23122 @var{cpu_type}, but do not set the instruction set or register set that the
23123 option @option{-mcpu=@var{cpu_type}} does.
23125 The same values for @option{-mcpu=@var{cpu_type}} can be used for
23126 @option{-mtune=@var{cpu_type}}, but the only useful values are those
23127 that select a particular CPU implementation. Those are
23128 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
23129 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
23130 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
23131 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
23132 @samp{niagara4} and @samp{niagara7}. With native Solaris and
23133 GNU/Linux toolchains, @samp{native} can also be used.
23138 @opindex mno-v8plus
23139 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
23140 difference from the V8 ABI is that the global and out registers are
23141 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
23142 mode for all SPARC-V9 processors.
23148 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
23149 Visual Instruction Set extensions. The default is @option{-mno-vis}.
23155 With @option{-mvis2}, GCC generates code that takes advantage of
23156 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
23157 default is @option{-mvis2} when targeting a cpu that supports such
23158 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
23159 also sets @option{-mvis}.
23165 With @option{-mvis3}, GCC generates code that takes advantage of
23166 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
23167 default is @option{-mvis3} when targeting a cpu that supports such
23168 instructions, such as niagara-3 and later. Setting @option{-mvis3}
23169 also sets @option{-mvis2} and @option{-mvis}.
23175 With @option{-mvis4}, GCC generates code that takes advantage of
23176 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
23177 default is @option{-mvis4} when targeting a cpu that supports such
23178 instructions, such as niagara-7 and later. Setting @option{-mvis4}
23179 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
23184 @opindex mno-cbcond
23185 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
23186 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
23187 when targeting a CPU that supports such instructions, such as Niagara-4 and
23194 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
23195 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
23196 when targeting a CPU that supports such instructions, such as Niagara-3 and
23203 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
23204 Population Count instruction. The default is @option{-mpopc}
23205 when targeting a CPU that supports such an instruction, such as Niagara-2 and
23212 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
23213 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
23214 when targeting a CPU that supports such an instruction, such as Niagara-7 and
23218 @opindex mfix-at697f
23219 Enable the documented workaround for the single erratum of the Atmel AT697F
23220 processor (which corresponds to erratum #13 of the AT697E processor).
23223 @opindex mfix-ut699
23224 Enable the documented workarounds for the floating-point errata and the data
23225 cache nullify errata of the UT699 processor.
23228 These @samp{-m} options are supported in addition to the above
23229 on SPARC-V9 processors in 64-bit environments:
23236 Generate code for a 32-bit or 64-bit environment.
23237 The 32-bit environment sets int, long and pointer to 32 bits.
23238 The 64-bit environment sets int to 32 bits and long and pointer
23241 @item -mcmodel=@var{which}
23243 Set the code model to one of
23247 The Medium/Low code model: 64-bit addresses, programs
23248 must be linked in the low 32 bits of memory. Programs can be statically
23249 or dynamically linked.
23252 The Medium/Middle code model: 64-bit addresses, programs
23253 must be linked in the low 44 bits of memory, the text and data segments must
23254 be less than 2GB in size and the data segment must be located within 2GB of
23258 The Medium/Anywhere code model: 64-bit addresses, programs
23259 may be linked anywhere in memory, the text and data segments must be less
23260 than 2GB in size and the data segment must be located within 2GB of the
23264 The Medium/Anywhere code model for embedded systems:
23265 64-bit addresses, the text and data segments must be less than 2GB in
23266 size, both starting anywhere in memory (determined at link time). The
23267 global register %g4 points to the base of the data segment. Programs
23268 are statically linked and PIC is not supported.
23271 @item -mmemory-model=@var{mem-model}
23272 @opindex mmemory-model
23273 Set the memory model in force on the processor to one of
23277 The default memory model for the processor and operating system.
23280 Relaxed Memory Order
23283 Partial Store Order
23289 Sequential Consistency
23292 These memory models are formally defined in Appendix D of the Sparc V9
23293 architecture manual, as set in the processor's @code{PSTATE.MM} field.
23296 @itemx -mno-stack-bias
23297 @opindex mstack-bias
23298 @opindex mno-stack-bias
23299 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
23300 frame pointer if present, are offset by @minus{}2047 which must be added back
23301 when making stack frame references. This is the default in 64-bit mode.
23302 Otherwise, assume no such offset is present.
23306 @subsection SPU Options
23307 @cindex SPU options
23309 These @samp{-m} options are supported on the SPU:
23313 @itemx -merror-reloc
23314 @opindex mwarn-reloc
23315 @opindex merror-reloc
23317 The loader for SPU does not handle dynamic relocations. By default, GCC
23318 gives an error when it generates code that requires a dynamic
23319 relocation. @option{-mno-error-reloc} disables the error,
23320 @option{-mwarn-reloc} generates a warning instead.
23323 @itemx -munsafe-dma
23325 @opindex munsafe-dma
23327 Instructions that initiate or test completion of DMA must not be
23328 reordered with respect to loads and stores of the memory that is being
23330 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
23331 memory accesses, but that can lead to inefficient code in places where the
23332 memory is known to not change. Rather than mark the memory as volatile,
23333 you can use @option{-msafe-dma} to tell the compiler to treat
23334 the DMA instructions as potentially affecting all memory.
23336 @item -mbranch-hints
23337 @opindex mbranch-hints
23339 By default, GCC generates a branch hint instruction to avoid
23340 pipeline stalls for always-taken or probably-taken branches. A hint
23341 is not generated closer than 8 instructions away from its branch.
23342 There is little reason to disable them, except for debugging purposes,
23343 or to make an object a little bit smaller.
23347 @opindex msmall-mem
23348 @opindex mlarge-mem
23350 By default, GCC generates code assuming that addresses are never larger
23351 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
23352 a full 32-bit address.
23357 By default, GCC links against startup code that assumes the SPU-style
23358 main function interface (which has an unconventional parameter list).
23359 With @option{-mstdmain}, GCC links your program against startup
23360 code that assumes a C99-style interface to @code{main}, including a
23361 local copy of @code{argv} strings.
23363 @item -mfixed-range=@var{register-range}
23364 @opindex mfixed-range
23365 Generate code treating the given register range as fixed registers.
23366 A fixed register is one that the register allocator cannot use. This is
23367 useful when compiling kernel code. A register range is specified as
23368 two registers separated by a dash. Multiple register ranges can be
23369 specified separated by a comma.
23375 Compile code assuming that pointers to the PPU address space accessed
23376 via the @code{__ea} named address space qualifier are either 32 or 64
23377 bits wide. The default is 32 bits. As this is an ABI-changing option,
23378 all object code in an executable must be compiled with the same setting.
23380 @item -maddress-space-conversion
23381 @itemx -mno-address-space-conversion
23382 @opindex maddress-space-conversion
23383 @opindex mno-address-space-conversion
23384 Allow/disallow treating the @code{__ea} address space as superset
23385 of the generic address space. This enables explicit type casts
23386 between @code{__ea} and generic pointer as well as implicit
23387 conversions of generic pointers to @code{__ea} pointers. The
23388 default is to allow address space pointer conversions.
23390 @item -mcache-size=@var{cache-size}
23391 @opindex mcache-size
23392 This option controls the version of libgcc that the compiler links to an
23393 executable and selects a software-managed cache for accessing variables
23394 in the @code{__ea} address space with a particular cache size. Possible
23395 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
23396 and @samp{128}. The default cache size is 64KB.
23398 @item -matomic-updates
23399 @itemx -mno-atomic-updates
23400 @opindex matomic-updates
23401 @opindex mno-atomic-updates
23402 This option controls the version of libgcc that the compiler links to an
23403 executable and selects whether atomic updates to the software-managed
23404 cache of PPU-side variables are used. If you use atomic updates, changes
23405 to a PPU variable from SPU code using the @code{__ea} named address space
23406 qualifier do not interfere with changes to other PPU variables residing
23407 in the same cache line from PPU code. If you do not use atomic updates,
23408 such interference may occur; however, writing back cache lines is
23409 more efficient. The default behavior is to use atomic updates.
23412 @itemx -mdual-nops=@var{n}
23413 @opindex mdual-nops
23414 By default, GCC inserts NOPs to increase dual issue when it expects
23415 it to increase performance. @var{n} can be a value from 0 to 10. A
23416 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
23417 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
23419 @item -mhint-max-nops=@var{n}
23420 @opindex mhint-max-nops
23421 Maximum number of NOPs to insert for a branch hint. A branch hint must
23422 be at least 8 instructions away from the branch it is affecting. GCC
23423 inserts up to @var{n} NOPs to enforce this, otherwise it does not
23424 generate the branch hint.
23426 @item -mhint-max-distance=@var{n}
23427 @opindex mhint-max-distance
23428 The encoding of the branch hint instruction limits the hint to be within
23429 256 instructions of the branch it is affecting. By default, GCC makes
23430 sure it is within 125.
23433 @opindex msafe-hints
23434 Work around a hardware bug that causes the SPU to stall indefinitely.
23435 By default, GCC inserts the @code{hbrp} instruction to make sure
23436 this stall won't happen.
23440 @node System V Options
23441 @subsection Options for System V
23443 These additional options are available on System V Release 4 for
23444 compatibility with other compilers on those systems:
23449 Create a shared object.
23450 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
23454 Identify the versions of each tool used by the compiler, in a
23455 @code{.ident} assembler directive in the output.
23459 Refrain from adding @code{.ident} directives to the output file (this is
23462 @item -YP,@var{dirs}
23464 Search the directories @var{dirs}, and no others, for libraries
23465 specified with @option{-l}.
23467 @item -Ym,@var{dir}
23469 Look in the directory @var{dir} to find the M4 preprocessor.
23470 The assembler uses this option.
23471 @c This is supposed to go with a -Yd for predefined M4 macro files, but
23472 @c the generic assembler that comes with Solaris takes just -Ym.
23475 @node TILE-Gx Options
23476 @subsection TILE-Gx Options
23477 @cindex TILE-Gx options
23479 These @samp{-m} options are supported on the TILE-Gx:
23482 @item -mcmodel=small
23483 @opindex mcmodel=small
23484 Generate code for the small model. The distance for direct calls is
23485 limited to 500M in either direction. PC-relative addresses are 32
23486 bits. Absolute addresses support the full address range.
23488 @item -mcmodel=large
23489 @opindex mcmodel=large
23490 Generate code for the large model. There is no limitation on call
23491 distance, pc-relative addresses, or absolute addresses.
23493 @item -mcpu=@var{name}
23495 Selects the type of CPU to be targeted. Currently the only supported
23496 type is @samp{tilegx}.
23502 Generate code for a 32-bit or 64-bit environment. The 32-bit
23503 environment sets int, long, and pointer to 32 bits. The 64-bit
23504 environment sets int to 32 bits and long and pointer to 64 bits.
23507 @itemx -mlittle-endian
23508 @opindex mbig-endian
23509 @opindex mlittle-endian
23510 Generate code in big/little endian mode, respectively.
23513 @node TILEPro Options
23514 @subsection TILEPro Options
23515 @cindex TILEPro options
23517 These @samp{-m} options are supported on the TILEPro:
23520 @item -mcpu=@var{name}
23522 Selects the type of CPU to be targeted. Currently the only supported
23523 type is @samp{tilepro}.
23527 Generate code for a 32-bit environment, which sets int, long, and
23528 pointer to 32 bits. This is the only supported behavior so the flag
23529 is essentially ignored.
23533 @subsection V850 Options
23534 @cindex V850 Options
23536 These @samp{-m} options are defined for V850 implementations:
23540 @itemx -mno-long-calls
23541 @opindex mlong-calls
23542 @opindex mno-long-calls
23543 Treat all calls as being far away (near). If calls are assumed to be
23544 far away, the compiler always loads the function's address into a
23545 register, and calls indirect through the pointer.
