1 @c Copyright (C) 1988-2017 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2017 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
176 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
178 @item C Language Options
179 @xref{C Dialect Options,,Options Controlling C Dialect}.
180 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
181 -fpermitted-flt-eval-methods=@var{standard} @gol
182 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
183 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
184 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
185 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
186 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
187 -fsigned-bitfields -fsigned-char @gol
188 -funsigned-bitfields -funsigned-char}
190 @item C++ Language Options
191 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
192 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
193 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
194 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
195 -ffriend-injection @gol
196 -fno-elide-constructors @gol
197 -fno-enforce-eh-specs @gol
198 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
199 -fno-implicit-templates @gol
200 -fno-implicit-inline-templates @gol
201 -fno-implement-inlines -fms-extensions @gol
202 -fnew-inheriting-ctors @gol
203 -fnew-ttp-matching @gol
204 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
205 -fno-optional-diags -fpermissive @gol
206 -fno-pretty-templates @gol
207 -frepo -fno-rtti -fsized-deallocation @gol
208 -ftemplate-backtrace-limit=@var{n} @gol
209 -ftemplate-depth=@var{n} @gol
210 -fno-threadsafe-statics -fuse-cxa-atexit @gol
211 -fno-weak -nostdinc++ @gol
212 -fvisibility-inlines-hidden @gol
213 -fvisibility-ms-compat @gol
214 -fext-numeric-literals @gol
215 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
216 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
217 -Wnamespaces -Wnarrowing @gol
218 -Wnoexcept -Wnon-virtual-dtor -Wreorder -Wregister @gol
219 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
220 -Wno-non-template-friend -Wold-style-cast @gol
221 -Woverloaded-virtual -Wno-pmf-conversions @gol
222 -Wsign-promo -Wvirtual-inheritance}
224 @item Objective-C and Objective-C++ Language Options
225 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
226 Objective-C and Objective-C++ Dialects}.
227 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
228 -fgnu-runtime -fnext-runtime @gol
229 -fno-nil-receivers @gol
230 -fobjc-abi-version=@var{n} @gol
231 -fobjc-call-cxx-cdtors @gol
232 -fobjc-direct-dispatch @gol
233 -fobjc-exceptions @gol
236 -fobjc-std=objc1 @gol
237 -fno-local-ivars @gol
238 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
239 -freplace-objc-classes @gol
242 -Wassign-intercept @gol
243 -Wno-protocol -Wselector @gol
244 -Wstrict-selector-match @gol
245 -Wundeclared-selector}
247 @item Diagnostic Message Formatting Options
248 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
249 @gccoptlist{-fmessage-length=@var{n} @gol
250 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
251 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
252 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
253 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
256 @item Warning Options
257 @xref{Warning Options,,Options to Request or Suppress Warnings}.
258 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
259 -pedantic-errors @gol
260 -w -Wextra -Wall -Waddress -Waggregate-return @gol
261 -Walloc-zero -Walloc-size-larger-than=@var{n}
262 -Walloca -Walloca-larger-than=@var{n} @gol
263 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
264 -Wno-attributes -Wbool-compare -Wbool-operation @gol
265 -Wno-builtin-declaration-mismatch @gol
266 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
267 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
268 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
269 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
270 -Wdelete-incomplete @gol
271 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
272 -Wdisabled-optimization @gol
273 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
274 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
275 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
276 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
277 -Wno-format-contains-nul -Wno-format-extra-args @gol
278 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
279 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
280 -Wformat-y2k -Wframe-address @gol
281 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
282 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
283 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
284 -Wimplicit-function-declaration -Wimplicit-int @gol
285 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
286 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
287 -Winvalid-pch -Wlarger-than=@var{len} @gol
288 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
289 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
290 -Wmisleading-indentation -Wmissing-braces @gol
291 -Wmissing-field-initializers -Wmissing-include-dirs @gol
292 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
293 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
294 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
295 -Woverride-init-side-effects -Woverlength-strings @gol
296 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
297 -Wparentheses -Wno-pedantic-ms-format @gol
298 -Wplacement-new -Wplacement-new=@var{n} @gol
299 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
300 -Wno-pragmas -Wredundant-decls -Wrestrict -Wno-return-local-addr @gol
301 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
302 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
303 -Wshift-overflow -Wshift-overflow=@var{n} @gol
304 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
305 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
306 -Wno-scalar-storage-order @gol
307 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
308 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
309 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
310 -Wstringop-overflow=@var{n} @gol
311 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
312 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
313 -Wmissing-format-attribute -Wsubobject-linkage @gol
314 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
315 -Wswitch-unreachable -Wsync-nand @gol
316 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
317 -Wtype-limits -Wundef @gol
318 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
319 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
320 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
321 -Wunused-parameter -Wno-unused-result @gol
322 -Wunused-value -Wunused-variable @gol
323 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
324 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
325 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
326 -Wvla -Wvla-larger-than=@var{n} -Wvolatile-register-var -Wwrite-strings @gol
327 -Wzero-as-null-pointer-constant -Whsa}
329 @item C and Objective-C-only Warning Options
330 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
331 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
332 -Wold-style-declaration -Wold-style-definition @gol
333 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
334 -Wdeclaration-after-statement -Wpointer-sign}
336 @item Debugging Options
337 @xref{Debugging Options,,Options for Debugging Your Program}.
338 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
339 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
340 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
341 -gcolumn-info -gno-column-info @gol
342 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
343 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
344 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
345 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
346 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
347 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
348 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
349 -fvar-tracking -fvar-tracking-assignments}
351 @item Optimization Options
352 @xref{Optimize Options,,Options that Control Optimization}.
353 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
354 -falign-jumps[=@var{n}] @gol
355 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
356 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
357 -fauto-inc-dec -fbranch-probabilities @gol
358 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
359 -fbtr-bb-exclusive -fcaller-saves @gol
360 -fcombine-stack-adjustments -fconserve-stack @gol
361 -fcompare-elim -fcprop-registers -fcrossjumping @gol
362 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
363 -fcx-limited-range @gol
364 -fdata-sections -fdce -fdelayed-branch @gol
365 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
366 -fdevirtualize-at-ltrans -fdse @gol
367 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
368 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
369 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
370 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
371 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
372 -fif-conversion2 -findirect-inlining @gol
373 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
374 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
375 -fipa-cp-alignment -fipa-bit-cp @gol
376 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
377 -fira-algorithm=@var{algorithm} @gol
378 -fira-region=@var{region} -fira-hoist-pressure @gol
379 -fira-loop-pressure -fno-ira-share-save-slots @gol
380 -fno-ira-share-spill-slots @gol
381 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
382 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
383 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
384 -floop-block -floop-interchange -floop-strip-mine @gol
385 -floop-unroll-and-jam -floop-nest-optimize @gol
386 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
387 -flto-partition=@var{alg} -fmerge-all-constants @gol
388 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
389 -fmove-loop-invariants -fno-branch-count-reg @gol
390 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
391 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
392 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
393 -fno-sched-spec -fno-signed-zeros @gol
394 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
395 -fomit-frame-pointer -foptimize-sibling-calls @gol
396 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
397 -fprefetch-loop-arrays @gol
398 -fprofile-correction @gol
399 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
400 -fprofile-reorder-functions @gol
401 -freciprocal-math -free -frename-registers -freorder-blocks @gol
402 -freorder-blocks-algorithm=@var{algorithm} @gol
403 -freorder-blocks-and-partition -freorder-functions @gol
404 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
405 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
406 -fsched-spec-load -fsched-spec-load-dangerous @gol
407 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
408 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
409 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
410 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
411 -fschedule-fusion @gol
412 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
413 -fselective-scheduling -fselective-scheduling2 @gol
414 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
415 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
416 -fsignaling-nans @gol
417 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
419 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
420 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
421 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
422 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
423 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
424 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
425 -ftree-loop-if-convert -ftree-loop-im @gol
426 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
427 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
428 -ftree-loop-vectorize @gol
429 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
430 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
431 -ftree-switch-conversion -ftree-tail-merge @gol
432 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
433 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
434 -funsafe-math-optimizations -funswitch-loops @gol
435 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
436 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
437 --param @var{name}=@var{value}
438 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
440 @item Program Instrumentation Options
441 @xref{Instrumentation Options,,Program Instrumentation Options}.
442 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
443 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
444 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
445 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
446 -fsanitize-undefined-trap-on-error -fbounds-check @gol
447 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
448 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
449 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
450 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
451 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
452 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
453 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
454 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
455 -fchkp-use-wrappers -fchkp-flexible-struct-trailing-arrays@gol
456 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
457 -fstack-protector-explicit -fstack-check @gol
458 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
459 -fno-stack-limit -fsplit-stack @gol
460 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
461 -fvtv-counts -fvtv-debug @gol
462 -finstrument-functions @gol
463 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
464 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
466 @item Preprocessor Options
467 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
468 @gccoptlist{-A@var{question}=@var{answer} @gol
469 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
470 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
471 -dD -dI -dM -dN -dU @gol
472 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
473 -fexec-charset=@var{charset} -fextended-identifiers @gol
474 -finput-charset=@var{charset} -fno-canonical-system-headers @gol
475 -fpch-deps -fpch-preprocess -fpreprocessed @gol
476 -ftabstop=@var{width} -ftrack-macro-expansion @gol
477 -fwide-exec-charset=@var{charset} -fworking-directory @gol
478 -H -imacros @var{file} -include @var{file} @gol
479 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
480 -no-integrated-cpp -P -pthread -remap @gol
481 -traditional -traditional-cpp -trigraphs @gol
482 -U@var{macro} -undef @gol
483 -Wp,@var{option} -Xpreprocessor @var{option}}
485 @item Assembler Options
486 @xref{Assembler Options,,Passing Options to the Assembler}.
487 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
490 @xref{Link Options,,Options for Linking}.
491 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
492 -nostartfiles -nodefaultlibs -nostdlib -pie -pthread -rdynamic @gol
493 -s -static -static-libgcc -static-libstdc++ @gol
494 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
495 -static-libmpx -static-libmpxwrappers @gol
496 -shared -shared-libgcc -symbolic @gol
497 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
498 -u @var{symbol} -z @var{keyword}}
500 @item Directory Options
501 @xref{Directory Options,,Options for Directory Search}.
502 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
503 -idirafter @var{dir} @gol
504 -imacros @var{file} -imultilib @var{dir} @gol
505 -iplugindir=@var{dir} -iprefix @var{file} @gol
506 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
507 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
508 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
509 -nostdinc -nostdinc++ --sysroot=@var{dir}}
511 @item Code Generation Options
512 @xref{Code Gen Options,,Options for Code Generation Conventions}.
513 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
514 -ffixed-@var{reg} -fexceptions @gol
515 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
516 -fasynchronous-unwind-tables @gol
518 -finhibit-size-directive -fno-common -fno-ident @gol
519 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
520 -fno-jump-tables @gol
521 -frecord-gcc-switches @gol
522 -freg-struct-return -fshort-enums -fshort-wchar @gol
523 -fverbose-asm -fpack-struct[=@var{n}] @gol
524 -fleading-underscore -ftls-model=@var{model} @gol
525 -fstack-reuse=@var{reuse_level} @gol
526 -ftrampolines -ftrapv -fwrapv @gol
527 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
528 -fstrict-volatile-bitfields -fsync-libcalls}
530 @item Developer Options
531 @xref{Developer Options,,GCC Developer Options}.
532 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
533 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
534 -fdbg-cnt=@var{counter-value-list} @gol
535 -fdisable-ipa-@var{pass_name} @gol
536 -fdisable-rtl-@var{pass_name} @gol
537 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
538 -fdisable-tree-@var{pass_name} @gol
539 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
540 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
541 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
542 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
543 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
545 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
546 -fdump-statistics @gol
547 -fdump-final-insns@r{[}=@var{file}@r{]}
549 -fdump-tree-@var{switch} @gol
550 -fdump-tree-@var{switch}-@var{options} @gol
551 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
552 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
553 -fenable-@var{kind}-@var{pass} @gol
554 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
555 -fira-verbose=@var{n} @gol
556 -flto-report -flto-report-wpa -fmem-report-wpa @gol
557 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
558 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
559 -fprofile-report @gol
560 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
561 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
562 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
563 -fvar-tracking-assignments-toggle -gtoggle @gol
564 -print-file-name=@var{library} -print-libgcc-file-name @gol
565 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
566 -print-prog-name=@var{program} -print-search-dirs -Q @gol
567 -print-sysroot -print-sysroot-headers-suffix @gol
568 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
570 @item Machine-Dependent Options
571 @xref{Submodel Options,,Machine-Dependent Options}.
572 @c This list is ordered alphanumerically by subsection name.
573 @c Try and put the significant identifier (CPU or system) first,
574 @c so users have a clue at guessing where the ones they want will be.
576 @emph{AArch64 Options}
577 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
578 -mgeneral-regs-only @gol
579 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
581 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
582 -mtls-dialect=desc -mtls-dialect=traditional @gol
583 -mtls-size=@var{size} @gol
584 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
585 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
586 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
587 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
588 -mlow-precision-div -mno-low-precision-div @gol
589 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
591 @emph{Adapteva Epiphany Options}
592 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
593 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
594 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
595 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
596 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
597 -msplit-vecmove-early -m1reg-@var{reg}}
600 @gccoptlist{-mbarrel-shifter @gol
601 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
602 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
603 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
604 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
605 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
606 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
607 -mlong-calls -mmedium-calls -msdata @gol
608 -mvolatile-cache -mtp-regno=@var{regno} @gol
609 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
610 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
611 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
612 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
613 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
614 -mtune=@var{cpu} -mmultcost=@var{num} @gol
615 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
616 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
619 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
620 -mabi=@var{name} @gol
621 -mapcs-stack-check -mno-apcs-stack-check @gol
622 -mapcs-reentrant -mno-apcs-reentrant @gol
623 -msched-prolog -mno-sched-prolog @gol
624 -mlittle-endian -mbig-endian @gol
625 -mfloat-abi=@var{name} @gol
626 -mfp16-format=@var{name}
627 -mthumb-interwork -mno-thumb-interwork @gol
628 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
629 -mtune=@var{name} -mprint-tune-info @gol
630 -mstructure-size-boundary=@var{n} @gol
631 -mabort-on-noreturn @gol
632 -mlong-calls -mno-long-calls @gol
633 -msingle-pic-base -mno-single-pic-base @gol
634 -mpic-register=@var{reg} @gol
635 -mnop-fun-dllimport @gol
636 -mpoke-function-name @gol
638 -mtpcs-frame -mtpcs-leaf-frame @gol
639 -mcaller-super-interworking -mcallee-super-interworking @gol
640 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
641 -mword-relocations @gol
642 -mfix-cortex-m3-ldrd @gol
643 -munaligned-access @gol
644 -mneon-for-64bits @gol
645 -mslow-flash-data @gol
646 -masm-syntax-unified @gol
652 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
653 -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 @gol
657 -Waddr-space-convert -Wmisspelled-isr}
659 @emph{Blackfin Options}
660 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
661 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
662 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
663 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
664 -mno-id-shared-library -mshared-library-id=@var{n} @gol
665 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
666 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
667 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
671 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
672 -msim -msdata=@var{sdata-type}}
675 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
676 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
677 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
678 -mstack-align -mdata-align -mconst-align @gol
679 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
680 -melf -maout -melinux -mlinux -sim -sim2 @gol
681 -mmul-bug-workaround -mno-mul-bug-workaround}
684 @gccoptlist{-mmac @gol
685 -mcr16cplus -mcr16c @gol
686 -msim -mint32 -mbit-ops
687 -mdata-model=@var{model}}
689 @emph{Darwin Options}
690 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
691 -arch_only -bind_at_load -bundle -bundle_loader @gol
692 -client_name -compatibility_version -current_version @gol
694 -dependency-file -dylib_file -dylinker_install_name @gol
695 -dynamic -dynamiclib -exported_symbols_list @gol
696 -filelist -flat_namespace -force_cpusubtype_ALL @gol
697 -force_flat_namespace -headerpad_max_install_names @gol
699 -image_base -init -install_name -keep_private_externs @gol
700 -multi_module -multiply_defined -multiply_defined_unused @gol
701 -noall_load -no_dead_strip_inits_and_terms @gol
702 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
703 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
704 -private_bundle -read_only_relocs -sectalign @gol
705 -sectobjectsymbols -whyload -seg1addr @gol
706 -sectcreate -sectobjectsymbols -sectorder @gol
707 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
708 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
709 -segprot -segs_read_only_addr -segs_read_write_addr @gol
710 -single_module -static -sub_library -sub_umbrella @gol
711 -twolevel_namespace -umbrella -undefined @gol
712 -unexported_symbols_list -weak_reference_mismatches @gol
713 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
714 -mkernel -mone-byte-bool}
716 @emph{DEC Alpha Options}
717 @gccoptlist{-mno-fp-regs -msoft-float @gol
718 -mieee -mieee-with-inexact -mieee-conformant @gol
719 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
720 -mtrap-precision=@var{mode} -mbuild-constants @gol
721 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
722 -mbwx -mmax -mfix -mcix @gol
723 -mfloat-vax -mfloat-ieee @gol
724 -mexplicit-relocs -msmall-data -mlarge-data @gol
725 -msmall-text -mlarge-text @gol
726 -mmemory-latency=@var{time}}
729 @gccoptlist{-msmall-model -mno-lsim}
732 @gccoptlist{-msim -mlra -mnodiv}
735 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
736 -mhard-float -msoft-float @gol
737 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
738 -mdouble -mno-double @gol
739 -mmedia -mno-media -mmuladd -mno-muladd @gol
740 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
741 -mlinked-fp -mlong-calls -malign-labels @gol
742 -mlibrary-pic -macc-4 -macc-8 @gol
743 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
744 -moptimize-membar -mno-optimize-membar @gol
745 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
746 -mvliw-branch -mno-vliw-branch @gol
747 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
748 -mno-nested-cond-exec -mtomcat-stats @gol
752 @emph{GNU/Linux Options}
753 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
754 -tno-android-cc -tno-android-ld}
756 @emph{H8/300 Options}
757 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
760 @gccoptlist{-march=@var{architecture-type} @gol
761 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
762 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
763 -mfixed-range=@var{register-range} @gol
764 -mjump-in-delay -mlinker-opt -mlong-calls @gol
765 -mlong-load-store -mno-disable-fpregs @gol
766 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
767 -mno-jump-in-delay -mno-long-load-store @gol
768 -mno-portable-runtime -mno-soft-float @gol
769 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
770 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
771 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
772 -munix=@var{unix-std} -nolibdld -static -threads}
775 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
776 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
777 -mconstant-gp -mauto-pic -mfused-madd @gol
778 -minline-float-divide-min-latency @gol
779 -minline-float-divide-max-throughput @gol
780 -mno-inline-float-divide @gol
781 -minline-int-divide-min-latency @gol
782 -minline-int-divide-max-throughput @gol
783 -mno-inline-int-divide @gol
784 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
785 -mno-inline-sqrt @gol
786 -mdwarf2-asm -mearly-stop-bits @gol
787 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
788 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
789 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
790 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
791 -msched-spec-ldc -msched-spec-control-ldc @gol
792 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
793 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
794 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
795 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
798 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
799 -msign-extend-enabled -muser-enabled}
801 @emph{M32R/D Options}
802 @gccoptlist{-m32r2 -m32rx -m32r @gol
804 -malign-loops -mno-align-loops @gol
805 -missue-rate=@var{number} @gol
806 -mbranch-cost=@var{number} @gol
807 -mmodel=@var{code-size-model-type} @gol
808 -msdata=@var{sdata-type} @gol
809 -mno-flush-func -mflush-func=@var{name} @gol
810 -mno-flush-trap -mflush-trap=@var{number} @gol
814 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
816 @emph{M680x0 Options}
817 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
818 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
819 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
820 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
821 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
822 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
823 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
824 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
825 -mxgot -mno-xgot -mlong-jump-table-offsets}
828 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
829 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
830 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
831 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
832 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
835 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
836 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
837 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
838 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
841 @emph{MicroBlaze Options}
842 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
843 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
844 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
845 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
846 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
849 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
850 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
851 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
852 -mips16 -mno-mips16 -mflip-mips16 @gol
853 -minterlink-compressed -mno-interlink-compressed @gol
854 -minterlink-mips16 -mno-interlink-mips16 @gol
855 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
856 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
857 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
858 -mno-float -msingle-float -mdouble-float @gol
859 -modd-spreg -mno-odd-spreg @gol
860 -mabs=@var{mode} -mnan=@var{encoding} @gol
861 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
864 -mvirt -mno-virt @gol
866 -mmicromips -mno-micromips @gol
868 -mfpu=@var{fpu-type} @gol
869 -msmartmips -mno-smartmips @gol
870 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
871 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
872 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
873 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
874 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
875 -membedded-data -mno-embedded-data @gol
876 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
877 -mcode-readable=@var{setting} @gol
878 -msplit-addresses -mno-split-addresses @gol
879 -mexplicit-relocs -mno-explicit-relocs @gol
880 -mcheck-zero-division -mno-check-zero-division @gol
881 -mdivide-traps -mdivide-breaks @gol
882 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
883 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
884 -mfix-24k -mno-fix-24k @gol
885 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
886 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
887 -mfix-vr4120 -mno-fix-vr4120 @gol
888 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
889 -mflush-func=@var{func} -mno-flush-func @gol
890 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
891 -mcompact-branches=@var{policy} @gol
892 -mfp-exceptions -mno-fp-exceptions @gol
893 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
894 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
895 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
896 -mframe-header-opt -mno-frame-header-opt}
899 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
900 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
901 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
902 -mno-base-addresses -msingle-exit -mno-single-exit}
904 @emph{MN10300 Options}
905 @gccoptlist{-mmult-bug -mno-mult-bug @gol
906 -mno-am33 -mam33 -mam33-2 -mam34 @gol
907 -mtune=@var{cpu-type} @gol
908 -mreturn-pointer-on-d0 @gol
909 -mno-crt0 -mrelax -mliw -msetlb}
912 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
914 @emph{MSP430 Options}
915 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
917 -mcode-region= -mdata-region= @gol
918 -msilicon-errata= -msilicon-errata-warn= @gol
922 @gccoptlist{-mbig-endian -mlittle-endian @gol
923 -mreduced-regs -mfull-regs @gol
924 -mcmov -mno-cmov @gol
925 -mperf-ext -mno-perf-ext @gol
926 -mv3push -mno-v3push @gol
927 -m16bit -mno-16bit @gol
928 -misr-vector-size=@var{num} @gol
929 -mcache-block-size=@var{num} @gol
930 -march=@var{arch} @gol
931 -mcmodel=@var{code-model} @gol
934 @emph{Nios II Options}
935 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
937 -mno-bypass-cache -mbypass-cache @gol
938 -mno-cache-volatile -mcache-volatile @gol
939 -mno-fast-sw-div -mfast-sw-div @gol
940 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
941 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
942 -mcustom-fpu-cfg=@var{name} @gol
943 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
944 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
946 @emph{Nvidia PTX Options}
947 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
949 @emph{PDP-11 Options}
950 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
951 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
952 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
953 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
954 -mbranch-expensive -mbranch-cheap @gol
955 -munix-asm -mdec-asm}
957 @emph{picoChip Options}
958 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
959 -msymbol-as-address -mno-inefficient-warnings}
961 @emph{PowerPC Options}
962 See RS/6000 and PowerPC Options.
964 @emph{RISC-V Options}
965 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
966 -mmemcpy -mno-memcpy @gol
968 -mabi=@var{ABI-string} @gol
969 -mfdiv -mno-fdiv @gol
971 -march=@var{ISA-string} @gol
972 -mtune=@var{processor-string} @gol
973 -msmall-data-limit=@var{N-bytes} @gol
974 -msave-restore -mno-save-restore @gol
975 -mcmodel=@var{code-model} @gol
976 -mexplicit-relocs -mno-explicit-relocs @gol}
979 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
980 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
981 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
983 @emph{RS/6000 and PowerPC Options}
984 @gccoptlist{-mcpu=@var{cpu-type} @gol
985 -mtune=@var{cpu-type} @gol
986 -mcmodel=@var{code-model} @gol
988 -maltivec -mno-altivec @gol
989 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
990 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
991 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
992 -mfprnd -mno-fprnd @gol
993 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
994 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
995 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
996 -malign-power -malign-natural @gol
997 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
998 -msingle-float -mdouble-float -msimple-fpu @gol
999 -mstring -mno-string -mupdate -mno-update @gol
1000 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1001 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1002 -mstrict-align -mno-strict-align -mrelocatable @gol
1003 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1004 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1005 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1006 -mprioritize-restricted-insns=@var{priority} @gol
1007 -msched-costly-dep=@var{dependence_type} @gol
1008 -minsert-sched-nops=@var{scheme} @gol
1009 -mcall-sysv -mcall-netbsd @gol
1010 -maix-struct-return -msvr4-struct-return @gol
1011 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1012 -mblock-move-inline-limit=@var{num} @gol
1013 -misel -mno-isel @gol
1014 -misel=yes -misel=no @gol
1016 -mspe=yes -mspe=no @gol
1018 -mgen-cell-microcode -mwarn-cell-microcode @gol
1019 -mvrsave -mno-vrsave @gol
1020 -mmulhw -mno-mulhw @gol
1021 -mdlmzb -mno-dlmzb @gol
1022 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1023 -mprototype -mno-prototype @gol
1024 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1025 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1026 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1027 -mno-recip-precision @gol
1028 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1029 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1030 -msave-toc-indirect -mno-save-toc-indirect @gol
1031 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1032 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1033 -mquad-memory -mno-quad-memory @gol
1034 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1035 -mcompat-align-parm -mno-compat-align-parm @gol
1036 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1037 -mupper-regs-di -mno-upper-regs-di @gol
1038 -mupper-regs -mno-upper-regs @gol
1039 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1040 -mgnu-attribute -mno-gnu-attribute @gol
1041 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1042 -mstack-protector-guard-offset=@var{offset} @gol
1046 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1048 -mbig-endian-data -mlittle-endian-data @gol
1051 -mas100-syntax -mno-as100-syntax@gol
1053 -mmax-constant-size=@gol
1056 -mallow-string-insns -mno-allow-string-insns@gol
1058 -mno-warn-multiple-fast-interrupts@gol
1059 -msave-acc-in-interrupts}
1061 @emph{S/390 and zSeries Options}
1062 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1063 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1064 -mlong-double-64 -mlong-double-128 @gol
1065 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1066 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1067 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1068 -mhtm -mvx -mzvector @gol
1069 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1070 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1071 -mhotpatch=@var{halfwords},@var{halfwords}}
1073 @emph{Score Options}
1074 @gccoptlist{-meb -mel @gol
1078 -mscore5 -mscore5u -mscore7 -mscore7d}
1081 @gccoptlist{-m1 -m2 -m2e @gol
1082 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1084 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1085 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1086 -mb -ml -mdalign -mrelax @gol
1087 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1088 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1089 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1090 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1091 -maccumulate-outgoing-args @gol
1092 -matomic-model=@var{atomic-model} @gol
1093 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1094 -mcbranch-force-delay-slot @gol
1095 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1096 -mpretend-cmove -mtas}
1098 @emph{Solaris 2 Options}
1099 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1102 @emph{SPARC Options}
1103 @gccoptlist{-mcpu=@var{cpu-type} @gol
1104 -mtune=@var{cpu-type} @gol
1105 -mcmodel=@var{code-model} @gol
1106 -mmemory-model=@var{mem-model} @gol
1107 -m32 -m64 -mapp-regs -mno-app-regs @gol
1108 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1109 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1110 -mhard-quad-float -msoft-quad-float @gol
1111 -mstack-bias -mno-stack-bias @gol
1112 -mstd-struct-return -mno-std-struct-return @gol
1113 -munaligned-doubles -mno-unaligned-doubles @gol
1114 -muser-mode -mno-user-mode @gol
1115 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1116 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1117 -mcbcond -mno-cbcond -mfmaf -mno-fmaf @gol
1118 -mpopc -mno-popc -msubxc -mno-subxc@gol
1119 -mfix-at697f -mfix-ut699 @gol
1123 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1124 -msafe-dma -munsafe-dma @gol
1126 -msmall-mem -mlarge-mem -mstdmain @gol
1127 -mfixed-range=@var{register-range} @gol
1129 -maddress-space-conversion -mno-address-space-conversion @gol
1130 -mcache-size=@var{cache-size} @gol
1131 -matomic-updates -mno-atomic-updates}
1133 @emph{System V Options}
1134 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1136 @emph{TILE-Gx Options}
1137 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1138 -mcmodel=@var{code-model}}
1140 @emph{TILEPro Options}
1141 @gccoptlist{-mcpu=@var{cpu} -m32}
1144 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1145 -mprolog-function -mno-prolog-function -mspace @gol
1146 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1147 -mapp-regs -mno-app-regs @gol
1148 -mdisable-callt -mno-disable-callt @gol
1149 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1150 -mv850e -mv850 -mv850e3v5 @gol
1161 @gccoptlist{-mg -mgnu -munix}
1163 @emph{Visium Options}
1164 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1165 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1168 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1169 -mpointer-size=@var{size}}
1171 @emph{VxWorks Options}
1172 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1173 -Xbind-lazy -Xbind-now}
1176 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1177 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1178 -mfpmath=@var{unit} @gol
1179 -masm=@var{dialect} -mno-fancy-math-387 @gol
1180 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1181 -mno-wide-multiply -mrtd -malign-double @gol
1182 -mpreferred-stack-boundary=@var{num} @gol
1183 -mincoming-stack-boundary=@var{num} @gol
1184 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1185 -mrecip -mrecip=@var{opt} @gol
1186 -mvzeroupper -mprefer-avx128 @gol
1187 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1188 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1189 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1190 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1191 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1192 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1193 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol
1194 -mmwaitx -mclzero -mpku -mthreads @gol
1195 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1196 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1197 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1198 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1199 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1200 -mregparm=@var{num} -msseregparm @gol
1201 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1202 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1203 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1204 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1205 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1206 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1207 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1208 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1209 -mmitigate-rop -mgeneral-regs-only}
1211 @emph{x86 Windows Options}
1212 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1213 -mnop-fun-dllimport -mthread @gol
1214 -municode -mwin32 -mwindows -fno-set-stack-executable}
1216 @emph{Xstormy16 Options}
1219 @emph{Xtensa Options}
1220 @gccoptlist{-mconst16 -mno-const16 @gol
1221 -mfused-madd -mno-fused-madd @gol
1223 -mserialize-volatile -mno-serialize-volatile @gol
1224 -mtext-section-literals -mno-text-section-literals @gol
1225 -mauto-litpools -mno-auto-litpools @gol
1226 -mtarget-align -mno-target-align @gol
1227 -mlongcalls -mno-longcalls}
1229 @emph{zSeries Options}
1230 See S/390 and zSeries Options.
1234 @node Overall Options
1235 @section Options Controlling the Kind of Output
1237 Compilation can involve up to four stages: preprocessing, compilation
1238 proper, assembly and linking, always in that order. GCC is capable of
1239 preprocessing and compiling several files either into several
1240 assembler input files, or into one assembler input file; then each
1241 assembler input file produces an object file, and linking combines all
1242 the object files (those newly compiled, and those specified as input)
1243 into an executable file.
1245 @cindex file name suffix
1246 For any given input file, the file name suffix determines what kind of
1247 compilation is done:
1251 C source code that must be preprocessed.
1254 C source code that should not be preprocessed.
1257 C++ source code that should not be preprocessed.
1260 Objective-C source code. Note that you must link with the @file{libobjc}
1261 library to make an Objective-C program work.
1264 Objective-C source code that should not be preprocessed.
1268 Objective-C++ source code. Note that you must link with the @file{libobjc}
1269 library to make an Objective-C++ program work. Note that @samp{.M} refers
1270 to a literal capital M@.
1272 @item @var{file}.mii
1273 Objective-C++ source code that should not be preprocessed.
1276 C, C++, Objective-C or Objective-C++ header file to be turned into a
1277 precompiled header (default), or C, C++ header file to be turned into an
1278 Ada spec (via the @option{-fdump-ada-spec} switch).
1281 @itemx @var{file}.cp
1282 @itemx @var{file}.cxx
1283 @itemx @var{file}.cpp
1284 @itemx @var{file}.CPP
1285 @itemx @var{file}.c++
1287 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1288 the last two letters must both be literally @samp{x}. Likewise,
1289 @samp{.C} refers to a literal capital C@.
1293 Objective-C++ source code that must be preprocessed.
1295 @item @var{file}.mii
1296 Objective-C++ source code that should not be preprocessed.
1300 @itemx @var{file}.hp
1301 @itemx @var{file}.hxx
1302 @itemx @var{file}.hpp
1303 @itemx @var{file}.HPP
1304 @itemx @var{file}.h++
1305 @itemx @var{file}.tcc
1306 C++ header file to be turned into a precompiled header or Ada spec.
1309 @itemx @var{file}.for
1310 @itemx @var{file}.ftn
1311 Fixed form Fortran source code that should not be preprocessed.
1314 @itemx @var{file}.FOR
1315 @itemx @var{file}.fpp
1316 @itemx @var{file}.FPP
1317 @itemx @var{file}.FTN
1318 Fixed form Fortran source code that must be preprocessed (with the traditional
1321 @item @var{file}.f90
1322 @itemx @var{file}.f95
1323 @itemx @var{file}.f03
1324 @itemx @var{file}.f08
1325 Free form Fortran source code that should not be preprocessed.
1327 @item @var{file}.F90
1328 @itemx @var{file}.F95
1329 @itemx @var{file}.F03
1330 @itemx @var{file}.F08
1331 Free form Fortran source code that must be preprocessed (with the
1332 traditional preprocessor).
1337 @item @var{file}.brig
1338 BRIG files (binary representation of HSAIL).
1340 @item @var{file}.ads
1341 Ada source code file that contains a library unit declaration (a
1342 declaration of a package, subprogram, or generic, or a generic
1343 instantiation), or a library unit renaming declaration (a package,
1344 generic, or subprogram renaming declaration). Such files are also
1347 @item @var{file}.adb
1348 Ada source code file containing a library unit body (a subprogram or
1349 package body). Such files are also called @dfn{bodies}.
1351 @c GCC also knows about some suffixes for languages not yet included:
1362 @itemx @var{file}.sx
1363 Assembler code that must be preprocessed.
1366 An object file to be fed straight into linking.
1367 Any file name with no recognized suffix is treated this way.
1371 You can specify the input language explicitly with the @option{-x} option:
1374 @item -x @var{language}
1375 Specify explicitly the @var{language} for the following input files
1376 (rather than letting the compiler choose a default based on the file
1377 name suffix). This option applies to all following input files until
1378 the next @option{-x} option. Possible values for @var{language} are:
1380 c c-header cpp-output
1381 c++ c++-header c++-cpp-output
1382 objective-c objective-c-header objective-c-cpp-output
1383 objective-c++ objective-c++-header objective-c++-cpp-output
1384 assembler assembler-with-cpp
1386 f77 f77-cpp-input f95 f95-cpp-input
1392 Turn off any specification of a language, so that subsequent files are
1393 handled according to their file name suffixes (as they are if @option{-x}
1394 has not been used at all).
1397 If you only want some of the stages of compilation, you can use
1398 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1399 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1400 @command{gcc} is to stop. Note that some combinations (for example,
1401 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1406 Compile or assemble the source files, but do not link. The linking
1407 stage simply is not done. The ultimate output is in the form of an
1408 object file for each source file.
1410 By default, the object file name for a source file is made by replacing
1411 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1413 Unrecognized input files, not requiring compilation or assembly, are
1418 Stop after the stage of compilation proper; do not assemble. The output
1419 is in the form of an assembler code file for each non-assembler input
1422 By default, the assembler file name for a source file is made by
1423 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1425 Input files that don't require compilation are ignored.
1429 Stop after the preprocessing stage; do not run the compiler proper. The
1430 output is in the form of preprocessed source code, which is sent to the
1433 Input files that don't require preprocessing are ignored.
1435 @cindex output file option
1438 Place output in file @var{file}. This applies to whatever
1439 sort of output is being produced, whether it be an executable file,
1440 an object file, an assembler file or preprocessed C code.
1442 If @option{-o} is not specified, the default is to put an executable
1443 file in @file{a.out}, the object file for
1444 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1445 assembler file in @file{@var{source}.s}, a precompiled header file in
1446 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1451 Print (on standard error output) the commands executed to run the stages
1452 of compilation. Also print the version number of the compiler driver
1453 program and of the preprocessor and the compiler proper.
1457 Like @option{-v} except the commands are not executed and arguments
1458 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1459 This is useful for shell scripts to capture the driver-generated command lines.
1463 Print (on the standard output) a description of the command-line options
1464 understood by @command{gcc}. If the @option{-v} option is also specified
1465 then @option{--help} is also passed on to the various processes
1466 invoked by @command{gcc}, so that they can display the command-line options
1467 they accept. If the @option{-Wextra} option has also been specified
1468 (prior to the @option{--help} option), then command-line options that
1469 have no documentation associated with them are also displayed.
1472 @opindex target-help
1473 Print (on the standard output) a description of target-specific command-line
1474 options for each tool. For some targets extra target-specific
1475 information may also be printed.
1477 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1478 Print (on the standard output) a description of the command-line
1479 options understood by the compiler that fit into all specified classes
1480 and qualifiers. These are the supported classes:
1483 @item @samp{optimizers}
1484 Display all of the optimization options supported by the
1487 @item @samp{warnings}
1488 Display all of the options controlling warning messages
1489 produced by the compiler.
1492 Display target-specific options. Unlike the
1493 @option{--target-help} option however, target-specific options of the
1494 linker and assembler are not displayed. This is because those
1495 tools do not currently support the extended @option{--help=} syntax.
1498 Display the values recognized by the @option{--param}
1501 @item @var{language}
1502 Display the options supported for @var{language}, where
1503 @var{language} is the name of one of the languages supported in this
1507 Display the options that are common to all languages.
1510 These are the supported qualifiers:
1513 @item @samp{undocumented}
1514 Display only those options that are undocumented.
1517 Display options taking an argument that appears after an equal
1518 sign in the same continuous piece of text, such as:
1519 @samp{--help=target}.
1521 @item @samp{separate}
1522 Display options taking an argument that appears as a separate word
1523 following the original option, such as: @samp{-o output-file}.
1526 Thus for example to display all the undocumented target-specific
1527 switches supported by the compiler, use:
1530 --help=target,undocumented
1533 The sense of a qualifier can be inverted by prefixing it with the
1534 @samp{^} character, so for example to display all binary warning
1535 options (i.e., ones that are either on or off and that do not take an
1536 argument) that have a description, use:
1539 --help=warnings,^joined,^undocumented
1542 The argument to @option{--help=} should not consist solely of inverted
1545 Combining several classes is possible, although this usually
1546 restricts the output so much that there is nothing to display. One
1547 case where it does work, however, is when one of the classes is
1548 @var{target}. For example, to display all the target-specific
1549 optimization options, use:
1552 --help=target,optimizers
1555 The @option{--help=} option can be repeated on the command line. Each
1556 successive use displays its requested class of options, skipping
1557 those that have already been displayed.
1559 If the @option{-Q} option appears on the command line before the
1560 @option{--help=} option, then the descriptive text displayed by
1561 @option{--help=} is changed. Instead of describing the displayed
1562 options, an indication is given as to whether the option is enabled,
1563 disabled or set to a specific value (assuming that the compiler
1564 knows this at the point where the @option{--help=} option is used).
1566 Here is a truncated example from the ARM port of @command{gcc}:
1569 % gcc -Q -mabi=2 --help=target -c
1570 The following options are target specific:
1572 -mabort-on-noreturn [disabled]
1576 The output is sensitive to the effects of previous command-line
1577 options, so for example it is possible to find out which optimizations
1578 are enabled at @option{-O2} by using:
1581 -Q -O2 --help=optimizers
1584 Alternatively you can discover which binary optimizations are enabled
1585 by @option{-O3} by using:
1588 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1589 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1590 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1595 Display the version number and copyrights of the invoked GCC@.
1597 @item -pass-exit-codes
1598 @opindex pass-exit-codes
1599 Normally the @command{gcc} program exits with the code of 1 if any
1600 phase of the compiler returns a non-success return code. If you specify
1601 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1602 the numerically highest error produced by any phase returning an error
1603 indication. The C, C++, and Fortran front ends return 4 if an internal
1604 compiler error is encountered.
1608 Use pipes rather than temporary files for communication between the
1609 various stages of compilation. This fails to work on some systems where
1610 the assembler is unable to read from a pipe; but the GNU assembler has
1613 @item -specs=@var{file}
1615 Process @var{file} after the compiler reads in the standard @file{specs}
1616 file, in order to override the defaults which the @command{gcc} driver
1617 program uses when determining what switches to pass to @command{cc1},
1618 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1619 @option{-specs=@var{file}} can be specified on the command line, and they
1620 are processed in order, from left to right. @xref{Spec Files}, for
1621 information about the format of the @var{file}.
1625 Invoke all subcommands under a wrapper program. The name of the
1626 wrapper program and its parameters are passed as a comma separated
1630 gcc -c t.c -wrapper gdb,--args
1634 This invokes all subprograms of @command{gcc} under
1635 @samp{gdb --args}, thus the invocation of @command{cc1} is
1636 @samp{gdb --args cc1 @dots{}}.
1638 @item -fplugin=@var{name}.so
1640 Load the plugin code in file @var{name}.so, assumed to be a
1641 shared object to be dlopen'd by the compiler. The base name of
1642 the shared object file is used to identify the plugin for the
1643 purposes of argument parsing (See
1644 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1645 Each plugin should define the callback functions specified in the
1648 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1649 @opindex fplugin-arg
1650 Define an argument called @var{key} with a value of @var{value}
1651 for the plugin called @var{name}.
1653 @item -fdump-ada-spec@r{[}-slim@r{]}
1654 @opindex fdump-ada-spec
1655 For C and C++ source and include files, generate corresponding Ada specs.
1656 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1657 GNAT User's Guide}, which provides detailed documentation on this feature.
1659 @item -fada-spec-parent=@var{unit}
1660 @opindex fada-spec-parent
1661 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1662 Ada specs as child units of parent @var{unit}.
1664 @item -fdump-go-spec=@var{file}
1665 @opindex fdump-go-spec
1666 For input files in any language, generate corresponding Go
1667 declarations in @var{file}. This generates Go @code{const},
1668 @code{type}, @code{var}, and @code{func} declarations which may be a
1669 useful way to start writing a Go interface to code written in some
1672 @include @value{srcdir}/../libiberty/at-file.texi
1676 @section Compiling C++ Programs
1678 @cindex suffixes for C++ source
1679 @cindex C++ source file suffixes
1680 C++ source files conventionally use one of the suffixes @samp{.C},
1681 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1682 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1683 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1684 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1685 files with these names and compiles them as C++ programs even if you
1686 call the compiler the same way as for compiling C programs (usually
1687 with the name @command{gcc}).
1691 However, the use of @command{gcc} does not add the C++ library.
1692 @command{g++} is a program that calls GCC and automatically specifies linking
1693 against the C++ library. It treats @samp{.c},
1694 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1695 files unless @option{-x} is used. This program is also useful when
1696 precompiling a C header file with a @samp{.h} extension for use in C++
1697 compilations. On many systems, @command{g++} is also installed with
1698 the name @command{c++}.
1700 @cindex invoking @command{g++}
1701 When you compile C++ programs, you may specify many of the same
1702 command-line options that you use for compiling programs in any
1703 language; or command-line options meaningful for C and related
1704 languages; or options that are meaningful only for C++ programs.
1705 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1706 explanations of options for languages related to C@.
1707 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1708 explanations of options that are meaningful only for C++ programs.
1710 @node C Dialect Options
1711 @section Options Controlling C Dialect
1712 @cindex dialect options
1713 @cindex language dialect options
1714 @cindex options, dialect
1716 The following options control the dialect of C (or languages derived
1717 from C, such as C++, Objective-C and Objective-C++) that the compiler
1721 @cindex ANSI support
1725 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1726 equivalent to @option{-std=c++98}.
1728 This turns off certain features of GCC that are incompatible with ISO
1729 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1730 such as the @code{asm} and @code{typeof} keywords, and
1731 predefined macros such as @code{unix} and @code{vax} that identify the
1732 type of system you are using. It also enables the undesirable and
1733 rarely used ISO trigraph feature. For the C compiler,
1734 it disables recognition of C++ style @samp{//} comments as well as
1735 the @code{inline} keyword.
1737 The alternate keywords @code{__asm__}, @code{__extension__},
1738 @code{__inline__} and @code{__typeof__} continue to work despite
1739 @option{-ansi}. You would not want to use them in an ISO C program, of
1740 course, but it is useful to put them in header files that might be included
1741 in compilations done with @option{-ansi}. Alternate predefined macros
1742 such as @code{__unix__} and @code{__vax__} are also available, with or
1743 without @option{-ansi}.
1745 The @option{-ansi} option does not cause non-ISO programs to be
1746 rejected gratuitously. For that, @option{-Wpedantic} is required in
1747 addition to @option{-ansi}. @xref{Warning Options}.
1749 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1750 option is used. Some header files may notice this macro and refrain
1751 from declaring certain functions or defining certain macros that the
1752 ISO standard doesn't call for; this is to avoid interfering with any
1753 programs that might use these names for other things.
1755 Functions that are normally built in but do not have semantics
1756 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1757 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1758 built-in functions provided by GCC}, for details of the functions
1763 Determine the language standard. @xref{Standards,,Language Standards
1764 Supported by GCC}, for details of these standard versions. This option
1765 is currently only supported when compiling C or C++.
1767 The compiler can accept several base standards, such as @samp{c90} or
1768 @samp{c++98}, and GNU dialects of those standards, such as
1769 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1770 compiler accepts all programs following that standard plus those
1771 using GNU extensions that do not contradict it. For example,
1772 @option{-std=c90} turns off certain features of GCC that are
1773 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1774 keywords, but not other GNU extensions that do not have a meaning in
1775 ISO C90, such as omitting the middle term of a @code{?:}
1776 expression. On the other hand, when a GNU dialect of a standard is
1777 specified, all features supported by the compiler are enabled, even when
1778 those features change the meaning of the base standard. As a result, some
1779 strict-conforming programs may be rejected. The particular standard
1780 is used by @option{-Wpedantic} to identify which features are GNU
1781 extensions given that version of the standard. For example
1782 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1783 comments, while @option{-std=gnu99 -Wpedantic} does not.
1785 A value for this option must be provided; possible values are
1791 Support all ISO C90 programs (certain GNU extensions that conflict
1792 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1794 @item iso9899:199409
1795 ISO C90 as modified in amendment 1.
1801 ISO C99. This standard is substantially completely supported, modulo
1802 bugs and floating-point issues
1803 (mainly but not entirely relating to optional C99 features from
1804 Annexes F and G). See
1805 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1806 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1811 ISO C11, the 2011 revision of the ISO C standard. This standard is
1812 substantially completely supported, modulo bugs, floating-point issues
1813 (mainly but not entirely relating to optional C11 features from
1814 Annexes F and G) and the optional Annexes K (Bounds-checking
1815 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1819 GNU dialect of ISO C90 (including some C99 features).
1823 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1827 GNU dialect of ISO C11. This is the default for C code.
1828 The name @samp{gnu1x} is deprecated.
1832 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1833 additional defect reports. Same as @option{-ansi} for C++ code.
1837 GNU dialect of @option{-std=c++98}.
1841 The 2011 ISO C++ standard plus amendments.
1842 The name @samp{c++0x} is deprecated.
1846 GNU dialect of @option{-std=c++11}.
1847 The name @samp{gnu++0x} is deprecated.
1851 The 2014 ISO C++ standard plus amendments.
1852 The name @samp{c++1y} is deprecated.
1856 GNU dialect of @option{-std=c++14}.
1857 This is the default for C++ code.
1858 The name @samp{gnu++1y} is deprecated.
1861 The next revision of the ISO C++ standard, tentatively planned for
1862 2017. Support is highly experimental, and will almost certainly
1863 change in incompatible ways in future releases.
1866 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1867 and will almost certainly change in incompatible ways in future
1871 @item -fgnu89-inline
1872 @opindex fgnu89-inline
1873 The option @option{-fgnu89-inline} tells GCC to use the traditional
1874 GNU semantics for @code{inline} functions when in C99 mode.
1875 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1876 Using this option is roughly equivalent to adding the
1877 @code{gnu_inline} function attribute to all inline functions
1878 (@pxref{Function Attributes}).
1880 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1881 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1882 specifies the default behavior).
1883 This option is not supported in @option{-std=c90} or
1884 @option{-std=gnu90} mode.
1886 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1887 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1888 in effect for @code{inline} functions. @xref{Common Predefined
1889 Macros,,,cpp,The C Preprocessor}.
1891 @item -fpermitted-flt-eval-methods=@var{style}
1892 @opindex fpermitted-flt-eval-methods
1893 @opindex fpermitted-flt-eval-methods=c11
1894 @opindex fpermitted-flt-eval-methods=ts-18661-3
1895 ISO/IEC TS 18661-3 defines new permissible values for
1896 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
1897 a semantic type that is an interchange or extended format should be
1898 evaluated to the precision and range of that type. These new values are
1899 a superset of those permitted under C99/C11, which does not specify the
1900 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
1901 conforming to C11 may not have been written expecting the possibility of
1904 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
1905 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
1906 or the extended set of values specified in ISO/IEC TS 18661-3.
1908 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
1910 The default when in a standards compliant mode (@option{-std=c11} or similar)
1911 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
1912 dialect (@option{-std=gnu11} or similar) is
1913 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
1915 @item -aux-info @var{filename}
1917 Output to the given filename prototyped declarations for all functions
1918 declared and/or defined in a translation unit, including those in header
1919 files. This option is silently ignored in any language other than C@.
1921 Besides declarations, the file indicates, in comments, the origin of
1922 each declaration (source file and line), whether the declaration was
1923 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1924 @samp{O} for old, respectively, in the first character after the line
1925 number and the colon), and whether it came from a declaration or a
1926 definition (@samp{C} or @samp{F}, respectively, in the following
1927 character). In the case of function definitions, a K&R-style list of
1928 arguments followed by their declarations is also provided, inside
1929 comments, after the declaration.
1931 @item -fallow-parameterless-variadic-functions
1932 @opindex fallow-parameterless-variadic-functions
1933 Accept variadic functions without named parameters.
1935 Although it is possible to define such a function, this is not very
1936 useful as it is not possible to read the arguments. This is only
1937 supported for C as this construct is allowed by C++.
1941 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1942 keyword, so that code can use these words as identifiers. You can use
1943 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1944 instead. @option{-ansi} implies @option{-fno-asm}.
1946 In C++, this switch only affects the @code{typeof} keyword, since
1947 @code{asm} and @code{inline} are standard keywords. You may want to
1948 use the @option{-fno-gnu-keywords} flag instead, which has the same
1949 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1950 switch only affects the @code{asm} and @code{typeof} keywords, since
1951 @code{inline} is a standard keyword in ISO C99.
1954 @itemx -fno-builtin-@var{function}
1955 @opindex fno-builtin
1956 @cindex built-in functions
1957 Don't recognize built-in functions that do not begin with
1958 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1959 functions provided by GCC}, for details of the functions affected,
1960 including those which are not built-in functions when @option{-ansi} or
1961 @option{-std} options for strict ISO C conformance are used because they
1962 do not have an ISO standard meaning.
1964 GCC normally generates special code to handle certain built-in functions
1965 more efficiently; for instance, calls to @code{alloca} may become single
1966 instructions which adjust the stack directly, and calls to @code{memcpy}
1967 may become inline copy loops. The resulting code is often both smaller
1968 and faster, but since the function calls no longer appear as such, you
1969 cannot set a breakpoint on those calls, nor can you change the behavior
1970 of the functions by linking with a different library. In addition,
1971 when a function is recognized as a built-in function, GCC may use
1972 information about that function to warn about problems with calls to
1973 that function, or to generate more efficient code, even if the
1974 resulting code still contains calls to that function. For example,
1975 warnings are given with @option{-Wformat} for bad calls to
1976 @code{printf} when @code{printf} is built in and @code{strlen} is
1977 known not to modify global memory.
1979 With the @option{-fno-builtin-@var{function}} option
1980 only the built-in function @var{function} is
1981 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1982 function is named that is not built-in in this version of GCC, this
1983 option is ignored. There is no corresponding
1984 @option{-fbuiltin-@var{function}} option; if you wish to enable
1985 built-in functions selectively when using @option{-fno-builtin} or
1986 @option{-ffreestanding}, you may define macros such as:
1989 #define abs(n) __builtin_abs ((n))
1990 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1996 Enable parsing of function definitions marked with @code{__GIMPLE}.
1997 This is an experimental feature that allows unit testing of GIMPLE
2002 @cindex hosted environment
2004 Assert that compilation targets a hosted environment. This implies
2005 @option{-fbuiltin}. A hosted environment is one in which the
2006 entire standard library is available, and in which @code{main} has a return
2007 type of @code{int}. Examples are nearly everything except a kernel.
2008 This is equivalent to @option{-fno-freestanding}.
2010 @item -ffreestanding
2011 @opindex ffreestanding
2012 @cindex hosted environment
2014 Assert that compilation targets a freestanding environment. This
2015 implies @option{-fno-builtin}. A freestanding environment
2016 is one in which the standard library may not exist, and program startup may
2017 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2018 This is equivalent to @option{-fno-hosted}.
2020 @xref{Standards,,Language Standards Supported by GCC}, for details of
2021 freestanding and hosted environments.
2025 @cindex OpenACC accelerator programming
2026 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2027 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2028 compiler generates accelerated code according to the OpenACC Application
2029 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
2030 implies @option{-pthread}, and thus is only supported on targets that
2031 have support for @option{-pthread}.
2033 @item -fopenacc-dim=@var{geom}
2034 @opindex fopenacc-dim
2035 @cindex OpenACC accelerator programming
2036 Specify default compute dimensions for parallel offload regions that do
2037 not explicitly specify. The @var{geom} value is a triple of
2038 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2039 can be omitted, to use a target-specific default value.
2043 @cindex OpenMP parallel
2044 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2045 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2046 compiler generates parallel code according to the OpenMP Application
2047 Program Interface v4.5 @w{@uref{http://www.openmp.org/}}. This option
2048 implies @option{-pthread}, and thus is only supported on targets that
2049 have support for @option{-pthread}. @option{-fopenmp} implies
2050 @option{-fopenmp-simd}.
2053 @opindex fopenmp-simd
2056 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2057 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2062 @cindex Enable Cilk Plus
2063 Enable the usage of Cilk Plus language extension features for C/C++.
2064 When the option @option{-fcilkplus} is specified, enable the usage of
2065 the Cilk Plus Language extension features for C/C++. The present
2066 implementation follows ABI version 1.2. This is an experimental
2067 feature that is only partially complete, and whose interface may
2068 change in future versions of GCC as the official specification
2069 changes. Currently, all features but @code{_Cilk_for} have been
2074 When the option @option{-fgnu-tm} is specified, the compiler
2075 generates code for the Linux variant of Intel's current Transactional
2076 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2077 an experimental feature whose interface may change in future versions
2078 of GCC, as the official specification changes. Please note that not
2079 all architectures are supported for this feature.
2081 For more information on GCC's support for transactional memory,
2082 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2083 Transactional Memory Library}.
2085 Note that the transactional memory feature is not supported with
2086 non-call exceptions (@option{-fnon-call-exceptions}).
2088 @item -fms-extensions
2089 @opindex fms-extensions
2090 Accept some non-standard constructs used in Microsoft header files.
2092 In C++ code, this allows member names in structures to be similar
2093 to previous types declarations.
2102 Some cases of unnamed fields in structures and unions are only
2103 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2104 fields within structs/unions}, for details.
2106 Note that this option is off for all targets but x86
2107 targets using ms-abi.
2109 @item -fplan9-extensions
2110 @opindex fplan9-extensions
2111 Accept some non-standard constructs used in Plan 9 code.
2113 This enables @option{-fms-extensions}, permits passing pointers to
2114 structures with anonymous fields to functions that expect pointers to
2115 elements of the type of the field, and permits referring to anonymous
2116 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2117 struct/union fields within structs/unions}, for details. This is only
2118 supported for C, not C++.
2120 @item -fcond-mismatch
2121 @opindex fcond-mismatch
2122 Allow conditional expressions with mismatched types in the second and
2123 third arguments. The value of such an expression is void. This option
2124 is not supported for C++.
2126 @item -flax-vector-conversions
2127 @opindex flax-vector-conversions
2128 Allow implicit conversions between vectors with differing numbers of
2129 elements and/or incompatible element types. This option should not be
2132 @item -funsigned-char
2133 @opindex funsigned-char
2134 Let the type @code{char} be unsigned, like @code{unsigned char}.
2136 Each kind of machine has a default for what @code{char} should
2137 be. It is either like @code{unsigned char} by default or like
2138 @code{signed char} by default.
2140 Ideally, a portable program should always use @code{signed char} or
2141 @code{unsigned char} when it depends on the signedness of an object.
2142 But many programs have been written to use plain @code{char} and
2143 expect it to be signed, or expect it to be unsigned, depending on the
2144 machines they were written for. This option, and its inverse, let you
2145 make such a program work with the opposite default.
2147 The type @code{char} is always a distinct type from each of
2148 @code{signed char} or @code{unsigned char}, even though its behavior
2149 is always just like one of those two.
2152 @opindex fsigned-char
2153 Let the type @code{char} be signed, like @code{signed char}.
2155 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2156 the negative form of @option{-funsigned-char}. Likewise, the option
2157 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2159 @item -fsigned-bitfields
2160 @itemx -funsigned-bitfields
2161 @itemx -fno-signed-bitfields
2162 @itemx -fno-unsigned-bitfields
2163 @opindex fsigned-bitfields
2164 @opindex funsigned-bitfields
2165 @opindex fno-signed-bitfields
2166 @opindex fno-unsigned-bitfields
2167 These options control whether a bit-field is signed or unsigned, when the
2168 declaration does not use either @code{signed} or @code{unsigned}. By
2169 default, such a bit-field is signed, because this is consistent: the
2170 basic integer types such as @code{int} are signed types.
2172 @item -fsso-struct=@var{endianness}
2173 @opindex fsso-struct
2174 Set the default scalar storage order of structures and unions to the
2175 specified endianness. The accepted values are @samp{big-endian},
2176 @samp{little-endian} and @samp{native} for the native endianness of
2177 the target (the default). This option is not supported for C++.
2179 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2180 code that is not binary compatible with code generated without it if the
2181 specified endianness is not the native endianness of the target.
2184 @node C++ Dialect Options
2185 @section Options Controlling C++ Dialect
2187 @cindex compiler options, C++
2188 @cindex C++ options, command-line
2189 @cindex options, C++
2190 This section describes the command-line options that are only meaningful
2191 for C++ programs. You can also use most of the GNU compiler options
2192 regardless of what language your program is in. For example, you
2193 might compile a file @file{firstClass.C} like this:
2196 g++ -g -fstrict-enums -O -c firstClass.C
2200 In this example, only @option{-fstrict-enums} is an option meant
2201 only for C++ programs; you can use the other options with any
2202 language supported by GCC@.
2204 Some options for compiling C programs, such as @option{-std}, are also
2205 relevant for C++ programs.
2206 @xref{C Dialect Options,,Options Controlling C Dialect}.
2208 Here is a list of options that are @emph{only} for compiling C++ programs:
2212 @item -fabi-version=@var{n}
2213 @opindex fabi-version
2214 Use version @var{n} of the C++ ABI@. The default is version 0.
2216 Version 0 refers to the version conforming most closely to
2217 the C++ ABI specification. Therefore, the ABI obtained using version 0
2218 will change in different versions of G++ as ABI bugs are fixed.
2220 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2222 Version 2 is the version of the C++ ABI that first appeared in G++
2223 3.4, and was the default through G++ 4.9.
2225 Version 3 corrects an error in mangling a constant address as a
2228 Version 4, which first appeared in G++ 4.5, implements a standard
2229 mangling for vector types.
2231 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2232 attribute const/volatile on function pointer types, decltype of a
2233 plain decl, and use of a function parameter in the declaration of
2236 Version 6, which first appeared in G++ 4.7, corrects the promotion
2237 behavior of C++11 scoped enums and the mangling of template argument
2238 packs, const/static_cast, prefix ++ and --, and a class scope function
2239 used as a template argument.
2241 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2242 builtin type and corrects the mangling of lambdas in default argument
2245 Version 8, which first appeared in G++ 4.9, corrects the substitution
2246 behavior of function types with function-cv-qualifiers.
2248 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2251 Version 10, which first appeared in G++ 6.1, adds mangling of
2252 attributes that affect type identity, such as ia32 calling convention
2253 attributes (e.g. @samp{stdcall}).
2255 Version 11, which first appeared in G++ 7, corrects the mangling of
2256 sizeof... expressions and operator names. For multiple entities with
2257 the same name within a function, that are declared in different scopes,
2258 the mangling now changes starting with the twelfth occurrence. It also
2259 implies @option{-fnew-inheriting-ctors}.
2261 See also @option{-Wabi}.
2263 @item -fabi-compat-version=@var{n}
2264 @opindex fabi-compat-version
2265 On targets that support strong aliases, G++
2266 works around mangling changes by creating an alias with the correct
2267 mangled name when defining a symbol with an incorrect mangled name.
2268 This switch specifies which ABI version to use for the alias.
2270 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2271 compatibility). If another ABI version is explicitly selected, this
2272 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2273 use @option{-fabi-compat-version=2}.
2275 If this option is not provided but @option{-Wabi=@var{n}} is, that
2276 version is used for compatibility aliases. If this option is provided
2277 along with @option{-Wabi} (without the version), the version from this
2278 option is used for the warning.
2280 @item -fno-access-control
2281 @opindex fno-access-control
2282 Turn off all access checking. This switch is mainly useful for working
2283 around bugs in the access control code.
2286 @opindex faligned-new
2287 Enable support for C++17 @code{new} of types that require more
2288 alignment than @code{void* ::operator new(std::size_t)} provides. A
2289 numeric argument such as @code{-faligned-new=32} can be used to
2290 specify how much alignment (in bytes) is provided by that function,
2291 but few users will need to override the default of
2292 @code{alignof(std::max_align_t)}.
2296 Check that the pointer returned by @code{operator new} is non-null
2297 before attempting to modify the storage allocated. This check is
2298 normally unnecessary because the C++ standard specifies that
2299 @code{operator new} only returns @code{0} if it is declared
2300 @code{throw()}, in which case the compiler always checks the
2301 return value even without this option. In all other cases, when
2302 @code{operator new} has a non-empty exception specification, memory
2303 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2304 @samp{new (nothrow)}.
2308 Enable support for the C++ Extensions for Concepts Technical
2309 Specification, ISO 19217 (2015), which allows code like
2312 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2313 template <Addable T> T add (T a, T b) @{ return a + b; @}
2316 @item -fconstexpr-depth=@var{n}
2317 @opindex fconstexpr-depth
2318 Set the maximum nested evaluation depth for C++11 constexpr functions
2319 to @var{n}. A limit is needed to detect endless recursion during
2320 constant expression evaluation. The minimum specified by the standard
2323 @item -fconstexpr-loop-limit=@var{n}
2324 @opindex fconstexpr-loop-limit
2325 Set the maximum number of iterations for a loop in C++14 constexpr functions
2326 to @var{n}. A limit is needed to detect infinite loops during
2327 constant expression evaluation. The default is 262144 (1<<18).
2329 @item -fdeduce-init-list
2330 @opindex fdeduce-init-list
2331 Enable deduction of a template type parameter as
2332 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2335 template <class T> auto forward(T t) -> decltype (realfn (t))
2342 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2346 This deduction was implemented as a possible extension to the
2347 originally proposed semantics for the C++11 standard, but was not part
2348 of the final standard, so it is disabled by default. This option is
2349 deprecated, and may be removed in a future version of G++.
2351 @item -ffriend-injection
2352 @opindex ffriend-injection
2353 Inject friend functions into the enclosing namespace, so that they are
2354 visible outside the scope of the class in which they are declared.
2355 Friend functions were documented to work this way in the old Annotated
2356 C++ Reference Manual.
2357 However, in ISO C++ a friend function that is not declared
2358 in an enclosing scope can only be found using argument dependent
2359 lookup. GCC defaults to the standard behavior.
2361 This option is for compatibility, and may be removed in a future
2364 @item -fno-elide-constructors
2365 @opindex fno-elide-constructors
2366 The C++ standard allows an implementation to omit creating a temporary
2367 that is only used to initialize another object of the same type.
2368 Specifying this option disables that optimization, and forces G++ to
2369 call the copy constructor in all cases. This option also causes G++
2370 to call trivial member functions which otherwise would be expanded inline.
2372 In C++17, the compiler is required to omit these temporaries, but this
2373 option still affects trivial member functions.
2375 @item -fno-enforce-eh-specs
2376 @opindex fno-enforce-eh-specs
2377 Don't generate code to check for violation of exception specifications
2378 at run time. This option violates the C++ standard, but may be useful
2379 for reducing code size in production builds, much like defining
2380 @code{NDEBUG}. This does not give user code permission to throw
2381 exceptions in violation of the exception specifications; the compiler
2382 still optimizes based on the specifications, so throwing an
2383 unexpected exception results in undefined behavior at run time.
2385 @item -fextern-tls-init
2386 @itemx -fno-extern-tls-init
2387 @opindex fextern-tls-init
2388 @opindex fno-extern-tls-init
2389 The C++11 and OpenMP standards allow @code{thread_local} and
2390 @code{threadprivate} variables to have dynamic (runtime)
2391 initialization. To support this, any use of such a variable goes
2392 through a wrapper function that performs any necessary initialization.
2393 When the use and definition of the variable are in the same
2394 translation unit, this overhead can be optimized away, but when the
2395 use is in a different translation unit there is significant overhead
2396 even if the variable doesn't actually need dynamic initialization. If
2397 the programmer can be sure that no use of the variable in a
2398 non-defining TU needs to trigger dynamic initialization (either
2399 because the variable is statically initialized, or a use of the
2400 variable in the defining TU will be executed before any uses in
2401 another TU), they can avoid this overhead with the
2402 @option{-fno-extern-tls-init} option.
2404 On targets that support symbol aliases, the default is
2405 @option{-fextern-tls-init}. On targets that do not support symbol
2406 aliases, the default is @option{-fno-extern-tls-init}.
2409 @itemx -fno-for-scope
2411 @opindex fno-for-scope
2412 If @option{-ffor-scope} is specified, the scope of variables declared in
2413 a @i{for-init-statement} is limited to the @code{for} loop itself,
2414 as specified by the C++ standard.