23551 Do not optimize (do optimize) basic blocks that use the same index
23552 pointer 4 or more times to copy pointer into the @code{ep} register, and
23553 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
23554 option is on by default if you optimize.
23556 @item -mno-prolog-function
23557 @itemx -mprolog-function
23558 @opindex mno-prolog-function
23559 @opindex mprolog-function
23560 Do not use (do use) external functions to save and restore registers
23561 at the prologue and epilogue of a function. The external functions
23562 are slower, but use less code space if more than one function saves
23563 the same number of registers. The @option{-mprolog-function} option
23564 is on by default if you optimize.
23568 Try to make the code as small as possible. At present, this just turns
23569 on the @option{-mep} and @option{-mprolog-function} options.
23571 @item -mtda=@var{n}
23573 Put static or global variables whose size is @var{n} bytes or less into
23574 the tiny data area that register @code{ep} points to. The tiny data
23575 area can hold up to 256 bytes in total (128 bytes for byte references).
23577 @item -msda=@var{n}
23579 Put static or global variables whose size is @var{n} bytes or less into
23580 the small data area that register @code{gp} points to. The small data
23581 area can hold up to 64 kilobytes.
23583 @item -mzda=@var{n}
23585 Put static or global variables whose size is @var{n} bytes or less into
23586 the first 32 kilobytes of memory.
23590 Specify that the target processor is the V850.
23594 Specify that the target processor is the V850E3V5. The preprocessor
23595 constant @code{__v850e3v5__} is defined if this option is used.
23599 Specify that the target processor is the V850E3V5. This is an alias for
23600 the @option{-mv850e3v5} option.
23604 Specify that the target processor is the V850E2V3. The preprocessor
23605 constant @code{__v850e2v3__} is defined if this option is used.
23609 Specify that the target processor is the V850E2. The preprocessor
23610 constant @code{__v850e2__} is defined if this option is used.
23614 Specify that the target processor is the V850E1. The preprocessor
23615 constants @code{__v850e1__} and @code{__v850e__} are defined if
23616 this option is used.
23620 Specify that the target processor is the V850ES. This is an alias for
23621 the @option{-mv850e1} option.
23625 Specify that the target processor is the V850E@. The preprocessor
23626 constant @code{__v850e__} is defined if this option is used.
23628 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
23629 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
23630 are defined then a default target processor is chosen and the
23631 relevant @samp{__v850*__} preprocessor constant is defined.
23633 The preprocessor constants @code{__v850} and @code{__v851__} are always
23634 defined, regardless of which processor variant is the target.
23636 @item -mdisable-callt
23637 @itemx -mno-disable-callt
23638 @opindex mdisable-callt
23639 @opindex mno-disable-callt
23640 This option suppresses generation of the @code{CALLT} instruction for the
23641 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
23644 This option is enabled by default when the RH850 ABI is
23645 in use (see @option{-mrh850-abi}), and disabled by default when the
23646 GCC ABI is in use. If @code{CALLT} instructions are being generated
23647 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
23653 Pass on (or do not pass on) the @option{-mrelax} command-line option
23657 @itemx -mno-long-jumps
23658 @opindex mlong-jumps
23659 @opindex mno-long-jumps
23660 Disable (or re-enable) the generation of PC-relative jump instructions.
23663 @itemx -mhard-float
23664 @opindex msoft-float
23665 @opindex mhard-float
23666 Disable (or re-enable) the generation of hardware floating point
23667 instructions. This option is only significant when the target
23668 architecture is @samp{V850E2V3} or higher. If hardware floating point
23669 instructions are being generated then the C preprocessor symbol
23670 @code{__FPU_OK__} is defined, otherwise the symbol
23671 @code{__NO_FPU__} is defined.
23675 Enables the use of the e3v5 LOOP instruction. The use of this
23676 instruction is not enabled by default when the e3v5 architecture is
23677 selected because its use is still experimental.
23681 @opindex mrh850-abi
23683 Enables support for the RH850 version of the V850 ABI. This is the
23684 default. With this version of the ABI the following rules apply:
23688 Integer sized structures and unions are returned via a memory pointer
23689 rather than a register.
23692 Large structures and unions (more than 8 bytes in size) are passed by
23696 Functions are aligned to 16-bit boundaries.
23699 The @option{-m8byte-align} command-line option is supported.
23702 The @option{-mdisable-callt} command-line option is enabled by
23703 default. The @option{-mno-disable-callt} command-line option is not
23707 When this version of the ABI is enabled the C preprocessor symbol
23708 @code{__V850_RH850_ABI__} is defined.
23712 Enables support for the old GCC version of the V850 ABI. With this
23713 version of the ABI the following rules apply:
23717 Integer sized structures and unions are returned in register @code{r10}.
23720 Large structures and unions (more than 8 bytes in size) are passed by
23724 Functions are aligned to 32-bit boundaries, unless optimizing for
23728 The @option{-m8byte-align} command-line option is not supported.
23731 The @option{-mdisable-callt} command-line option is supported but not
23732 enabled by default.
23735 When this version of the ABI is enabled the C preprocessor symbol
23736 @code{__V850_GCC_ABI__} is defined.
23738 @item -m8byte-align
23739 @itemx -mno-8byte-align
23740 @opindex m8byte-align
23741 @opindex mno-8byte-align
23742 Enables support for @code{double} and @code{long long} types to be
23743 aligned on 8-byte boundaries. The default is to restrict the
23744 alignment of all objects to at most 4-bytes. When
23745 @option{-m8byte-align} is in effect the C preprocessor symbol
23746 @code{__V850_8BYTE_ALIGN__} is defined.
23749 @opindex mbig-switch
23750 Generate code suitable for big switch tables. Use this option only if
23751 the assembler/linker complain about out of range branches within a switch
23756 This option causes r2 and r5 to be used in the code generated by
23757 the compiler. This setting is the default.
23759 @item -mno-app-regs
23760 @opindex mno-app-regs
23761 This option causes r2 and r5 to be treated as fixed registers.
23766 @subsection VAX Options
23767 @cindex VAX options
23769 These @samp{-m} options are defined for the VAX:
23774 Do not output certain jump instructions (@code{aobleq} and so on)
23775 that the Unix assembler for the VAX cannot handle across long
23780 Do output those jump instructions, on the assumption that the
23781 GNU assembler is being used.
23785 Output code for G-format floating-point numbers instead of D-format.
23788 @node Visium Options
23789 @subsection Visium Options
23790 @cindex Visium options
23796 A program which performs file I/O and is destined to run on an MCM target
23797 should be linked with this option. It causes the libraries libc.a and
23798 libdebug.a to be linked. The program should be run on the target under
23799 the control of the GDB remote debugging stub.
23803 A program which performs file I/O and is destined to run on the simulator
23804 should be linked with option. This causes libraries libc.a and libsim.a to
23808 @itemx -mhard-float
23810 @opindex mhard-float
23811 Generate code containing floating-point instructions. This is the
23815 @itemx -msoft-float
23817 @opindex msoft-float
23818 Generate code containing library calls for floating-point.
23820 @option{-msoft-float} changes the calling convention in the output file;
23821 therefore, it is only useful if you compile @emph{all} of a program with
23822 this option. In particular, you need to compile @file{libgcc.a}, the
23823 library that comes with GCC, with @option{-msoft-float} in order for
23826 @item -mcpu=@var{cpu_type}
23828 Set the instruction set, register set, and instruction scheduling parameters
23829 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23830 @samp{mcm}, @samp{gr5} and @samp{gr6}.
23832 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
23834 By default (unless configured otherwise), GCC generates code for the GR5
23835 variant of the Visium architecture.
23837 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
23838 architecture. The only difference from GR5 code is that the compiler will
23839 generate block move instructions.
23841 @item -mtune=@var{cpu_type}
23843 Set the instruction scheduling parameters for machine type @var{cpu_type},
23844 but do not set the instruction set or register set that the option
23845 @option{-mcpu=@var{cpu_type}} would.
23849 Generate code for the supervisor mode, where there are no restrictions on
23850 the access to general registers. This is the default.
23853 @opindex muser-mode
23854 Generate code for the user mode, where the access to some general registers
23855 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
23856 mode; on the GR6, only registers r29 to r31 are affected.
23860 @subsection VMS Options
23862 These @samp{-m} options are defined for the VMS implementations:
23865 @item -mvms-return-codes
23866 @opindex mvms-return-codes
23867 Return VMS condition codes from @code{main}. The default is to return POSIX-style
23868 condition (e.g.@ error) codes.
23870 @item -mdebug-main=@var{prefix}
23871 @opindex mdebug-main=@var{prefix}
23872 Flag the first routine whose name starts with @var{prefix} as the main
23873 routine for the debugger.
23877 Default to 64-bit memory allocation routines.
23879 @item -mpointer-size=@var{size}
23880 @opindex mpointer-size=@var{size}
23881 Set the default size of pointers. Possible options for @var{size} are
23882 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
23883 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
23884 The later option disables @code{pragma pointer_size}.
23887 @node VxWorks Options
23888 @subsection VxWorks Options
23889 @cindex VxWorks Options
23891 The options in this section are defined for all VxWorks targets.
23892 Options specific to the target hardware are listed with the other
23893 options for that target.
23898 GCC can generate code for both VxWorks kernels and real time processes
23899 (RTPs). This option switches from the former to the latter. It also
23900 defines the preprocessor macro @code{__RTP__}.
23903 @opindex non-static
23904 Link an RTP executable against shared libraries rather than static
23905 libraries. The options @option{-static} and @option{-shared} can
23906 also be used for RTPs (@pxref{Link Options}); @option{-static}
23913 These options are passed down to the linker. They are defined for
23914 compatibility with Diab.
23917 @opindex Xbind-lazy
23918 Enable lazy binding of function calls. This option is equivalent to
23919 @option{-Wl,-z,now} and is defined for compatibility with Diab.
23923 Disable lazy binding of function calls. This option is the default and
23924 is defined for compatibility with Diab.
23928 @subsection x86 Options
23929 @cindex x86 Options
23931 These @samp{-m} options are defined for the x86 family of computers.
23935 @item -march=@var{cpu-type}
23937 Generate instructions for the machine type @var{cpu-type}. In contrast to
23938 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
23939 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
23940 to generate code that may not run at all on processors other than the one
23941 indicated. Specifying @option{-march=@var{cpu-type}} implies
23942 @option{-mtune=@var{cpu-type}}.
23944 The choices for @var{cpu-type} are:
23948 This selects the CPU to generate code for at compilation time by determining
23949 the processor type of the compiling machine. Using @option{-march=native}
23950 enables all instruction subsets supported by the local machine (hence
23951 the result might not run on different machines). Using @option{-mtune=native}
23952 produces code optimized for the local machine under the constraints
23953 of the selected instruction set.
23956 Original Intel i386 CPU@.
23959 Intel i486 CPU@. (No scheduling is implemented for this chip.)
23963 Intel Pentium CPU with no MMX support.
23966 Intel Lakemont MCU, based on Intel Pentium CPU.
23969 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
23972 Intel Pentium Pro CPU@.
23975 When used with @option{-march}, the Pentium Pro
23976 instruction set is used, so the code runs on all i686 family chips.
23977 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
23980 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
23985 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
23989 Intel Pentium M; low-power version of Intel Pentium III CPU
23990 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
23994 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
23997 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
24001 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
24002 SSE2 and SSE3 instruction set support.
24005 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
24006 instruction set support.
24009 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24010 SSE4.1, SSE4.2 and POPCNT instruction set support.
24013 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24014 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
24017 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24018 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
24021 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24022 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
24023 instruction set support.