2415 If @option{-fno-for-scope} is specified, the scope of variables declared in
2416 a @i{for-init-statement} extends to the end of the enclosing scope,
2417 as was the case in old versions of G++, and other (traditional)
2418 implementations of C++.
2420 If neither flag is given, the default is to follow the standard,
2421 but to allow and give a warning for old-style code that would
2422 otherwise be invalid, or have different behavior.
2424 @item -fno-gnu-keywords
2425 @opindex fno-gnu-keywords
2426 Do not recognize @code{typeof} as a keyword, so that code can use this
2427 word as an identifier. You can use the keyword @code{__typeof__} instead.
2428 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2429 @option{-std=c++98}, @option{-std=c++11}, etc.
2431 @item -fno-implicit-templates
2432 @opindex fno-implicit-templates
2433 Never emit code for non-inline templates that are instantiated
2434 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2435 @xref{Template Instantiation}, for more information.
2437 @item -fno-implicit-inline-templates
2438 @opindex fno-implicit-inline-templates
2439 Don't emit code for implicit instantiations of inline templates, either.
2440 The default is to handle inlines differently so that compiles with and
2441 without optimization need the same set of explicit instantiations.
2443 @item -fno-implement-inlines
2444 @opindex fno-implement-inlines
2445 To save space, do not emit out-of-line copies of inline functions
2446 controlled by @code{#pragma implementation}. This causes linker
2447 errors if these functions are not inlined everywhere they are called.
2449 @item -fms-extensions
2450 @opindex fms-extensions
2451 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2452 int and getting a pointer to member function via non-standard syntax.
2454 @item -fnew-inheriting-ctors
2455 @opindex fnew-inheriting-ctors
2456 Enable the P0136 adjustment to the semantics of C++11 constructor
2457 inheritance. This is part of C++17 but also considered to be a Defect
2458 Report against C++11 and C++14. This flag is enabled by default
2459 unless @option{-fabi-version=10} or lower is specified.
2461 @item -fnew-ttp-matching
2462 @opindex fnew-ttp-matching
2463 Enable the P0522 resolution to Core issue 150, template template
2464 parameters and default arguments: this allows a template with default
2465 template arguments as an argument for a template template parameter
2466 with fewer template parameters. This flag is enabled by default for
2467 @option{-std=c++1z}.
2469 @item -fno-nonansi-builtins
2470 @opindex fno-nonansi-builtins
2471 Disable built-in declarations of functions that are not mandated by
2472 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2473 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2476 @opindex fnothrow-opt
2477 Treat a @code{throw()} exception specification as if it were a
2478 @code{noexcept} specification to reduce or eliminate the text size
2479 overhead relative to a function with no exception specification. If
2480 the function has local variables of types with non-trivial
2481 destructors, the exception specification actually makes the
2482 function smaller because the EH cleanups for those variables can be
2483 optimized away. The semantic effect is that an exception thrown out of
2484 a function with such an exception specification results in a call
2485 to @code{terminate} rather than @code{unexpected}.
2487 @item -fno-operator-names
2488 @opindex fno-operator-names
2489 Do not treat the operator name keywords @code{and}, @code{bitand},
2490 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2491 synonyms as keywords.
2493 @item -fno-optional-diags
2494 @opindex fno-optional-diags
2495 Disable diagnostics that the standard says a compiler does not need to
2496 issue. Currently, the only such diagnostic issued by G++ is the one for
2497 a name having multiple meanings within a class.
2500 @opindex fpermissive
2501 Downgrade some diagnostics about nonconformant code from errors to
2502 warnings. Thus, using @option{-fpermissive} allows some
2503 nonconforming code to compile.
2505 @item -fno-pretty-templates
2506 @opindex fno-pretty-templates
2507 When an error message refers to a specialization of a function
2508 template, the compiler normally prints the signature of the
2509 template followed by the template arguments and any typedefs or
2510 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2511 rather than @code{void f(int)}) so that it's clear which template is
2512 involved. When an error message refers to a specialization of a class
2513 template, the compiler omits any template arguments that match
2514 the default template arguments for that template. If either of these
2515 behaviors make it harder to understand the error message rather than
2516 easier, you can use @option{-fno-pretty-templates} to disable them.
2520 Enable automatic template instantiation at link time. This option also
2521 implies @option{-fno-implicit-templates}. @xref{Template
2522 Instantiation}, for more information.
2526 Disable generation of information about every class with virtual
2527 functions for use by the C++ run-time type identification features
2528 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2529 of the language, you can save some space by using this flag. Note that
2530 exception handling uses the same information, but G++ generates it as
2531 needed. The @code{dynamic_cast} operator can still be used for casts that
2532 do not require run-time type information, i.e.@: casts to @code{void *} or to
2533 unambiguous base classes.
2535 @item -fsized-deallocation
2536 @opindex fsized-deallocation
2537 Enable the built-in global declarations
2539 void operator delete (void *, std::size_t) noexcept;
2540 void operator delete[] (void *, std::size_t) noexcept;
2542 as introduced in C++14. This is useful for user-defined replacement
2543 deallocation functions that, for example, use the size of the object
2544 to make deallocation faster. Enabled by default under
2545 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2546 warns about places that might want to add a definition.
2548 @item -fstrict-enums
2549 @opindex fstrict-enums
2550 Allow the compiler to optimize using the assumption that a value of
2551 enumerated type can only be one of the values of the enumeration (as
2552 defined in the C++ standard; basically, a value that can be
2553 represented in the minimum number of bits needed to represent all the
2554 enumerators). This assumption may not be valid if the program uses a
2555 cast to convert an arbitrary integer value to the enumerated type.
2557 @item -fstrong-eval-order
2558 @opindex fstrong-eval-order
2559 Evaluate member access, array subscripting, and shift expressions in
2560 left-to-right order, and evaluate assignment in right-to-left order,
2561 as adopted for C++17. Enabled by default with @option{-std=c++1z}.
2562 @option{-fstrong-eval-order=some} enables just the ordering of member
2563 access and shift expressions, and is the default without
2564 @option{-std=c++1z}.
2566 @item -ftemplate-backtrace-limit=@var{n}
2567 @opindex ftemplate-backtrace-limit
2568 Set the maximum number of template instantiation notes for a single
2569 warning or error to @var{n}. The default value is 10.
2571 @item -ftemplate-depth=@var{n}
2572 @opindex ftemplate-depth
2573 Set the maximum instantiation depth for template classes to @var{n}.
2574 A limit on the template instantiation depth is needed to detect
2575 endless recursions during template class instantiation. ANSI/ISO C++
2576 conforming programs must not rely on a maximum depth greater than 17
2577 (changed to 1024 in C++11). The default value is 900, as the compiler
2578 can run out of stack space before hitting 1024 in some situations.
2580 @item -fno-threadsafe-statics
2581 @opindex fno-threadsafe-statics
2582 Do not emit the extra code to use the routines specified in the C++
2583 ABI for thread-safe initialization of local statics. You can use this
2584 option to reduce code size slightly in code that doesn't need to be
2587 @item -fuse-cxa-atexit
2588 @opindex fuse-cxa-atexit
2589 Register destructors for objects with static storage duration with the
2590 @code{__cxa_atexit} function rather than the @code{atexit} function.
2591 This option is required for fully standards-compliant handling of static
2592 destructors, but only works if your C library supports
2593 @code{__cxa_atexit}.
2595 @item -fno-use-cxa-get-exception-ptr
2596 @opindex fno-use-cxa-get-exception-ptr
2597 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2598 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2599 if the runtime routine is not available.
2601 @item -fvisibility-inlines-hidden
2602 @opindex fvisibility-inlines-hidden
2603 This switch declares that the user does not attempt to compare
2604 pointers to inline functions or methods where the addresses of the two functions
2605 are taken in different shared objects.
2607 The effect of this is that GCC may, effectively, mark inline methods with
2608 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2609 appear in the export table of a DSO and do not require a PLT indirection
2610 when used within the DSO@. Enabling this option can have a dramatic effect
2611 on load and link times of a DSO as it massively reduces the size of the
2612 dynamic export table when the library makes heavy use of templates.
2614 The behavior of this switch is not quite the same as marking the
2615 methods as hidden directly, because it does not affect static variables
2616 local to the function or cause the compiler to deduce that
2617 the function is defined in only one shared object.
2619 You may mark a method as having a visibility explicitly to negate the
2620 effect of the switch for that method. For example, if you do want to
2621 compare pointers to a particular inline method, you might mark it as
2622 having default visibility. Marking the enclosing class with explicit
2623 visibility has no effect.
2625 Explicitly instantiated inline methods are unaffected by this option
2626 as their linkage might otherwise cross a shared library boundary.
2627 @xref{Template Instantiation}.
2629 @item -fvisibility-ms-compat
2630 @opindex fvisibility-ms-compat
2631 This flag attempts to use visibility settings to make GCC's C++
2632 linkage model compatible with that of Microsoft Visual Studio.
2634 The flag makes these changes to GCC's linkage model:
2638 It sets the default visibility to @code{hidden}, like
2639 @option{-fvisibility=hidden}.
2642 Types, but not their members, are not hidden by default.
2645 The One Definition Rule is relaxed for types without explicit
2646 visibility specifications that are defined in more than one
2647 shared object: those declarations are permitted if they are
2648 permitted when this option is not used.
2651 In new code it is better to use @option{-fvisibility=hidden} and
2652 export those classes that are intended to be externally visible.
2653 Unfortunately it is possible for code to rely, perhaps accidentally,
2654 on the Visual Studio behavior.
2656 Among the consequences of these changes are that static data members
2657 of the same type with the same name but defined in different shared
2658 objects are different, so changing one does not change the other;
2659 and that pointers to function members defined in different shared
2660 objects may not compare equal. When this flag is given, it is a
2661 violation of the ODR to define types with the same name differently.
2665 Do not use weak symbol support, even if it is provided by the linker.
2666 By default, G++ uses weak symbols if they are available. This
2667 option exists only for testing, and should not be used by end-users;
2668 it results in inferior code and has no benefits. This option may
2669 be removed in a future release of G++.
2673 Do not search for header files in the standard directories specific to
2674 C++, but do still search the other standard directories. (This option
2675 is used when building the C++ library.)
2678 In addition, these optimization, warning, and code generation options
2679 have meanings only for C++ programs:
2682 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2685 Warn when G++ it generates code that is probably not compatible with
2686 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2687 ABI with each major release, normally @option{-Wabi} will warn only if
2688 there is a check added later in a release series for an ABI issue
2689 discovered since the initial release. @option{-Wabi} will warn about
2690 more things if an older ABI version is selected (with
2691 @option{-fabi-version=@var{n}}).
2693 @option{-Wabi} can also be used with an explicit version number to
2694 warn about compatibility with a particular @option{-fabi-version}
2695 level, e.g. @option{-Wabi=2} to warn about changes relative to
2696 @option{-fabi-version=2}.
2698 If an explicit version number is provided and
2699 @option{-fabi-compat-version} is not specified, the version number
2700 from this option is used for compatibility aliases. If no explicit
2701 version number is provided with this option, but
2702 @option{-fabi-compat-version} is specified, that version number is
2703 used for ABI warnings.
2705 Although an effort has been made to warn about
2706 all such cases, there are probably some cases that are not warned about,
2707 even though G++ is generating incompatible code. There may also be
2708 cases where warnings are emitted even though the code that is generated
2711 You should rewrite your code to avoid these warnings if you are
2712 concerned about the fact that code generated by G++ may not be binary
2713 compatible with code generated by other compilers.
2715 Known incompatibilities in @option{-fabi-version=2} (which was the
2716 default from GCC 3.4 to 4.9) include:
2721 A template with a non-type template parameter of reference type was
2722 mangled incorrectly:
2725 template <int &> struct S @{@};
2729 This was fixed in @option{-fabi-version=3}.
2732 SIMD vector types declared using @code{__attribute ((vector_size))} were
2733 mangled in a non-standard way that does not allow for overloading of
2734 functions taking vectors of different sizes.
2736 The mangling was changed in @option{-fabi-version=4}.
2739 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2740 qualifiers, and @code{decltype} of a plain declaration was folded away.
2742 These mangling issues were fixed in @option{-fabi-version=5}.
2745 Scoped enumerators passed as arguments to a variadic function are
2746 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2747 On most targets this does not actually affect the parameter passing
2748 ABI, as there is no way to pass an argument smaller than @code{int}.
2750 Also, the ABI changed the mangling of template argument packs,
2751 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2752 a class scope function used as a template argument.
2754 These issues were corrected in @option{-fabi-version=6}.
2757 Lambdas in default argument scope were mangled incorrectly, and the
2758 ABI changed the mangling of @code{nullptr_t}.
2760 These issues were corrected in @option{-fabi-version=7}.
2763 When mangling a function type with function-cv-qualifiers, the
2764 un-qualified function type was incorrectly treated as a substitution
2767 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2770 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2771 unaligned accesses. Note that this did not affect the ABI of a
2772 function with a @code{nullptr_t} parameter, as parameters have a
2775 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2778 Target-specific attributes that affect the identity of a type, such as
2779 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2780 did not affect the mangled name, leading to name collisions when
2781 function pointers were used as template arguments.
2783 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2787 It also warns about psABI-related changes. The known psABI changes at this
2793 For SysV/x86-64, unions with @code{long double} members are
2794 passed in memory as specified in psABI. For example:
2804 @code{union U} is always passed in memory.
2808 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2811 Warn when a type with an ABI tag is used in a context that does not
2812 have that ABI tag. See @ref{C++ Attributes} for more information
2815 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2816 @opindex Wctor-dtor-privacy
2817 @opindex Wno-ctor-dtor-privacy
2818 Warn when a class seems unusable because all the constructors or
2819 destructors in that class are private, and it has neither friends nor
2820 public static member functions. Also warn if there are no non-private
2821 methods, and there's at least one private member function that isn't
2822 a constructor or destructor.
2824 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2825 @opindex Wdelete-non-virtual-dtor
2826 @opindex Wno-delete-non-virtual-dtor
2827 Warn when @code{delete} is used to destroy an instance of a class that
2828 has virtual functions and non-virtual destructor. It is unsafe to delete
2829 an instance of a derived class through a pointer to a base class if the
2830 base class does not have a virtual destructor. This warning is enabled
2833 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2834 @opindex Wliteral-suffix
2835 @opindex Wno-literal-suffix
2836 Warn when a string or character literal is followed by a ud-suffix which does
2837 not begin with an underscore. As a conforming extension, GCC treats such
2838 suffixes as separate preprocessing tokens in order to maintain backwards
2839 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2843 #define __STDC_FORMAT_MACROS
2844 #include <inttypes.h>
2849 printf("My int64: %" PRId64"\n", i64);
2853 In this case, @code{PRId64} is treated as a separate preprocessing token.
2855 Additionally, warn when a user-defined literal operator is declared with
2856 a literal suffix identifier that doesn't begin with an underscore. Literal
2857 suffix identifiers that don't begin with an underscore are reserved for
2858 future standardization.
2860 This warning is enabled by default.
2862 @item -Wlto-type-mismatch
2863 @opindex Wlto-type-mismatch
2864 @opindex Wno-lto-type-mismatch
2866 During the link-time optimization warn about type mismatches in
2867 global declarations from different compilation units.
2868 Requires @option{-flto} to be enabled. Enabled by default.
2870 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2872 @opindex Wno-narrowing
2873 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2874 conversion prohibited by C++11 occurs within
2878 int i = @{ 2.2 @}; // error: narrowing from double to int
2881 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2883 When a later standard is in effect, e.g. when using @option{-std=c++11},
2884 narrowing conversions are diagnosed by default, as required by the standard.
2885 A narrowing conversion from a constant produces an error,
2886 and a narrowing conversion from a non-constant produces a warning,
2887 but @option{-Wno-narrowing} suppresses the diagnostic.
2888 Note that this does not affect the meaning of well-formed code;
2889 narrowing conversions are still considered ill-formed in SFINAE contexts.
2891 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2893 @opindex Wno-noexcept
2894 Warn when a noexcept-expression evaluates to false because of a call
2895 to a function that does not have a non-throwing exception
2896 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2897 the compiler to never throw an exception.
2899 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2900 @opindex Wnon-virtual-dtor
2901 @opindex Wno-non-virtual-dtor
2902 Warn when a class has virtual functions and an accessible non-virtual
2903 destructor itself or in an accessible polymorphic base class, in which
2904 case it is possible but unsafe to delete an instance of a derived
2905 class through a pointer to the class itself or base class. This
2906 warning is automatically enabled if @option{-Weffc++} is specified.
2908 @item -Wregister @r{(C++ and Objective-C++ only)}
2910 @opindex Wno-register
2911 Warn on uses of the @code{register} storage class specifier, except
2912 when it is part of the GNU @ref{Explicit Register Variables} extension.
2913 The use of the @code{register} keyword as storage class specifier has
2914 been deprecated in C++11 and removed in C++17.
2915 Enabled by default with @option{-std=c++1z}.
2917 @item -Wreorder @r{(C++ and Objective-C++ only)}
2919 @opindex Wno-reorder
2920 @cindex reordering, warning
2921 @cindex warning for reordering of member initializers
2922 Warn when the order of member initializers given in the code does not
2923 match the order in which they must be executed. For instance:
2929 A(): j (0), i (1) @{ @}
2934 The compiler rearranges the member initializers for @code{i}
2935 and @code{j} to match the declaration order of the members, emitting
2936 a warning to that effect. This warning is enabled by @option{-Wall}.
2938 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2939 @opindex fext-numeric-literals
2940 @opindex fno-ext-numeric-literals
2941 Accept imaginary, fixed-point, or machine-defined
2942 literal number suffixes as GNU extensions.
2943 When this option is turned off these suffixes are treated
2944 as C++11 user-defined literal numeric suffixes.
2945 This is on by default for all pre-C++11 dialects and all GNU dialects:
2946 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2947 @option{-std=gnu++14}.
2948 This option is off by default
2949 for ISO C++11 onwards (@option{-std=c++11}, ...).
2952 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2955 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2958 Warn about violations of the following style guidelines from Scott Meyers'
2959 @cite{Effective C++} series of books:
2963 Define a copy constructor and an assignment operator for classes
2964 with dynamically-allocated memory.
2967 Prefer initialization to assignment in constructors.
2970 Have @code{operator=} return a reference to @code{*this}.
2973 Don't try to return a reference when you must return an object.
2976 Distinguish between prefix and postfix forms of increment and
2977 decrement operators.
2980 Never overload @code{&&}, @code{||}, or @code{,}.
2984 This option also enables @option{-Wnon-virtual-dtor}, which is also
2985 one of the effective C++ recommendations. However, the check is
2986 extended to warn about the lack of virtual destructor in accessible
2987 non-polymorphic bases classes too.
2989 When selecting this option, be aware that the standard library
2990 headers do not obey all of these guidelines; use @samp{grep -v}
2991 to filter out those warnings.
2993 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2994 @opindex Wstrict-null-sentinel
2995 @opindex Wno-strict-null-sentinel
2996 Warn about the use of an uncasted @code{NULL} as sentinel. When
2997 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2998 to @code{__null}. Although it is a null pointer constant rather than a
2999 null pointer, it is guaranteed to be of the same size as a pointer.
3000 But this use is not portable across different compilers.
3002 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3003 @opindex Wno-non-template-friend
3004 @opindex Wnon-template-friend
3005 Disable warnings when non-template friend functions are declared
3006 within a template. In very old versions of GCC that predate implementation
3007 of the ISO standard, declarations such as
3008 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3009 could be interpreted as a particular specialization of a template
3010 function; the warning exists to diagnose compatibility problems,
3011 and is enabled by default.
3013 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3014 @opindex Wold-style-cast
3015 @opindex Wno-old-style-cast
3016 Warn if an old-style (C-style) cast to a non-void type is used within
3017 a C++ program. The new-style casts (@code{dynamic_cast},
3018 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3019 less vulnerable to unintended effects and much easier to search for.
3021 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3022 @opindex Woverloaded-virtual
3023 @opindex Wno-overloaded-virtual
3024 @cindex overloaded virtual function, warning
3025 @cindex warning for overloaded virtual function
3026 Warn when a function declaration hides virtual functions from a
3027 base class. For example, in:
3034 struct B: public A @{
3039 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3050 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3051 @opindex Wno-pmf-conversions
3052 @opindex Wpmf-conversions
3053 Disable the diagnostic for converting a bound pointer to member function
3056 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3057 @opindex Wsign-promo
3058 @opindex Wno-sign-promo
3059 Warn when overload resolution chooses a promotion from unsigned or
3060 enumerated type to a signed type, over a conversion to an unsigned type of
3061 the same size. Previous versions of G++ tried to preserve
3062 unsignedness, but the standard mandates the current behavior.
3064 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3066 Warn when a primary template declaration is encountered. Some coding
3067 rules disallow templates, and this may be used to enforce that rule.
3068 The warning is inactive inside a system header file, such as the STL, so
3069 one can still use the STL. One may also instantiate or specialize
3072 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3073 @opindex Wmultiple-inheritance
3074 Warn when a class is defined with multiple direct base classes. Some
3075 coding rules disallow multiple inheritance, and this may be used to
3076 enforce that rule. The warning is inactive inside a system header file,
3077 such as the STL, so one can still use the STL. One may also define
3078 classes that indirectly use multiple inheritance.
3080 @item -Wvirtual-inheritance
3081 @opindex Wvirtual-inheritance
3082 Warn when a class is defined with a virtual direct base class. Some
3083 coding rules disallow multiple inheritance, and this may be used to
3084 enforce that rule. The warning is inactive inside a system header file,
3085 such as the STL, so one can still use the STL. One may also define
3086 classes that indirectly use virtual inheritance.
3089 @opindex Wnamespaces
3090 Warn when a namespace definition is opened. Some coding rules disallow
3091 namespaces, and this may be used to enforce that rule. The warning is
3092 inactive inside a system header file, such as the STL, so one can still
3093 use the STL. One may also use using directives and qualified names.
3095 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3097 @opindex Wno-terminate
3098 Disable the warning about a throw-expression that will immediately
3099 result in a call to @code{terminate}.
3102 @node Objective-C and Objective-C++ Dialect Options
3103 @section Options Controlling Objective-C and Objective-C++ Dialects
3105 @cindex compiler options, Objective-C and Objective-C++
3106 @cindex Objective-C and Objective-C++ options, command-line
3107 @cindex options, Objective-C and Objective-C++
3108 (NOTE: This manual does not describe the Objective-C and Objective-C++
3109 languages themselves. @xref{Standards,,Language Standards
3110 Supported by GCC}, for references.)
3112 This section describes the command-line options that are only meaningful
3113 for Objective-C and Objective-C++ programs. You can also use most of
3114 the language-independent GNU compiler options.
3115 For example, you might compile a file @file{some_class.m} like this:
3118 gcc -g -fgnu-runtime -O -c some_class.m
3122 In this example, @option{-fgnu-runtime} is an option meant only for
3123 Objective-C and Objective-C++ programs; you can use the other options with
3124 any language supported by GCC@.
3126 Note that since Objective-C is an extension of the C language, Objective-C
3127 compilations may also use options specific to the C front-end (e.g.,
3128 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3129 C++-specific options (e.g., @option{-Wabi}).
3131 Here is a list of options that are @emph{only} for compiling Objective-C
3132 and Objective-C++ programs:
3135 @item -fconstant-string-class=@var{class-name}
3136 @opindex fconstant-string-class
3137 Use @var{class-name} as the name of the class to instantiate for each
3138 literal string specified with the syntax @code{@@"@dots{}"}. The default
3139 class name is @code{NXConstantString} if the GNU runtime is being used, and
3140 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3141 @option{-fconstant-cfstrings} option, if also present, overrides the
3142 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3143 to be laid out as constant CoreFoundation strings.
3146 @opindex fgnu-runtime
3147 Generate object code compatible with the standard GNU Objective-C
3148 runtime. This is the default for most types of systems.
3150 @item -fnext-runtime
3151 @opindex fnext-runtime
3152 Generate output compatible with the NeXT runtime. This is the default
3153 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3154 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3157 @item -fno-nil-receivers
3158 @opindex fno-nil-receivers
3159 Assume that all Objective-C message dispatches (@code{[receiver
3160 message:arg]}) in this translation unit ensure that the receiver is
3161 not @code{nil}. This allows for more efficient entry points in the
3162 runtime to be used. This option is only available in conjunction with
3163 the NeXT runtime and ABI version 0 or 1.
3165 @item -fobjc-abi-version=@var{n}
3166 @opindex fobjc-abi-version
3167 Use version @var{n} of the Objective-C ABI for the selected runtime.
3168 This option is currently supported only for the NeXT runtime. In that
3169 case, Version 0 is the traditional (32-bit) ABI without support for
3170 properties and other Objective-C 2.0 additions. Version 1 is the
3171 traditional (32-bit) ABI with support for properties and other
3172 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3173 nothing is specified, the default is Version 0 on 32-bit target
3174 machines, and Version 2 on 64-bit target machines.
3176 @item -fobjc-call-cxx-cdtors
3177 @opindex fobjc-call-cxx-cdtors
3178 For each Objective-C class, check if any of its instance variables is a
3179 C++ object with a non-trivial default constructor. If so, synthesize a
3180 special @code{- (id) .cxx_construct} instance method which runs
3181 non-trivial default constructors on any such instance variables, in order,
3182 and then return @code{self}. Similarly, check if any instance variable
3183 is a C++ object with a non-trivial destructor, and if so, synthesize a
3184 special @code{- (void) .cxx_destruct} method which runs
3185 all such default destructors, in reverse order.
3187 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3188 methods thusly generated only operate on instance variables
3189 declared in the current Objective-C class, and not those inherited
3190 from superclasses. It is the responsibility of the Objective-C
3191 runtime to invoke all such methods in an object's inheritance
3192 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3193 by the runtime immediately after a new object instance is allocated;
3194 the @code{- (void) .cxx_destruct} methods are invoked immediately
3195 before the runtime deallocates an object instance.
3197 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3198 support for invoking the @code{- (id) .cxx_construct} and
3199 @code{- (void) .cxx_destruct} methods.
3201 @item -fobjc-direct-dispatch
3202 @opindex fobjc-direct-dispatch
3203 Allow fast jumps to the message dispatcher. On Darwin this is
3204 accomplished via the comm page.
3206 @item -fobjc-exceptions
3207 @opindex fobjc-exceptions
3208 Enable syntactic support for structured exception handling in
3209 Objective-C, similar to what is offered by C++. This option
3210 is required to use the Objective-C keywords @code{@@try},
3211 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3212 @code{@@synchronized}. This option is available with both the GNU
3213 runtime and the NeXT runtime (but not available in conjunction with
3214 the NeXT runtime on Mac OS X 10.2 and earlier).
3218 Enable garbage collection (GC) in Objective-C and Objective-C++
3219 programs. This option is only available with the NeXT runtime; the
3220 GNU runtime has a different garbage collection implementation that
3221 does not require special compiler flags.
3223 @item -fobjc-nilcheck
3224 @opindex fobjc-nilcheck
3225 For the NeXT runtime with version 2 of the ABI, check for a nil
3226 receiver in method invocations before doing the actual method call.
3227 This is the default and can be disabled using
3228 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3229 checked for nil in this way no matter what this flag is set to.
3230 Currently this flag does nothing when the GNU runtime, or an older
3231 version of the NeXT runtime ABI, is used.
3233 @item -fobjc-std=objc1
3235 Conform to the language syntax of Objective-C 1.0, the language
3236 recognized by GCC 4.0. This only affects the Objective-C additions to
3237 the C/C++ language; it does not affect conformance to C/C++ standards,
3238 which is controlled by the separate C/C++ dialect option flags. When
3239 this option is used with the Objective-C or Objective-C++ compiler,
3240 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3241 This is useful if you need to make sure that your Objective-C code can
3242 be compiled with older versions of GCC@.
3244 @item -freplace-objc-classes
3245 @opindex freplace-objc-classes
3246 Emit a special marker instructing @command{ld(1)} not to statically link in
3247 the resulting object file, and allow @command{dyld(1)} to load it in at
3248 run time instead. This is used in conjunction with the Fix-and-Continue
3249 debugging mode, where the object file in question may be recompiled and
3250 dynamically reloaded in the course of program execution, without the need
3251 to restart the program itself. Currently, Fix-and-Continue functionality
3252 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3257 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3258 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3259 compile time) with static class references that get initialized at load time,
3260 which improves run-time performance. Specifying the @option{-fzero-link} flag
3261 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3262 to be retained. This is useful in Zero-Link debugging mode, since it allows
3263 for individual class implementations to be modified during program execution.
3264 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3265 regardless of command-line options.
3267 @item -fno-local-ivars
3268 @opindex fno-local-ivars
3269 @opindex flocal-ivars
3270 By default instance variables in Objective-C can be accessed as if
3271 they were local variables from within the methods of the class they're
3272 declared in. This can lead to shadowing between instance variables
3273 and other variables declared either locally inside a class method or
3274 globally with the same name. Specifying the @option{-fno-local-ivars}
3275 flag disables this behavior thus avoiding variable shadowing issues.
3277 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3278 @opindex fivar-visibility
3279 Set the default instance variable visibility to the specified option
3280 so that instance variables declared outside the scope of any access
3281 modifier directives default to the specified visibility.
3285 Dump interface declarations for all classes seen in the source file to a
3286 file named @file{@var{sourcename}.decl}.
3288 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3289 @opindex Wassign-intercept
3290 @opindex Wno-assign-intercept
3291 Warn whenever an Objective-C assignment is being intercepted by the
3294 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3295 @opindex Wno-protocol
3297 If a class is declared to implement a protocol, a warning is issued for
3298 every method in the protocol that is not implemented by the class. The
3299 default behavior is to issue a warning for every method not explicitly
3300 implemented in the class, even if a method implementation is inherited
3301 from the superclass. If you use the @option{-Wno-protocol} option, then
3302 methods inherited from the superclass are considered to be implemented,
3303 and no warning is issued for them.
3305 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3307 @opindex Wno-selector
3308 Warn if multiple methods of different types for the same selector are
3309 found during compilation. The check is performed on the list of methods
3310 in the final stage of compilation. Additionally, a check is performed
3311 for each selector appearing in a @code{@@selector(@dots{})}
3312 expression, and a corresponding method for that selector has been found
3313 during compilation. Because these checks scan the method table only at
3314 the end of compilation, these warnings are not produced if the final
3315 stage of compilation is not reached, for example because an error is
3316 found during compilation, or because the @option{-fsyntax-only} option is
3319 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3320 @opindex Wstrict-selector-match
3321 @opindex Wno-strict-selector-match
3322 Warn if multiple methods with differing argument and/or return types are
3323 found for a given selector when attempting to send a message using this
3324 selector to a receiver of type @code{id} or @code{Class}. When this flag
3325 is off (which is the default behavior), the compiler omits such warnings
3326 if any differences found are confined to types that share the same size
3329 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3330 @opindex Wundeclared-selector
3331 @opindex Wno-undeclared-selector
3332 Warn if a @code{@@selector(@dots{})} expression referring to an
3333 undeclared selector is found. A selector is considered undeclared if no
3334 method with that name has been declared before the
3335 @code{@@selector(@dots{})} expression, either explicitly in an
3336 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3337 an @code{@@implementation} section. This option always performs its
3338 checks as soon as a @code{@@selector(@dots{})} expression is found,
3339 while @option{-Wselector} only performs its checks in the final stage of
3340 compilation. This also enforces the coding style convention
3341 that methods and selectors must be declared before being used.
3343 @item -print-objc-runtime-info
3344 @opindex print-objc-runtime-info
3345 Generate C header describing the largest structure that is passed by
3350 @node Diagnostic Message Formatting Options
3351 @section Options to Control Diagnostic Messages Formatting
3352 @cindex options to control diagnostics formatting
3353 @cindex diagnostic messages
3354 @cindex message formatting
3356 Traditionally, diagnostic messages have been formatted irrespective of
3357 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3358 options described below
3359 to control the formatting algorithm for diagnostic messages,
3360 e.g.@: how many characters per line, how often source location
3361 information should be reported. Note that some language front ends may not
3362 honor these options.
3365 @item -fmessage-length=@var{n}
3366 @opindex fmessage-length
3367 Try to format error messages so that they fit on lines of about
3368 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3369 done; each error message appears on a single line. This is the
3370 default for all front ends.
3372 @item -fdiagnostics-show-location=once
3373 @opindex fdiagnostics-show-location
3374 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3375 reporter to emit source location information @emph{once}; that is, in
3376 case the message is too long to fit on a single physical line and has to
3377 be wrapped, the source location won't be emitted (as prefix) again,
3378 over and over, in subsequent continuation lines. This is the default
3381 @item -fdiagnostics-show-location=every-line
3382 Only meaningful in line-wrapping mode. Instructs the diagnostic
3383 messages reporter to emit the same source location information (as
3384 prefix) for physical lines that result from the process of breaking
3385 a message which is too long to fit on a single line.
3387 @item -fdiagnostics-color[=@var{WHEN}]
3388 @itemx -fno-diagnostics-color
3389 @opindex fdiagnostics-color
3390 @cindex highlight, color
3391 @vindex GCC_COLORS @r{environment variable}
3392 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3393 or @samp{auto}. The default depends on how the compiler has been configured,
3394 it can be any of the above @var{WHEN} options or also @samp{never}
3395 if @env{GCC_COLORS} environment variable isn't present in the environment,
3396 and @samp{auto} otherwise.
3397 @samp{auto} means to use color only when the standard error is a terminal.
3398 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3399 aliases for @option{-fdiagnostics-color=always} and
3400 @option{-fdiagnostics-color=never}, respectively.
3402 The colors are defined by the environment variable @env{GCC_COLORS}.
3403 Its value is a colon-separated list of capabilities and Select Graphic
3404 Rendition (SGR) substrings. SGR commands are interpreted by the
3405 terminal or terminal emulator. (See the section in the documentation
3406 of your text terminal for permitted values and their meanings as
3407 character attributes.) These substring values are integers in decimal
3408 representation and can be concatenated with semicolons.
3409 Common values to concatenate include
3411 @samp{4} for underline,
3413 @samp{7} for inverse,
3414 @samp{39} for default foreground color,
3415 @samp{30} to @samp{37} for foreground colors,
3416 @samp{90} to @samp{97} for 16-color mode foreground colors,
3417 @samp{38;5;0} to @samp{38;5;255}
3418 for 88-color and 256-color modes foreground colors,
3419 @samp{49} for default background color,
3420 @samp{40} to @samp{47} for background colors,
3421 @samp{100} to @samp{107} for 16-color mode background colors,
3422 and @samp{48;5;0} to @samp{48;5;255}
3423 for 88-color and 256-color modes background colors.
3425 The default @env{GCC_COLORS} is
3427 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3428 quote=01:fixit-insert=32:fixit-delete=31:\
3429 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32
3432 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3433 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3434 @samp{01} is bold, and @samp{31} is red.
3435 Setting @env{GCC_COLORS} to the empty string disables colors.
3436 Supported capabilities are as follows.
3440 @vindex error GCC_COLORS @r{capability}
3441 SGR substring for error: markers.
3444 @vindex warning GCC_COLORS @r{capability}
3445 SGR substring for warning: markers.
3448 @vindex note GCC_COLORS @r{capability}
3449 SGR substring for note: markers.
3452 @vindex range1 GCC_COLORS @r{capability}
3453 SGR substring for first additional range.
3456 @vindex range2 GCC_COLORS @r{capability}
3457 SGR substring for second additional range.
3460 @vindex locus GCC_COLORS @r{capability}
3461 SGR substring for location information, @samp{file:line} or
3462 @samp{file:line:column} etc.
3465 @vindex quote GCC_COLORS @r{capability}
3466 SGR substring for information printed within quotes.
3469 @vindex fixit-insert GCC_COLORS @r{capability}
3470 SGR substring for fix-it hints suggesting text to
3471 be inserted or replaced.
3474 @vindex fixit-delete GCC_COLORS @r{capability}
3475 SGR substring for fix-it hints suggesting text to
3478 @item diff-filename=
3479 @vindex diff-filename GCC_COLORS @r{capability}
3480 SGR substring for filename headers within generated patches.
3483 @vindex diff-hunk GCC_COLORS @r{capability}
3484 SGR substring for the starts of hunks within generated patches.
3487 @vindex diff-delete GCC_COLORS @r{capability}
3488 SGR substring for deleted lines within generated patches.
3491 @vindex diff-insert GCC_COLORS @r{capability}
3492 SGR substring for inserted lines within generated patches.
3495 @item -fno-diagnostics-show-option
3496 @opindex fno-diagnostics-show-option
3497 @opindex fdiagnostics-show-option
3498 By default, each diagnostic emitted includes text indicating the
3499 command-line option that directly controls the diagnostic (if such an
3500 option is known to the diagnostic machinery). Specifying the
3501 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3503 @item -fno-diagnostics-show-caret
3504 @opindex fno-diagnostics-show-caret
3505 @opindex fdiagnostics-show-caret
3506 By default, each diagnostic emitted includes the original source line
3507 and a caret @samp{^} indicating the column. This option suppresses this
3508 information. The source line is truncated to @var{n} characters, if
3509 the @option{-fmessage-length=n} option is given. When the output is done
3510 to the terminal, the width is limited to the width given by the
3511 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3513 @item -fdiagnostics-parseable-fixits
3514 @opindex fdiagnostics-parseable-fixits
3515 Emit fix-it hints in a machine-parseable format, suitable for consumption
3516 by IDEs. For each fix-it, a line will be printed after the relevant
3517 diagnostic, starting with the string ``fix-it:''. For example:
3520 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3523 The location is expressed as a half-open range, expressed as a count of
3524 bytes, starting at byte 1 for the initial column. In the above example,
3525 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3529 00000000011111111112222222222
3530 12345678901234567890123456789
3531 gtk_widget_showall (dlg);
3536 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3537 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3538 (e.g. vertical tab as ``\013'').
3540 An empty replacement string indicates that the given range is to be removed.
3541 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3542 be inserted at the given position.
3544 @item -fdiagnostics-generate-patch
3545 @opindex fdiagnostics-generate-patch
3546 Print fix-it hints to stderr in unified diff format, after any diagnostics
3547 are printed. For example:
3554 void show_cb(GtkDialog *dlg)
3556 - gtk_widget_showall(dlg);
3557 + gtk_widget_show_all(dlg);
3562 The diff may or may not be colorized, following the same rules
3563 as for diagnostics (see @option{-fdiagnostics-color}).
3565 @item -fno-show-column
3566 @opindex fno-show-column
3567 Do not print column numbers in diagnostics. This may be necessary if
3568 diagnostics are being scanned by a program that does not understand the
3569 column numbers, such as @command{dejagnu}.
3573 @node Warning Options
3574 @section Options to Request or Suppress Warnings
3575 @cindex options to control warnings
3576 @cindex warning messages
3577 @cindex messages, warning
3578 @cindex suppressing warnings
3580 Warnings are diagnostic messages that report constructions that
3581 are not inherently erroneous but that are risky or suggest there
3582 may have been an error.
3584 The following language-independent options do not enable specific
3585 warnings but control the kinds of diagnostics produced by GCC@.
3588 @cindex syntax checking
3590 @opindex fsyntax-only
3591 Check the code for syntax errors, but don't do anything beyond that.
3593 @item -fmax-errors=@var{n}
3594 @opindex fmax-errors
3595 Limits the maximum number of error messages to @var{n}, at which point
3596 GCC bails out rather than attempting to continue processing the source
3597 code. If @var{n} is 0 (the default), there is no limit on the number
3598 of error messages produced. If @option{-Wfatal-errors} is also
3599 specified, then @option{-Wfatal-errors} takes precedence over this
3604 Inhibit all warning messages.
3609 Make all warnings into errors.
3614 Make the specified warning into an error. The specifier for a warning
3615 is appended; for example @option{-Werror=switch} turns the warnings
3616 controlled by @option{-Wswitch} into errors. This switch takes a
3617 negative form, to be used to negate @option{-Werror} for specific
3618 warnings; for example @option{-Wno-error=switch} makes
3619 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3622 The warning message for each controllable warning includes the
3623 option that controls the warning. That option can then be used with
3624 @option{-Werror=} and @option{-Wno-error=} as described above.
3625 (Printing of the option in the warning message can be disabled using the
3626 @option{-fno-diagnostics-show-option} flag.)
3628 Note that specifying @option{-Werror=}@var{foo} automatically implies
3629 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3632 @item -Wfatal-errors
3633 @opindex Wfatal-errors
3634 @opindex Wno-fatal-errors
3635 This option causes the compiler to abort compilation on the first error
3636 occurred rather than trying to keep going and printing further error
3641 You can request many specific warnings with options beginning with
3642 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3643 implicit declarations. Each of these specific warning options also
3644 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3645 example, @option{-Wno-implicit}. This manual lists only one of the
3646 two forms, whichever is not the default. For further
3647 language-specific options also refer to @ref{C++ Dialect Options} and
3648 @ref{Objective-C and Objective-C++ Dialect Options}.
3650 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3651 options, such as @option{-Wunused}, which may turn on further options,
3652 such as @option{-Wunused-value}. The combined effect of positive and
3653 negative forms is that more specific options have priority over less
3654 specific ones, independently of their position in the command-line. For
3655 options of the same specificity, the last one takes effect. Options
3656 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3657 as if they appeared at the end of the command-line.
3659 When an unrecognized warning option is requested (e.g.,
3660 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3661 that the option is not recognized. However, if the @option{-Wno-} form
3662 is used, the behavior is slightly different: no diagnostic is
3663 produced for @option{-Wno-unknown-warning} unless other diagnostics
3664 are being produced. This allows the use of new @option{-Wno-} options
3665 with old compilers, but if something goes wrong, the compiler
3666 warns that an unrecognized option is present.
3673 Issue all the warnings demanded by strict ISO C and ISO C++;
3674 reject all programs that use forbidden extensions, and some other
3675 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3676 version of the ISO C standard specified by any @option{-std} option used.
3678 Valid ISO C and ISO C++ programs should compile properly with or without
3679 this option (though a rare few require @option{-ansi} or a
3680 @option{-std} option specifying the required version of ISO C)@. However,
3681 without this option, certain GNU extensions and traditional C and C++
3682 features are supported as well. With this option, they are rejected.
3684 @option{-Wpedantic} does not cause warning messages for use of the
3685 alternate keywords whose names begin and end with @samp{__}. Pedantic
3686 warnings are also disabled in the expression that follows
3687 @code{__extension__}. However, only system header files should use
3688 these escape routes; application programs should avoid them.
3689 @xref{Alternate Keywords}.
3691 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3692 C conformance. They soon find that it does not do quite what they want:
3693 it finds some non-ISO practices, but not all---only those for which
3694 ISO C @emph{requires} a diagnostic, and some others for which
3695 diagnostics have been added.
3697 A feature to report any failure to conform to ISO C might be useful in
3698 some instances, but would require considerable additional work and would
3699 be quite different from @option{-Wpedantic}. We don't have plans to
3700 support such a feature in the near future.
3702 Where the standard specified with @option{-std} represents a GNU
3703 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3704 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3705 extended dialect is based. Warnings from @option{-Wpedantic} are given
3706 where they are required by the base standard. (It does not make sense
3707 for such warnings to be given only for features not in the specified GNU
3708 C dialect, since by definition the GNU dialects of C include all
3709 features the compiler supports with the given option, and there would be
3710 nothing to warn about.)
3712 @item -pedantic-errors
3713 @opindex pedantic-errors
3714 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3715 requires a diagnostic, in some cases where there is undefined behavior
3716 at compile-time and in some other cases that do not prevent compilation
3717 of programs that are valid according to the standard. This is not
3718 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3719 by this option and not enabled by the latter and vice versa.
3724 This enables all the warnings about constructions that some users
3725 consider questionable, and that are easy to avoid (or modify to
3726 prevent the warning), even in conjunction with macros. This also
3727 enables some language-specific warnings described in @ref{C++ Dialect
3728 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3730 @option{-Wall} turns on the following warning flags:
3732 @gccoptlist{-Waddress @gol
3733 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3735 -Wbool-operation @gol
3736 -Wc++11-compat -Wc++14-compat @gol
3737 -Wchar-subscripts @gol
3739 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3740 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3742 -Wint-in-bool-context @gol
3743 -Wimplicit @r{(C and Objective-C only)} @gol
3744 -Wimplicit-int @r{(C and Objective-C only)} @gol
3745 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3746 -Winit-self @r{(only for C++)} @gol
3747 -Wlogical-not-parentheses @gol
3748 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3749 -Wmaybe-uninitialized @gol
3750 -Wmemset-elt-size @gol
3751 -Wmemset-transposed-args @gol
3752 -Wmisleading-indentation @r{(only for C/C++)} @gol
3753 -Wmissing-braces @r{(only for C/ObjC)} @gol
3754 -Wnarrowing @r{(only for C++)} @gol
3756 -Wnonnull-compare @gol
3762 -Wsequence-point @gol
3763 -Wsign-compare @r{(only in C++)} @gol
3764 -Wsizeof-pointer-memaccess @gol
3765 -Wstrict-aliasing @gol
3766 -Wstrict-overflow=1 @gol
3768 -Wtautological-compare @gol
3770 -Wuninitialized @gol
3771 -Wunknown-pragmas @gol
3772 -Wunused-function @gol
3775 -Wunused-variable @gol
3776 -Wvolatile-register-var @gol
3779 Note that some warning flags are not implied by @option{-Wall}. Some of
3780 them warn about constructions that users generally do not consider
3781 questionable, but which occasionally you might wish to check for;
3782 others warn about constructions that are necessary or hard to avoid in
3783 some cases, and there is no simple way to modify the code to suppress
3784 the warning. Some of them are enabled by @option{-Wextra} but many of
3785 them must be enabled individually.
3791 This enables some extra warning flags that are not enabled by
3792 @option{-Wall}. (This option used to be called @option{-W}. The older
3793 name is still supported, but the newer name is more descriptive.)
3795 @gccoptlist{-Wclobbered @gol
3797 -Wignored-qualifiers @gol
3798 -Wimplicit-fallthrough=3 @gol
3799 -Wmissing-field-initializers @gol
3800 -Wmissing-parameter-type @r{(C only)} @gol
3801 -Wold-style-declaration @r{(C only)} @gol
3802 -Woverride-init @gol
3803 -Wsign-compare @r{(C only)} @gol
3805 -Wuninitialized @gol
3806 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3807 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3808 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3811 The option @option{-Wextra} also prints warning messages for the
3817 A pointer is compared against integer zero with @code{<}, @code{<=},
3818 @code{>}, or @code{>=}.
3821 (C++ only) An enumerator and a non-enumerator both appear in a
3822 conditional expression.
3825 (C++ only) Ambiguous virtual bases.
3828 (C++ only) Subscripting an array that has been declared @code{register}.
3831 (C++ only) Taking the address of a variable that has been declared
3835 (C++ only) A base class is not initialized in the copy constructor
3840 @item -Wchar-subscripts
3841 @opindex Wchar-subscripts
3842 @opindex Wno-char-subscripts
3843 Warn if an array subscript has type @code{char}. This is a common cause
3844 of error, as programmers often forget that this type is signed on some
3846 This warning is enabled by @option{-Wall}.
3848 @item -Wno-coverage-mismatch
3849 @opindex Wno-coverage-mismatch
3850 Warn if feedback profiles do not match when using the
3851 @option{-fprofile-use} option.
3852 If a source file is changed between compiling with @option{-fprofile-gen} and
3853 with @option{-fprofile-use}, the files with the profile feedback can fail
3854 to match the source file and GCC cannot use the profile feedback
3855 information. By default, this warning is enabled and is treated as an
3856 error. @option{-Wno-coverage-mismatch} can be used to disable the
3857 warning or @option{-Wno-error=coverage-mismatch} can be used to
3858 disable the error. Disabling the error for this warning can result in
3859 poorly optimized code and is useful only in the
3860 case of very minor changes such as bug fixes to an existing code-base.
3861 Completely disabling the warning is not recommended.
3864 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3866 Suppress warning messages emitted by @code{#warning} directives.
3868 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3869 @opindex Wdouble-promotion
3870 @opindex Wno-double-promotion
3871 Give a warning when a value of type @code{float} is implicitly
3872 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3873 floating-point unit implement @code{float} in hardware, but emulate
3874 @code{double} in software. On such a machine, doing computations
3875 using @code{double} values is much more expensive because of the
3876 overhead required for software emulation.
3878 It is easy to accidentally do computations with @code{double} because
3879 floating-point literals are implicitly of type @code{double}. For
3883 float area(float radius)
3885 return 3.14159 * radius * radius;
3889 the compiler performs the entire computation with @code{double}
3890 because the floating-point literal is a @code{double}.
3892 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3893 @opindex Wduplicate-decl-specifier
3894 @opindex Wno-duplicate-decl-specifier
3895 Warn if a declaration has duplicate @code{const}, @code{volatile},
3896 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3900 @itemx -Wformat=@var{n}
3903 @opindex ffreestanding
3904 @opindex fno-builtin
3906 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3907 the arguments supplied have types appropriate to the format string
3908 specified, and that the conversions specified in the format string make
3909 sense. This includes standard functions, and others specified by format
3910 attributes (@pxref{Function Attributes}), in the @code{printf},
3911 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3912 not in the C standard) families (or other target-specific families).
3913 Which functions are checked without format attributes having been
3914 specified depends on the standard version selected, and such checks of
3915 functions without the attribute specified are disabled by
3916 @option{-ffreestanding} or @option{-fno-builtin}.
3918 The formats are checked against the format features supported by GNU
3919 libc version 2.2. These include all ISO C90 and C99 features, as well
3920 as features from the Single Unix Specification and some BSD and GNU
3921 extensions. Other library implementations may not support all these
3922 features; GCC does not support warning about features that go beyond a
3923 particular library's limitations. However, if @option{-Wpedantic} is used
3924 with @option{-Wformat}, warnings are given about format features not
3925 in the selected standard version (but not for @code{strfmon} formats,
3926 since those are not in any version of the C standard). @xref{C Dialect
3927 Options,,Options Controlling C Dialect}.
3934 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3935 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3936 @option{-Wformat} also checks for null format arguments for several
3937 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3938 aspects of this level of format checking can be disabled by the
3939 options: @option{-Wno-format-contains-nul},
3940 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3941 @option{-Wformat} is enabled by @option{-Wall}.
3943 @item -Wno-format-contains-nul
3944 @opindex Wno-format-contains-nul
3945 @opindex Wformat-contains-nul
3946 If @option{-Wformat} is specified, do not warn about format strings that
3949 @item -Wno-format-extra-args
3950 @opindex Wno-format-extra-args
3951 @opindex Wformat-extra-args
3952 If @option{-Wformat} is specified, do not warn about excess arguments to a
3953 @code{printf} or @code{scanf} format function. The C standard specifies
3954 that such arguments are ignored.
3956 Where the unused arguments lie between used arguments that are
3957 specified with @samp{$} operand number specifications, normally
3958 warnings are still given, since the implementation could not know what
3959 type to pass to @code{va_arg} to skip the unused arguments. However,
3960 in the case of @code{scanf} formats, this option suppresses the
3961 warning if the unused arguments are all pointers, since the Single
3962 Unix Specification says that such unused arguments are allowed.
3964 @item -Wformat-overflow
3965 @itemx -Wformat-overflow=@var{level}
3966 @opindex Wformat-overflow
3967 @opindex Wno-format-overflow
3968 Warn about calls to formatted input/output functions such as @code{sprintf}
3969 and @code{vsprintf} that might overflow the destination buffer. When the
3970 exact number of bytes written by a format directive cannot be determined
3971 at compile-time it is estimated based on heuristics that depend on the
3972 @var{level} argument and on optimization. While enabling optimization
3973 will in most cases improve the accuracy of the warning, it may also
3974 result in false positives.
3977 @item -Wformat-overflow
3978 @item -Wformat-overflow=1
3979 @opindex Wformat-overflow
3980 @opindex Wno-format-overflow
3981 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
3982 employs a conservative approach that warns only about calls that most
3983 likely overflow the buffer. At this level, numeric arguments to format
3984 directives with unknown values are assumed to have the value of one, and
3985 strings of unknown length to be empty. Numeric arguments that are known
3986 to be bounded to a subrange of their type, or string arguments whose output
3987 is bounded either by their directive's precision or by a finite set of
3988 string literals, are assumed to take on the value within the range that
3989 results in the most bytes on output. For example, the call to @code{sprintf}
3990 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
3991 the terminating NUL character (@code{'\0'}) appended by the function
3992 to the destination buffer will be written past its end. Increasing
3993 the size of the buffer by a single byte is sufficient to avoid the
3994 warning, though it may not be sufficient to avoid the overflow.
3997 void f (int a, int b)
4000 sprintf (buf, "a = %i, b = %i\n", a, b);
4004 @item -Wformat-overflow=2
4005 Level @var{2} warns also about calls that might overflow the destination
4006 buffer given an argument of sufficient length or magnitude. At level
4007 @var{2}, unknown numeric arguments are assumed to have the minimum
4008 representable value for signed types with a precision greater than 1, and
4009 the maximum representable value otherwise. Unknown string arguments whose
4010 length cannot be assumed to be bounded either by the directive's precision,
4011 or by a finite set of string literals they may evaluate to, or the character
4012 array they may point to, are assumed to be 1 character long.
4014 At level @var{2}, the call in the example above is again diagnosed, but
4015 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4016 @code{%i} directive will write some of its digits beyond the end of
4017 the destination buffer. To make the call safe regardless of the values
4018 of the two variables, the size of the destination buffer must be increased
4019 to at least 34 bytes. GCC includes the minimum size of the buffer in
4020 an informational note following the warning.
4022 An alternative to increasing the size of the destination buffer is to
4023 constrain the range of formatted values. The maximum length of string
4024 arguments can be bounded by specifying the precision in the format
4025 directive. When numeric arguments of format directives can be assumed
4026 to be bounded by less than the precision of their type, choosing
4027 an appropriate length modifier to the format specifier will reduce
4028 the required buffer size. For example, if @var{a} and @var{b} in the
4029 example above can be assumed to be within the precision of
4030 the @code{short int} type then using either the @code{%hi} format
4031 directive or casting the argument to @code{short} reduces the maximum
4032 required size of the buffer to 24 bytes.
4035 void f (int a, int b)
4038 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4043 @item -Wno-format-zero-length
4044 @opindex Wno-format-zero-length
4045 @opindex Wformat-zero-length
4046 If @option{-Wformat} is specified, do not warn about zero-length formats.
4047 The C standard specifies that zero-length formats are allowed.
4052 Enable @option{-Wformat} plus additional format checks. Currently
4053 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4056 @item -Wformat-nonliteral
4057 @opindex Wformat-nonliteral
4058 @opindex Wno-format-nonliteral
4059 If @option{-Wformat} is specified, also warn if the format string is not a
4060 string literal and so cannot be checked, unless the format function
4061 takes its format arguments as a @code{va_list}.
4063 @item -Wformat-security
4064 @opindex Wformat-security
4065 @opindex Wno-format-security
4066 If @option{-Wformat} is specified, also warn about uses of format
4067 functions that represent possible security problems. At present, this
4068 warns about calls to @code{printf} and @code{scanf} functions where the
4069 format string is not a string literal and there are no format arguments,
4070 as in @code{printf (foo);}. This may be a security hole if the format
4071 string came from untrusted input and contains @samp{%n}. (This is
4072 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4073 in future warnings may be added to @option{-Wformat-security} that are not
4074 included in @option{-Wformat-nonliteral}.)
4076 @item -Wformat-signedness
4077 @opindex Wformat-signedness
4078 @opindex Wno-format-signedness
4079 If @option{-Wformat} is specified, also warn if the format string
4080 requires an unsigned argument and the argument is signed and vice versa.
4082 @item -Wformat-truncation
4083 @itemx -Wformat-truncation=@var{level}
4084 @opindex Wformat-truncation
4085 @opindex Wno-format-truncation
4086 Warn about calls to formatted input/output functions such as @code{snprintf}
4087 and @code{vsnprintf} that might result in output truncation. When the exact
4088 number of bytes written by a format directive cannot be determined at
4089 compile-time it is estimated based on heuristics that depend on
4090 the @var{level} argument and on optimization. While enabling optimization
4091 will in most cases improve the accuracy of the warning, it may also result
4092 in false positives. Except as noted otherwise, the option uses the same
4093 logic @option{-Wformat-overflow}.
4096 @item -Wformat-truncation
4097 @item -Wformat-truncation=1
4098 @opindex Wformat-truncation
4099 @opindex Wno-format-overflow
4100 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4101 employs a conservative approach that warns only about calls to bounded
4102 functions whose return value is unused and that will most likely result
4103 in output truncation.
4105 @item -Wformat-truncation=2
4106 Level @var{2} warns also about calls to bounded functions whose return
4107 value is used and that might result in truncation given an argument of
4108 sufficient length or magnitude.
4112 @opindex Wformat-y2k
4113 @opindex Wno-format-y2k
4114 If @option{-Wformat} is specified, also warn about @code{strftime}
4115 formats that may yield only a two-digit year.
4120 @opindex Wno-nonnull
4121 Warn about passing a null pointer for arguments marked as
4122 requiring a non-null value by the @code{nonnull} function attribute.
4124 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4125 can be disabled with the @option{-Wno-nonnull} option.
4127 @item -Wnonnull-compare
4128 @opindex Wnonnull-compare
4129 @opindex Wno-nonnull-compare
4130 Warn when comparing an argument marked with the @code{nonnull}
4131 function attribute against null inside the function.
4133 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4134 can be disabled with the @option{-Wno-nonnull-compare} option.
4136 @item -Wnull-dereference
4137 @opindex Wnull-dereference
4138 @opindex Wno-null-dereference
4139 Warn if the compiler detects paths that trigger erroneous or
4140 undefined behavior due to dereferencing a null pointer. This option
4141 is only active when @option{-fdelete-null-pointer-checks} is active,
4142 which is enabled by optimizations in most targets. The precision of
4143 the warnings depends on the optimization options used.
4145 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4147 @opindex Wno-init-self
4148 Warn about uninitialized variables that are initialized with themselves.
4149 Note this option can only be used with the @option{-Wuninitialized} option.
4151 For example, GCC warns about @code{i} being uninitialized in the
4152 following snippet only when @option{-Winit-self} has been specified:
4163 This warning is enabled by @option{-Wall} in C++.
4165 @item -Wimplicit-int @r{(C and Objective-C only)}
4166 @opindex Wimplicit-int
4167 @opindex Wno-implicit-int
4168 Warn when a declaration does not specify a type.
4169 This warning is enabled by @option{-Wall}.
4171 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4172 @opindex Wimplicit-function-declaration
4173 @opindex Wno-implicit-function-declaration
4174 Give a warning whenever a function is used before being declared. In
4175 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4176 enabled by default and it is made into an error by
4177 @option{-pedantic-errors}. This warning is also enabled by
4180 @item -Wimplicit @r{(C and Objective-C only)}
4182 @opindex Wno-implicit
4183 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4184 This warning is enabled by @option{-Wall}.
4186 @item -Wimplicit-fallthrough
4187 @opindex Wimplicit-fallthrough
4188 @opindex Wno-implicit-fallthrough
4189 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4190 and @option{-Wno-implicit-fallthrough} is the same as
4191 @option{-Wimplicit-fallthrough=0}.
4193 @item -Wimplicit-fallthrough=@var{n}
4194 @opindex Wimplicit-fallthrough=
4195 Warn when a switch case falls through. For example:
4213 This warning does not warn when the last statement of a case cannot
4214 fall through, e.g. when there is a return statement or a call to function
4215 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4216 also takes into account control flow statements, such as ifs, and only
4217 warns when appropriate. E.g.@:
4227 @} else if (i < 1) @{
4237 Since there are occasions where a switch case fall through is desirable,
4238 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4239 to be used along with a null statement to suppress this warning that
4240 would normally occur:
4248 __attribute__ ((fallthrough));
4255 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4256 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4257 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4258 Instead of the these attributes, it is also possible to add a fallthrough
4259 comment to silence the warning. The whole body of the C or C++ style comment
4260 should match the given regular expressions listed below. The option argument
4261 @var{n} specifies what kind of comments are accepted:
4265 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4267 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4268 expression, any comment is used as fallthrough comment.
4270 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4271 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4273 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4274 following regular expressions:
4278 @item @code{-fallthrough}
4280 @item @code{@@fallthrough@@}
4282 @item @code{lint -fallthrough[ \t]*}
4284 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4286 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4288 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4292 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4293 following regular expressions:
4297 @item @code{-fallthrough}
4299 @item @code{@@fallthrough@@}
4301 @item @code{lint -fallthrough[ \t]*}
4303 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4307 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4308 fallthrough comments, only attributes disable the warning.
4312 The comment needs to be followed after optional whitespace and other comments
4313 by @code{case} or @code{default} keywords or by a user label that precedes some
4314 @code{case} or @code{default} label.
4329 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4331 @item -Wignored-qualifiers @r{(C and C++ only)}
4332 @opindex Wignored-qualifiers
4333 @opindex Wno-ignored-qualifiers
4334 Warn if the return type of a function has a type qualifier
4335 such as @code{const}. For ISO C such a type qualifier has no effect,
4336 since the value returned by a function is not an lvalue.
4337 For C++, the warning is only emitted for scalar types or @code{void}.
4338 ISO C prohibits qualified @code{void} return types on function
4339 definitions, so such return types always receive a warning
4340 even without this option.
4342 This warning is also enabled by @option{-Wextra}.
4344 @item -Wignored-attributes @r{(C and C++ only)}
4345 @opindex Wignored-attributes
4346 @opindex Wno-ignored-attributes
4347 Warn when an attribute is ignored. This is different from the
4348 @option{-Wattributes} option in that it warns whenever the compiler decides
4349 to drop an attribute, not that the attribute is either unknown, used in a
4350 wrong place, etc. This warning is enabled by default.
4355 Warn if the type of @code{main} is suspicious. @code{main} should be
4356 a function with external linkage, returning int, taking either zero
4357 arguments, two, or three arguments of appropriate types. This warning
4358 is enabled by default in C++ and is enabled by either @option{-Wall}
4359 or @option{-Wpedantic}.
4361 @item -Wmisleading-indentation @r{(C and C++ only)}
4362 @opindex Wmisleading-indentation
4363 @opindex Wno-misleading-indentation
4364 Warn when the indentation of the code does not reflect the block structure.
4365 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4366 @code{for} clauses with a guarded statement that does not use braces,
4367 followed by an unguarded statement with the same indentation.
4369 In the following example, the call to ``bar'' is misleadingly indented as
4370 if it were guarded by the ``if'' conditional.
4373 if (some_condition ())
4375 bar (); /* Gotcha: this is not guarded by the "if". */
4378 In the case of mixed tabs and spaces, the warning uses the
4379 @option{-ftabstop=} option to determine if the statements line up
4382 The warning is not issued for code involving multiline preprocessor logic
4383 such as the following example.
4388 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4394 The warning is not issued after a @code{#line} directive, since this
4395 typically indicates autogenerated code, and no assumptions can be made
4396 about the layout of the file that the directive references.
4398 This warning is enabled by @option{-Wall} in C and C++.
4400 @item -Wmissing-braces
4401 @opindex Wmissing-braces
4402 @opindex Wno-missing-braces
4403 Warn if an aggregate or union initializer is not fully bracketed. In
4404 the following example, the initializer for @code{a} is not fully
4405 bracketed, but that for @code{b} is fully bracketed. This warning is
4406 enabled by @option{-Wall} in C.
4409 int a[2][2] = @{ 0, 1, 2, 3 @};
4410 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4413 This warning is enabled by @option{-Wall}.
4415 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4416 @opindex Wmissing-include-dirs
4417 @opindex Wno-missing-include-dirs
4418 Warn if a user-supplied include directory does not exist.
4421 @opindex Wparentheses
4422 @opindex Wno-parentheses
4423 Warn if parentheses are omitted in certain contexts, such
4424 as when there is an assignment in a context where a truth value
4425 is expected, or when operators are nested whose precedence people
4426 often get confused about.
4428 Also warn if a comparison like @code{x<=y<=z} appears; this is
4429 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4430 interpretation from that of ordinary mathematical notation.
4432 Also warn for dangerous uses of the GNU extension to
4433 @code{?:} with omitted middle operand. When the condition
4434 in the @code{?}: operator is a boolean expression, the omitted value is
4435 always 1. Often programmers expect it to be a value computed
4436 inside the conditional expression instead.
4438 This warning is enabled by @option{-Wall}.
4440 @item -Wsequence-point
4441 @opindex Wsequence-point
4442 @opindex Wno-sequence-point
4443 Warn about code that may have undefined semantics because of violations
4444 of sequence point rules in the C and C++ standards.
4446 The C and C++ standards define the order in which expressions in a C/C++
4447 program are evaluated in terms of @dfn{sequence points}, which represent
4448 a partial ordering between the execution of parts of the program: those
4449 executed before the sequence point, and those executed after it. These
4450 occur after the evaluation of a full expression (one which is not part
4451 of a larger expression), after the evaluation of the first operand of a
4452 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4453 function is called (but after the evaluation of its arguments and the
4454 expression denoting the called function), and in certain other places.
4455 Other than as expressed by the sequence point rules, the order of
4456 evaluation of subexpressions of an expression is not specified. All
4457 these rules describe only a partial order rather than a total order,
4458 since, for example, if two functions are called within one expression
4459 with no sequence point between them, the order in which the functions
4460 are called is not specified. However, the standards committee have
4461 ruled that function calls do not overlap.
4463 It is not specified when between sequence points modifications to the
4464 values of objects take effect. Programs whose behavior depends on this
4465 have undefined behavior; the C and C++ standards specify that ``Between
4466 the previous and next sequence point an object shall have its stored
4467 value modified at most once by the evaluation of an expression.
4468 Furthermore, the prior value shall be read only to determine the value
4469 to be stored.''. If a program breaks these rules, the results on any
4470 particular implementation are entirely unpredictable.
4472 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4473 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4474 diagnosed by this option, and it may give an occasional false positive
4475 result, but in general it has been found fairly effective at detecting
4476 this sort of problem in programs.
4478 The C++17 standard will define the order of evaluation of operands in
4479 more cases: in particular it requires that the right-hand side of an
4480 assignment be evaluated before the left-hand side, so the above
4481 examples are no longer undefined. But this warning will still warn
4482 about them, to help people avoid writing code that is undefined in C
4483 and earlier revisions of C++.
4485 The standard is worded confusingly, therefore there is some debate
4486 over the precise meaning of the sequence point rules in subtle cases.
4487 Links to discussions of the problem, including proposed formal
4488 definitions, may be found on the GCC readings page, at
4489 @uref{http://gcc.gnu.org/@/readings.html}.