24026 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24027 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24028 BMI, BMI2 and F16C instruction set support.
24031 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24032 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24033 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
24036 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24037 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24038 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
24039 XSAVES instruction set support.
24042 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
24043 instruction set support.
24046 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24047 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
24050 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24051 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24052 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
24053 AVX512CD instruction set support.
24055 @item skylake-avx512
24056 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24057 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24058 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
24059 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
24062 AMD K6 CPU with MMX instruction set support.
24066 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
24069 @itemx athlon-tbird
24070 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
24076 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
24077 instruction set support.
24083 Processors based on the AMD K8 core with x86-64 instruction set support,
24084 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
24085 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
24086 instruction set extensions.)
24089 @itemx opteron-sse3
24090 @itemx athlon64-sse3
24091 Improved versions of AMD K8 cores with SSE3 instruction set support.
24095 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
24096 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
24097 instruction set extensions.)
24100 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
24101 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
24102 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
24104 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24105 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
24106 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
24109 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24110 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
24111 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
24112 64-bit instruction set extensions.
24114 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24115 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
24116 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
24117 SSE4.2, ABM and 64-bit instruction set extensions.
24120 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
24121 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
24122 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
24123 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
24124 instruction set extensions.
24127 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
24128 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
24129 instruction set extensions.)
24132 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
24133 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
24134 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
24137 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
24141 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
24142 instruction set support.
24145 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
24146 (No scheduling is implemented for this chip.)
24149 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
24150 (No scheduling is implemented for this chip.)
24153 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24154 (No scheduling is implemented for this chip.)
24157 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
24158 (No scheduling is implemented for this chip.)
24161 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
24162 (No scheduling is implemented for this chip.)
24165 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24166 (No scheduling is implemented for this chip.)
24169 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
24170 (No scheduling is implemented for this chip.)
24173 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
24174 AVX and AVX2 instruction set support.
24175 (No scheduling is implemented for this chip.)
24178 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24179 instruction set support.
24180 (No scheduling is implemented for this chip.)
24183 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24184 instruction set support.
24185 (No scheduling is implemented for this chip.)
24188 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24189 instruction set support.
24190 (No scheduling is implemented for this chip.)
24193 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24194 instruction set support.
24195 (No scheduling is implemented for this chip.)
24198 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24199 instruction set support.
24200 (No scheduling is implemented for this chip.)
24203 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24204 instruction set support.
24205 (No scheduling is implemented for this chip.)
24208 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
24211 @item -mtune=@var{cpu-type}
24213 Tune to @var{cpu-type} everything applicable about the generated code, except
24214 for the ABI and the set of available instructions.
24215 While picking a specific @var{cpu-type} schedules things appropriately
24216 for that particular chip, the compiler does not generate any code that
24217 cannot run on the default machine type unless you use a
24218 @option{-march=@var{cpu-type}} option.
24219 For example, if GCC is configured for i686-pc-linux-gnu
24220 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
24221 but still runs on i686 machines.
24223 The choices for @var{cpu-type} are the same as for @option{-march}.
24224 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
24228 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
24229 If you know the CPU on which your code will run, then you should use
24230 the corresponding @option{-mtune} or @option{-march} option instead of
24231 @option{-mtune=generic}. But, if you do not know exactly what CPU users
24232 of your application will have, then you should use this option.
24234 As new processors are deployed in the marketplace, the behavior of this
24235 option will change. Therefore, if you upgrade to a newer version of
24236 GCC, code generation controlled by this option will change to reflect
24238 that are most common at the time that version of GCC is released.
24240 There is no @option{-march=generic} option because @option{-march}
24241 indicates the instruction set the compiler can use, and there is no
24242 generic instruction set applicable to all processors. In contrast,
24243 @option{-mtune} indicates the processor (or, in this case, collection of
24244 processors) for which the code is optimized.
24247 Produce code optimized for the most current Intel processors, which are
24248 Haswell and Silvermont for this version of GCC. If you know the CPU
24249 on which your code will run, then you should use the corresponding
24250 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
24251 But, if you want your application performs better on both Haswell and
24252 Silvermont, then you should use this option.
24254 As new Intel processors are deployed in the marketplace, the behavior of
24255 this option will change. Therefore, if you upgrade to a newer version of
24256 GCC, code generation controlled by this option will change to reflect
24257 the most current Intel processors at the time that version of GCC is
24260 There is no @option{-march=intel} option because @option{-march} indicates
24261 the instruction set the compiler can use, and there is no common
24262 instruction set applicable to all processors. In contrast,
24263 @option{-mtune} indicates the processor (or, in this case, collection of
24264 processors) for which the code is optimized.
24267 @item -mcpu=@var{cpu-type}
24269 A deprecated synonym for @option{-mtune}.
24271 @item -mfpmath=@var{unit}
24273 Generate floating-point arithmetic for selected unit @var{unit}. The choices
24274 for @var{unit} are:
24278 Use the standard 387 floating-point coprocessor present on the majority of chips and
24279 emulated otherwise. Code compiled with this option runs almost everywhere.
24280 The temporary results are computed in 80-bit precision instead of the precision
24281 specified by the type, resulting in slightly different results compared to most
24282 of other chips. See @option{-ffloat-store} for more detailed description.
24284 This is the default choice for x86-32 targets.
24287 Use scalar floating-point instructions present in the SSE instruction set.
24288 This instruction set is supported by Pentium III and newer chips,
24289 and in the AMD line
24290 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
24291 instruction set supports only single-precision arithmetic, thus the double and
24292 extended-precision arithmetic are still done using 387. A later version, present
24293 only in Pentium 4 and AMD x86-64 chips, supports double-precision
24296 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
24297 or @option{-msse2} switches to enable SSE extensions and make this option
24298 effective. For the x86-64 compiler, these extensions are enabled by default.
24300 The resulting code should be considerably faster in the majority of cases and avoid
24301 the numerical instability problems of 387 code, but may break some existing
24302 code that expects temporaries to be 80 bits.
24304 This is the default choice for the x86-64 compiler.
24309 Attempt to utilize both instruction sets at once. This effectively doubles the
24310 amount of available registers, and on chips with separate execution units for
24311 387 and SSE the execution resources too. Use this option with care, as it is
24312 still experimental, because the GCC register allocator does not model separate
24313 functional units well, resulting in unstable performance.
24316 @item -masm=@var{dialect}
24317 @opindex masm=@var{dialect}
24318 Output assembly instructions using selected @var{dialect}. Also affects
24319 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
24320 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
24321 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
24322 not support @samp{intel}.
24325 @itemx -mno-ieee-fp
24327 @opindex mno-ieee-fp
24328 Control whether or not the compiler uses IEEE floating-point
24329 comparisons. These correctly handle the case where the result of a
24330 comparison is unordered.
24335 @opindex mhard-float
24336 Generate output containing 80387 instructions for floating point.
24341 @opindex msoft-float
24342 Generate output containing library calls for floating point.
24344 @strong{Warning:} the requisite libraries are not part of GCC@.
24345 Normally the facilities of the machine's usual C compiler are used, but
24346 this can't be done directly in cross-compilation. You must make your
24347 own arrangements to provide suitable library functions for
24350 On machines where a function returns floating-point results in the 80387
24351 register stack, some floating-point opcodes may be emitted even if
24352 @option{-msoft-float} is used.
24354 @item -mno-fp-ret-in-387
24355 @opindex mno-fp-ret-in-387
24356 Do not use the FPU registers for return values of functions.
24358 The usual calling convention has functions return values of types
24359 @code{float} and @code{double} in an FPU register, even if there
24360 is no FPU@. The idea is that the operating system should emulate
24363 The option @option{-mno-fp-ret-in-387} causes such values to be returned
24364 in ordinary CPU registers instead.
24366 @item -mno-fancy-math-387
24367 @opindex mno-fancy-math-387
24368 Some 387 emulators do not support the @code{sin}, @code{cos} and
24369 @code{sqrt} instructions for the 387. Specify this option to avoid
24370 generating those instructions. This option is the default on
24371 OpenBSD and NetBSD@. This option is overridden when @option{-march}
24372 indicates that the target CPU always has an FPU and so the
24373 instruction does not need emulation. These
24374 instructions are not generated unless you also use the
24375 @option{-funsafe-math-optimizations} switch.
24377 @item -malign-double
24378 @itemx -mno-align-double
24379 @opindex malign-double
24380 @opindex mno-align-double
24381 Control whether GCC aligns @code{double}, @code{long double}, and
24382 @code{long long} variables on a two-word boundary or a one-word
24383 boundary. Aligning @code{double} variables on a two-word boundary
24384 produces code that runs somewhat faster on a Pentium at the
24385 expense of more memory.
24387 On x86-64, @option{-malign-double} is enabled by default.
24389 @strong{Warning:} if you use the @option{-malign-double} switch,
24390 structures containing the above types are aligned differently than
24391 the published application binary interface specifications for the x86-32
24392 and are not binary compatible with structures in code compiled
24393 without that switch.
24395 @item -m96bit-long-double
24396 @itemx -m128bit-long-double
24397 @opindex m96bit-long-double
24398 @opindex m128bit-long-double
24399 These switches control the size of @code{long double} type. The x86-32
24400 application binary interface specifies the size to be 96 bits,
24401 so @option{-m96bit-long-double} is the default in 32-bit mode.
24403 Modern architectures (Pentium and newer) prefer @code{long double}
24404 to be aligned to an 8- or 16-byte boundary. In arrays or structures
24405 conforming to the ABI, this is not possible. So specifying
24406 @option{-m128bit-long-double} aligns @code{long double}
24407 to a 16-byte boundary by padding the @code{long double} with an additional
24410 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
24411 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
24413 Notice that neither of these options enable any extra precision over the x87
24414 standard of 80 bits for a @code{long double}.
24416 @strong{Warning:} if you override the default value for your target ABI, this
24417 changes the size of
24418 structures and arrays containing @code{long double} variables,
24419 as well as modifying the function calling convention for functions taking
24420 @code{long double}. Hence they are not binary-compatible
24421 with code compiled without that switch.
24423 @item -mlong-double-64
24424 @itemx -mlong-double-80
24425 @itemx -mlong-double-128
24426 @opindex mlong-double-64
24427 @opindex mlong-double-80
24428 @opindex mlong-double-128
24429 These switches control the size of @code{long double} type. A size
24430 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24431 type. This is the default for 32-bit Bionic C library. A size
24432 of 128 bits makes the @code{long double} type equivalent to the
24433 @code{__float128} type. This is the default for 64-bit Bionic C library.
24435 @strong{Warning:} if you override the default value for your target ABI, this
24436 changes the size of
24437 structures and arrays containing @code{long double} variables,
24438 as well as modifying the function calling convention for functions taking
24439 @code{long double}. Hence they are not binary-compatible
24440 with code compiled without that switch.
24442 @item -malign-data=@var{type}
24443 @opindex malign-data
24444 Control how GCC aligns variables. Supported values for @var{type} are
24445 @samp{compat} uses increased alignment value compatible uses GCC 4.8
24446 and earlier, @samp{abi} uses alignment value as specified by the
24447 psABI, and @samp{cacheline} uses increased alignment value to match
24448 the cache line size. @samp{compat} is the default.
24450 @item -mlarge-data-threshold=@var{threshold}
24451 @opindex mlarge-data-threshold
24452 When @option{-mcmodel=medium} is specified, data objects larger than
24453 @var{threshold} are placed in the large data section. This value must be the
24454 same across all objects linked into the binary, and defaults to 65535.