4491 This warning is enabled by @option{-Wall} for C and C++.
4493 @item -Wno-return-local-addr
4494 @opindex Wno-return-local-addr
4495 @opindex Wreturn-local-addr
4496 Do not warn about returning a pointer (or in C++, a reference) to a
4497 variable that goes out of scope after the function returns.
4500 @opindex Wreturn-type
4501 @opindex Wno-return-type
4502 Warn whenever a function is defined with a return type that defaults
4503 to @code{int}. Also warn about any @code{return} statement with no
4504 return value in a function whose return type is not @code{void}
4505 (falling off the end of the function body is considered returning
4508 For C only, warn about a @code{return} statement with an expression in a
4509 function whose return type is @code{void}, unless the expression type is
4510 also @code{void}. As a GNU extension, the latter case is accepted
4511 without a warning unless @option{-Wpedantic} is used.
4513 For C++, a function without return type always produces a diagnostic
4514 message, even when @option{-Wno-return-type} is specified. The only
4515 exceptions are @code{main} and functions defined in system headers.
4517 This warning is enabled by @option{-Wall}.
4519 @item -Wshift-count-negative
4520 @opindex Wshift-count-negative
4521 @opindex Wno-shift-count-negative
4522 Warn if shift count is negative. This warning is enabled by default.
4524 @item -Wshift-count-overflow
4525 @opindex Wshift-count-overflow
4526 @opindex Wno-shift-count-overflow
4527 Warn if shift count >= width of type. This warning is enabled by default.
4529 @item -Wshift-negative-value
4530 @opindex Wshift-negative-value
4531 @opindex Wno-shift-negative-value
4532 Warn if left shifting a negative value. This warning is enabled by
4533 @option{-Wextra} in C99 and C++11 modes (and newer).
4535 @item -Wshift-overflow
4536 @itemx -Wshift-overflow=@var{n}
4537 @opindex Wshift-overflow
4538 @opindex Wno-shift-overflow
4539 Warn about left shift overflows. This warning is enabled by
4540 default in C99 and C++11 modes (and newer).
4543 @item -Wshift-overflow=1
4544 This is the warning level of @option{-Wshift-overflow} and is enabled
4545 by default in C99 and C++11 modes (and newer). This warning level does
4546 not warn about left-shifting 1 into the sign bit. (However, in C, such
4547 an overflow is still rejected in contexts where an integer constant expression
4550 @item -Wshift-overflow=2
4551 This warning level also warns about left-shifting 1 into the sign bit,
4552 unless C++14 mode is active.
4558 Warn whenever a @code{switch} statement has an index of enumerated type
4559 and lacks a @code{case} for one or more of the named codes of that
4560 enumeration. (The presence of a @code{default} label prevents this
4561 warning.) @code{case} labels outside the enumeration range also
4562 provoke warnings when this option is used (even if there is a
4563 @code{default} label).
4564 This warning is enabled by @option{-Wall}.
4566 @item -Wswitch-default
4567 @opindex Wswitch-default
4568 @opindex Wno-switch-default
4569 Warn whenever a @code{switch} statement does not have a @code{default}
4573 @opindex Wswitch-enum
4574 @opindex Wno-switch-enum
4575 Warn whenever a @code{switch} statement has an index of enumerated type
4576 and lacks a @code{case} for one or more of the named codes of that
4577 enumeration. @code{case} labels outside the enumeration range also
4578 provoke warnings when this option is used. The only difference
4579 between @option{-Wswitch} and this option is that this option gives a
4580 warning about an omitted enumeration code even if there is a
4581 @code{default} label.
4584 @opindex Wswitch-bool
4585 @opindex Wno-switch-bool
4586 Warn whenever a @code{switch} statement has an index of boolean type
4587 and the case values are outside the range of a boolean type.
4588 It is possible to suppress this warning by casting the controlling
4589 expression to a type other than @code{bool}. For example:
4592 switch ((int) (a == 4))
4598 This warning is enabled by default for C and C++ programs.
4600 @item -Wswitch-unreachable
4601 @opindex Wswitch-unreachable
4602 @opindex Wno-switch-unreachable
4603 Warn whenever a @code{switch} statement contains statements between the
4604 controlling expression and the first case label, which will never be
4605 executed. For example:
4617 @option{-Wswitch-unreachable} does not warn if the statement between the
4618 controlling expression and the first case label is just a declaration:
4631 This warning is enabled by default for C and C++ programs.
4633 @item -Wsync-nand @r{(C and C++ only)}
4635 @opindex Wno-sync-nand
4636 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4637 built-in functions are used. These functions changed semantics in GCC 4.4.
4639 @item -Wunused-but-set-parameter
4640 @opindex Wunused-but-set-parameter
4641 @opindex Wno-unused-but-set-parameter
4642 Warn whenever a function parameter is assigned to, but otherwise unused
4643 (aside from its declaration).
4645 To suppress this warning use the @code{unused} attribute
4646 (@pxref{Variable Attributes}).
4648 This warning is also enabled by @option{-Wunused} together with
4651 @item -Wunused-but-set-variable
4652 @opindex Wunused-but-set-variable
4653 @opindex Wno-unused-but-set-variable
4654 Warn whenever a local variable is assigned to, but otherwise unused
4655 (aside from its declaration).
4656 This warning is enabled by @option{-Wall}.
4658 To suppress this warning use the @code{unused} attribute
4659 (@pxref{Variable Attributes}).
4661 This warning is also enabled by @option{-Wunused}, which is enabled
4664 @item -Wunused-function
4665 @opindex Wunused-function
4666 @opindex Wno-unused-function
4667 Warn whenever a static function is declared but not defined or a
4668 non-inline static function is unused.
4669 This warning is enabled by @option{-Wall}.
4671 @item -Wunused-label
4672 @opindex Wunused-label
4673 @opindex Wno-unused-label
4674 Warn whenever a label is declared but not used.
4675 This warning is enabled by @option{-Wall}.
4677 To suppress this warning use the @code{unused} attribute
4678 (@pxref{Variable Attributes}).
4680 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4681 @opindex Wunused-local-typedefs
4682 Warn when a typedef locally defined in a function is not used.
4683 This warning is enabled by @option{-Wall}.
4685 @item -Wunused-parameter
4686 @opindex Wunused-parameter
4687 @opindex Wno-unused-parameter
4688 Warn whenever a function parameter is unused aside from its declaration.
4690 To suppress this warning use the @code{unused} attribute
4691 (@pxref{Variable Attributes}).
4693 @item -Wno-unused-result
4694 @opindex Wunused-result
4695 @opindex Wno-unused-result
4696 Do not warn if a caller of a function marked with attribute
4697 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4698 its return value. The default is @option{-Wunused-result}.
4700 @item -Wunused-variable
4701 @opindex Wunused-variable
4702 @opindex Wno-unused-variable
4703 Warn whenever a local or static variable is unused aside from its
4704 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4705 but not for C++. This warning is enabled by @option{-Wall}.
4707 To suppress this warning use the @code{unused} attribute
4708 (@pxref{Variable Attributes}).
4710 @item -Wunused-const-variable
4711 @itemx -Wunused-const-variable=@var{n}
4712 @opindex Wunused-const-variable
4713 @opindex Wno-unused-const-variable
4714 Warn whenever a constant static variable is unused aside from its declaration.
4715 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4716 for C, but not for C++. In C this declares variable storage, but in C++ this
4717 is not an error since const variables take the place of @code{#define}s.
4719 To suppress this warning use the @code{unused} attribute
4720 (@pxref{Variable Attributes}).
4723 @item -Wunused-const-variable=1
4724 This is the warning level that is enabled by @option{-Wunused-variable} for
4725 C. It warns only about unused static const variables defined in the main
4726 compilation unit, but not about static const variables declared in any
4729 @item -Wunused-const-variable=2
4730 This warning level also warns for unused constant static variables in
4731 headers (excluding system headers). This is the warning level of
4732 @option{-Wunused-const-variable} and must be explicitly requested since
4733 in C++ this isn't an error and in C it might be harder to clean up all
4737 @item -Wunused-value
4738 @opindex Wunused-value
4739 @opindex Wno-unused-value
4740 Warn whenever a statement computes a result that is explicitly not
4741 used. To suppress this warning cast the unused expression to
4742 @code{void}. This includes an expression-statement or the left-hand
4743 side of a comma expression that contains no side effects. For example,
4744 an expression such as @code{x[i,j]} causes a warning, while
4745 @code{x[(void)i,j]} does not.
4747 This warning is enabled by @option{-Wall}.
4752 All the above @option{-Wunused} options combined.
4754 In order to get a warning about an unused function parameter, you must
4755 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4756 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4758 @item -Wuninitialized
4759 @opindex Wuninitialized
4760 @opindex Wno-uninitialized
4761 Warn if an automatic variable is used without first being initialized
4762 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4763 warn if a non-static reference or non-static @code{const} member
4764 appears in a class without constructors.
4766 If you want to warn about code that uses the uninitialized value of the
4767 variable in its own initializer, use the @option{-Winit-self} option.
4769 These warnings occur for individual uninitialized or clobbered
4770 elements of structure, union or array variables as well as for
4771 variables that are uninitialized or clobbered as a whole. They do
4772 not occur for variables or elements declared @code{volatile}. Because
4773 these warnings depend on optimization, the exact variables or elements
4774 for which there are warnings depends on the precise optimization
4775 options and version of GCC used.
4777 Note that there may be no warning about a variable that is used only
4778 to compute a value that itself is never used, because such
4779 computations may be deleted by data flow analysis before the warnings
4782 @item -Winvalid-memory-model
4783 @opindex Winvalid-memory-model
4784 @opindex Wno-invalid-memory-model
4785 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4786 and the C11 atomic generic functions with a memory consistency argument
4787 that is either invalid for the operation or outside the range of values
4788 of the @code{memory_order} enumeration. For example, since the
4789 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4790 defined for the relaxed, release, and sequentially consistent memory
4791 orders the following code is diagnosed:
4796 __atomic_store_n (i, 0, memory_order_consume);
4800 @option{-Winvalid-memory-model} is enabled by default.
4802 @item -Wmaybe-uninitialized
4803 @opindex Wmaybe-uninitialized
4804 @opindex Wno-maybe-uninitialized
4805 For an automatic variable, if there exists a path from the function
4806 entry to a use of the variable that is initialized, but there exist
4807 some other paths for which the variable is not initialized, the compiler
4808 emits a warning if it cannot prove the uninitialized paths are not
4809 executed at run time. These warnings are made optional because GCC is
4810 not smart enough to see all the reasons why the code might be correct
4811 in spite of appearing to have an error. Here is one example of how
4832 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4833 always initialized, but GCC doesn't know this. To suppress the
4834 warning, you need to provide a default case with assert(0) or
4837 @cindex @code{longjmp} warnings
4838 This option also warns when a non-volatile automatic variable might be
4839 changed by a call to @code{longjmp}. These warnings as well are possible
4840 only in optimizing compilation.
4842 The compiler sees only the calls to @code{setjmp}. It cannot know
4843 where @code{longjmp} will be called; in fact, a signal handler could
4844 call it at any point in the code. As a result, you may get a warning
4845 even when there is in fact no problem because @code{longjmp} cannot
4846 in fact be called at the place that would cause a problem.
4848 Some spurious warnings can be avoided if you declare all the functions
4849 you use that never return as @code{noreturn}. @xref{Function
4852 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4854 @item -Wunknown-pragmas
4855 @opindex Wunknown-pragmas
4856 @opindex Wno-unknown-pragmas
4857 @cindex warning for unknown pragmas
4858 @cindex unknown pragmas, warning
4859 @cindex pragmas, warning of unknown
4860 Warn when a @code{#pragma} directive is encountered that is not understood by
4861 GCC@. If this command-line option is used, warnings are even issued
4862 for unknown pragmas in system header files. This is not the case if
4863 the warnings are only enabled by the @option{-Wall} command-line option.
4866 @opindex Wno-pragmas
4868 Do not warn about misuses of pragmas, such as incorrect parameters,
4869 invalid syntax, or conflicts between pragmas. See also
4870 @option{-Wunknown-pragmas}.
4872 @item -Wstrict-aliasing
4873 @opindex Wstrict-aliasing
4874 @opindex Wno-strict-aliasing
4875 This option is only active when @option{-fstrict-aliasing} is active.
4876 It warns about code that might break the strict aliasing rules that the
4877 compiler is using for optimization. The warning does not catch all
4878 cases, but does attempt to catch the more common pitfalls. It is
4879 included in @option{-Wall}.
4880 It is equivalent to @option{-Wstrict-aliasing=3}
4882 @item -Wstrict-aliasing=n
4883 @opindex Wstrict-aliasing=n
4884 This option is only active when @option{-fstrict-aliasing} is active.
4885 It warns about code that might break the strict aliasing rules that the
4886 compiler is using for optimization.
4887 Higher levels correspond to higher accuracy (fewer false positives).
4888 Higher levels also correspond to more effort, similar to the way @option{-O}
4890 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4892 Level 1: Most aggressive, quick, least accurate.
4893 Possibly useful when higher levels
4894 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4895 false negatives. However, it has many false positives.
4896 Warns for all pointer conversions between possibly incompatible types,
4897 even if never dereferenced. Runs in the front end only.
4899 Level 2: Aggressive, quick, not too precise.
4900 May still have many false positives (not as many as level 1 though),
4901 and few false negatives (but possibly more than level 1).
4902 Unlike level 1, it only warns when an address is taken. Warns about
4903 incomplete types. Runs in the front end only.
4905 Level 3 (default for @option{-Wstrict-aliasing}):
4906 Should have very few false positives and few false
4907 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4908 Takes care of the common pun+dereference pattern in the front end:
4909 @code{*(int*)&some_float}.
4910 If optimization is enabled, it also runs in the back end, where it deals
4911 with multiple statement cases using flow-sensitive points-to information.
4912 Only warns when the converted pointer is dereferenced.
4913 Does not warn about incomplete types.
4915 @item -Wstrict-overflow
4916 @itemx -Wstrict-overflow=@var{n}
4917 @opindex Wstrict-overflow
4918 @opindex Wno-strict-overflow
4919 This option is only active when @option{-fstrict-overflow} is active.
4920 It warns about cases where the compiler optimizes based on the
4921 assumption that signed overflow does not occur. Note that it does not
4922 warn about all cases where the code might overflow: it only warns
4923 about cases where the compiler implements some optimization. Thus
4924 this warning depends on the optimization level.
4926 An optimization that assumes that signed overflow does not occur is
4927 perfectly safe if the values of the variables involved are such that
4928 overflow never does, in fact, occur. Therefore this warning can
4929 easily give a false positive: a warning about code that is not
4930 actually a problem. To help focus on important issues, several
4931 warning levels are defined. No warnings are issued for the use of
4932 undefined signed overflow when estimating how many iterations a loop
4933 requires, in particular when determining whether a loop will be
4937 @item -Wstrict-overflow=1
4938 Warn about cases that are both questionable and easy to avoid. For
4939 example, with @option{-fstrict-overflow}, the compiler simplifies
4940 @code{x + 1 > x} to @code{1}. This level of
4941 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4942 are not, and must be explicitly requested.
4944 @item -Wstrict-overflow=2
4945 Also warn about other cases where a comparison is simplified to a
4946 constant. For example: @code{abs (x) >= 0}. This can only be
4947 simplified when @option{-fstrict-overflow} is in effect, because
4948 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4949 zero. @option{-Wstrict-overflow} (with no level) is the same as
4950 @option{-Wstrict-overflow=2}.
4952 @item -Wstrict-overflow=3
4953 Also warn about other cases where a comparison is simplified. For
4954 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4956 @item -Wstrict-overflow=4
4957 Also warn about other simplifications not covered by the above cases.
4958 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4960 @item -Wstrict-overflow=5
4961 Also warn about cases where the compiler reduces the magnitude of a
4962 constant involved in a comparison. For example: @code{x + 2 > y} is
4963 simplified to @code{x + 1 >= y}. This is reported only at the
4964 highest warning level because this simplification applies to many
4965 comparisons, so this warning level gives a very large number of
4969 @item -Wstringop-overflow
4970 @itemx -Wstringop-overflow=@var{type}
4971 @opindex Wstringop-overflow
4972 @opindex Wno-stringop-overflow
4973 Warn for calls to string manipulation functions such as @code{memcpy} and
4974 @code{strcpy} that are determined to overflow the destination buffer. The
4975 optional argument is one greater than the type of Object Size Checking to
4976 perform to determine the size of the destination. @xref{Object Size Checking}.
4977 The argument is meaningful only for functions that operate on character arrays
4978 but not for raw memory functions like @code{memcpy} which always make use
4979 of Object Size type-0. The option also warns for calls that specify a size
4980 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
4981 The option produces the best results with optimization enabled but can detect
4982 a small subset of simple buffer overflows even without optimization in
4983 calls to the GCC built-in functions like @code{__builtin_memcpy} that
4984 correspond to the standard functions. In any case, the option warns about
4985 just a subset of buffer overflows detected by the corresponding overflow
4986 checking built-ins. For example, the option will issue a warning for
4987 the @code{strcpy} call below because it copies at least 5 characters
4988 (the string @code{"blue"} including the terminating NUL) into the buffer
4992 enum Color @{ blue, purple, yellow @};
4993 const char* f (enum Color clr)
4995 static char buf [4];
4999 case blue: str = "blue"; break;
5000 case purple: str = "purple"; break;
5001 case yellow: str = "yellow"; break;
5004 return strcpy (buf, str); // warning here
5008 Option @option{-Wstringop-overflow=2} is enabled by default.
5011 @item -Wstringop-overflow
5012 @item -Wstringop-overflow=1
5013 @opindex Wstringop-overflow
5014 @opindex Wno-stringop-overflow
5015 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5016 to determine the sizes of destination objects. This is the default setting
5017 of the option. At this setting the option will not warn for writes past
5018 the end of subobjects of larger objects accessed by pointers unless the
5019 size of the largest surrounding object is known. When the destination may
5020 be one of several objects it is assumed to be the largest one of them. On
5021 Linux systems, when optimization is enabled at this setting the option warns
5022 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5025 @item -Wstringop-overflow=2
5026 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5027 to determine the sizes of destination objects. At this setting the option
5028 will warn about overflows when writing to members of the largest complete
5029 objects whose exact size is known. It will, however, not warn for excessive
5030 writes to the same members of unknown objects referenced by pointers since
5031 they may point to arrays containing unknown numbers of elements.
5033 @item -Wstringop-overflow=3
5034 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5035 to determine the sizes of destination objects. At this setting the option
5036 warns about overflowing the smallest object or data member. This is the
5037 most restrictive setting of the option that may result in warnings for safe
5040 @item -Wstringop-overflow=4
5041 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5042 to determine the sizes of destination objects. At this setting the option
5043 will warn about overflowing any data members, and when the destination is
5044 one of several objects it uses the size of the largest of them to decide
5045 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5046 setting of the option may result in warnings for benign code.
5049 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
5050 @opindex Wsuggest-attribute=
5051 @opindex Wno-suggest-attribute=
5052 Warn for cases where adding an attribute may be beneficial. The
5053 attributes currently supported are listed below.
5056 @item -Wsuggest-attribute=pure
5057 @itemx -Wsuggest-attribute=const
5058 @itemx -Wsuggest-attribute=noreturn
5059 @opindex Wsuggest-attribute=pure
5060 @opindex Wno-suggest-attribute=pure
5061 @opindex Wsuggest-attribute=const
5062 @opindex Wno-suggest-attribute=const
5063 @opindex Wsuggest-attribute=noreturn
5064 @opindex Wno-suggest-attribute=noreturn
5066 Warn about functions that might be candidates for attributes
5067 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
5068 functions visible in other compilation units or (in the case of @code{pure} and
5069 @code{const}) if it cannot prove that the function returns normally. A function
5070 returns normally if it doesn't contain an infinite loop or return abnormally
5071 by throwing, calling @code{abort} or trapping. This analysis requires option
5072 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
5073 higher. Higher optimization levels improve the accuracy of the analysis.
5075 @item -Wsuggest-attribute=format
5076 @itemx -Wmissing-format-attribute
5077 @opindex Wsuggest-attribute=format
5078 @opindex Wmissing-format-attribute
5079 @opindex Wno-suggest-attribute=format
5080 @opindex Wno-missing-format-attribute
5084 Warn about function pointers that might be candidates for @code{format}
5085 attributes. Note these are only possible candidates, not absolute ones.
5086 GCC guesses that function pointers with @code{format} attributes that
5087 are used in assignment, initialization, parameter passing or return
5088 statements should have a corresponding @code{format} attribute in the
5089 resulting type. I.e.@: the left-hand side of the assignment or
5090 initialization, the type of the parameter variable, or the return type
5091 of the containing function respectively should also have a @code{format}
5092 attribute to avoid the warning.
5094 GCC also warns about function definitions that might be
5095 candidates for @code{format} attributes. Again, these are only
5096 possible candidates. GCC guesses that @code{format} attributes
5097 might be appropriate for any function that calls a function like
5098 @code{vprintf} or @code{vscanf}, but this might not always be the
5099 case, and some functions for which @code{format} attributes are
5100 appropriate may not be detected.
5103 @item -Wsuggest-final-types
5104 @opindex Wno-suggest-final-types
5105 @opindex Wsuggest-final-types
5106 Warn about types with virtual methods where code quality would be improved
5107 if the type were declared with the C++11 @code{final} specifier,
5109 declared in an anonymous namespace. This allows GCC to more aggressively
5110 devirtualize the polymorphic calls. This warning is more effective with link
5111 time optimization, where the information about the class hierarchy graph is
5114 @item -Wsuggest-final-methods
5115 @opindex Wno-suggest-final-methods
5116 @opindex Wsuggest-final-methods
5117 Warn about virtual methods where code quality would be improved if the method
5118 were declared with the C++11 @code{final} specifier,
5119 or, if possible, its type were
5120 declared in an anonymous namespace or with the @code{final} specifier.
5122 more effective with link time optimization, where the information about the
5123 class hierarchy graph is more complete. It is recommended to first consider
5124 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5127 @item -Wsuggest-override
5128 Warn about overriding virtual functions that are not marked with the override
5132 @opindex Wno-alloc-zero
5133 @opindex Walloc-zero
5134 Warn about calls to allocation functions decorated with attribute
5135 @code{alloc_size} that specify zero bytes, including those to the built-in
5136 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5137 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5138 when called with a zero size differs among implementations (and in the case
5139 of @code{realloc} has been deprecated) relying on it may result in subtle
5140 portability bugs and should be avoided.
5142 @item -Walloc-size-larger-than=@var{n}
5143 Warn about calls to functions decorated with attribute @code{alloc_size}
5144 that attempt to allocate objects larger than the specified number of bytes,
5145 or where the result of the size computation in an integer type with infinite
5146 precision would exceed @code{SIZE_MAX / 2}. The option argument @var{n}
5147 may end in one of the standard suffixes designating a multiple of bytes
5148 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
5149 @code{MB} and @code{MiB} for megabyte and mebibyte, and so on.
5150 @xref{Function Attributes}.
5155 This option warns on all uses of @code{alloca} in the source.
5157 @item -Walloca-larger-than=@var{n}
5158 This option warns on calls to @code{alloca} that are not bounded by a
5159 controlling predicate limiting its argument of integer type to at most
5160 @var{n} bytes, or calls to @code{alloca} where the bound is unknown.
5161 Arguments of non-integer types are considered unbounded even if they
5162 appear to be constrained to the expected range.
5164 For example, a bounded case of @code{alloca} could be:
5167 void func (size_t n)
5178 In the above example, passing @code{-Walloca-larger-than=1000} would not
5179 issue a warning because the call to @code{alloca} is known to be at most
5180 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5181 the compiler would emit a warning.
5183 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5184 controlling predicate constraining its integer argument. For example:
5189 void *p = alloca (n);
5194 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5195 a warning, but this time because of the lack of bounds checking.
5197 Note, that even seemingly correct code involving signed integers could
5201 void func (signed int n)
5211 In the above example, @var{n} could be negative, causing a larger than
5212 expected argument to be implicitly cast into the @code{alloca} call.
5214 This option also warns when @code{alloca} is used in a loop.
5216 This warning is not enabled by @option{-Wall}, and is only active when
5217 @option{-ftree-vrp} is active (default for @option{-O2} and above).
5219 See also @option{-Wvla-larger-than=@var{n}}.
5221 @item -Warray-bounds
5222 @itemx -Warray-bounds=@var{n}
5223 @opindex Wno-array-bounds
5224 @opindex Warray-bounds
5225 This option is only active when @option{-ftree-vrp} is active
5226 (default for @option{-O2} and above). It warns about subscripts to arrays
5227 that are always out of bounds. This warning is enabled by @option{-Wall}.
5230 @item -Warray-bounds=1
5231 This is the warning level of @option{-Warray-bounds} and is enabled
5232 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5234 @item -Warray-bounds=2
5235 This warning level also warns about out of bounds access for
5236 arrays at the end of a struct and for arrays accessed through
5237 pointers. This warning level may give a larger number of
5238 false positives and is deactivated by default.
5241 @item -Wbool-compare
5242 @opindex Wno-bool-compare
5243 @opindex Wbool-compare
5244 Warn about boolean expression compared with an integer value different from
5245 @code{true}/@code{false}. For instance, the following comparison is
5250 if ((n > 1) == 2) @{ @dots{} @}
5252 This warning is enabled by @option{-Wall}.
5254 @item -Wbool-operation
5255 @opindex Wno-bool-operation
5256 @opindex Wbool-operation
5257 Warn about suspicious operations on expressions of a boolean type. For
5258 instance, bitwise negation of a boolean is very likely a bug in the program.
5259 For C, this warning also warns about incrementing or decrementing a boolean,
5260 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5261 Incrementing a boolean is invalid in C++1z, and deprecated otherwise.)
5263 This warning is enabled by @option{-Wall}.
5265 @item -Wduplicated-branches
5266 @opindex Wno-duplicated-branches
5267 @opindex Wduplicated-branches
5268 Warn when an if-else has identical branches. This warning detects cases like
5275 It doesn't warn when both branches contain just a null statement. This warning
5276 also warn for conditional operators:
5278 int i = x ? *p : *p;
5281 @item -Wduplicated-cond
5282 @opindex Wno-duplicated-cond
5283 @opindex Wduplicated-cond
5284 Warn about duplicated conditions in an if-else-if chain. For instance,
5285 warn for the following code:
5287 if (p->q != NULL) @{ @dots{} @}
5288 else if (p->q != NULL) @{ @dots{} @}
5291 @item -Wframe-address
5292 @opindex Wno-frame-address
5293 @opindex Wframe-address
5294 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5295 is called with an argument greater than 0. Such calls may return indeterminate
5296 values or crash the program. The warning is included in @option{-Wall}.
5298 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5299 @opindex Wno-discarded-qualifiers
5300 @opindex Wdiscarded-qualifiers
5301 Do not warn if type qualifiers on pointers are being discarded.
5302 Typically, the compiler warns if a @code{const char *} variable is
5303 passed to a function that takes a @code{char *} parameter. This option
5304 can be used to suppress such a warning.
5306 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5307 @opindex Wno-discarded-array-qualifiers
5308 @opindex Wdiscarded-array-qualifiers
5309 Do not warn if type qualifiers on arrays which are pointer targets
5310 are being discarded. Typically, the compiler warns if a
5311 @code{const int (*)[]} variable is passed to a function that
5312 takes a @code{int (*)[]} parameter. This option can be used to
5313 suppress such a warning.
5315 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5316 @opindex Wno-incompatible-pointer-types
5317 @opindex Wincompatible-pointer-types
5318 Do not warn when there is a conversion between pointers that have incompatible
5319 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5320 which warns for pointer argument passing or assignment with different
5323 @item -Wno-int-conversion @r{(C and Objective-C only)}
5324 @opindex Wno-int-conversion
5325 @opindex Wint-conversion
5326 Do not warn about incompatible integer to pointer and pointer to integer
5327 conversions. This warning is about implicit conversions; for explicit
5328 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5329 @option{-Wno-pointer-to-int-cast} may be used.
5331 @item -Wno-div-by-zero
5332 @opindex Wno-div-by-zero
5333 @opindex Wdiv-by-zero
5334 Do not warn about compile-time integer division by zero. Floating-point
5335 division by zero is not warned about, as it can be a legitimate way of
5336 obtaining infinities and NaNs.
5338 @item -Wsystem-headers
5339 @opindex Wsystem-headers
5340 @opindex Wno-system-headers
5341 @cindex warnings from system headers
5342 @cindex system headers, warnings from
5343 Print warning messages for constructs found in system header files.
5344 Warnings from system headers are normally suppressed, on the assumption
5345 that they usually do not indicate real problems and would only make the
5346 compiler output harder to read. Using this command-line option tells
5347 GCC to emit warnings from system headers as if they occurred in user
5348 code. However, note that using @option{-Wall} in conjunction with this
5349 option does @emph{not} warn about unknown pragmas in system
5350 headers---for that, @option{-Wunknown-pragmas} must also be used.
5352 @item -Wtautological-compare
5353 @opindex Wtautological-compare
5354 @opindex Wno-tautological-compare
5355 Warn if a self-comparison always evaluates to true or false. This
5356 warning detects various mistakes such as:
5360 if (i > i) @{ @dots{} @}
5362 This warning is enabled by @option{-Wall}.
5365 @opindex Wtrampolines
5366 @opindex Wno-trampolines
5367 Warn about trampolines generated for pointers to nested functions.
5368 A trampoline is a small piece of data or code that is created at run
5369 time on the stack when the address of a nested function is taken, and is
5370 used to call the nested function indirectly. For some targets, it is
5371 made up of data only and thus requires no special treatment. But, for
5372 most targets, it is made up of code and thus requires the stack to be
5373 made executable in order for the program to work properly.
5376 @opindex Wfloat-equal
5377 @opindex Wno-float-equal
5378 Warn if floating-point values are used in equality comparisons.
5380 The idea behind this is that sometimes it is convenient (for the
5381 programmer) to consider floating-point values as approximations to
5382 infinitely precise real numbers. If you are doing this, then you need
5383 to compute (by analyzing the code, or in some other way) the maximum or
5384 likely maximum error that the computation introduces, and allow for it
5385 when performing comparisons (and when producing output, but that's a
5386 different problem). In particular, instead of testing for equality, you
5387 should check to see whether the two values have ranges that overlap; and
5388 this is done with the relational operators, so equality comparisons are
5391 @item -Wtraditional @r{(C and Objective-C only)}
5392 @opindex Wtraditional
5393 @opindex Wno-traditional
5394 Warn about certain constructs that behave differently in traditional and
5395 ISO C@. Also warn about ISO C constructs that have no traditional C
5396 equivalent, and/or problematic constructs that should be avoided.
5400 Macro parameters that appear within string literals in the macro body.
5401 In traditional C macro replacement takes place within string literals,
5402 but in ISO C it does not.
5405 In traditional C, some preprocessor directives did not exist.
5406 Traditional preprocessors only considered a line to be a directive
5407 if the @samp{#} appeared in column 1 on the line. Therefore
5408 @option{-Wtraditional} warns about directives that traditional C
5409 understands but ignores because the @samp{#} does not appear as the
5410 first character on the line. It also suggests you hide directives like
5411 @code{#pragma} not understood by traditional C by indenting them. Some
5412 traditional implementations do not recognize @code{#elif}, so this option
5413 suggests avoiding it altogether.
5416 A function-like macro that appears without arguments.
5419 The unary plus operator.
5422 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
5423 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
5424 constants.) Note, these suffixes appear in macros defined in the system
5425 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
5426 Use of these macros in user code might normally lead to spurious
5427 warnings, however GCC's integrated preprocessor has enough context to
5428 avoid warning in these cases.
5431 A function declared external in one block and then used after the end of
5435 A @code{switch} statement has an operand of type @code{long}.
5438 A non-@code{static} function declaration follows a @code{static} one.
5439 This construct is not accepted by some traditional C compilers.
5442 The ISO type of an integer constant has a different width or
5443 signedness from its traditional type. This warning is only issued if
5444 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
5445 typically represent bit patterns, are not warned about.
5448 Usage of ISO string concatenation is detected.
5451 Initialization of automatic aggregates.
5454 Identifier conflicts with labels. Traditional C lacks a separate
5455 namespace for labels.
5458 Initialization of unions. If the initializer is zero, the warning is
5459 omitted. This is done under the assumption that the zero initializer in
5460 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
5461 initializer warnings and relies on default initialization to zero in the
5465 Conversions by prototypes between fixed/floating-point values and vice
5466 versa. The absence of these prototypes when compiling with traditional
5467 C causes serious problems. This is a subset of the possible
5468 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
5471 Use of ISO C style function definitions. This warning intentionally is
5472 @emph{not} issued for prototype declarations or variadic functions
5473 because these ISO C features appear in your code when using
5474 libiberty's traditional C compatibility macros, @code{PARAMS} and
5475 @code{VPARAMS}. This warning is also bypassed for nested functions
5476 because that feature is already a GCC extension and thus not relevant to
5477 traditional C compatibility.
5480 @item -Wtraditional-conversion @r{(C and Objective-C only)}
5481 @opindex Wtraditional-conversion
5482 @opindex Wno-traditional-conversion
5483 Warn if a prototype causes a type conversion that is different from what
5484 would happen to the same argument in the absence of a prototype. This
5485 includes conversions of fixed point to floating and vice versa, and
5486 conversions changing the width or signedness of a fixed-point argument
5487 except when the same as the default promotion.