24458 Use a different function-calling convention, in which functions that
24459 take a fixed number of arguments return with the @code{ret @var{num}}
24460 instruction, which pops their arguments while returning. This saves one
24461 instruction in the caller since there is no need to pop the arguments
24464 You can specify that an individual function is called with this calling
24465 sequence with the function attribute @code{stdcall}. You can also
24466 override the @option{-mrtd} option by using the function attribute
24467 @code{cdecl}. @xref{Function Attributes}.
24469 @strong{Warning:} this calling convention is incompatible with the one
24470 normally used on Unix, so you cannot use it if you need to call
24471 libraries compiled with the Unix compiler.
24473 Also, you must provide function prototypes for all functions that
24474 take variable numbers of arguments (including @code{printf});
24475 otherwise incorrect code is generated for calls to those
24478 In addition, seriously incorrect code results if you call a
24479 function with too many arguments. (Normally, extra arguments are
24480 harmlessly ignored.)
24482 @item -mregparm=@var{num}
24484 Control how many registers are used to pass integer arguments. By
24485 default, no registers are used to pass arguments, and at most 3
24486 registers can be used. You can control this behavior for a specific
24487 function by using the function attribute @code{regparm}.
24488 @xref{Function Attributes}.
24490 @strong{Warning:} if you use this switch, and
24491 @var{num} is nonzero, then you must build all modules with the same
24492 value, including any libraries. This includes the system libraries and
24496 @opindex msseregparm
24497 Use SSE register passing conventions for float and double arguments
24498 and return values. You can control this behavior for a specific
24499 function by using the function attribute @code{sseregparm}.
24500 @xref{Function Attributes}.
24502 @strong{Warning:} if you use this switch then you must build all
24503 modules with the same value, including any libraries. This includes
24504 the system libraries and startup modules.
24506 @item -mvect8-ret-in-mem
24507 @opindex mvect8-ret-in-mem
24508 Return 8-byte vectors in memory instead of MMX registers. This is the
24509 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
24510 Studio compilers until version 12. Later compiler versions (starting
24511 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
24512 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
24513 you need to remain compatible with existing code produced by those
24514 previous compiler versions or older versions of GCC@.
24523 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
24524 is specified, the significands of results of floating-point operations are
24525 rounded to 24 bits (single precision); @option{-mpc64} rounds the
24526 significands of results of floating-point operations to 53 bits (double
24527 precision) and @option{-mpc80} rounds the significands of results of
24528 floating-point operations to 64 bits (extended double precision), which is
24529 the default. When this option is used, floating-point operations in higher
24530 precisions are not available to the programmer without setting the FPU
24531 control word explicitly.
24533 Setting the rounding of floating-point operations to less than the default
24534 80 bits can speed some programs by 2% or more. Note that some mathematical
24535 libraries assume that extended-precision (80-bit) floating-point operations
24536 are enabled by default; routines in such libraries could suffer significant
24537 loss of accuracy, typically through so-called ``catastrophic cancellation'',
24538 when this option is used to set the precision to less than extended precision.
24540 @item -mstackrealign
24541 @opindex mstackrealign
24542 Realign the stack at entry. On the x86, the @option{-mstackrealign}
24543 option generates an alternate prologue and epilogue that realigns the
24544 run-time stack if necessary. This supports mixing legacy codes that keep
24545 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
24546 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
24547 applicable to individual functions.
24549 @item -mpreferred-stack-boundary=@var{num}
24550 @opindex mpreferred-stack-boundary
24551 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24552 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24553 the default is 4 (16 bytes or 128 bits).
24555 @strong{Warning:} When generating code for the x86-64 architecture with
24556 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
24557 used to keep the stack boundary aligned to 8 byte boundary. Since
24558 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
24559 intended to be used in controlled environment where stack space is
24560 important limitation. This option leads to wrong code when functions
24561 compiled with 16 byte stack alignment (such as functions from a standard
24562 library) are called with misaligned stack. In this case, SSE
24563 instructions may lead to misaligned memory access traps. In addition,
24564 variable arguments are handled incorrectly for 16 byte aligned
24565 objects (including x87 long double and __int128), leading to wrong
24566 results. You must build all modules with
24567 @option{-mpreferred-stack-boundary=3}, including any libraries. This
24568 includes the system libraries and startup modules.
24570 @item -mincoming-stack-boundary=@var{num}
24571 @opindex mincoming-stack-boundary
24572 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
24573 boundary. If @option{-mincoming-stack-boundary} is not specified,
24574 the one specified by @option{-mpreferred-stack-boundary} is used.
24576 On Pentium and Pentium Pro, @code{double} and @code{long double} values
24577 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
24578 suffer significant run time performance penalties. On Pentium III, the
24579 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
24580 properly if it is not 16-byte aligned.
24582 To ensure proper alignment of this values on the stack, the stack boundary
24583 must be as aligned as that required by any value stored on the stack.
24584 Further, every function must be generated such that it keeps the stack
24585 aligned. Thus calling a function compiled with a higher preferred
24586 stack boundary from a function compiled with a lower preferred stack
24587 boundary most likely misaligns the stack. It is recommended that
24588 libraries that use callbacks always use the default setting.
24590 This extra alignment does consume extra stack space, and generally
24591 increases code size. Code that is sensitive to stack space usage, such
24592 as embedded systems and operating system kernels, may want to reduce the
24593 preferred alignment to @option{-mpreferred-stack-boundary=2}.
24650 @itemx -mavx512ifma
24651 @opindex mavx512ifma
24653 @itemx -mavx512vbmi
24654 @opindex mavx512vbmi
24666 @opindex mclfushopt
24683 @itemx -mprefetchwt1
24684 @opindex mprefetchwt1
24740 These switches enable the use of instructions in the MMX, SSE,
24741 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
24742 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
24743 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
24744 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
24745 extended instruction sets. Each has a corresponding @option{-mno-} option
24746 to disable use of these instructions.
24748 These extensions are also available as built-in functions: see
24749 @ref{x86 Built-in Functions}, for details of the functions enabled and
24750 disabled by these switches.
24752 To generate SSE/SSE2 instructions automatically from floating-point
24753 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
24755 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
24756 generates new AVX instructions or AVX equivalence for all SSEx instructions
24759 These options enable GCC to use these extended instructions in
24760 generated code, even without @option{-mfpmath=sse}. Applications that
24761 perform run-time CPU detection must compile separate files for each
24762 supported architecture, using the appropriate flags. In particular,
24763 the file containing the CPU detection code should be compiled without
24766 @item -mdump-tune-features
24767 @opindex mdump-tune-features
24768 This option instructs GCC to dump the names of the x86 performance
24769 tuning features and default settings. The names can be used in
24770 @option{-mtune-ctrl=@var{feature-list}}.
24772 @item -mtune-ctrl=@var{feature-list}
24773 @opindex mtune-ctrl=@var{feature-list}
24774 This option is used to do fine grain control of x86 code generation features.
24775 @var{feature-list} is a comma separated list of @var{feature} names. See also
24776 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
24777 on if it is not preceded with @samp{^}, otherwise, it is turned off.
24778 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
24779 developers. Using it may lead to code paths not covered by testing and can
24780 potentially result in compiler ICEs or runtime errors.
24783 @opindex mno-default
24784 This option instructs GCC to turn off all tunable features. See also
24785 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
24789 This option instructs GCC to emit a @code{cld} instruction in the prologue
24790 of functions that use string instructions. String instructions depend on
24791 the DF flag to select between autoincrement or autodecrement mode. While the
24792 ABI specifies the DF flag to be cleared on function entry, some operating
24793 systems violate this specification by not clearing the DF flag in their
24794 exception dispatchers. The exception handler can be invoked with the DF flag
24795 set, which leads to wrong direction mode when string instructions are used.
24796 This option can be enabled by default on 32-bit x86 targets by configuring
24797 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
24798 instructions can be suppressed with the @option{-mno-cld} compiler option
24802 @opindex mvzeroupper
24803 This option instructs GCC to emit a @code{vzeroupper} instruction
24804 before a transfer of control flow out of the function to minimize
24805 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
24808 @item -mprefer-avx128
24809 @opindex mprefer-avx128
24810 This option instructs GCC to use 128-bit AVX instructions instead of
24811 256-bit AVX instructions in the auto-vectorizer.
24815 This option enables GCC to generate @code{CMPXCHG16B} instructions.
24816 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
24817 (or oword) data types.
24818 This is useful for high-resolution counters that can be updated
24819 by multiple processors (or cores). This instruction is generated as part of
24820 atomic built-in functions: see @ref{__sync Builtins} or
24821 @ref{__atomic Builtins} for details.
24825 This option enables generation of @code{SAHF} instructions in 64-bit code.
24826 Early Intel Pentium 4 CPUs with Intel 64 support,
24827 prior to the introduction of Pentium 4 G1 step in December 2005,
24828 lacked the @code{LAHF} and @code{SAHF} instructions
24829 which are supported by AMD64.
24830 These are load and store instructions, respectively, for certain status flags.
24831 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
24832 @code{drem}, and @code{remainder} built-in functions;
24833 see @ref{Other Builtins} for details.
24837 This option enables use of the @code{movbe} instruction to implement
24838 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
24842 This option enables built-in functions @code{__builtin_ia32_crc32qi},
24843 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
24844 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
24848 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
24849 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
24850 with an additional Newton-Raphson step
24851 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
24852 (and their vectorized
24853 variants) for single-precision floating-point arguments. These instructions
24854 are generated only when @option{-funsafe-math-optimizations} is enabled
24855 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
24856 Note that while the throughput of the sequence is higher than the throughput
24857 of the non-reciprocal instruction, the precision of the sequence can be
24858 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
24860 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
24861 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
24862 combination), and doesn't need @option{-mrecip}.
24864 Also note that GCC emits the above sequence with additional Newton-Raphson step
24865 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
24866 already with @option{-ffast-math} (or the above option combination), and
24867 doesn't need @option{-mrecip}.
24869 @item -mrecip=@var{opt}
24870 @opindex mrecip=opt
24871 This option controls which reciprocal estimate instructions
24872 may be used. @var{opt} is a comma-separated list of options, which may
24873 be preceded by a @samp{!} to invert the option:
24877 Enable all estimate instructions.
24880 Enable the default instructions, equivalent to @option{-mrecip}.
24883 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24886 Enable the approximation for scalar division.
24889 Enable the approximation for vectorized division.
24892 Enable the approximation for scalar square root.
24895 Enable the approximation for vectorized square root.
24898 So, for example, @option{-mrecip=all,!sqrt} enables
24899 all of the reciprocal approximations, except for square root.
24901 @item -mveclibabi=@var{type}
24902 @opindex mveclibabi
24903 Specifies the ABI type to use for vectorizing intrinsics using an
24904 external library. Supported values for @var{type} are @samp{svml}
24905 for the Intel short
24906 vector math library and @samp{acml} for the AMD math core library.
24907 To use this option, both @option{-ftree-vectorize} and
24908 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
24909 ABI-compatible library must be specified at link time.
24911 GCC currently emits calls to @code{vmldExp2},
24912 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
24913 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
24914 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
24915 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
24916 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
24917 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
24918 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
24919 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
24920 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
24921 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
24922 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
24923 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
24924 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
24925 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
24926 when @option{-mveclibabi=acml} is used.
24928 @item -mabi=@var{name}
24930 Generate code for the specified calling convention. Permissible values
24931 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
24932 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
24933 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
24934 You can control this behavior for specific functions by
24935 using the function attributes @code{ms_abi} and @code{sysv_abi}.
24936 @xref{Function Attributes}.
24938 @item -mtls-dialect=@var{type}
24939 @opindex mtls-dialect
24940 Generate code to access thread-local storage using the @samp{gnu} or
24941 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
24942 @samp{gnu2} is more efficient, but it may add compile- and run-time
24943 requirements that cannot be satisfied on all systems.
24946 @itemx -mno-push-args
24947 @opindex mpush-args
24948 @opindex mno-push-args
24949 Use PUSH operations to store outgoing parameters. This method is shorter
24950 and usually equally fast as method using SUB/MOV operations and is enabled
24951 by default. In some cases disabling it may improve performance because of
24952 improved scheduling and reduced dependencies.
24954 @item -maccumulate-outgoing-args
24955 @opindex maccumulate-outgoing-args
24956 If enabled, the maximum amount of space required for outgoing arguments is
24957 computed in the function prologue. This is faster on most modern CPUs
24958 because of reduced dependencies, improved scheduling and reduced stack usage
24959 when the preferred stack boundary is not equal to 2. The drawback is a notable
24960 increase in code size. This switch implies @option{-mno-push-args}.
24964 Support thread-safe exception handling on MinGW. Programs that rely
24965 on thread-safe exception handling must compile and link all code with the
24966 @option{-mthreads} option. When compiling, @option{-mthreads} defines
24967 @option{-D_MT}; when linking, it links in a special thread helper library
24968 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
24970 @item -mms-bitfields
24971 @itemx -mno-ms-bitfields
24972 @opindex mms-bitfields
24973 @opindex mno-ms-bitfields
24975 Enable/disable bit-field layout compatible with the native Microsoft
24978 If @code{packed} is used on a structure, or if bit-fields are used,
24979 it may be that the Microsoft ABI lays out the structure differently
24980 than the way GCC normally does. Particularly when moving packed
24981 data between functions compiled with GCC and the native Microsoft compiler
24982 (either via function call or as data in a file), it may be necessary to access
24985 This option is enabled by default for Microsoft Windows
24986 targets. This behavior can also be controlled locally by use of variable
24987 or type attributes. For more information, see @ref{x86 Variable Attributes}
24988 and @ref{x86 Type Attributes}.
24990 The Microsoft structure layout algorithm is fairly simple with the exception
24991 of the bit-field packing.
24992 The padding and alignment of members of structures and whether a bit-field
24993 can straddle a storage-unit boundary are determine by these rules:
24996 @item Structure members are stored sequentially in the order in which they are
24997 declared: the first member has the lowest memory address and the last member
25000 @item Every data object has an alignment requirement. The alignment requirement
25001 for all data except structures, unions, and arrays is either the size of the
25002 object or the current packing size (specified with either the
25003 @code{aligned} attribute or the @code{pack} pragma),
25004 whichever is less. For structures, unions, and arrays,
25005 the alignment requirement is the largest alignment requirement of its members.
25006 Every object is allocated an offset so that:
25009 offset % alignment_requirement == 0
25012 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
25013 unit if the integral types are the same size and if the next bit-field fits
25014 into the current allocation unit without crossing the boundary imposed by the
25015 common alignment requirements of the bit-fields.
25018 MSVC interprets zero-length bit-fields in the following ways:
25021 @item If a zero-length bit-field is inserted between two bit-fields that
25022 are normally coalesced, the bit-fields are not coalesced.
25029 unsigned long bf_1 : 12;
25031 unsigned long bf_2 : 12;
25036 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
25037 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
25039 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
25040 alignment of the zero-length bit-field is greater than the member that follows it,
25041 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
25062 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
25063 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
25064 bit-field does not affect the alignment of @code{bar} or, as a result, the size
25067 Taking this into account, it is important to note the following:
25070 @item If a zero-length bit-field follows a normal bit-field, the type of the
25071 zero-length bit-field may affect the alignment of the structure as whole. For
25072 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
25073 normal bit-field, and is of type short.
25075 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
25076 still affect the alignment of the structure:
25087 Here, @code{t4} takes up 4 bytes.
25090 @item Zero-length bit-fields following non-bit-field members are ignored:
25102 Here, @code{t5} takes up 2 bytes.
25106 @item -mno-align-stringops
25107 @opindex mno-align-stringops
25108 Do not align the destination of inlined string operations. This switch reduces
25109 code size and improves performance in case the destination is already aligned,
25110 but GCC doesn't know about it.
25112 @item -minline-all-stringops
25113 @opindex minline-all-stringops
25114 By default GCC inlines string operations only when the destination is
25115 known to be aligned to least a 4-byte boundary.
25116 This enables more inlining and increases code
25117 size, but may improve performance of code that depends on fast
25118 @code{memcpy}, @code{strlen},
25119 and @code{memset} for short lengths.
25121 @item -minline-stringops-dynamically
25122 @opindex minline-stringops-dynamically
25123 For string operations of unknown size, use run-time checks with
25124 inline code for small blocks and a library call for large blocks.
25126 @item -mstringop-strategy=@var{alg}
25127 @opindex mstringop-strategy=@var{alg}
25128 Override the internal decision heuristic for the particular algorithm to use
25129 for inlining string operations. The allowed values for @var{alg} are:
25135 Expand using i386 @code{rep} prefix of the specified size.
25139 @itemx unrolled_loop
25140 Expand into an inline loop.
25143 Always use a library call.
25146 @item -mmemcpy-strategy=@var{strategy}
25147 @opindex mmemcpy-strategy=@var{strategy}
25148 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
25149 should be inlined and what inline algorithm to use when the expected size
25150 of the copy operation is known. @var{strategy}
25151 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
25152 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
25153 the max byte size with which inline algorithm @var{alg} is allowed. For the last
25154 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
25155 in the list must be specified in increasing order. The minimal byte size for
25156 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
25159 @item -mmemset-strategy=@var{strategy}
25160 @opindex mmemset-strategy=@var{strategy}
25161 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
25162 @code{__builtin_memset} expansion.
25164 @item -momit-leaf-frame-pointer
25165 @opindex momit-leaf-frame-pointer
25166 Don't keep the frame pointer in a register for leaf functions. This
25167 avoids the instructions to save, set up, and restore frame pointers and
25168 makes an extra register available in leaf functions. The option
25169 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
25170 which might make debugging harder.
25172 @item -mtls-direct-seg-refs
25173 @itemx -mno-tls-direct-seg-refs
25174 @opindex mtls-direct-seg-refs
25175 Controls whether TLS variables may be accessed with offsets from the
25176 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
25177 or whether the thread base pointer must be added. Whether or not this
25178 is valid depends on the operating system, and whether it maps the
25179 segment to cover the entire TLS area.
25181 For systems that use the GNU C Library, the default is on.
25184 @itemx -mno-sse2avx
25186 Specify that the assembler should encode SSE instructions with VEX
25187 prefix. The option @option{-mavx} turns this on by default.
25192 If profiling is active (@option{-pg}), put the profiling
25193 counter call before the prologue.
25194 Note: On x86 architectures the attribute @code{ms_hook_prologue}
25195 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
25197 @item -mrecord-mcount
25198 @itemx -mno-record-mcount
25199 @opindex mrecord-mcount
25200 If profiling is active (@option{-pg}), generate a __mcount_loc section
25201 that contains pointers to each profiling call. This is useful for
25202 automatically patching and out calls.
25205 @itemx -mno-nop-mcount
25206 @opindex mnop-mcount
25207 If profiling is active (@option{-pg}), generate the calls to
25208 the profiling functions as NOPs. This is useful when they
25209 should be patched in later dynamically. This is likely only
25210 useful together with @option{-mrecord-mcount}.
25212 @item -mskip-rax-setup
25213 @itemx -mno-skip-rax-setup
25214 @opindex mskip-rax-setup
25215 When generating code for the x86-64 architecture with SSE extensions
25216 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
25217 register when there are no variable arguments passed in vector registers.
25219 @strong{Warning:} Since RAX register is used to avoid unnecessarily
25220 saving vector registers on stack when passing variable arguments, the
25221 impacts of this option are callees may waste some stack space,
25222 misbehave or jump to a random location. GCC 4.4 or newer don't have
25223 those issues, regardless the RAX register value.
25226 @itemx -mno-8bit-idiv
25227 @opindex m8bit-idiv
25228 On some processors, like Intel Atom, 8-bit unsigned integer divide is
25229 much faster than 32-bit/64-bit integer divide. This option generates a
25230 run-time check. If both dividend and divisor are within range of 0
25231 to 255, 8-bit unsigned integer divide is used instead of
25232 32-bit/64-bit integer divide.
25234 @item -mavx256-split-unaligned-load
25235 @itemx -mavx256-split-unaligned-store
25236 @opindex mavx256-split-unaligned-load
25237 @opindex mavx256-split-unaligned-store
25238 Split 32-byte AVX unaligned load and store.
25240 @item -mstack-protector-guard=@var{guard}
25241 @opindex mstack-protector-guard=@var{guard}
25242 Generate stack protection code using canary at @var{guard}. Supported
25243 locations are @samp{global} for global canary or @samp{tls} for per-thread
25244 canary in the TLS block (the default). This option has effect only when
25245 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
25247 @item -mmitigate-rop
25248 @opindex mmitigate-rop
25249 Try to avoid generating code sequences that contain unintended return
25250 opcodes, to mitigate against certain forms of attack. At the moment,
25251 this option is limited in what it can do and should not be relied
25252 on to provide serious protection.
25254 @item -mgeneral-regs-only
25255 @opindex mgeneral-regs-only
25256 Generate code that uses only the general-purpose registers. This
25257 prevents the compiler from using floating-point, vector, mask and bound
25262 These @samp{-m} switches are supported in addition to the above
25263 on x86-64 processors in 64-bit environments.
25276 Generate code for a 16-bit, 32-bit or 64-bit environment.
25277 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
25279 generates code that runs on any i386 system.
25281 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
25282 types to 64 bits, and generates code for the x86-64 architecture.
25283 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
25284 and @option{-mdynamic-no-pic} options.
25286 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
25288 generates code for the x86-64 architecture.
25290 The @option{-m16} option is the same as @option{-m32}, except for that
25291 it outputs the @code{.code16gcc} assembly directive at the beginning of
25292 the assembly output so that the binary can run in 16-bit mode.
25294 The @option{-miamcu} option generates code which conforms to Intel MCU
25295 psABI. It requires the @option{-m32} option to be turned on.
25297 @item -mno-red-zone
25298 @opindex mno-red-zone
25299 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
25300 by the x86-64 ABI; it is a 128-byte area beyond the location of the
25301 stack pointer that is not modified by signal or interrupt handlers
25302 and therefore can be used for temporary data without adjusting the stack
25303 pointer. The flag @option{-mno-red-zone} disables this red zone.
25305 @item -mcmodel=small
25306 @opindex mcmodel=small
25307 Generate code for the small code model: the program and its symbols must
25308 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
25309 Programs can be statically or dynamically linked. This is the default
25312 @item -mcmodel=kernel
25313 @opindex mcmodel=kernel
25314 Generate code for the kernel code model. The kernel runs in the
25315 negative 2 GB of the address space.
25316 This model has to be used for Linux kernel code.
25318 @item -mcmodel=medium
25319 @opindex mcmodel=medium
25320 Generate code for the medium model: the program is linked in the lower 2
25321 GB of the address space. Small symbols are also placed there. Symbols
25322 with sizes larger than @option{-mlarge-data-threshold} are put into
25323 large data or BSS sections and can be located above 2GB. Programs can
25324 be statically or dynamically linked.
25326 @item -mcmodel=large
25327 @opindex mcmodel=large
25328 Generate code for the large model. This model makes no assumptions
25329 about addresses and sizes of sections.