5489 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
5490 @opindex Wdeclaration-after-statement
5491 @opindex Wno-declaration-after-statement
5492 Warn when a declaration is found after a statement in a block. This
5493 construct, known from C++, was introduced with ISO C99 and is by default
5494 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
5499 Warn whenever a local variable or type declaration shadows another
5500 variable, parameter, type, class member (in C++), or instance variable
5501 (in Objective-C) or whenever a built-in function is shadowed. Note
5502 that in C++, the compiler warns if a local variable shadows an
5503 explicit typedef, but not if it shadows a struct/class/enum.
5504 Same as @option{-Wshadow=global}.
5506 @item -Wno-shadow-ivar @r{(Objective-C only)}
5507 @opindex Wno-shadow-ivar
5508 @opindex Wshadow-ivar
5509 Do not warn whenever a local variable shadows an instance variable in an
5512 @item -Wshadow=global
5513 @opindex Wshadow=local
5514 The default for @option{-Wshadow}. Warns for any (global) shadowing.
5516 @item -Wshadow=local
5517 @opindex Wshadow=local
5518 Warn when a local variable shadows another local variable or parameter.
5519 This warning is enabled by @option{-Wshadow=global}.
5521 @item -Wshadow=compatible-local
5522 @opindex Wshadow=compatible-local
5523 Warn when a local variable shadows another local variable or parameter
5524 whose type is compatible with that of the shadowing variable. In C++,
5525 type compatibility here means the type of the shadowing variable can be
5526 converted to that of the shadowed variable. The creation of this flag
5527 (in addition to @option{-Wshadow=local}) is based on the idea that when
5528 a local variable shadows another one of incompatible type, it is most
5529 likely intentional, not a bug or typo, as shown in the following example:
5533 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
5535 for (int i = 0; i < N; ++i)
5544 Since the two variable @code{i} in the example above have incompatible types,
5545 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
5546 Because their types are incompatible, if a programmer accidentally uses one
5547 in place of the other, type checking will catch that and emit an error or
5548 warning. So not warning (about shadowing) in this case will not lead to
5549 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
5550 possibly reduce the number of warnings triggered by intentional shadowing.
5552 This warning is enabled by @option{-Wshadow=local}.
5554 @item -Wlarger-than=@var{len}
5555 @opindex Wlarger-than=@var{len}
5556 @opindex Wlarger-than-@var{len}
5557 Warn whenever an object of larger than @var{len} bytes is defined.
5559 @item -Wframe-larger-than=@var{len}
5560 @opindex Wframe-larger-than
5561 Warn if the size of a function frame is larger than @var{len} bytes.
5562 The computation done to determine the stack frame size is approximate
5563 and not conservative.
5564 The actual requirements may be somewhat greater than @var{len}
5565 even if you do not get a warning. In addition, any space allocated
5566 via @code{alloca}, variable-length arrays, or related constructs
5567 is not included by the compiler when determining
5568 whether or not to issue a warning.
5570 @item -Wno-free-nonheap-object
5571 @opindex Wno-free-nonheap-object
5572 @opindex Wfree-nonheap-object
5573 Do not warn when attempting to free an object that was not allocated
5576 @item -Wstack-usage=@var{len}
5577 @opindex Wstack-usage
5578 Warn if the stack usage of a function might be larger than @var{len} bytes.
5579 The computation done to determine the stack usage is conservative.
5580 Any space allocated via @code{alloca}, variable-length arrays, or related
5581 constructs is included by the compiler when determining whether or not to
5584 The message is in keeping with the output of @option{-fstack-usage}.
5588 If the stack usage is fully static but exceeds the specified amount, it's:
5591 warning: stack usage is 1120 bytes
5594 If the stack usage is (partly) dynamic but bounded, it's:
5597 warning: stack usage might be 1648 bytes
5600 If the stack usage is (partly) dynamic and not bounded, it's:
5603 warning: stack usage might be unbounded
5607 @item -Wunsafe-loop-optimizations
5608 @opindex Wunsafe-loop-optimizations
5609 @opindex Wno-unsafe-loop-optimizations
5610 Warn if the loop cannot be optimized because the compiler cannot
5611 assume anything on the bounds of the loop indices. With
5612 @option{-funsafe-loop-optimizations} warn if the compiler makes
5615 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
5616 @opindex Wno-pedantic-ms-format
5617 @opindex Wpedantic-ms-format
5618 When used in combination with @option{-Wformat}
5619 and @option{-pedantic} without GNU extensions, this option
5620 disables the warnings about non-ISO @code{printf} / @code{scanf} format
5621 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
5622 which depend on the MS runtime.
5625 @opindex Waligned-new
5626 @opindex Wno-aligned-new
5627 Warn about a new-expression of a type that requires greater alignment
5628 than the @code{alignof(std::max_align_t)} but uses an allocation
5629 function without an explicit alignment parameter. This option is
5630 enabled by @option{-Wall}.
5632 Normally this only warns about global allocation functions, but
5633 @option{-Waligned-new=all} also warns about class member allocation
5636 @item -Wplacement-new
5637 @itemx -Wplacement-new=@var{n}
5638 @opindex Wplacement-new
5639 @opindex Wno-placement-new
5640 Warn about placement new expressions with undefined behavior, such as
5641 constructing an object in a buffer that is smaller than the type of
5642 the object. For example, the placement new expression below is diagnosed
5643 because it attempts to construct an array of 64 integers in a buffer only
5649 This warning is enabled by default.
5652 @item -Wplacement-new=1
5653 This is the default warning level of @option{-Wplacement-new}. At this
5654 level the warning is not issued for some strictly undefined constructs that
5655 GCC allows as extensions for compatibility with legacy code. For example,
5656 the following @code{new} expression is not diagnosed at this level even
5657 though it has undefined behavior according to the C++ standard because
5658 it writes past the end of the one-element array.
5660 struct S @{ int n, a[1]; @};
5661 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5662 new (s->a)int [32]();
5665 @item -Wplacement-new=2
5666 At this level, in addition to diagnosing all the same constructs as at level
5667 1, a diagnostic is also issued for placement new expressions that construct
5668 an object in the last member of structure whose type is an array of a single
5669 element and whose size is less than the size of the object being constructed.
5670 While the previous example would be diagnosed, the following construct makes
5671 use of the flexible member array extension to avoid the warning at level 2.
5673 struct S @{ int n, a[]; @};
5674 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5675 new (s->a)int [32]();
5680 @item -Wpointer-arith
5681 @opindex Wpointer-arith
5682 @opindex Wno-pointer-arith
5683 Warn about anything that depends on the ``size of'' a function type or
5684 of @code{void}. GNU C assigns these types a size of 1, for
5685 convenience in calculations with @code{void *} pointers and pointers
5686 to functions. In C++, warn also when an arithmetic operation involves
5687 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5689 @item -Wpointer-compare
5690 @opindex Wpointer-compare
5691 @opindex Wno-pointer-compare
5692 Warn if a pointer is compared with a zero character constant. This usually
5693 means that the pointer was meant to be dereferenced. For example:
5696 const char *p = foo ();
5701 Note that the code above is invalid in C++11.
5703 This warning is enabled by default.
5706 @opindex Wtype-limits
5707 @opindex Wno-type-limits
5708 Warn if a comparison is always true or always false due to the limited
5709 range of the data type, but do not warn for constant expressions. For
5710 example, warn if an unsigned variable is compared against zero with
5711 @code{<} or @code{>=}. This warning is also enabled by
5714 @include cppwarnopts.texi
5716 @item -Wbad-function-cast @r{(C and Objective-C only)}
5717 @opindex Wbad-function-cast
5718 @opindex Wno-bad-function-cast
5719 Warn when a function call is cast to a non-matching type.
5720 For example, warn if a call to a function returning an integer type
5721 is cast to a pointer type.
5723 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5724 @opindex Wc90-c99-compat
5725 @opindex Wno-c90-c99-compat
5726 Warn about features not present in ISO C90, but present in ISO C99.
5727 For instance, warn about use of variable length arrays, @code{long long}
5728 type, @code{bool} type, compound literals, designated initializers, and so
5729 on. This option is independent of the standards mode. Warnings are disabled
5730 in the expression that follows @code{__extension__}.
5732 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5733 @opindex Wc99-c11-compat
5734 @opindex Wno-c99-c11-compat
5735 Warn about features not present in ISO C99, but present in ISO C11.
5736 For instance, warn about use of anonymous structures and unions,
5737 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5738 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5739 and so on. This option is independent of the standards mode. Warnings are
5740 disabled in the expression that follows @code{__extension__}.
5742 @item -Wc++-compat @r{(C and Objective-C only)}
5743 @opindex Wc++-compat
5744 Warn about ISO C constructs that are outside of the common subset of
5745 ISO C and ISO C++, e.g.@: request for implicit conversion from
5746 @code{void *} to a pointer to non-@code{void} type.
5748 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5749 @opindex Wc++11-compat
5750 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5751 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5752 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5753 enabled by @option{-Wall}.
5755 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5756 @opindex Wc++14-compat
5757 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5758 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5760 @item -Wc++1z-compat @r{(C++ and Objective-C++ only)}
5761 @opindex Wc++1z-compat
5762 Warn about C++ constructs whose meaning differs between ISO C++ 2014
5763 and the forthoming ISO C++ 2017(?). This warning is enabled by @option{-Wall}.
5767 @opindex Wno-cast-qual
5768 Warn whenever a pointer is cast so as to remove a type qualifier from
5769 the target type. For example, warn if a @code{const char *} is cast
5770 to an ordinary @code{char *}.
5772 Also warn when making a cast that introduces a type qualifier in an
5773 unsafe way. For example, casting @code{char **} to @code{const char **}
5774 is unsafe, as in this example:
5777 /* p is char ** value. */
5778 const char **q = (const char **) p;
5779 /* Assignment of readonly string to const char * is OK. */
5781 /* Now char** pointer points to read-only memory. */
5786 @opindex Wcast-align
5787 @opindex Wno-cast-align
5788 Warn whenever a pointer is cast such that the required alignment of the
5789 target is increased. For example, warn if a @code{char *} is cast to
5790 an @code{int *} on machines where integers can only be accessed at
5791 two- or four-byte boundaries.
5793 @item -Wwrite-strings
5794 @opindex Wwrite-strings
5795 @opindex Wno-write-strings
5796 When compiling C, give string constants the type @code{const
5797 char[@var{length}]} so that copying the address of one into a
5798 non-@code{const} @code{char *} pointer produces a warning. These
5799 warnings help you find at compile time code that can try to write
5800 into a string constant, but only if you have been very careful about
5801 using @code{const} in declarations and prototypes. Otherwise, it is
5802 just a nuisance. This is why we did not make @option{-Wall} request
5805 When compiling C++, warn about the deprecated conversion from string
5806 literals to @code{char *}. This warning is enabled by default for C++
5811 @opindex Wno-clobbered
5812 Warn for variables that might be changed by @code{longjmp} or
5813 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5815 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5816 @opindex Wconditionally-supported
5817 @opindex Wno-conditionally-supported
5818 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5821 @opindex Wconversion
5822 @opindex Wno-conversion
5823 Warn for implicit conversions that may alter a value. This includes
5824 conversions between real and integer, like @code{abs (x)} when
5825 @code{x} is @code{double}; conversions between signed and unsigned,
5826 like @code{unsigned ui = -1}; and conversions to smaller types, like
5827 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5828 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5829 changed by the conversion like in @code{abs (2.0)}. Warnings about
5830 conversions between signed and unsigned integers can be disabled by
5831 using @option{-Wno-sign-conversion}.
5833 For C++, also warn for confusing overload resolution for user-defined
5834 conversions; and conversions that never use a type conversion
5835 operator: conversions to @code{void}, the same type, a base class or a
5836 reference to them. Warnings about conversions between signed and
5837 unsigned integers are disabled by default in C++ unless
5838 @option{-Wsign-conversion} is explicitly enabled.
5840 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5841 @opindex Wconversion-null
5842 @opindex Wno-conversion-null
5843 Do not warn for conversions between @code{NULL} and non-pointer
5844 types. @option{-Wconversion-null} is enabled by default.
5846 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5847 @opindex Wzero-as-null-pointer-constant
5848 @opindex Wno-zero-as-null-pointer-constant
5849 Warn when a literal @samp{0} is used as null pointer constant. This can
5850 be useful to facilitate the conversion to @code{nullptr} in C++11.
5852 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5853 @opindex Wsubobject-linkage
5854 @opindex Wno-subobject-linkage
5855 Warn if a class type has a base or a field whose type uses the anonymous
5856 namespace or depends on a type with no linkage. If a type A depends on
5857 a type B with no or internal linkage, defining it in multiple
5858 translation units would be an ODR violation because the meaning of B
5859 is different in each translation unit. If A only appears in a single
5860 translation unit, the best way to silence the warning is to give it
5861 internal linkage by putting it in an anonymous namespace as well. The
5862 compiler doesn't give this warning for types defined in the main .C
5863 file, as those are unlikely to have multiple definitions.
5864 @option{-Wsubobject-linkage} is enabled by default.
5866 @item -Wdangling-else
5867 @opindex Wdangling-else
5868 @opindex Wno-dangling-else
5869 Warn about constructions where there may be confusion to which
5870 @code{if} statement an @code{else} branch belongs. Here is an example of
5885 In C/C++, every @code{else} branch belongs to the innermost possible
5886 @code{if} statement, which in this example is @code{if (b)}. This is
5887 often not what the programmer expected, as illustrated in the above
5888 example by indentation the programmer chose. When there is the
5889 potential for this confusion, GCC issues a warning when this flag
5890 is specified. To eliminate the warning, add explicit braces around
5891 the innermost @code{if} statement so there is no way the @code{else}
5892 can belong to the enclosing @code{if}. The resulting code
5909 This warning is enabled by @option{-Wparentheses}.
5913 @opindex Wno-date-time
5914 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5915 are encountered as they might prevent bit-wise-identical reproducible
5918 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5919 @opindex Wdelete-incomplete
5920 @opindex Wno-delete-incomplete
5921 Warn when deleting a pointer to incomplete type, which may cause
5922 undefined behavior at runtime. This warning is enabled by default.
5924 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5925 @opindex Wuseless-cast
5926 @opindex Wno-useless-cast
5927 Warn when an expression is casted to its own type.
5930 @opindex Wempty-body
5931 @opindex Wno-empty-body
5932 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5933 while} statement. This warning is also enabled by @option{-Wextra}.
5935 @item -Wenum-compare
5936 @opindex Wenum-compare
5937 @opindex Wno-enum-compare
5938 Warn about a comparison between values of different enumerated types.
5939 In C++ enumerated type mismatches in conditional expressions are also
5940 diagnosed and the warning is enabled by default. In C this warning is
5941 enabled by @option{-Wall}.
5943 @item -Wjump-misses-init @r{(C, Objective-C only)}
5944 @opindex Wjump-misses-init
5945 @opindex Wno-jump-misses-init
5946 Warn if a @code{goto} statement or a @code{switch} statement jumps
5947 forward across the initialization of a variable, or jumps backward to a
5948 label after the variable has been initialized. This only warns about
5949 variables that are initialized when they are declared. This warning is
5950 only supported for C and Objective-C; in C++ this sort of branch is an
5953 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5954 can be disabled with the @option{-Wno-jump-misses-init} option.
5956 @item -Wsign-compare
5957 @opindex Wsign-compare
5958 @opindex Wno-sign-compare
5959 @cindex warning for comparison of signed and unsigned values
5960 @cindex comparison of signed and unsigned values, warning
5961 @cindex signed and unsigned values, comparison warning
5962 Warn when a comparison between signed and unsigned values could produce
5963 an incorrect result when the signed value is converted to unsigned.
5964 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5965 also enabled by @option{-Wextra}.
5967 @item -Wsign-conversion
5968 @opindex Wsign-conversion
5969 @opindex Wno-sign-conversion
5970 Warn for implicit conversions that may change the sign of an integer
5971 value, like assigning a signed integer expression to an unsigned
5972 integer variable. An explicit cast silences the warning. In C, this
5973 option is enabled also by @option{-Wconversion}.
5975 @item -Wfloat-conversion
5976 @opindex Wfloat-conversion
5977 @opindex Wno-float-conversion
5978 Warn for implicit conversions that reduce the precision of a real value.
5979 This includes conversions from real to integer, and from higher precision
5980 real to lower precision real values. This option is also enabled by
5981 @option{-Wconversion}.
5983 @item -Wno-scalar-storage-order
5984 @opindex -Wno-scalar-storage-order
5985 @opindex -Wscalar-storage-order
5986 Do not warn on suspicious constructs involving reverse scalar storage order.
5988 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5989 @opindex Wsized-deallocation
5990 @opindex Wno-sized-deallocation
5991 Warn about a definition of an unsized deallocation function
5993 void operator delete (void *) noexcept;
5994 void operator delete[] (void *) noexcept;
5996 without a definition of the corresponding sized deallocation function
5998 void operator delete (void *, std::size_t) noexcept;
5999 void operator delete[] (void *, std::size_t) noexcept;
6001 or vice versa. Enabled by @option{-Wextra} along with
6002 @option{-fsized-deallocation}.
6004 @item -Wsizeof-pointer-memaccess
6005 @opindex Wsizeof-pointer-memaccess
6006 @opindex Wno-sizeof-pointer-memaccess
6007 Warn for suspicious length parameters to certain string and memory built-in
6008 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
6009 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
6010 but a pointer, and suggests a possible fix, or about
6011 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
6014 @item -Wsizeof-array-argument
6015 @opindex Wsizeof-array-argument
6016 @opindex Wno-sizeof-array-argument
6017 Warn when the @code{sizeof} operator is applied to a parameter that is
6018 declared as an array in a function definition. This warning is enabled by
6019 default for C and C++ programs.
6021 @item -Wmemset-elt-size
6022 @opindex Wmemset-elt-size
6023 @opindex Wno-memset-elt-size
6024 Warn for suspicious calls to the @code{memset} built-in function, if the
6025 first argument references an array, and the third argument is a number
6026 equal to the number of elements, but not equal to the size of the array
6027 in memory. This indicates that the user has omitted a multiplication by
6028 the element size. This warning is enabled by @option{-Wall}.
6030 @item -Wmemset-transposed-args
6031 @opindex Wmemset-transposed-args
6032 @opindex Wno-memset-transposed-args
6033 Warn for suspicious calls to the @code{memset} built-in function, if the
6034 second argument is not zero and the third argument is zero. This warns e.g.@
6035 about @code{memset (buf, sizeof buf, 0)} where most probably
6036 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6037 is only emitted if the third argument is literal zero. If it is some
6038 expression that is folded to zero, a cast of zero to some type, etc.,
6039 it is far less likely that the user has mistakenly exchanged the arguments
6040 and no warning is emitted. This warning is enabled by @option{-Wall}.
6044 @opindex Wno-address
6045 Warn about suspicious uses of memory addresses. These include using
6046 the address of a function in a conditional expression, such as
6047 @code{void func(void); if (func)}, and comparisons against the memory
6048 address of a string literal, such as @code{if (x == "abc")}. Such
6049 uses typically indicate a programmer error: the address of a function
6050 always evaluates to true, so their use in a conditional usually
6051 indicate that the programmer forgot the parentheses in a function
6052 call; and comparisons against string literals result in unspecified
6053 behavior and are not portable in C, so they usually indicate that the
6054 programmer intended to use @code{strcmp}. This warning is enabled by
6058 @opindex Wlogical-op
6059 @opindex Wno-logical-op
6060 Warn about suspicious uses of logical operators in expressions.
6061 This includes using logical operators in contexts where a
6062 bit-wise operator is likely to be expected. Also warns when
6063 the operands of a logical operator are the same:
6066 if (a < 0 && a < 0) @{ @dots{} @}
6069 @item -Wlogical-not-parentheses
6070 @opindex Wlogical-not-parentheses
6071 @opindex Wno-logical-not-parentheses
6072 Warn about logical not used on the left hand side operand of a comparison.
6073 This option does not warn if the right operand is considered to be a boolean
6074 expression. Its purpose is to detect suspicious code like the following:
6078 if (!a > 1) @{ @dots{} @}
6081 It is possible to suppress the warning by wrapping the LHS into
6084 if ((!a) > 1) @{ @dots{} @}
6087 This warning is enabled by @option{-Wall}.
6089 @item -Waggregate-return
6090 @opindex Waggregate-return
6091 @opindex Wno-aggregate-return
6092 Warn if any functions that return structures or unions are defined or
6093 called. (In languages where you can return an array, this also elicits
6096 @item -Wno-aggressive-loop-optimizations
6097 @opindex Wno-aggressive-loop-optimizations
6098 @opindex Waggressive-loop-optimizations
6099 Warn if in a loop with constant number of iterations the compiler detects
6100 undefined behavior in some statement during one or more of the iterations.
6102 @item -Wno-attributes
6103 @opindex Wno-attributes
6104 @opindex Wattributes
6105 Do not warn if an unexpected @code{__attribute__} is used, such as
6106 unrecognized attributes, function attributes applied to variables,
6107 etc. This does not stop errors for incorrect use of supported
6110 @item -Wno-builtin-declaration-mismatch
6111 @opindex Wno-builtin-declaration-mismatch
6112 @opindex Wbuiltin-declaration-mismatch
6113 Warn if a built-in function is declared with the wrong signature.
6114 This warning is enabled by default.
6116 @item -Wno-builtin-macro-redefined
6117 @opindex Wno-builtin-macro-redefined
6118 @opindex Wbuiltin-macro-redefined
6119 Do not warn if certain built-in macros are redefined. This suppresses
6120 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6121 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6123 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6124 @opindex Wstrict-prototypes
6125 @opindex Wno-strict-prototypes
6126 Warn if a function is declared or defined without specifying the
6127 argument types. (An old-style function definition is permitted without
6128 a warning if preceded by a declaration that specifies the argument
6131 @item -Wold-style-declaration @r{(C and Objective-C only)}
6132 @opindex Wold-style-declaration
6133 @opindex Wno-old-style-declaration
6134 Warn for obsolescent usages, according to the C Standard, in a
6135 declaration. For example, warn if storage-class specifiers like
6136 @code{static} are not the first things in a declaration. This warning
6137 is also enabled by @option{-Wextra}.
6139 @item -Wold-style-definition @r{(C and Objective-C only)}
6140 @opindex Wold-style-definition
6141 @opindex Wno-old-style-definition
6142 Warn if an old-style function definition is used. A warning is given
6143 even if there is a previous prototype.
6145 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6146 @opindex Wmissing-parameter-type
6147 @opindex Wno-missing-parameter-type
6148 A function parameter is declared without a type specifier in K&R-style
6155 This warning is also enabled by @option{-Wextra}.
6157 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6158 @opindex Wmissing-prototypes
6159 @opindex Wno-missing-prototypes
6160 Warn if a global function is defined without a previous prototype
6161 declaration. This warning is issued even if the definition itself
6162 provides a prototype. Use this option to detect global functions
6163 that do not have a matching prototype declaration in a header file.
6164 This option is not valid for C++ because all function declarations
6165 provide prototypes and a non-matching declaration declares an
6166 overload rather than conflict with an earlier declaration.
6167 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6169 @item -Wmissing-declarations
6170 @opindex Wmissing-declarations
6171 @opindex Wno-missing-declarations
6172 Warn if a global function is defined without a previous declaration.
6173 Do so even if the definition itself provides a prototype.
6174 Use this option to detect global functions that are not declared in
6175 header files. In C, no warnings are issued for functions with previous
6176 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6177 missing prototypes. In C++, no warnings are issued for function templates,
6178 or for inline functions, or for functions in anonymous namespaces.
6180 @item -Wmissing-field-initializers
6181 @opindex Wmissing-field-initializers
6182 @opindex Wno-missing-field-initializers
6186 Warn if a structure's initializer has some fields missing. For
6187 example, the following code causes such a warning, because
6188 @code{x.h} is implicitly zero:
6191 struct s @{ int f, g, h; @};
6192 struct s x = @{ 3, 4 @};
6195 This option does not warn about designated initializers, so the following
6196 modification does not trigger a warning:
6199 struct s @{ int f, g, h; @};
6200 struct s x = @{ .f = 3, .g = 4 @};
6203 In C++ this option does not warn either about the empty @{ @}
6204 initializer, for example:
6207 struct s @{ int f, g, h; @};
6211 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6212 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6214 @item -Wno-multichar
6215 @opindex Wno-multichar
6217 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6218 Usually they indicate a typo in the user's code, as they have
6219 implementation-defined values, and should not be used in portable code.
6221 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
6222 @opindex Wnormalized=
6223 @opindex Wnormalized
6224 @opindex Wno-normalized
6227 @cindex character set, input normalization
6228 In ISO C and ISO C++, two identifiers are different if they are
6229 different sequences of characters. However, sometimes when characters
6230 outside the basic ASCII character set are used, you can have two
6231 different character sequences that look the same. To avoid confusion,
6232 the ISO 10646 standard sets out some @dfn{normalization rules} which
6233 when applied ensure that two sequences that look the same are turned into
6234 the same sequence. GCC can warn you if you are using identifiers that
6235 have not been normalized; this option controls that warning.
6237 There are four levels of warning supported by GCC@. The default is
6238 @option{-Wnormalized=nfc}, which warns about any identifier that is
6239 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6240 recommended form for most uses. It is equivalent to
6241 @option{-Wnormalized}.
6243 Unfortunately, there are some characters allowed in identifiers by
6244 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6245 identifiers. That is, there's no way to use these symbols in portable
6246 ISO C or C++ and have all your identifiers in NFC@.
6247 @option{-Wnormalized=id} suppresses the warning for these characters.
6248 It is hoped that future versions of the standards involved will correct
6249 this, which is why this option is not the default.
6251 You can switch the warning off for all characters by writing
6252 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6253 only do this if you are using some other normalization scheme (like
6254 ``D''), because otherwise you can easily create bugs that are
6255 literally impossible to see.
6257 Some characters in ISO 10646 have distinct meanings but look identical
6258 in some fonts or display methodologies, especially once formatting has
6259 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6260 LETTER N'', displays just like a regular @code{n} that has been
6261 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6262 normalization scheme to convert all these into a standard form as
6263 well, and GCC warns if your code is not in NFKC if you use
6264 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6265 about every identifier that contains the letter O because it might be
6266 confused with the digit 0, and so is not the default, but may be
6267 useful as a local coding convention if the programming environment
6268 cannot be fixed to display these characters distinctly.
6270 @item -Wno-deprecated
6271 @opindex Wno-deprecated
6272 @opindex Wdeprecated
6273 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
6275 @item -Wno-deprecated-declarations
6276 @opindex Wno-deprecated-declarations
6277 @opindex Wdeprecated-declarations
6278 Do not warn about uses of functions (@pxref{Function Attributes}),
6279 variables (@pxref{Variable Attributes}), and types (@pxref{Type
6280 Attributes}) marked as deprecated by using the @code{deprecated}
6284 @opindex Wno-overflow
6286 Do not warn about compile-time overflow in constant expressions.
6291 Warn about One Definition Rule violations during link-time optimization.
6292 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
6295 @opindex Wopenm-simd
6296 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
6297 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
6298 option can be used to relax the cost model.
6300 @item -Woverride-init @r{(C and Objective-C only)}
6301 @opindex Woverride-init
6302 @opindex Wno-override-init
6306 Warn if an initialized field without side effects is overridden when
6307 using designated initializers (@pxref{Designated Inits, , Designated
6310 This warning is included in @option{-Wextra}. To get other
6311 @option{-Wextra} warnings without this one, use @option{-Wextra
6312 -Wno-override-init}.
6314 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
6315 @opindex Woverride-init-side-effects
6316 @opindex Wno-override-init-side-effects
6317 Warn if an initialized field with side effects is overridden when
6318 using designated initializers (@pxref{Designated Inits, , Designated
6319 Initializers}). This warning is enabled by default.
6324 Warn if a structure is given the packed attribute, but the packed
6325 attribute has no effect on the layout or size of the structure.
6326 Such structures may be mis-aligned for little benefit. For
6327 instance, in this code, the variable @code{f.x} in @code{struct bar}
6328 is misaligned even though @code{struct bar} does not itself
6329 have the packed attribute:
6336 @} __attribute__((packed));
6344 @item -Wpacked-bitfield-compat
6345 @opindex Wpacked-bitfield-compat
6346 @opindex Wno-packed-bitfield-compat
6347 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
6348 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
6349 the change can lead to differences in the structure layout. GCC
6350 informs you when the offset of such a field has changed in GCC 4.4.
6351 For example there is no longer a 4-bit padding between field @code{a}
6352 and @code{b} in this structure:
6359 @} __attribute__ ((packed));
6362 This warning is enabled by default. Use
6363 @option{-Wno-packed-bitfield-compat} to disable this warning.
6368 Warn if padding is included in a structure, either to align an element
6369 of the structure or to align the whole structure. Sometimes when this
6370 happens it is possible to rearrange the fields of the structure to
6371 reduce the padding and so make the structure smaller.
6373 @item -Wredundant-decls
6374 @opindex Wredundant-decls
6375 @opindex Wno-redundant-decls
6376 Warn if anything is declared more than once in the same scope, even in
6377 cases where multiple declaration is valid and changes nothing.
6381 @opindex Wno-restrict
6382 Warn when an argument passed to a restrict-qualified parameter
6383 aliases with another argument.
6385 @item -Wnested-externs @r{(C and Objective-C only)}
6386 @opindex Wnested-externs
6387 @opindex Wno-nested-externs
6388 Warn if an @code{extern} declaration is encountered within a function.
6390 @item -Wno-inherited-variadic-ctor
6391 @opindex Winherited-variadic-ctor
6392 @opindex Wno-inherited-variadic-ctor
6393 Suppress warnings about use of C++11 inheriting constructors when the
6394 base class inherited from has a C variadic constructor; the warning is
6395 on by default because the ellipsis is not inherited.
6400 Warn if a function that is declared as inline cannot be inlined.
6401 Even with this option, the compiler does not warn about failures to
6402 inline functions declared in system headers.
6404 The compiler uses a variety of heuristics to determine whether or not
6405 to inline a function. For example, the compiler takes into account
6406 the size of the function being inlined and the amount of inlining
6407 that has already been done in the current function. Therefore,
6408 seemingly insignificant changes in the source program can cause the
6409 warnings produced by @option{-Winline} to appear or disappear.
6411 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
6412 @opindex Wno-invalid-offsetof
6413 @opindex Winvalid-offsetof
6414 Suppress warnings from applying the @code{offsetof} macro to a non-POD
6415 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
6416 to a non-standard-layout type is undefined. In existing C++ implementations,
6417 however, @code{offsetof} typically gives meaningful results.
6418 This flag is for users who are aware that they are
6419 writing nonportable code and who have deliberately chosen to ignore the
6422 The restrictions on @code{offsetof} may be relaxed in a future version
6423 of the C++ standard.
6425 @item -Wint-in-bool-context
6426 @opindex Wint-in-bool-context
6427 @opindex Wno-int-in-bool-context
6428 Warn for suspicious use of integer values where boolean values are expected,
6429 such as conditional expressions (?:) using non-boolean integer constants in
6430 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
6431 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
6432 for all kinds of multiplications regardless of the data type.
6433 This warning is enabled by @option{-Wall}.
6435 @item -Wno-int-to-pointer-cast
6436 @opindex Wno-int-to-pointer-cast
6437 @opindex Wint-to-pointer-cast
6438 Suppress warnings from casts to pointer type of an integer of a
6439 different size. In C++, casting to a pointer type of smaller size is
6440 an error. @option{Wint-to-pointer-cast} is enabled by default.
6443 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
6444 @opindex Wno-pointer-to-int-cast
6445 @opindex Wpointer-to-int-cast
6446 Suppress warnings from casts from a pointer to an integer type of a
6450 @opindex Winvalid-pch
6451 @opindex Wno-invalid-pch
6452 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
6453 the search path but can't be used.
6457 @opindex Wno-long-long
6458 Warn if @code{long long} type is used. This is enabled by either
6459 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
6460 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
6462 @item -Wvariadic-macros
6463 @opindex Wvariadic-macros
6464 @opindex Wno-variadic-macros
6465 Warn if variadic macros are used in ISO C90 mode, or if the GNU
6466 alternate syntax is used in ISO C99 mode. This is enabled by either
6467 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
6468 messages, use @option{-Wno-variadic-macros}.
6472 @opindex Wno-varargs
6473 Warn upon questionable usage of the macros used to handle variable
6474 arguments like @code{va_start}. This is default. To inhibit the
6475 warning messages, use @option{-Wno-varargs}.
6477 @item -Wvector-operation-performance
6478 @opindex Wvector-operation-performance
6479 @opindex Wno-vector-operation-performance
6480 Warn if vector operation is not implemented via SIMD capabilities of the
6481 architecture. Mainly useful for the performance tuning.
6482 Vector operation can be implemented @code{piecewise}, which means that the
6483 scalar operation is performed on every vector element;
6484 @code{in parallel}, which means that the vector operation is implemented
6485 using scalars of wider type, which normally is more performance efficient;
6486 and @code{as a single scalar}, which means that vector fits into a
6489 @item -Wno-virtual-move-assign
6490 @opindex Wvirtual-move-assign
6491 @opindex Wno-virtual-move-assign
6492 Suppress warnings about inheriting from a virtual base with a
6493 non-trivial C++11 move assignment operator. This is dangerous because
6494 if the virtual base is reachable along more than one path, it is
6495 moved multiple times, which can mean both objects end up in the
6496 moved-from state. If the move assignment operator is written to avoid
6497 moving from a moved-from object, this warning can be disabled.