25331 @item -maddress-mode=long
25332 @opindex maddress-mode=long
25333 Generate code for long address mode. This is only supported for 64-bit
25334 and x32 environments. It is the default address mode for 64-bit
25337 @item -maddress-mode=short
25338 @opindex maddress-mode=short
25339 Generate code for short address mode. This is only supported for 32-bit
25340 and x32 environments. It is the default address mode for 32-bit and
25344 @node x86 Windows Options
25345 @subsection x86 Windows Options
25346 @cindex x86 Windows Options
25347 @cindex Windows Options for x86
25349 These additional options are available for Microsoft Windows targets:
25355 specifies that a console application is to be generated, by
25356 instructing the linker to set the PE header subsystem type
25357 required for console applications.
25358 This option is available for Cygwin and MinGW targets and is
25359 enabled by default on those targets.
25363 This option is available for Cygwin and MinGW targets. It
25364 specifies that a DLL---a dynamic link library---is to be
25365 generated, enabling the selection of the required runtime
25366 startup object and entry point.
25368 @item -mnop-fun-dllimport
25369 @opindex mnop-fun-dllimport
25370 This option is available for Cygwin and MinGW targets. It
25371 specifies that the @code{dllimport} attribute should be ignored.
25375 This option is available for MinGW targets. It specifies
25376 that MinGW-specific thread support is to be used.
25380 This option is available for MinGW-w64 targets. It causes
25381 the @code{UNICODE} preprocessor macro to be predefined, and
25382 chooses Unicode-capable runtime startup code.
25386 This option is available for Cygwin and MinGW targets. It
25387 specifies that the typical Microsoft Windows predefined macros are to
25388 be set in the pre-processor, but does not influence the choice
25389 of runtime library/startup code.
25393 This option is available for Cygwin and MinGW targets. It
25394 specifies that a GUI application is to be generated by
25395 instructing the linker to set the PE header subsystem type
25398 @item -fno-set-stack-executable
25399 @opindex fno-set-stack-executable
25400 This option is available for MinGW targets. It specifies that
25401 the executable flag for the stack used by nested functions isn't
25402 set. This is necessary for binaries running in kernel mode of
25403 Microsoft Windows, as there the User32 API, which is used to set executable
25404 privileges, isn't available.
25406 @item -fwritable-relocated-rdata
25407 @opindex fno-writable-relocated-rdata
25408 This option is available for MinGW and Cygwin targets. It specifies
25409 that relocated-data in read-only section is put into the @code{.data}
25410 section. This is a necessary for older runtimes not supporting
25411 modification of @code{.rdata} sections for pseudo-relocation.
25413 @item -mpe-aligned-commons
25414 @opindex mpe-aligned-commons
25415 This option is available for Cygwin and MinGW targets. It
25416 specifies that the GNU extension to the PE file format that
25417 permits the correct alignment of COMMON variables should be
25418 used when generating code. It is enabled by default if
25419 GCC detects that the target assembler found during configuration
25420 supports the feature.
25423 See also under @ref{x86 Options} for standard options.
25425 @node Xstormy16 Options
25426 @subsection Xstormy16 Options
25427 @cindex Xstormy16 Options
25429 These options are defined for Xstormy16:
25434 Choose startup files and linker script suitable for the simulator.
25437 @node Xtensa Options
25438 @subsection Xtensa Options
25439 @cindex Xtensa Options
25441 These options are supported for Xtensa targets:
25445 @itemx -mno-const16
25447 @opindex mno-const16
25448 Enable or disable use of @code{CONST16} instructions for loading
25449 constant values. The @code{CONST16} instruction is currently not a
25450 standard option from Tensilica. When enabled, @code{CONST16}
25451 instructions are always used in place of the standard @code{L32R}
25452 instructions. The use of @code{CONST16} is enabled by default only if
25453 the @code{L32R} instruction is not available.
25456 @itemx -mno-fused-madd
25457 @opindex mfused-madd
25458 @opindex mno-fused-madd
25459 Enable or disable use of fused multiply/add and multiply/subtract
25460 instructions in the floating-point option. This has no effect if the
25461 floating-point option is not also enabled. Disabling fused multiply/add
25462 and multiply/subtract instructions forces the compiler to use separate
25463 instructions for the multiply and add/subtract operations. This may be
25464 desirable in some cases where strict IEEE 754-compliant results are
25465 required: the fused multiply add/subtract instructions do not round the
25466 intermediate result, thereby producing results with @emph{more} bits of
25467 precision than specified by the IEEE standard. Disabling fused multiply
25468 add/subtract instructions also ensures that the program output is not
25469 sensitive to the compiler's ability to combine multiply and add/subtract
25472 @item -mserialize-volatile
25473 @itemx -mno-serialize-volatile
25474 @opindex mserialize-volatile
25475 @opindex mno-serialize-volatile
25476 When this option is enabled, GCC inserts @code{MEMW} instructions before
25477 @code{volatile} memory references to guarantee sequential consistency.
25478 The default is @option{-mserialize-volatile}. Use
25479 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
25481 @item -mforce-no-pic
25482 @opindex mforce-no-pic
25483 For targets, like GNU/Linux, where all user-mode Xtensa code must be
25484 position-independent code (PIC), this option disables PIC for compiling
25487 @item -mtext-section-literals
25488 @itemx -mno-text-section-literals
25489 @opindex mtext-section-literals
25490 @opindex mno-text-section-literals
25491 These options control the treatment of literal pools. The default is
25492 @option{-mno-text-section-literals}, which places literals in a separate
25493 section in the output file. This allows the literal pool to be placed
25494 in a data RAM/ROM, and it also allows the linker to combine literal
25495 pools from separate object files to remove redundant literals and
25496 improve code size. With @option{-mtext-section-literals}, the literals
25497 are interspersed in the text section in order to keep them as close as
25498 possible to their references. This may be necessary for large assembly
25499 files. Literals for each function are placed right before that function.
25501 @item -mauto-litpools
25502 @itemx -mno-auto-litpools
25503 @opindex mauto-litpools
25504 @opindex mno-auto-litpools
25505 These options control the treatment of literal pools. The default is
25506 @option{-mno-auto-litpools}, which places literals in a separate
25507 section in the output file unless @option{-mtext-section-literals} is
25508 used. With @option{-mauto-litpools} the literals are interspersed in
25509 the text section by the assembler. Compiler does not produce explicit
25510 @code{.literal} directives and loads literals into registers with
25511 @code{MOVI} instructions instead of @code{L32R} to let the assembler
25512 do relaxation and place literals as necessary. This option allows
25513 assembler to create several literal pools per function and assemble
25514 very big functions, which may not be possible with
25515 @option{-mtext-section-literals}.
25517 @item -mtarget-align
25518 @itemx -mno-target-align
25519 @opindex mtarget-align
25520 @opindex mno-target-align
25521 When this option is enabled, GCC instructs the assembler to
25522 automatically align instructions to reduce branch penalties at the
25523 expense of some code density. The assembler attempts to widen density
25524 instructions to align branch targets and the instructions following call
25525 instructions. If there are not enough preceding safe density
25526 instructions to align a target, no widening is performed. The
25527 default is @option{-mtarget-align}. These options do not affect the
25528 treatment of auto-aligned instructions like @code{LOOP}, which the
25529 assembler always aligns, either by widening density instructions or
25530 by inserting NOP instructions.
25533 @itemx -mno-longcalls
25534 @opindex mlongcalls
25535 @opindex mno-longcalls
25536 When this option is enabled, GCC instructs the assembler to translate
25537 direct calls to indirect calls unless it can determine that the target
25538 of a direct call is in the range allowed by the call instruction. This
25539 translation typically occurs for calls to functions in other source
25540 files. Specifically, the assembler translates a direct @code{CALL}
25541 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
25542 The default is @option{-mno-longcalls}. This option should be used in
25543 programs where the call target can potentially be out of range. This
25544 option is implemented in the assembler, not the compiler, so the
25545 assembly code generated by GCC still shows direct call
25546 instructions---look at the disassembled object code to see the actual
25547 instructions. Note that the assembler uses an indirect call for
25548 every cross-file call, not just those that really are out of range.
25551 @node zSeries Options
25552 @subsection zSeries Options
25553 @cindex zSeries options
25555 These are listed under @xref{S/390 and zSeries Options}.
25561 @section Specifying Subprocesses and the Switches to Pass to Them
25564 @command{gcc} is a driver program. It performs its job by invoking a
25565 sequence of other programs to do the work of compiling, assembling and
25566 linking. GCC interprets its command-line parameters and uses these to
25567 deduce which programs it should invoke, and which command-line options
25568 it ought to place on their command lines. This behavior is controlled
25569 by @dfn{spec strings}. In most cases there is one spec string for each
25570 program that GCC can invoke, but a few programs have multiple spec
25571 strings to control their behavior. The spec strings built into GCC can
25572 be overridden by using the @option{-specs=} command-line switch to specify
25575 @dfn{Spec files} are plain-text files that are used to construct spec
25576 strings. They consist of a sequence of directives separated by blank
25577 lines. The type of directive is determined by the first non-whitespace
25578 character on the line, which can be one of the following:
25581 @item %@var{command}
25582 Issues a @var{command} to the spec file processor. The commands that can
25586 @item %include <@var{file}>
25587 @cindex @code{%include}
25588 Search for @var{file} and insert its text at the current point in the
25591 @item %include_noerr <@var{file}>
25592 @cindex @code{%include_noerr}
25593 Just like @samp{%include}, but do not generate an error message if the include
25594 file cannot be found.
25596 @item %rename @var{old_name} @var{new_name}
25597 @cindex @code{%rename}
25598 Rename the spec string @var{old_name} to @var{new_name}.
25602 @item *[@var{spec_name}]:
25603 This tells the compiler to create, override or delete the named spec
25604 string. All lines after this directive up to the next directive or
25605 blank line are considered to be the text for the spec string. If this
25606 results in an empty string then the spec is deleted. (Or, if the
25607 spec did not exist, then nothing happens.) Otherwise, if the spec
25608 does not currently exist a new spec is created. If the spec does
25609 exist then its contents are overridden by the text of this
25610 directive, unless the first character of that text is the @samp{+}
25611 character, in which case the text is appended to the spec.
25613 @item [@var{suffix}]:
25614 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
25615 and up to the next directive or blank line are considered to make up the
25616 spec string for the indicated suffix. When the compiler encounters an
25617 input file with the named suffix, it processes the spec string in
25618 order to work out how to compile that file. For example:
25622 z-compile -input %i
25625 This says that any input file whose name ends in @samp{.ZZ} should be
25626 passed to the program @samp{z-compile}, which should be invoked with the
25627 command-line switch @option{-input} and with the result of performing the
25628 @samp{%i} substitution. (See below.)
25630 As an alternative to providing a spec string, the text following a
25631 suffix directive can be one of the following:
25634 @item @@@var{language}
25635 This says that the suffix is an alias for a known @var{language}. This is
25636 similar to using the @option{-x} command-line switch to GCC to specify a
25637 language explicitly. For example:
25644 Says that .ZZ files are, in fact, C++ source files.
25647 This causes an error messages saying:
25650 @var{name} compiler not installed on this system.
25654 GCC already has an extensive list of suffixes built into it.
25655 This directive adds an entry to the end of the list of suffixes, but
25656 since the list is searched from the end backwards, it is effectively
25657 possible to override earlier entries using this technique.
25661 GCC has the following spec strings built into it. Spec files can
25662 override these strings or create their own. Note that individual
25663 targets can also add their own spec strings to this list.
25666 asm Options to pass to the assembler
25667 asm_final Options to pass to the assembler post-processor
25668 cpp Options to pass to the C preprocessor
25669 cc1 Options to pass to the C compiler
25670 cc1plus Options to pass to the C++ compiler
25671 endfile Object files to include at the end of the link
25672 link Options to pass to the linker
25673 lib Libraries to include on the command line to the linker
25674 libgcc Decides which GCC support library to pass to the linker
25675 linker Sets the name of the linker
25676 predefines Defines to be passed to the C preprocessor
25677 signed_char Defines to pass to CPP to say whether @code{char} is signed
25679 startfile Object files to include at the start of the link
25682 Here is a small example of a spec file:
25685 %rename lib old_lib
25688 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
25691 This example renames the spec called @samp{lib} to @samp{old_lib} and
25692 then overrides the previous definition of @samp{lib} with a new one.