6502 Warn if a variable-length array is used in the code.
6503 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
6504 the variable-length array.
6506 @item -Wvla-larger-than=@var{n}
6507 If this option is used, the compiler will warn on uses of
6508 variable-length arrays where the size is either unbounded, or bounded
6509 by an argument that can be larger than @var{n} bytes. This is similar
6510 to how @option{-Walloca-larger-than=@var{n}} works, but with
6511 variable-length arrays.
6513 Note that GCC may optimize small variable-length arrays of a known
6514 value into plain arrays, so this warning may not get triggered for
6517 This warning is not enabled by @option{-Wall}, and is only active when
6518 @option{-ftree-vrp} is active (default for @option{-O2} and above).
6520 See also @option{-Walloca-larger-than=@var{n}}.
6522 @item -Wvolatile-register-var
6523 @opindex Wvolatile-register-var
6524 @opindex Wno-volatile-register-var
6525 Warn if a register variable is declared volatile. The volatile
6526 modifier does not inhibit all optimizations that may eliminate reads
6527 and/or writes to register variables. This warning is enabled by
6530 @item -Wdisabled-optimization
6531 @opindex Wdisabled-optimization
6532 @opindex Wno-disabled-optimization
6533 Warn if a requested optimization pass is disabled. This warning does
6534 not generally indicate that there is anything wrong with your code; it
6535 merely indicates that GCC's optimizers are unable to handle the code
6536 effectively. Often, the problem is that your code is too big or too
6537 complex; GCC refuses to optimize programs when the optimization
6538 itself is likely to take inordinate amounts of time.
6540 @item -Wpointer-sign @r{(C and Objective-C only)}
6541 @opindex Wpointer-sign
6542 @opindex Wno-pointer-sign
6543 Warn for pointer argument passing or assignment with different signedness.
6544 This option is only supported for C and Objective-C@. It is implied by
6545 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
6546 @option{-Wno-pointer-sign}.
6548 @item -Wstack-protector
6549 @opindex Wstack-protector
6550 @opindex Wno-stack-protector
6551 This option is only active when @option{-fstack-protector} is active. It
6552 warns about functions that are not protected against stack smashing.
6554 @item -Woverlength-strings
6555 @opindex Woverlength-strings
6556 @opindex Wno-overlength-strings
6557 Warn about string constants that are longer than the ``minimum
6558 maximum'' length specified in the C standard. Modern compilers
6559 generally allow string constants that are much longer than the
6560 standard's minimum limit, but very portable programs should avoid
6561 using longer strings.
6563 The limit applies @emph{after} string constant concatenation, and does
6564 not count the trailing NUL@. In C90, the limit was 509 characters; in
6565 C99, it was raised to 4095. C++98 does not specify a normative
6566 minimum maximum, so we do not diagnose overlength strings in C++@.
6568 This option is implied by @option{-Wpedantic}, and can be disabled with
6569 @option{-Wno-overlength-strings}.
6571 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
6572 @opindex Wunsuffixed-float-constants
6574 Issue a warning for any floating constant that does not have
6575 a suffix. When used together with @option{-Wsystem-headers} it
6576 warns about such constants in system header files. This can be useful
6577 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
6578 from the decimal floating-point extension to C99.
6580 @item -Wno-designated-init @r{(C and Objective-C only)}
6581 Suppress warnings when a positional initializer is used to initialize
6582 a structure that has been marked with the @code{designated_init}
6586 Issue a warning when HSAIL cannot be emitted for the compiled function or
6591 @node Debugging Options
6592 @section Options for Debugging Your Program
6593 @cindex options, debugging
6594 @cindex debugging information options
6596 To tell GCC to emit extra information for use by a debugger, in almost
6597 all cases you need only to add @option{-g} to your other options.
6599 GCC allows you to use @option{-g} with
6600 @option{-O}. The shortcuts taken by optimized code may occasionally
6601 be surprising: some variables you declared may not exist
6602 at all; flow of control may briefly move where you did not expect it;
6603 some statements may not be executed because they compute constant
6604 results or their values are already at hand; some statements may
6605 execute in different places because they have been moved out of loops.
6606 Nevertheless it is possible to debug optimized output. This makes
6607 it reasonable to use the optimizer for programs that might have bugs.
6609 If you are not using some other optimization option, consider
6610 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
6611 With no @option{-O} option at all, some compiler passes that collect
6612 information useful for debugging do not run at all, so that
6613 @option{-Og} may result in a better debugging experience.
6618 Produce debugging information in the operating system's native format
6619 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
6622 On most systems that use stabs format, @option{-g} enables use of extra
6623 debugging information that only GDB can use; this extra information
6624 makes debugging work better in GDB but probably makes other debuggers
6626 refuse to read the program. If you want to control for certain whether
6627 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
6628 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
6632 Produce debugging information for use by GDB@. This means to use the
6633 most expressive format available (DWARF, stabs, or the native format
6634 if neither of those are supported), including GDB extensions if at all
6638 @itemx -gdwarf-@var{version}
6640 Produce debugging information in DWARF format (if that is supported).
6641 The value of @var{version} may be either 2, 3, 4 or 5; the default version
6642 for most targets is 4. DWARF Version 5 is only experimental.
6644 Note that with DWARF Version 2, some ports require and always
6645 use some non-conflicting DWARF 3 extensions in the unwind tables.
6647 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
6648 for maximum benefit.
6650 GCC no longer supports DWARF Version 1, which is substantially
6651 different than Version 2 and later. For historical reasons, some
6652 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
6653 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
6654 in their names, but apply to all currently-supported versions of DWARF.
6658 Produce debugging information in stabs format (if that is supported),
6659 without GDB extensions. This is the format used by DBX on most BSD
6660 systems. On MIPS, Alpha and System V Release 4 systems this option
6661 produces stabs debugging output that is not understood by DBX or SDB@.
6662 On System V Release 4 systems this option requires the GNU assembler.
6666 Produce debugging information in stabs format (if that is supported),
6667 using GNU extensions understood only by the GNU debugger (GDB)@. The
6668 use of these extensions is likely to make other debuggers crash or
6669 refuse to read the program.
6673 Produce debugging information in COFF format (if that is supported).
6674 This is the format used by SDB on most System V systems prior to
6679 Produce debugging information in XCOFF format (if that is supported).
6680 This is the format used by the DBX debugger on IBM RS/6000 systems.
6684 Produce debugging information in XCOFF format (if that is supported),
6685 using GNU extensions understood only by the GNU debugger (GDB)@. The
6686 use of these extensions is likely to make other debuggers crash or
6687 refuse to read the program, and may cause assemblers other than the GNU
6688 assembler (GAS) to fail with an error.
6692 Produce debugging information in Alpha/VMS debug format (if that is
6693 supported). This is the format used by DEBUG on Alpha/VMS systems.
6696 @itemx -ggdb@var{level}
6697 @itemx -gstabs@var{level}
6698 @itemx -gcoff@var{level}
6699 @itemx -gxcoff@var{level}
6700 @itemx -gvms@var{level}
6701 Request debugging information and also use @var{level} to specify how
6702 much information. The default level is 2.
6704 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6707 Level 1 produces minimal information, enough for making backtraces in
6708 parts of the program that you don't plan to debug. This includes
6709 descriptions of functions and external variables, and line number
6710 tables, but no information about local variables.
6712 Level 3 includes extra information, such as all the macro definitions
6713 present in the program. Some debuggers support macro expansion when
6714 you use @option{-g3}.
6716 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6717 confusion with @option{-gdwarf-@var{level}}.
6718 Instead use an additional @option{-g@var{level}} option to change the
6719 debug level for DWARF.
6721 @item -feliminate-unused-debug-symbols
6722 @opindex feliminate-unused-debug-symbols
6723 Produce debugging information in stabs format (if that is supported),
6724 for only symbols that are actually used.
6726 @item -femit-class-debug-always
6727 @opindex femit-class-debug-always
6728 Instead of emitting debugging information for a C++ class in only one
6729 object file, emit it in all object files using the class. This option
6730 should be used only with debuggers that are unable to handle the way GCC
6731 normally emits debugging information for classes because using this
6732 option increases the size of debugging information by as much as a
6735 @item -fno-merge-debug-strings
6736 @opindex fmerge-debug-strings
6737 @opindex fno-merge-debug-strings
6738 Direct the linker to not merge together strings in the debugging
6739 information that are identical in different object files. Merging is
6740 not supported by all assemblers or linkers. Merging decreases the size
6741 of the debug information in the output file at the cost of increasing
6742 link processing time. Merging is enabled by default.
6744 @item -fdebug-prefix-map=@var{old}=@var{new}
6745 @opindex fdebug-prefix-map
6746 When compiling files in directory @file{@var{old}}, record debugging
6747 information describing them as in @file{@var{new}} instead.
6749 @item -fvar-tracking
6750 @opindex fvar-tracking
6751 Run variable tracking pass. It computes where variables are stored at each
6752 position in code. Better debugging information is then generated
6753 (if the debugging information format supports this information).
6755 It is enabled by default when compiling with optimization (@option{-Os},
6756 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6757 the debug info format supports it.
6759 @item -fvar-tracking-assignments
6760 @opindex fvar-tracking-assignments
6761 @opindex fno-var-tracking-assignments
6762 Annotate assignments to user variables early in the compilation and
6763 attempt to carry the annotations over throughout the compilation all the
6764 way to the end, in an attempt to improve debug information while
6765 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6767 It can be enabled even if var-tracking is disabled, in which case
6768 annotations are created and maintained, but discarded at the end.
6769 By default, this flag is enabled together with @option{-fvar-tracking},
6770 except when selective scheduling is enabled.
6773 @opindex gsplit-dwarf
6774 Separate as much DWARF debugging information as possible into a
6775 separate output file with the extension @file{.dwo}. This option allows
6776 the build system to avoid linking files with debug information. To
6777 be useful, this option requires a debugger capable of reading @file{.dwo}
6782 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6784 @item -ggnu-pubnames
6785 @opindex ggnu-pubnames
6786 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6787 suitable for conversion into a GDB@ index. This option is only useful
6788 with a linker that can produce GDB@ index version 7.
6790 @item -fdebug-types-section
6791 @opindex fdebug-types-section
6792 @opindex fno-debug-types-section
6793 When using DWARF Version 4 or higher, type DIEs can be put into
6794 their own @code{.debug_types} section instead of making them part of the
6795 @code{.debug_info} section. It is more efficient to put them in a separate
6796 comdat sections since the linker can then remove duplicates.
6797 But not all DWARF consumers support @code{.debug_types} sections yet
6798 and on some objects @code{.debug_types} produces larger instead of smaller
6799 debugging information.
6801 @item -grecord-gcc-switches
6802 @item -gno-record-gcc-switches
6803 @opindex grecord-gcc-switches
6804 @opindex gno-record-gcc-switches
6805 This switch causes the command-line options used to invoke the
6806 compiler that may affect code generation to be appended to the
6807 DW_AT_producer attribute in DWARF debugging information. The options
6808 are concatenated with spaces separating them from each other and from
6809 the compiler version.
6810 It is enabled by default.
6811 See also @option{-frecord-gcc-switches} for another
6812 way of storing compiler options into the object file.
6814 @item -gstrict-dwarf
6815 @opindex gstrict-dwarf
6816 Disallow using extensions of later DWARF standard version than selected
6817 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6818 DWARF extensions from later standard versions is allowed.
6820 @item -gno-strict-dwarf
6821 @opindex gno-strict-dwarf
6822 Allow using extensions of later DWARF standard version than selected with
6823 @option{-gdwarf-@var{version}}.
6826 @item -gno-column-info
6827 @opindex gcolumn-info
6828 @opindex gno-column-info
6829 Emit location column information into DWARF debugging information, rather
6830 than just file and line.
6831 This option is disabled by default.
6833 @item -gz@r{[}=@var{type}@r{]}
6835 Produce compressed debug sections in DWARF format, if that is supported.
6836 If @var{type} is not given, the default type depends on the capabilities
6837 of the assembler and linker used. @var{type} may be one of
6838 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6839 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6840 compression in traditional GNU format). If the linker doesn't support
6841 writing compressed debug sections, the option is rejected. Otherwise,
6842 if the assembler does not support them, @option{-gz} is silently ignored
6843 when producing object files.
6845 @item -feliminate-dwarf2-dups
6846 @opindex feliminate-dwarf2-dups
6847 Compress DWARF debugging information by eliminating duplicated
6848 information about each symbol. This option only makes sense when
6849 generating DWARF debugging information.
6851 @item -femit-struct-debug-baseonly
6852 @opindex femit-struct-debug-baseonly
6853 Emit debug information for struct-like types
6854 only when the base name of the compilation source file
6855 matches the base name of file in which the struct is defined.
6857 This option substantially reduces the size of debugging information,
6858 but at significant potential loss in type information to the debugger.
6859 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6860 See @option{-femit-struct-debug-detailed} for more detailed control.
6862 This option works only with DWARF debug output.
6864 @item -femit-struct-debug-reduced
6865 @opindex femit-struct-debug-reduced
6866 Emit debug information for struct-like types
6867 only when the base name of the compilation source file
6868 matches the base name of file in which the type is defined,
6869 unless the struct is a template or defined in a system header.
6871 This option significantly reduces the size of debugging information,
6872 with some potential loss in type information to the debugger.
6873 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6874 See @option{-femit-struct-debug-detailed} for more detailed control.
6876 This option works only with DWARF debug output.
6878 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6879 @opindex femit-struct-debug-detailed
6880 Specify the struct-like types
6881 for which the compiler generates debug information.
6882 The intent is to reduce duplicate struct debug information
6883 between different object files within the same program.
6885 This option is a detailed version of
6886 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6887 which serves for most needs.
6889 A specification has the syntax@*
6890 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6892 The optional first word limits the specification to
6893 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6894 A struct type is used directly when it is the type of a variable, member.
6895 Indirect uses arise through pointers to structs.
6896 That is, when use of an incomplete struct is valid, the use is indirect.
6898 @samp{struct one direct; struct two * indirect;}.
6900 The optional second word limits the specification to
6901 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6902 Generic structs are a bit complicated to explain.
6903 For C++, these are non-explicit specializations of template classes,
6904 or non-template classes within the above.
6905 Other programming languages have generics,
6906 but @option{-femit-struct-debug-detailed} does not yet implement them.
6908 The third word specifies the source files for those
6909 structs for which the compiler should emit debug information.
6910 The values @samp{none} and @samp{any} have the normal meaning.
6911 The value @samp{base} means that
6912 the base of name of the file in which the type declaration appears
6913 must match the base of the name of the main compilation file.
6914 In practice, this means that when compiling @file{foo.c}, debug information
6915 is generated for types declared in that file and @file{foo.h},
6916 but not other header files.
6917 The value @samp{sys} means those types satisfying @samp{base}
6918 or declared in system or compiler headers.
6920 You may need to experiment to determine the best settings for your application.
6922 The default is @option{-femit-struct-debug-detailed=all}.
6924 This option works only with DWARF debug output.
6926 @item -fno-dwarf2-cfi-asm
6927 @opindex fdwarf2-cfi-asm
6928 @opindex fno-dwarf2-cfi-asm
6929 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6930 instead of using GAS @code{.cfi_*} directives.
6932 @item -fno-eliminate-unused-debug-types
6933 @opindex feliminate-unused-debug-types
6934 @opindex fno-eliminate-unused-debug-types
6935 Normally, when producing DWARF output, GCC avoids producing debug symbol
6936 output for types that are nowhere used in the source file being compiled.
6937 Sometimes it is useful to have GCC emit debugging
6938 information for all types declared in a compilation
6939 unit, regardless of whether or not they are actually used
6940 in that compilation unit, for example
6941 if, in the debugger, you want to cast a value to a type that is
6942 not actually used in your program (but is declared). More often,
6943 however, this results in a significant amount of wasted space.
6946 @node Optimize Options
6947 @section Options That Control Optimization
6948 @cindex optimize options
6949 @cindex options, optimization
6951 These options control various sorts of optimizations.
6953 Without any optimization option, the compiler's goal is to reduce the
6954 cost of compilation and to make debugging produce the expected
6955 results. Statements are independent: if you stop the program with a
6956 breakpoint between statements, you can then assign a new value to any
6957 variable or change the program counter to any other statement in the
6958 function and get exactly the results you expect from the source
6961 Turning on optimization flags makes the compiler attempt to improve
6962 the performance and/or code size at the expense of compilation time
6963 and possibly the ability to debug the program.
6965 The compiler performs optimization based on the knowledge it has of the
6966 program. Compiling multiple files at once to a single output file mode allows
6967 the compiler to use information gained from all of the files when compiling
6970 Not all optimizations are controlled directly by a flag. Only
6971 optimizations that have a flag are listed in this section.
6973 Most optimizations are only enabled if an @option{-O} level is set on
6974 the command line. Otherwise they are disabled, even if individual
6975 optimization flags are specified.
6977 Depending on the target and how GCC was configured, a slightly different
6978 set of optimizations may be enabled at each @option{-O} level than
6979 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6980 to find out the exact set of optimizations that are enabled at each level.
6981 @xref{Overall Options}, for examples.
6988 Optimize. Optimizing compilation takes somewhat more time, and a lot
6989 more memory for a large function.
6991 With @option{-O}, the compiler tries to reduce code size and execution
6992 time, without performing any optimizations that take a great deal of
6995 @option{-O} turns on the following optimization flags:
6998 -fbranch-count-reg @gol
6999 -fcombine-stack-adjustments @gol
7001 -fcprop-registers @gol
7004 -fdelayed-branch @gol
7006 -fforward-propagate @gol
7007 -fguess-branch-probability @gol
7008 -fif-conversion2 @gol
7009 -fif-conversion @gol
7010 -finline-functions-called-once @gol
7011 -fipa-pure-const @gol
7013 -fipa-reference @gol
7014 -fmerge-constants @gol
7015 -fmove-loop-invariants @gol
7016 -freorder-blocks @gol
7018 -fshrink-wrap-separate @gol
7019 -fsplit-wide-types @gol
7025 -ftree-coalesce-vars @gol
7026 -ftree-copy-prop @gol
7028 -ftree-dominator-opts @gol
7030 -ftree-forwprop @gol
7040 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7041 where doing so does not interfere with debugging.
7045 Optimize even more. GCC performs nearly all supported optimizations
7046 that do not involve a space-speed tradeoff.
7047 As compared to @option{-O}, this option increases both compilation time
7048 and the performance of the generated code.
7050 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7051 also turns on the following optimization flags:
7052 @gccoptlist{-fthread-jumps @gol
7053 -falign-functions -falign-jumps @gol
7054 -falign-loops -falign-labels @gol
7057 -fcse-follow-jumps -fcse-skip-blocks @gol
7058 -fdelete-null-pointer-checks @gol
7059 -fdevirtualize -fdevirtualize-speculatively @gol
7060 -fexpensive-optimizations @gol
7061 -fgcse -fgcse-lm @gol
7062 -fhoist-adjacent-loads @gol
7063 -finline-small-functions @gol
7064 -findirect-inlining @gol
7066 -fipa-cp-alignment @gol
7070 -fisolate-erroneous-paths-dereference @gol
7072 -foptimize-sibling-calls @gol
7073 -foptimize-strlen @gol
7074 -fpartial-inlining @gol
7076 -freorder-blocks-algorithm=stc @gol
7077 -freorder-blocks-and-partition -freorder-functions @gol
7078 -frerun-cse-after-loop @gol
7079 -fsched-interblock -fsched-spec @gol
7080 -fschedule-insns -fschedule-insns2 @gol
7081 -fstore-merging @gol
7082 -fstrict-aliasing -fstrict-overflow @gol
7083 -ftree-builtin-call-dce @gol
7084 -ftree-switch-conversion -ftree-tail-merge @gol
7085 -fcode-hoisting @gol
7090 Please note the warning under @option{-fgcse} about
7091 invoking @option{-O2} on programs that use computed gotos.
7095 Optimize yet more. @option{-O3} turns on all optimizations specified
7096 by @option{-O2} and also turns on the @option{-finline-functions},
7097 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7098 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7099 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
7100 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7101 @option{-ftree-partial-pre}, @option{-fpeel-loops}
7102 and @option{-fipa-cp-clone} options.
7106 Reduce compilation time and make debugging produce the expected
7107 results. This is the default.
7111 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7112 do not typically increase code size. It also performs further
7113 optimizations designed to reduce code size.
7115 @option{-Os} disables the following optimization flags:
7116 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7117 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7118 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7122 Disregard strict standards compliance. @option{-Ofast} enables all
7123 @option{-O3} optimizations. It also enables optimizations that are not
7124 valid for all standard-compliant programs.
7125 It turns on @option{-ffast-math} and the Fortran-specific
7126 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7130 Optimize debugging experience. @option{-Og} enables optimizations
7131 that do not interfere with debugging. It should be the optimization
7132 level of choice for the standard edit-compile-debug cycle, offering
7133 a reasonable level of optimization while maintaining fast compilation
7134 and a good debugging experience.
7137 If you use multiple @option{-O} options, with or without level numbers,
7138 the last such option is the one that is effective.
7140 Options of the form @option{-f@var{flag}} specify machine-independent
7141 flags. Most flags have both positive and negative forms; the negative
7142 form of @option{-ffoo} is @option{-fno-foo}. In the table
7143 below, only one of the forms is listed---the one you typically
7144 use. You can figure out the other form by either removing @samp{no-}
7147 The following options control specific optimizations. They are either
7148 activated by @option{-O} options or are related to ones that are. You
7149 can use the following flags in the rare cases when ``fine-tuning'' of
7150 optimizations to be performed is desired.
7153 @item -fno-defer-pop
7154 @opindex fno-defer-pop
7155 Always pop the arguments to each function call as soon as that function
7156 returns. For machines that must pop arguments after a function call,
7157 the compiler normally lets arguments accumulate on the stack for several
7158 function calls and pops them all at once.
7160 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7162 @item -fforward-propagate
7163 @opindex fforward-propagate
7164 Perform a forward propagation pass on RTL@. The pass tries to combine two
7165 instructions and checks if the result can be simplified. If loop unrolling
7166 is active, two passes are performed and the second is scheduled after
7169 This option is enabled by default at optimization levels @option{-O},
7170 @option{-O2}, @option{-O3}, @option{-Os}.
7172 @item -ffp-contract=@var{style}
7173 @opindex ffp-contract
7174 @option{-ffp-contract=off} disables floating-point expression contraction.
7175 @option{-ffp-contract=fast} enables floating-point expression contraction
7176 such as forming of fused multiply-add operations if the target has
7177 native support for them.
7178 @option{-ffp-contract=on} enables floating-point expression contraction
7179 if allowed by the language standard. This is currently not implemented
7180 and treated equal to @option{-ffp-contract=off}.
7182 The default is @option{-ffp-contract=fast}.
7184 @item -fomit-frame-pointer
7185 @opindex fomit-frame-pointer
7186 Don't keep the frame pointer in a register for functions that
7187 don't need one. This avoids the instructions to save, set up and
7188 restore frame pointers; it also makes an extra register available
7189 in many functions. @strong{It also makes debugging impossible on
7192 On some machines, such as the VAX, this flag has no effect, because
7193 the standard calling sequence automatically handles the frame pointer
7194 and nothing is saved by pretending it doesn't exist. The
7195 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7196 whether a target machine supports this flag. @xref{Registers,,Register
7197 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7199 The default setting (when not optimizing for
7200 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7201 @option{-fomit-frame-pointer}. You can configure GCC with the
7202 @option{--enable-frame-pointer} configure option to change the default.
7204 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7206 @item -foptimize-sibling-calls
7207 @opindex foptimize-sibling-calls
7208 Optimize sibling and tail recursive calls.
7210 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7212 @item -foptimize-strlen
7213 @opindex foptimize-strlen
7214 Optimize various standard C string functions (e.g. @code{strlen},
7215 @code{strchr} or @code{strcpy}) and
7216 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7218 Enabled at levels @option{-O2}, @option{-O3}.
7222 Do not expand any functions inline apart from those marked with
7223 the @code{always_inline} attribute. This is the default when not
7226 Single functions can be exempted from inlining by marking them
7227 with the @code{noinline} attribute.
7229 @item -finline-small-functions
7230 @opindex finline-small-functions
7231 Integrate functions into their callers when their body is smaller than expected
7232 function call code (so overall size of program gets smaller). The compiler
7233 heuristically decides which functions are simple enough to be worth integrating
7234 in this way. This inlining applies to all functions, even those not declared
7237 Enabled at level @option{-O2}.
7239 @item -findirect-inlining
7240 @opindex findirect-inlining
7241 Inline also indirect calls that are discovered to be known at compile
7242 time thanks to previous inlining. This option has any effect only
7243 when inlining itself is turned on by the @option{-finline-functions}
7244 or @option{-finline-small-functions} options.
7246 Enabled at level @option{-O2}.
7248 @item -finline-functions
7249 @opindex finline-functions
7250 Consider all functions for inlining, even if they are not declared inline.
7251 The compiler heuristically decides which functions are worth integrating
7254 If all calls to a given function are integrated, and the function is
7255 declared @code{static}, then the function is normally not output as
7256 assembler code in its own right.
7258 Enabled at level @option{-O3}.
7260 @item -finline-functions-called-once
7261 @opindex finline-functions-called-once
7262 Consider all @code{static} functions called once for inlining into their
7263 caller even if they are not marked @code{inline}. If a call to a given
7264 function is integrated, then the function is not output as assembler code
7267 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7269 @item -fearly-inlining
7270 @opindex fearly-inlining
7271 Inline functions marked by @code{always_inline} and functions whose body seems
7272 smaller than the function call overhead early before doing
7273 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7274 makes profiling significantly cheaper and usually inlining faster on programs
7275 having large chains of nested wrapper functions.
7281 Perform interprocedural scalar replacement of aggregates, removal of
7282 unused parameters and replacement of parameters passed by reference
7283 by parameters passed by value.
7285 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7287 @item -finline-limit=@var{n}
7288 @opindex finline-limit
7289 By default, GCC limits the size of functions that can be inlined. This flag
7290 allows coarse control of this limit. @var{n} is the size of functions that
7291 can be inlined in number of pseudo instructions.
7293 Inlining is actually controlled by a number of parameters, which may be
7294 specified individually by using @option{--param @var{name}=@var{value}}.
7295 The @option{-finline-limit=@var{n}} option sets some of these parameters
7299 @item max-inline-insns-single
7300 is set to @var{n}/2.
7301 @item max-inline-insns-auto
7302 is set to @var{n}/2.
7305 See below for a documentation of the individual
7306 parameters controlling inlining and for the defaults of these parameters.
7308 @emph{Note:} there may be no value to @option{-finline-limit} that results
7309 in default behavior.
7311 @emph{Note:} pseudo instruction represents, in this particular context, an
7312 abstract measurement of function's size. In no way does it represent a count
7313 of assembly instructions and as such its exact meaning might change from one
7314 release to an another.
7316 @item -fno-keep-inline-dllexport
7317 @opindex fno-keep-inline-dllexport
7318 This is a more fine-grained version of @option{-fkeep-inline-functions},
7319 which applies only to functions that are declared using the @code{dllexport}
7320 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
7323 @item -fkeep-inline-functions
7324 @opindex fkeep-inline-functions
7325 In C, emit @code{static} functions that are declared @code{inline}
7326 into the object file, even if the function has been inlined into all
7327 of its callers. This switch does not affect functions using the
7328 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7329 inline functions into the object file.
7331 @item -fkeep-static-functions
7332 @opindex fkeep-static-functions
7333 Emit @code{static} functions into the object file, even if the function
7336 @item -fkeep-static-consts
7337 @opindex fkeep-static-consts
7338 Emit variables declared @code{static const} when optimization isn't turned
7339 on, even if the variables aren't referenced.
7341 GCC enables this option by default. If you want to force the compiler to
7342 check if a variable is referenced, regardless of whether or not
7343 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7345 @item -fmerge-constants
7346 @opindex fmerge-constants
7347 Attempt to merge identical constants (string constants and floating-point
7348 constants) across compilation units.
7350 This option is the default for optimized compilation if the assembler and
7351 linker support it. Use @option{-fno-merge-constants} to inhibit this
7354 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7356 @item -fmerge-all-constants
7357 @opindex fmerge-all-constants
7358 Attempt to merge identical constants and identical variables.
7360 This option implies @option{-fmerge-constants}. In addition to
7361 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7362 arrays or initialized constant variables with integral or floating-point
7363 types. Languages like C or C++ require each variable, including multiple
7364 instances of the same variable in recursive calls, to have distinct locations,
7365 so using this option results in non-conforming
7368 @item -fmodulo-sched
7369 @opindex fmodulo-sched
7370 Perform swing modulo scheduling immediately before the first scheduling
7371 pass. This pass looks at innermost loops and reorders their
7372 instructions by overlapping different iterations.
7374 @item -fmodulo-sched-allow-regmoves
7375 @opindex fmodulo-sched-allow-regmoves
7376 Perform more aggressive SMS-based modulo scheduling with register moves
7377 allowed. By setting this flag certain anti-dependences edges are
7378 deleted, which triggers the generation of reg-moves based on the
7379 life-range analysis. This option is effective only with
7380 @option{-fmodulo-sched} enabled.
7382 @item -fno-branch-count-reg
7383 @opindex fno-branch-count-reg
7384 Avoid running a pass scanning for opportunities to use ``decrement and
7385 branch'' instructions on a count register instead of generating sequences
7386 of instructions that decrement a register, compare it against zero, and
7387 then branch based upon the result. This option is only meaningful on
7388 architectures that support such instructions, which include x86, PowerPC,
7389 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
7390 doesn't remove the decrement and branch instructions from the generated
7391 instruction stream introduced by other optimization passes.
7393 Enabled by default at @option{-O1} and higher.
7395 The default is @option{-fbranch-count-reg}.
7397 @item -fno-function-cse
7398 @opindex fno-function-cse
7399 Do not put function addresses in registers; make each instruction that
7400 calls a constant function contain the function's address explicitly.
7402 This option results in less efficient code, but some strange hacks
7403 that alter the assembler output may be confused by the optimizations
7404 performed when this option is not used.
7406 The default is @option{-ffunction-cse}
7408 @item -fno-zero-initialized-in-bss
7409 @opindex fno-zero-initialized-in-bss
7410 If the target supports a BSS section, GCC by default puts variables that
7411 are initialized to zero into BSS@. This can save space in the resulting
7414 This option turns off this behavior because some programs explicitly
7415 rely on variables going to the data section---e.g., so that the
7416 resulting executable can find the beginning of that section and/or make
7417 assumptions based on that.
7419 The default is @option{-fzero-initialized-in-bss}.
7421 @item -fthread-jumps
7422 @opindex fthread-jumps
7423 Perform optimizations that check to see if a jump branches to a
7424 location where another comparison subsumed by the first is found. If
7425 so, the first branch is redirected to either the destination of the
7426 second branch or a point immediately following it, depending on whether
7427 the condition is known to be true or false.
7429 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7431 @item -fsplit-wide-types
7432 @opindex fsplit-wide-types
7433 When using a type that occupies multiple registers, such as @code{long
7434 long} on a 32-bit system, split the registers apart and allocate them
7435 independently. This normally generates better code for those types,
7436 but may make debugging more difficult.
7438 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7441 @item -fcse-follow-jumps
7442 @opindex fcse-follow-jumps
7443 In common subexpression elimination (CSE), scan through jump instructions
7444 when the target of the jump is not reached by any other path. For
7445 example, when CSE encounters an @code{if} statement with an
7446 @code{else} clause, CSE follows the jump when the condition
7449 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7451 @item -fcse-skip-blocks
7452 @opindex fcse-skip-blocks
7453 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7454 follow jumps that conditionally skip over blocks. When CSE
7455 encounters a simple @code{if} statement with no else clause,
7456 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7457 body of the @code{if}.
7459 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7461 @item -frerun-cse-after-loop
7462 @opindex frerun-cse-after-loop
7463 Re-run common subexpression elimination after loop optimizations are
7466 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7470 Perform a global common subexpression elimination pass.
7471 This pass also performs global constant and copy propagation.
7473 @emph{Note:} When compiling a program using computed gotos, a GCC
7474 extension, you may get better run-time performance if you disable
7475 the global common subexpression elimination pass by adding
7476 @option{-fno-gcse} to the command line.
7478 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7482 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7483 attempts to move loads that are only killed by stores into themselves. This
7484 allows a loop containing a load/store sequence to be changed to a load outside
7485 the loop, and a copy/store within the loop.
7487 Enabled by default when @option{-fgcse} is enabled.
7491 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7492 global common subexpression elimination. This pass attempts to move
7493 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7494 loops containing a load/store sequence can be changed to a load before
7495 the loop and a store after the loop.
7497 Not enabled at any optimization level.
7501 When @option{-fgcse-las} is enabled, the global common subexpression
7502 elimination pass eliminates redundant loads that come after stores to the
7503 same memory location (both partial and full redundancies).
7505 Not enabled at any optimization level.
7507 @item -fgcse-after-reload
7508 @opindex fgcse-after-reload
7509 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7510 pass is performed after reload. The purpose of this pass is to clean up
7513 @item -faggressive-loop-optimizations
7514 @opindex faggressive-loop-optimizations
7515 This option tells the loop optimizer to use language constraints to
7516 derive bounds for the number of iterations of a loop. This assumes that
7517 loop code does not invoke undefined behavior by for example causing signed
7518 integer overflows or out-of-bound array accesses. The bounds for the
7519 number of iterations of a loop are used to guide loop unrolling and peeling
7520 and loop exit test optimizations.