25693 The new definition adds in some extra command-line options before
25694 including the text of the old definition.
25696 @dfn{Spec strings} are a list of command-line options to be passed to their
25697 corresponding program. In addition, the spec strings can contain
25698 @samp{%}-prefixed sequences to substitute variable text or to
25699 conditionally insert text into the command line. Using these constructs
25700 it is possible to generate quite complex command lines.
25702 Here is a table of all defined @samp{%}-sequences for spec
25703 strings. Note that spaces are not generated automatically around the
25704 results of expanding these sequences. Therefore you can concatenate them
25705 together or combine them with constant text in a single argument.
25709 Substitute one @samp{%} into the program name or argument.
25712 Substitute the name of the input file being processed.
25715 Substitute the basename of the input file being processed.
25716 This is the substring up to (and not including) the last period
25717 and not including the directory.
25720 This is the same as @samp{%b}, but include the file suffix (text after
25724 Marks the argument containing or following the @samp{%d} as a
25725 temporary file name, so that that file is deleted if GCC exits
25726 successfully. Unlike @samp{%g}, this contributes no text to the
25729 @item %g@var{suffix}
25730 Substitute a file name that has suffix @var{suffix} and is chosen
25731 once per compilation, and mark the argument in the same way as
25732 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
25733 name is now chosen in a way that is hard to predict even when previously
25734 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
25735 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
25736 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
25737 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
25738 was simply substituted with a file name chosen once per compilation,
25739 without regard to any appended suffix (which was therefore treated
25740 just like ordinary text), making such attacks more likely to succeed.
25742 @item %u@var{suffix}
25743 Like @samp{%g}, but generates a new temporary file name
25744 each time it appears instead of once per compilation.
25746 @item %U@var{suffix}
25747 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
25748 new one if there is no such last file name. In the absence of any
25749 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
25750 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
25751 involves the generation of two distinct file names, one
25752 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
25753 simply substituted with a file name chosen for the previous @samp{%u},
25754 without regard to any appended suffix.
25756 @item %j@var{suffix}
25757 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
25758 writable, and if @option{-save-temps} is not used;
25759 otherwise, substitute the name
25760 of a temporary file, just like @samp{%u}. This temporary file is not
25761 meant for communication between processes, but rather as a junk
25762 disposal mechanism.
25764 @item %|@var{suffix}
25765 @itemx %m@var{suffix}
25766 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
25767 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
25768 all. These are the two most common ways to instruct a program that it
25769 should read from standard input or write to standard output. If you
25770 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
25771 construct: see for example @file{f/lang-specs.h}.
25773 @item %.@var{SUFFIX}
25774 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
25775 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
25776 terminated by the next space or %.
25779 Marks the argument containing or following the @samp{%w} as the
25780 designated output file of this compilation. This puts the argument
25781 into the sequence of arguments that @samp{%o} substitutes.
25784 Substitutes the names of all the output files, with spaces
25785 automatically placed around them. You should write spaces
25786 around the @samp{%o} as well or the results are undefined.
25787 @samp{%o} is for use in the specs for running the linker.
25788 Input files whose names have no recognized suffix are not compiled
25789 at all, but they are included among the output files, so they are
25793 Substitutes the suffix for object files. Note that this is
25794 handled specially when it immediately follows @samp{%g, %u, or %U},
25795 because of the need for those to form complete file names. The
25796 handling is such that @samp{%O} is treated exactly as if it had already
25797 been substituted, except that @samp{%g, %u, and %U} do not currently
25798 support additional @var{suffix} characters following @samp{%O} as they do
25799 following, for example, @samp{.o}.
25802 Substitutes the standard macro predefinitions for the
25803 current target machine. Use this when running @command{cpp}.
25806 Like @samp{%p}, but puts @samp{__} before and after the name of each
25807 predefined macro, except for macros that start with @samp{__} or with
25808 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
25812 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
25813 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
25814 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
25815 and @option{-imultilib} as necessary.
25818 Current argument is the name of a library or startup file of some sort.
25819 Search for that file in a standard list of directories and substitute
25820 the full name found. The current working directory is included in the
25821 list of directories scanned.
25824 Current argument is the name of a linker script. Search for that file
25825 in the current list of directories to scan for libraries. If the file
25826 is located insert a @option{--script} option into the command line
25827 followed by the full path name found. If the file is not found then
25828 generate an error message. Note: the current working directory is not
25832 Print @var{str} as an error message. @var{str} is terminated by a newline.
25833 Use this when inconsistent options are detected.
25835 @item %(@var{name})
25836 Substitute the contents of spec string @var{name} at this point.
25838 @item %x@{@var{option}@}
25839 Accumulate an option for @samp{%X}.
25842 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
25846 Output the accumulated assembler options specified by @option{-Wa}.
25849 Output the accumulated preprocessor options specified by @option{-Wp}.
25852 Process the @code{asm} spec. This is used to compute the
25853 switches to be passed to the assembler.
25856 Process the @code{asm_final} spec. This is a spec string for
25857 passing switches to an assembler post-processor, if such a program is
25861 Process the @code{link} spec. This is the spec for computing the
25862 command line passed to the linker. Typically it makes use of the
25863 @samp{%L %G %S %D and %E} sequences.
25866 Dump out a @option{-L} option for each directory that GCC believes might
25867 contain startup files. If the target supports multilibs then the
25868 current multilib directory is prepended to each of these paths.
25871 Process the @code{lib} spec. This is a spec string for deciding which
25872 libraries are included on the command line to the linker.
25875 Process the @code{libgcc} spec. This is a spec string for deciding
25876 which GCC support library is included on the command line to the linker.
25879 Process the @code{startfile} spec. This is a spec for deciding which
25880 object files are the first ones passed to the linker. Typically
25881 this might be a file named @file{crt0.o}.
25884 Process the @code{endfile} spec. This is a spec string that specifies
25885 the last object files that are passed to the linker.
25888 Process the @code{cpp} spec. This is used to construct the arguments
25889 to be passed to the C preprocessor.
25892 Process the @code{cc1} spec. This is used to construct the options to be
25893 passed to the actual C compiler (@command{cc1}).
25896 Process the @code{cc1plus} spec. This is used to construct the options to be
25897 passed to the actual C++ compiler (@command{cc1plus}).
25900 Substitute the variable part of a matched option. See below.
25901 Note that each comma in the substituted string is replaced by
25905 Remove all occurrences of @code{-S} from the command line. Note---this
25906 command is position dependent. @samp{%} commands in the spec string
25907 before this one see @code{-S}, @samp{%} commands in the spec string
25908 after this one do not.
25910 @item %:@var{function}(@var{args})
25911 Call the named function @var{function}, passing it @var{args}.
25912 @var{args} is first processed as a nested spec string, then split
25913 into an argument vector in the usual fashion. The function returns
25914 a string which is processed as if it had appeared literally as part
25915 of the current spec.
25917 The following built-in spec functions are provided:
25920 @item @code{getenv}
25921 The @code{getenv} spec function takes two arguments: an environment
25922 variable name and a string. If the environment variable is not
25923 defined, a fatal error is issued. Otherwise, the return value is the
25924 value of the environment variable concatenated with the string. For
25925 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
25928 %:getenv(TOPDIR /include)
25931 expands to @file{/path/to/top/include}.
25933 @item @code{if-exists}
25934 The @code{if-exists} spec function takes one argument, an absolute
25935 pathname to a file. If the file exists, @code{if-exists} returns the
25936 pathname. Here is a small example of its usage:
25940 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
25943 @item @code{if-exists-else}
25944 The @code{if-exists-else} spec function is similar to the @code{if-exists}
25945 spec function, except that it takes two arguments. The first argument is
25946 an absolute pathname to a file. If the file exists, @code{if-exists-else}
25947 returns the pathname. If it does not exist, it returns the second argument.
25948 This way, @code{if-exists-else} can be used to select one file or another,
25949 based on the existence of the first. Here is a small example of its usage:
25953 crt0%O%s %:if-exists(crti%O%s) \
25954 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
25957 @item @code{replace-outfile}
25958 The @code{replace-outfile} spec function takes two arguments. It looks for the
25959 first argument in the outfiles array and replaces it with the second argument. Here
25960 is a small example of its usage:
25963 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
25966 @item @code{remove-outfile}
25967 The @code{remove-outfile} spec function takes one argument. It looks for the
25968 first argument in the outfiles array and removes it. Here is a small example
25972 %:remove-outfile(-lm)
25975 @item @code{pass-through-libs}
25976 The @code{pass-through-libs} spec function takes any number of arguments. It
25977 finds any @option{-l} options and any non-options ending in @file{.a} (which it
25978 assumes are the names of linker input library archive files) and returns a
25979 result containing all the found arguments each prepended by
25980 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
25981 intended to be passed to the LTO linker plugin.
25984 %:pass-through-libs(%G %L %G)
25987 @item @code{print-asm-header}
25988 The @code{print-asm-header} function takes no arguments and simply
25989 prints a banner like:
25995 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
25998 It is used to separate compiler options from assembler options
25999 in the @option{--target-help} output.
26002 @item %@{@code{S}@}
26003 Substitutes the @code{-S} switch, if that switch is given to GCC@.
26004 If that switch is not specified, this substitutes nothing. Note that
26005 the leading dash is omitted when specifying this option, and it is
26006 automatically inserted if the substitution is performed. Thus the spec
26007 string @samp{%@{foo@}} matches the command-line option @option{-foo}
26008 and outputs the command-line option @option{-foo}.
26010 @item %W@{@code{S}@}
26011 Like %@{@code{S}@} but mark last argument supplied within as a file to be
26012 deleted on failure.
26014 @item %@{@code{S}*@}
26015 Substitutes all the switches specified to GCC whose names start
26016 with @code{-S}, but which also take an argument. This is used for
26017 switches like @option{-o}, @option{-D}, @option{-I}, etc.
26018 GCC considers @option{-o foo} as being
26019 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
26020 text, including the space. Thus two arguments are generated.
26022 @item %@{@code{S}*&@code{T}*@}
26023 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
26024 (the order of @code{S} and @code{T} in the spec is not significant).
26025 There can be any number of ampersand-separated variables; for each the
26026 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
26028 @item %@{@code{S}:@code{X}@}
26029 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
26031 @item %@{!@code{S}:@code{X}@}
26032 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
26034 @item %@{@code{S}*:@code{X}@}
26035 Substitutes @code{X} if one or more switches whose names start with
26036 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
26037 once, no matter how many such switches appeared. However, if @code{%*}
26038 appears somewhere in @code{X}, then @code{X} is substituted once
26039 for each matching switch, with the @code{%*} replaced by the part of
26040 that switch matching the @code{*}.
26042 If @code{%*} appears as the last part of a spec sequence then a space
26043 is added after the end of the last substitution. If there is more
26044 text in the sequence, however, then a space is not generated. This
26045 allows the @code{%*} substitution to be used as part of a larger
26046 string. For example, a spec string like this:
26049 %@{mcu=*:--script=%*/memory.ld@}
26053 when matching an option like @option{-mcu=newchip} produces:
26056 --script=newchip/memory.ld
26059 @item %@{.@code{S}:@code{X}@}
26060 Substitutes @code{X}, if processing a file with suffix @code{S}.
26062 @item %@{!.@code{S}:@code{X}@}
26063 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
26065 @item %@{,@code{S}:@code{X}@}
26066 Substitutes @code{X}, if processing a file for language @code{S}.
26068 @item %@{!,@code{S}:@code{X}@}
26069 Substitutes @code{X}, if not processing a file for language @code{S}.
26071 @item %@{@code{S}|@code{P}:@code{X}@}
26072 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
26073 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
26074 @code{*} sequences as well, although they have a stronger binding than
26075 the @samp{|}. If @code{%*} appears in @code{X}, all of the
26076 alternatives must be starred, and only the first matching alternative
26079 For example, a spec string like this:
26082 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
26086 outputs the following command-line options from the following input
26087 command-line options:
26092 -d fred.c -foo -baz -boggle
26093 -d jim.d -bar -baz -boggle
26096 @item %@{S:X; T:Y; :D@}
26098 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
26099 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
26100 be as many clauses as you need. This may be combined with @code{.},
26101 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
26106 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
26107 construct may contain other nested @samp{%} constructs or spaces, or
26108 even newlines. They are processed as usual, as described above.
26109 Trailing white space in @code{X} is ignored. White space may also
26110 appear anywhere on the left side of the colon in these constructs,
26111 except between @code{.} or @code{*} and the corresponding word.
26113 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
26114 handled specifically in these constructs. If another value of
26115 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
26116 @option{-W} switch is found later in the command line, the earlier
26117 switch value is ignored, except with @{@code{S}*@} where @code{S} is
26118 just one letter, which passes all matching options.
26120 The character @samp{|} at the beginning of the predicate text is used to
26121 indicate that a command should be piped to the following command, but
26122 only if @option{-pipe} is specified.
26124 It is built into GCC which switches take arguments and which do not.
26125 (You might think it would be useful to generalize this to allow each
26126 compiler's spec to say which switches take arguments. But this cannot
26127 be done in a consistent fashion. GCC cannot even decide which input
26128 files have been specified without knowing which switches take arguments,
26129 and it must know which input files to compile in order to tell which
26132 GCC also knows implicitly that arguments starting in @option{-l} are to be
26133 treated as compiler output files, and passed to the linker in their
26134 proper position among the other output files.
26136 @node Environment Variables
26137 @section Environment Variables Affecting GCC
26138 @cindex environment variables
26140 @c man begin ENVIRONMENT
26141 This section describes several environment variables that affect how GCC
26142 operates. Some of them work by specifying directories or prefixes to use
26143 when searching for various kinds of files. Some are used to specify other
26144 aspects of the compilation environment.
26146 Note that you can also specify places to search using options such as
26147 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
26148 take precedence over places specified using environment variables, which
26149 in turn take precedence over those specified by the configuration of GCC@.
26150 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
26151 GNU Compiler Collection (GCC) Internals}.
26156 @c @itemx LC_COLLATE
26158 @c @itemx LC_MONETARY
26159 @c @itemx LC_NUMERIC
26164 @c @findex LC_COLLATE
26165 @findex LC_MESSAGES
26166 @c @findex LC_MONETARY
26167 @c @findex LC_NUMERIC
26171 These environment variables control the way that GCC uses
26172 localization information which allows GCC to work with different
26173 national conventions. GCC inspects the locale categories
26174 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
26175 so. These locale categories can be set to any value supported by your
26176 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
26177 Kingdom encoded in UTF-8.
26179 The @env{LC_CTYPE} environment variable specifies character
26180 classification. GCC uses it to determine the character boundaries in
26181 a string; this is needed for some multibyte encodings that contain quote
26182 and escape characters that are otherwise interpreted as a string
26185 The @env{LC_MESSAGES} environment variable specifies the language to
26186 use in diagnostic messages.
26188 If the @env{LC_ALL} environment variable is set, it overrides the value
26189 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
26190 and @env{LC_MESSAGES} default to the value of the @env{LANG}
26191 environment variable. If none of these variables are set, GCC
26192 defaults to traditional C English behavior.
26196 If @env{TMPDIR} is set, it specifies the directory to use for temporary
26197 files. GCC uses temporary files to hold the output of one stage of
26198 compilation which is to be used as input to the next stage: for example,
26199 the output of the preprocessor, which is the input to the compiler
26202 @item GCC_COMPARE_DEBUG
26203 @findex GCC_COMPARE_DEBUG
26204 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
26205 @option{-fcompare-debug} to the compiler driver. See the documentation
26206 of this option for more details.
26208 @item GCC_EXEC_PREFIX
26209 @findex GCC_EXEC_PREFIX
26210 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
26211 names of the subprograms executed by the compiler. No slash is added
26212 when this prefix is combined with the name of a subprogram, but you can
26213 specify a prefix that ends with a slash if you wish.
26215 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
26216 an appropriate prefix to use based on the pathname it is invoked with.
26218 If GCC cannot find the subprogram using the specified prefix, it
26219 tries looking in the usual places for the subprogram.
26221 The default value of @env{GCC_EXEC_PREFIX} is
26222 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
26223 the installed compiler. In many cases @var{prefix} is the value
26224 of @code{prefix} when you ran the @file{configure} script.
26226 Other prefixes specified with @option{-B} take precedence over this prefix.
26228 This prefix is also used for finding files such as @file{crt0.o} that are
26231 In addition, the prefix is used in an unusual way in finding the
26232 directories to search for header files. For each of the standard
26233 directories whose name normally begins with @samp{/usr/local/lib/gcc}
26234 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
26235 replacing that beginning with the specified prefix to produce an
26236 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
26237 @file{foo/bar} just before it searches the standard directory
26238 @file{/usr/local/lib/bar}.
26239 If a standard directory begins with the configured
26240 @var{prefix} then the value of @var{prefix} is replaced by
26241 @env{GCC_EXEC_PREFIX} when looking for header files.
26243 @item COMPILER_PATH
26244 @findex COMPILER_PATH
26245 The value of @env{COMPILER_PATH} is a colon-separated list of
26246 directories, much like @env{PATH}. GCC tries the directories thus
26247 specified when searching for subprograms, if it can't find the
26248 subprograms using @env{GCC_EXEC_PREFIX}.
26251 @findex LIBRARY_PATH
26252 The value of @env{LIBRARY_PATH} is a colon-separated list of
26253 directories, much like @env{PATH}. When configured as a native compiler,
26254 GCC tries the directories thus specified when searching for special
26255 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
26256 using GCC also uses these directories when searching for ordinary
26257 libraries for the @option{-l} option (but directories specified with
26258 @option{-L} come first).
26262 @cindex locale definition
26263 This variable is used to pass locale information to the compiler. One way in
26264 which this information is used is to determine the character set to be used
26265 when character literals, string literals and comments are parsed in C and C++.
26266 When the compiler is configured to allow multibyte characters,
26267 the following values for @env{LANG} are recognized:
26271 Recognize JIS characters.
26273 Recognize SJIS characters.
26275 Recognize EUCJP characters.
26278 If @env{LANG} is not defined, or if it has some other value, then the
26279 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
26280 recognize and translate multibyte characters.
26284 Some additional environment variables affect the behavior of the
26287 @include cppenv.texi
26291 @node Precompiled Headers
26292 @section Using Precompiled Headers
26293 @cindex precompiled headers
26294 @cindex speed of compilation
26296 Often large projects have many header files that are included in every
26297 source file. The time the compiler takes to process these header files
26298 over and over again can account for nearly all of the time required to
26299 build the project. To make builds faster, GCC allows you to
26300 @dfn{precompile} a header file.
26302 To create a precompiled header file, simply compile it as you would any
26303 other file, if necessary using the @option{-x} option to make the driver
26304 treat it as a C or C++ header file. You may want to use a
26305 tool like @command{make} to keep the precompiled header up-to-date when
26306 the headers it contains change.
26308 A precompiled header file is searched for when @code{#include} is
26309 seen in the compilation. As it searches for the included file
26310 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
26311 compiler looks for a precompiled header in each directory just before it
26312 looks for the include file in that directory. The name searched for is
26313 the name specified in the @code{#include} with @samp{.gch} appended. If
26314 the precompiled header file can't be used, it is ignored.
26316 For instance, if you have @code{#include "all.h"}, and you have
26317 @file{all.h.gch} in the same directory as @file{all.h}, then the
26318 precompiled header file is used if possible, and the original
26319 header is used otherwise.
26321 Alternatively, you might decide to put the precompiled header file in a
26322 directory and use @option{-I} to ensure that directory is searched
26323 before (or instead of) the directory containing the original header.
26324 Then, if you want to check that the precompiled header file is always
26325 used, you can put a file of the same name as the original header in this
26326 directory containing an @code{#error} command.
26328 This also works with @option{-include}. So yet another way to use
26329 precompiled headers, good for projects not designed with precompiled
26330 header files in mind, is to simply take most of the header files used by
26331 a project, include them from another header file, precompile that header
26332 file, and @option{-include} the precompiled header. If the header files
26333 have guards against multiple inclusion, they are skipped because
26334 they've already been included (in the precompiled header).
26336 If you need to precompile the same header file for different
26337 languages, targets, or compiler options, you can instead make a
26338 @emph{directory} named like @file{all.h.gch}, and put each precompiled
26339 header in the directory, perhaps using @option{-o}. It doesn't matter
26340 what you call the files in the directory; every precompiled header in
26341 the directory is considered. The first precompiled header
26342 encountered in the directory that is valid for this compilation is
26343 used; they're searched in no particular order.
26345 There are many other possibilities, limited only by your imagination,
26346 good sense, and the constraints of your build system.
26348 A precompiled header file can be used only when these conditions apply:
26352 Only one precompiled header can be used in a particular compilation.
26355 A precompiled header can't be used once the first C token is seen. You
26356 can have preprocessor directives before a precompiled header; you cannot
26357 include a precompiled header from inside another header.
26360 The precompiled header file must be produced for the same language as
26361 the current compilation. You can't use a C precompiled header for a C++
26365 The precompiled header file must have been produced by the same compiler
26366 binary as the current compilation is using.
26369 Any macros defined before the precompiled header is included must
26370 either be defined in the same way as when the precompiled header was
26371 generated, or must not affect the precompiled header, which usually
26372 means that they don't appear in the precompiled header at all.
26374 The @option{-D} option is one way to define a macro before a
26375 precompiled header is included; using a @code{#define} can also do it.
26376 There are also some options that define macros implicitly, like
26377 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
26380 @item If debugging information is output when using the precompiled
26381 header, using @option{-g} or similar, the same kind of debugging information
26382 must have been output when building the precompiled header. However,
26383 a precompiled header built using @option{-g} can be used in a compilation
26384 when no debugging information is being output.
26386 @item The same @option{-m} options must generally be used when building
26387 and using the precompiled header. @xref{Submodel Options},
26388 for any cases where this rule is relaxed.
26390 @item Each of the following options must be the same when building and using
26391 the precompiled header:
26393 @gccoptlist{-fexceptions}
26396 Some other command-line options starting with @option{-f},
26397 @option{-p}, or @option{-O} must be defined in the same way as when
26398 the precompiled header was generated. At present, it's not clear
26399 which options are safe to change and which are not; the safest choice
26400 is to use exactly the same options when generating and using the
26401 precompiled header. The following are known to be safe:
26403 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
26404 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
26405 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
26410 For all of these except the last, the compiler automatically
26411 ignores the precompiled header if the conditions aren't met. If you
26412 find an option combination that doesn't work and doesn't cause the
26413 precompiled header to be ignored, please consider filing a bug report,
26416 If you do use differing options when generating and using the
26417 precompiled header, the actual behavior is a mixture of the
26418 behavior for the options. For instance, if you use @option{-g} to
26419 generate the precompiled header but not when using it, you may or may
26420 not get debugging information for routines in the precompiled header.