7521 This option is enabled by default.
7523 @item -funconstrained-commons
7524 @opindex funconstrained-commons
7525 This option tells the compiler that variables declared in common blocks
7526 (e.g. Fortran) may later be overridden with longer trailing arrays. This
7527 prevents certain optimizations that depend on knowing the array bounds.
7529 @item -fcrossjumping
7530 @opindex fcrossjumping
7531 Perform cross-jumping transformation.
7532 This transformation unifies equivalent code and saves code size. The
7533 resulting code may or may not perform better than without cross-jumping.
7535 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7537 @item -fauto-inc-dec
7538 @opindex fauto-inc-dec
7539 Combine increments or decrements of addresses with memory accesses.
7540 This pass is always skipped on architectures that do not have
7541 instructions to support this. Enabled by default at @option{-O} and
7542 higher on architectures that support this.
7546 Perform dead code elimination (DCE) on RTL@.
7547 Enabled by default at @option{-O} and higher.
7551 Perform dead store elimination (DSE) on RTL@.
7552 Enabled by default at @option{-O} and higher.
7554 @item -fif-conversion
7555 @opindex fif-conversion
7556 Attempt to transform conditional jumps into branch-less equivalents. This
7557 includes use of conditional moves, min, max, set flags and abs instructions, and
7558 some tricks doable by standard arithmetics. The use of conditional execution
7559 on chips where it is available is controlled by @option{-fif-conversion2}.
7561 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7563 @item -fif-conversion2
7564 @opindex fif-conversion2
7565 Use conditional execution (where available) to transform conditional jumps into
7566 branch-less equivalents.
7568 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7570 @item -fdeclone-ctor-dtor
7571 @opindex fdeclone-ctor-dtor
7572 The C++ ABI requires multiple entry points for constructors and
7573 destructors: one for a base subobject, one for a complete object, and
7574 one for a virtual destructor that calls operator delete afterwards.
7575 For a hierarchy with virtual bases, the base and complete variants are
7576 clones, which means two copies of the function. With this option, the
7577 base and complete variants are changed to be thunks that call a common
7580 Enabled by @option{-Os}.
7582 @item -fdelete-null-pointer-checks
7583 @opindex fdelete-null-pointer-checks
7584 Assume that programs cannot safely dereference null pointers, and that
7585 no code or data element resides at address zero.
7586 This option enables simple constant
7587 folding optimizations at all optimization levels. In addition, other
7588 optimization passes in GCC use this flag to control global dataflow
7589 analyses that eliminate useless checks for null pointers; these assume
7590 that a memory access to address zero always results in a trap, so
7591 that if a pointer is checked after it has already been dereferenced,
7594 Note however that in some environments this assumption is not true.
7595 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7596 for programs that depend on that behavior.
7598 This option is enabled by default on most targets. On Nios II ELF, it
7599 defaults to off. On AVR and CR16, this option is completely disabled.
7601 Passes that use the dataflow information
7602 are enabled independently at different optimization levels.
7604 @item -fdevirtualize
7605 @opindex fdevirtualize
7606 Attempt to convert calls to virtual functions to direct calls. This
7607 is done both within a procedure and interprocedurally as part of
7608 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7609 propagation (@option{-fipa-cp}).
7610 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7612 @item -fdevirtualize-speculatively
7613 @opindex fdevirtualize-speculatively
7614 Attempt to convert calls to virtual functions to speculative direct calls.
7615 Based on the analysis of the type inheritance graph, determine for a given call
7616 the set of likely targets. If the set is small, preferably of size 1, change
7617 the call into a conditional deciding between direct and indirect calls. The
7618 speculative calls enable more optimizations, such as inlining. When they seem
7619 useless after further optimization, they are converted back into original form.
7621 @item -fdevirtualize-at-ltrans
7622 @opindex fdevirtualize-at-ltrans
7623 Stream extra information needed for aggressive devirtualization when running
7624 the link-time optimizer in local transformation mode.
7625 This option enables more devirtualization but
7626 significantly increases the size of streamed data. For this reason it is
7627 disabled by default.
7629 @item -fexpensive-optimizations
7630 @opindex fexpensive-optimizations
7631 Perform a number of minor optimizations that are relatively expensive.
7633 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7637 Attempt to remove redundant extension instructions. This is especially
7638 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7639 registers after writing to their lower 32-bit half.
7641 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7642 @option{-O3}, @option{-Os}.
7644 @item -fno-lifetime-dse
7645 @opindex fno-lifetime-dse
7646 In C++ the value of an object is only affected by changes within its
7647 lifetime: when the constructor begins, the object has an indeterminate
7648 value, and any changes during the lifetime of the object are dead when
7649 the object is destroyed. Normally dead store elimination will take
7650 advantage of this; if your code relies on the value of the object
7651 storage persisting beyond the lifetime of the object, you can use this
7652 flag to disable this optimization. To preserve stores before the
7653 constructor starts (e.g. because your operator new clears the object
7654 storage) but still treat the object as dead after the destructor you,
7655 can use @option{-flifetime-dse=1}. The default behavior can be
7656 explicitly selected with @option{-flifetime-dse=2}.
7657 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
7659 @item -flive-range-shrinkage
7660 @opindex flive-range-shrinkage
7661 Attempt to decrease register pressure through register live range
7662 shrinkage. This is helpful for fast processors with small or moderate
7665 @item -fira-algorithm=@var{algorithm}
7666 @opindex fira-algorithm
7667 Use the specified coloring algorithm for the integrated register
7668 allocator. The @var{algorithm} argument can be @samp{priority}, which
7669 specifies Chow's priority coloring, or @samp{CB}, which specifies
7670 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7671 for all architectures, but for those targets that do support it, it is
7672 the default because it generates better code.
7674 @item -fira-region=@var{region}
7675 @opindex fira-region
7676 Use specified regions for the integrated register allocator. The
7677 @var{region} argument should be one of the following:
7682 Use all loops as register allocation regions.
7683 This can give the best results for machines with a small and/or
7684 irregular register set.
7687 Use all loops except for loops with small register pressure
7688 as the regions. This value usually gives
7689 the best results in most cases and for most architectures,
7690 and is enabled by default when compiling with optimization for speed
7691 (@option{-O}, @option{-O2}, @dots{}).
7694 Use all functions as a single region.
7695 This typically results in the smallest code size, and is enabled by default for
7696 @option{-Os} or @option{-O0}.
7700 @item -fira-hoist-pressure
7701 @opindex fira-hoist-pressure
7702 Use IRA to evaluate register pressure in the code hoisting pass for
7703 decisions to hoist expressions. This option usually results in smaller
7704 code, but it can slow the compiler down.
7706 This option is enabled at level @option{-Os} for all targets.
7708 @item -fira-loop-pressure
7709 @opindex fira-loop-pressure
7710 Use IRA to evaluate register pressure in loops for decisions to move
7711 loop invariants. This option usually results in generation
7712 of faster and smaller code on machines with large register files (>= 32
7713 registers), but it can slow the compiler down.
7715 This option is enabled at level @option{-O3} for some targets.
7717 @item -fno-ira-share-save-slots
7718 @opindex fno-ira-share-save-slots
7719 Disable sharing of stack slots used for saving call-used hard
7720 registers living through a call. Each hard register gets a
7721 separate stack slot, and as a result function stack frames are
7724 @item -fno-ira-share-spill-slots
7725 @opindex fno-ira-share-spill-slots
7726 Disable sharing of stack slots allocated for pseudo-registers. Each
7727 pseudo-register that does not get a hard register gets a separate
7728 stack slot, and as a result function stack frames are larger.
7732 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7733 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7734 values if it is profitable.
7736 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7738 @item -fdelayed-branch
7739 @opindex fdelayed-branch
7740 If supported for the target machine, attempt to reorder instructions
7741 to exploit instruction slots available after delayed branch
7744 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7746 @item -fschedule-insns
7747 @opindex fschedule-insns
7748 If supported for the target machine, attempt to reorder instructions to
7749 eliminate execution stalls due to required data being unavailable. This
7750 helps machines that have slow floating point or memory load instructions
7751 by allowing other instructions to be issued until the result of the load
7752 or floating-point instruction is required.
7754 Enabled at levels @option{-O2}, @option{-O3}.
7756 @item -fschedule-insns2
7757 @opindex fschedule-insns2
7758 Similar to @option{-fschedule-insns}, but requests an additional pass of
7759 instruction scheduling after register allocation has been done. This is
7760 especially useful on machines with a relatively small number of
7761 registers and where memory load instructions take more than one cycle.
7763 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7765 @item -fno-sched-interblock
7766 @opindex fno-sched-interblock
7767 Don't schedule instructions across basic blocks. This is normally
7768 enabled by default when scheduling before register allocation, i.e.@:
7769 with @option{-fschedule-insns} or at @option{-O2} or higher.
7771 @item -fno-sched-spec
7772 @opindex fno-sched-spec
7773 Don't allow speculative motion of non-load instructions. This is normally
7774 enabled by default when scheduling before register allocation, i.e.@:
7775 with @option{-fschedule-insns} or at @option{-O2} or higher.
7777 @item -fsched-pressure
7778 @opindex fsched-pressure
7779 Enable register pressure sensitive insn scheduling before register
7780 allocation. This only makes sense when scheduling before register
7781 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7782 @option{-O2} or higher. Usage of this option can improve the
7783 generated code and decrease its size by preventing register pressure
7784 increase above the number of available hard registers and subsequent
7785 spills in register allocation.
7787 @item -fsched-spec-load
7788 @opindex fsched-spec-load
7789 Allow speculative motion of some load instructions. This only makes
7790 sense when scheduling before register allocation, i.e.@: with
7791 @option{-fschedule-insns} or at @option{-O2} or higher.
7793 @item -fsched-spec-load-dangerous
7794 @opindex fsched-spec-load-dangerous
7795 Allow speculative motion of more load instructions. This only makes
7796 sense when scheduling before register allocation, i.e.@: with
7797 @option{-fschedule-insns} or at @option{-O2} or higher.
7799 @item -fsched-stalled-insns
7800 @itemx -fsched-stalled-insns=@var{n}
7801 @opindex fsched-stalled-insns
7802 Define how many insns (if any) can be moved prematurely from the queue
7803 of stalled insns into the ready list during the second scheduling pass.
7804 @option{-fno-sched-stalled-insns} means that no insns are moved
7805 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7806 on how many queued insns can be moved prematurely.
7807 @option{-fsched-stalled-insns} without a value is equivalent to
7808 @option{-fsched-stalled-insns=1}.
7810 @item -fsched-stalled-insns-dep
7811 @itemx -fsched-stalled-insns-dep=@var{n}
7812 @opindex fsched-stalled-insns-dep
7813 Define how many insn groups (cycles) are examined for a dependency
7814 on a stalled insn that is a candidate for premature removal from the queue
7815 of stalled insns. This has an effect only during the second scheduling pass,
7816 and only if @option{-fsched-stalled-insns} is used.
7817 @option{-fno-sched-stalled-insns-dep} is equivalent to
7818 @option{-fsched-stalled-insns-dep=0}.
7819 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7820 @option{-fsched-stalled-insns-dep=1}.
7822 @item -fsched2-use-superblocks
7823 @opindex fsched2-use-superblocks
7824 When scheduling after register allocation, use superblock scheduling.
7825 This allows motion across basic block boundaries,
7826 resulting in faster schedules. This option is experimental, as not all machine
7827 descriptions used by GCC model the CPU closely enough to avoid unreliable
7828 results from the algorithm.
7830 This only makes sense when scheduling after register allocation, i.e.@: with
7831 @option{-fschedule-insns2} or at @option{-O2} or higher.
7833 @item -fsched-group-heuristic
7834 @opindex fsched-group-heuristic
7835 Enable the group heuristic in the scheduler. This heuristic favors
7836 the instruction that belongs to a schedule group. This is enabled
7837 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7838 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7840 @item -fsched-critical-path-heuristic
7841 @opindex fsched-critical-path-heuristic
7842 Enable the critical-path heuristic in the scheduler. This heuristic favors
7843 instructions on the critical path. This is enabled by default when
7844 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7845 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7847 @item -fsched-spec-insn-heuristic
7848 @opindex fsched-spec-insn-heuristic
7849 Enable the speculative instruction heuristic in the scheduler. This
7850 heuristic favors speculative instructions with greater dependency weakness.
7851 This is enabled by default when scheduling is enabled, i.e.@:
7852 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7853 or at @option{-O2} or higher.
7855 @item -fsched-rank-heuristic
7856 @opindex fsched-rank-heuristic
7857 Enable the rank heuristic in the scheduler. This heuristic favors
7858 the instruction belonging to a basic block with greater size or frequency.
7859 This is enabled by default when scheduling is enabled, i.e.@:
7860 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7861 at @option{-O2} or higher.
7863 @item -fsched-last-insn-heuristic
7864 @opindex fsched-last-insn-heuristic
7865 Enable the last-instruction heuristic in the scheduler. This heuristic
7866 favors the instruction that is less dependent on the last instruction
7867 scheduled. This is enabled by default when scheduling is enabled,
7868 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7869 at @option{-O2} or higher.
7871 @item -fsched-dep-count-heuristic
7872 @opindex fsched-dep-count-heuristic
7873 Enable the dependent-count heuristic in the scheduler. This heuristic
7874 favors the instruction that has more instructions depending on it.
7875 This is enabled by default when scheduling is enabled, i.e.@:
7876 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7877 at @option{-O2} or higher.
7879 @item -freschedule-modulo-scheduled-loops
7880 @opindex freschedule-modulo-scheduled-loops
7881 Modulo scheduling is performed before traditional scheduling. If a loop
7882 is modulo scheduled, later scheduling passes may change its schedule.
7883 Use this option to control that behavior.
7885 @item -fselective-scheduling
7886 @opindex fselective-scheduling
7887 Schedule instructions using selective scheduling algorithm. Selective
7888 scheduling runs instead of the first scheduler pass.
7890 @item -fselective-scheduling2
7891 @opindex fselective-scheduling2
7892 Schedule instructions using selective scheduling algorithm. Selective
7893 scheduling runs instead of the second scheduler pass.
7895 @item -fsel-sched-pipelining
7896 @opindex fsel-sched-pipelining
7897 Enable software pipelining of innermost loops during selective scheduling.
7898 This option has no effect unless one of @option{-fselective-scheduling} or
7899 @option{-fselective-scheduling2} is turned on.
7901 @item -fsel-sched-pipelining-outer-loops
7902 @opindex fsel-sched-pipelining-outer-loops
7903 When pipelining loops during selective scheduling, also pipeline outer loops.
7904 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7906 @item -fsemantic-interposition
7907 @opindex fsemantic-interposition
7908 Some object formats, like ELF, allow interposing of symbols by the
7910 This means that for symbols exported from the DSO, the compiler cannot perform
7911 interprocedural propagation, inlining and other optimizations in anticipation
7912 that the function or variable in question may change. While this feature is
7913 useful, for example, to rewrite memory allocation functions by a debugging
7914 implementation, it is expensive in the terms of code quality.
7915 With @option{-fno-semantic-interposition} the compiler assumes that
7916 if interposition happens for functions the overwriting function will have
7917 precisely the same semantics (and side effects).
7918 Similarly if interposition happens
7919 for variables, the constructor of the variable will be the same. The flag
7920 has no effect for functions explicitly declared inline
7921 (where it is never allowed for interposition to change semantics)
7922 and for symbols explicitly declared weak.
7925 @opindex fshrink-wrap
7926 Emit function prologues only before parts of the function that need it,
7927 rather than at the top of the function. This flag is enabled by default at
7928 @option{-O} and higher.
7930 @item -fshrink-wrap-separate
7931 @opindex fshrink-wrap-separate
7932 Shrink-wrap separate parts of the prologue and epilogue separately, so that
7933 those parts are only executed when needed.
7934 This option is on by default, but has no effect unless @option{-fshrink-wrap}
7935 is also turned on and the target supports this.
7937 @item -fcaller-saves
7938 @opindex fcaller-saves
7939 Enable allocation of values to registers that are clobbered by
7940 function calls, by emitting extra instructions to save and restore the
7941 registers around such calls. Such allocation is done only when it
7942 seems to result in better code.
7944 This option is always enabled by default on certain machines, usually
7945 those which have no call-preserved registers to use instead.
7947 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7949 @item -fcombine-stack-adjustments
7950 @opindex fcombine-stack-adjustments
7951 Tracks stack adjustments (pushes and pops) and stack memory references
7952 and then tries to find ways to combine them.
7954 Enabled by default at @option{-O1} and higher.
7958 Use caller save registers for allocation if those registers are not used by
7959 any called function. In that case it is not necessary to save and restore
7960 them around calls. This is only possible if called functions are part of
7961 same compilation unit as current function and they are compiled before it.
7963 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
7964 is disabled if generated code will be instrumented for profiling
7965 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
7966 exactly (this happens on targets that do not expose prologues
7967 and epilogues in RTL).
7969 @item -fconserve-stack
7970 @opindex fconserve-stack
7971 Attempt to minimize stack usage. The compiler attempts to use less
7972 stack space, even if that makes the program slower. This option
7973 implies setting the @option{large-stack-frame} parameter to 100
7974 and the @option{large-stack-frame-growth} parameter to 400.
7976 @item -ftree-reassoc
7977 @opindex ftree-reassoc
7978 Perform reassociation on trees. This flag is enabled by default
7979 at @option{-O} and higher.
7981 @item -fcode-hoisting
7982 @opindex fcode-hoisting
7983 Perform code hoisting. Code hoisting tries to move the
7984 evaluation of expressions executed on all paths to the function exit
7985 as early as possible. This is especially useful as a code size
7986 optimization, but it often helps for code speed as well.
7987 This flag is enabled by default at @option{-O2} and higher.
7991 Perform partial redundancy elimination (PRE) on trees. This flag is
7992 enabled by default at @option{-O2} and @option{-O3}.
7994 @item -ftree-partial-pre
7995 @opindex ftree-partial-pre
7996 Make partial redundancy elimination (PRE) more aggressive. This flag is
7997 enabled by default at @option{-O3}.
7999 @item -ftree-forwprop
8000 @opindex ftree-forwprop
8001 Perform forward propagation on trees. This flag is enabled by default
8002 at @option{-O} and higher.
8006 Perform full redundancy elimination (FRE) on trees. The difference
8007 between FRE and PRE is that FRE only considers expressions
8008 that are computed on all paths leading to the redundant computation.
8009 This analysis is faster than PRE, though it exposes fewer redundancies.
8010 This flag is enabled by default at @option{-O} and higher.
8012 @item -ftree-phiprop
8013 @opindex ftree-phiprop
8014 Perform hoisting of loads from conditional pointers on trees. This
8015 pass is enabled by default at @option{-O} and higher.
8017 @item -fhoist-adjacent-loads
8018 @opindex fhoist-adjacent-loads
8019 Speculatively hoist loads from both branches of an if-then-else if the
8020 loads are from adjacent locations in the same structure and the target
8021 architecture has a conditional move instruction. This flag is enabled
8022 by default at @option{-O2} and higher.
8024 @item -ftree-copy-prop
8025 @opindex ftree-copy-prop
8026 Perform copy propagation on trees. This pass eliminates unnecessary
8027 copy operations. This flag is enabled by default at @option{-O} and
8030 @item -fipa-pure-const
8031 @opindex fipa-pure-const
8032 Discover which functions are pure or constant.
8033 Enabled by default at @option{-O} and higher.
8035 @item -fipa-reference
8036 @opindex fipa-reference
8037 Discover which static variables do not escape the
8039 Enabled by default at @option{-O} and higher.
8043 Perform interprocedural pointer analysis and interprocedural modification
8044 and reference analysis. This option can cause excessive memory and
8045 compile-time usage on large compilation units. It is not enabled by
8046 default at any optimization level.
8049 @opindex fipa-profile
8050 Perform interprocedural profile propagation. The functions called only from
8051 cold functions are marked as cold. Also functions executed once (such as
8052 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8053 functions and loop less parts of functions executed once are then optimized for
8055 Enabled by default at @option{-O} and higher.
8059 Perform interprocedural constant propagation.
8060 This optimization analyzes the program to determine when values passed
8061 to functions are constants and then optimizes accordingly.
8062 This optimization can substantially increase performance
8063 if the application has constants passed to functions.
8064 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8066 @item -fipa-cp-clone
8067 @opindex fipa-cp-clone
8068 Perform function cloning to make interprocedural constant propagation stronger.
8069 When enabled, interprocedural constant propagation performs function cloning
8070 when externally visible function can be called with constant arguments.
8071 Because this optimization can create multiple copies of functions,
8072 it may significantly increase code size
8073 (see @option{--param ipcp-unit-growth=@var{value}}).
8074 This flag is enabled by default at @option{-O3}.
8076 @item -fipa-cp-alignment
8077 @opindex -fipa-cp-alignment
8078 When enabled, this optimization propagates alignment of function
8079 parameters to support better vectorization and string operations.
8081 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8082 requires that @option{-fipa-cp} is enabled.
8083 @option{-fipa-cp-alignment} is obsolete, use @option{-fipa-bit-cp} instead.
8086 @opindex -fipa-bit-cp
8087 When enabled, perform ipa bitwise constant propagation. This flag is
8088 enabled by default at @option{-O2}. It requires that @option{-fipa-cp}
8093 Perform Identical Code Folding for functions and read-only variables.
8094 The optimization reduces code size and may disturb unwind stacks by replacing
8095 a function by equivalent one with a different name. The optimization works
8096 more effectively with link time optimization enabled.
8098 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8099 works on different levels and thus the optimizations are not same - there are
8100 equivalences that are found only by GCC and equivalences found only by Gold.
8102 This flag is enabled by default at @option{-O2} and @option{-Os}.
8104 @item -fisolate-erroneous-paths-dereference
8105 @opindex fisolate-erroneous-paths-dereference
8106 Detect paths that trigger erroneous or undefined behavior due to
8107 dereferencing a null pointer. Isolate those paths from the main control
8108 flow and turn the statement with erroneous or undefined behavior into a trap.
8109 This flag is enabled by default at @option{-O2} and higher and depends on
8110 @option{-fdelete-null-pointer-checks} also being enabled.
8112 @item -fisolate-erroneous-paths-attribute
8113 @opindex fisolate-erroneous-paths-attribute
8114 Detect paths that trigger erroneous or undefined behavior due a null value
8115 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8116 attribute. Isolate those paths from the main control flow and turn the
8117 statement with erroneous or undefined behavior into a trap. This is not
8118 currently enabled, but may be enabled by @option{-O2} in the future.
8122 Perform forward store motion on trees. This flag is
8123 enabled by default at @option{-O} and higher.
8125 @item -ftree-bit-ccp
8126 @opindex ftree-bit-ccp
8127 Perform sparse conditional bit constant propagation on trees and propagate
8128 pointer alignment information.
8129 This pass only operates on local scalar variables and is enabled by default
8130 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8134 Perform sparse conditional constant propagation (CCP) on trees. This
8135 pass only operates on local scalar variables and is enabled by default
8136 at @option{-O} and higher.
8138 @item -fssa-backprop
8139 @opindex fssa-backprop
8140 Propagate information about uses of a value up the definition chain
8141 in order to simplify the definitions. For example, this pass strips
8142 sign operations if the sign of a value never matters. The flag is
8143 enabled by default at @option{-O} and higher.
8146 @opindex fssa-phiopt
8147 Perform pattern matching on SSA PHI nodes to optimize conditional
8148 code. This pass is enabled by default at @option{-O} and higher.
8150 @item -ftree-switch-conversion
8151 @opindex ftree-switch-conversion
8152 Perform conversion of simple initializations in a switch to
8153 initializations from a scalar array. This flag is enabled by default
8154 at @option{-O2} and higher.
8156 @item -ftree-tail-merge
8157 @opindex ftree-tail-merge
8158 Look for identical code sequences. When found, replace one with a jump to the
8159 other. This optimization is known as tail merging or cross jumping. This flag
8160 is enabled by default at @option{-O2} and higher. The compilation time
8162 be limited using @option{max-tail-merge-comparisons} parameter and
8163 @option{max-tail-merge-iterations} parameter.
8167 Perform dead code elimination (DCE) on trees. This flag is enabled by
8168 default at @option{-O} and higher.
8170 @item -ftree-builtin-call-dce
8171 @opindex ftree-builtin-call-dce
8172 Perform conditional dead code elimination (DCE) for calls to built-in functions
8173 that may set @code{errno} but are otherwise side-effect free. This flag is
8174 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8177 @item -ftree-dominator-opts
8178 @opindex ftree-dominator-opts
8179 Perform a variety of simple scalar cleanups (constant/copy
8180 propagation, redundancy elimination, range propagation and expression
8181 simplification) based on a dominator tree traversal. This also
8182 performs jump threading (to reduce jumps to jumps). This flag is
8183 enabled by default at @option{-O} and higher.
8187 Perform dead store elimination (DSE) on trees. A dead store is a store into
8188 a memory location that is later overwritten by another store without
8189 any intervening loads. In this case the earlier store can be deleted. This
8190 flag is enabled by default at @option{-O} and higher.
8194 Perform loop header copying on trees. This is beneficial since it increases
8195 effectiveness of code motion optimizations. It also saves one jump. This flag
8196 is enabled by default at @option{-O} and higher. It is not enabled
8197 for @option{-Os}, since it usually increases code size.
8199 @item -ftree-loop-optimize
8200 @opindex ftree-loop-optimize
8201 Perform loop optimizations on trees. This flag is enabled by default
8202 at @option{-O} and higher.
8204 @item -ftree-loop-linear
8205 @itemx -floop-interchange
8206 @itemx -floop-strip-mine
8208 @itemx -floop-unroll-and-jam
8209 @opindex ftree-loop-linear
8210 @opindex floop-interchange
8211 @opindex floop-strip-mine
8212 @opindex floop-block
8213 @opindex floop-unroll-and-jam
8214 Perform loop nest optimizations. Same as
8215 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8216 to be configured with @option{--with-isl} to enable the Graphite loop
8217 transformation infrastructure.
8219 @item -fgraphite-identity
8220 @opindex fgraphite-identity
8221 Enable the identity transformation for graphite. For every SCoP we generate
8222 the polyhedral representation and transform it back to gimple. Using
8223 @option{-fgraphite-identity} we can check the costs or benefits of the
8224 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8225 are also performed by the code generator isl, like index splitting and
8226 dead code elimination in loops.
8228 @item -floop-nest-optimize
8229 @opindex floop-nest-optimize
8230 Enable the isl based loop nest optimizer. This is a generic loop nest
8231 optimizer based on the Pluto optimization algorithms. It calculates a loop
8232 structure optimized for data-locality and parallelism. This option
8235 @item -floop-parallelize-all
8236 @opindex floop-parallelize-all
8237 Use the Graphite data dependence analysis to identify loops that can
8238 be parallelized. Parallelize all the loops that can be analyzed to
8239 not contain loop carried dependences without checking that it is
8240 profitable to parallelize the loops.
8242 @item -ftree-coalesce-vars
8243 @opindex ftree-coalesce-vars
8244 While transforming the program out of the SSA representation, attempt to
8245 reduce copying by coalescing versions of different user-defined
8246 variables, instead of just compiler temporaries. This may severely
8247 limit the ability to debug an optimized program compiled with
8248 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8249 prevents SSA coalescing of user variables. This option is enabled by
8250 default if optimization is enabled, and it does very little otherwise.
8252 @item -ftree-loop-if-convert
8253 @opindex ftree-loop-if-convert
8254 Attempt to transform conditional jumps in the innermost loops to
8255 branch-less equivalents. The intent is to remove control-flow from
8256 the innermost loops in order to improve the ability of the
8257 vectorization pass to handle these loops. This is enabled by default
8258 if vectorization is enabled.
8260 @item -ftree-loop-distribution
8261 @opindex ftree-loop-distribution
8262 Perform loop distribution. This flag can improve cache performance on
8263 big loop bodies and allow further loop optimizations, like
8264 parallelization or vectorization, to take place. For example, the loop
8281 @item -ftree-loop-distribute-patterns
8282 @opindex ftree-loop-distribute-patterns
8283 Perform loop distribution of patterns that can be code generated with
8284 calls to a library. This flag is enabled by default at @option{-O3}.
8286 This pass distributes the initialization loops and generates a call to
8287 memset zero. For example, the loop
8303 and the initialization loop is transformed into a call to memset zero.
8305 @item -ftree-loop-im
8306 @opindex ftree-loop-im
8307 Perform loop invariant motion on trees. This pass moves only invariants that
8308 are hard to handle at RTL level (function calls, operations that expand to
8309 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8310 operands of conditions that are invariant out of the loop, so that we can use
8311 just trivial invariantness analysis in loop unswitching. The pass also includes
8314 @item -ftree-loop-ivcanon
8315 @opindex ftree-loop-ivcanon
8316 Create a canonical counter for number of iterations in loops for which
8317 determining number of iterations requires complicated analysis. Later
8318 optimizations then may determine the number easily. Useful especially
8319 in connection with unrolling.
8323 Perform induction variable optimizations (strength reduction, induction
8324 variable merging and induction variable elimination) on trees.
8326 @item -ftree-parallelize-loops=n
8327 @opindex ftree-parallelize-loops
8328 Parallelize loops, i.e., split their iteration space to run in n threads.
8329 This is only possible for loops whose iterations are independent
8330 and can be arbitrarily reordered. The optimization is only
8331 profitable on multiprocessor machines, for loops that are CPU-intensive,
8332 rather than constrained e.g.@: by memory bandwidth. This option
8333 implies @option{-pthread}, and thus is only supported on targets
8334 that have support for @option{-pthread}.
8338 Perform function-local points-to analysis on trees. This flag is
8339 enabled by default at @option{-O} and higher.
8343 Perform scalar replacement of aggregates. This pass replaces structure
8344 references with scalars to prevent committing structures to memory too
8345 early. This flag is enabled by default at @option{-O} and higher.
8347 @item -fstore-merging
8348 @opindex fstore-merging
8349 Perform merging of narrow stores to consecutive memory addresses. This pass
8350 merges contiguous stores of immediate values narrower than a word into fewer
8351 wider stores to reduce the number of instructions. This is enabled by default
8352 at @option{-O2} and higher as well as @option{-Os}.
8356 Perform temporary expression replacement during the SSA->normal phase. Single
8357 use/single def temporaries are replaced at their use location with their
8358 defining expression. This results in non-GIMPLE code, but gives the expanders
8359 much more complex trees to work on resulting in better RTL generation. This is
8360 enabled by default at @option{-O} and higher.
8364 Perform straight-line strength reduction on trees. This recognizes related
8365 expressions involving multiplications and replaces them by less expensive
8366 calculations when possible. This is enabled by default at @option{-O} and
8369 @item -ftree-vectorize
8370 @opindex ftree-vectorize
8371 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8372 and @option{-ftree-slp-vectorize} if not explicitly specified.
8374 @item -ftree-loop-vectorize
8375 @opindex ftree-loop-vectorize
8376 Perform loop vectorization on trees. This flag is enabled by default at
8377 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8379 @item -ftree-slp-vectorize
8380 @opindex ftree-slp-vectorize
8381 Perform basic block vectorization on trees. This flag is enabled by default at
8382 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8384 @item -fvect-cost-model=@var{model}
8385 @opindex fvect-cost-model
8386 Alter the cost model used for vectorization. The @var{model} argument
8387 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8388 With the @samp{unlimited} model the vectorized code-path is assumed
8389 to be profitable while with the @samp{dynamic} model a runtime check
8390 guards the vectorized code-path to enable it only for iteration
8391 counts that will likely execute faster than when executing the original
8392 scalar loop. The @samp{cheap} model disables vectorization of
8393 loops where doing so would be cost prohibitive for example due to
8394 required runtime checks for data dependence or alignment but otherwise
8395 is equal to the @samp{dynamic} model.
8396 The default cost model depends on other optimization flags and is
8397 either @samp{dynamic} or @samp{cheap}.
8399 @item -fsimd-cost-model=@var{model}
8400 @opindex fsimd-cost-model
8401 Alter the cost model used for vectorization of loops marked with the OpenMP
8402 or Cilk Plus simd directive. The @var{model} argument should be one of
8403 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8404 have the same meaning as described in @option{-fvect-cost-model} and by
8405 default a cost model defined with @option{-fvect-cost-model} is used.
8409 Perform Value Range Propagation on trees. This is similar to the
8410 constant propagation pass, but instead of values, ranges of values are
8411 propagated. This allows the optimizers to remove unnecessary range
8412 checks like array bound checks and null pointer checks. This is
8413 enabled by default at @option{-O2} and higher. Null pointer check
8414 elimination is only done if @option{-fdelete-null-pointer-checks} is
8418 @opindex fsplit-paths
8419 Split paths leading to loop backedges. This can improve dead code
8420 elimination and common subexpression elimination. This is enabled by
8421 default at @option{-O2} and above.
8423 @item -fsplit-ivs-in-unroller
8424 @opindex fsplit-ivs-in-unroller
8425 Enables expression of values of induction variables in later iterations
8426 of the unrolled loop using the value in the first iteration. This breaks
8427 long dependency chains, thus improving efficiency of the scheduling passes.
8429 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8430 same effect. However, that is not reliable in cases where the loop body
8431 is more complicated than a single basic block. It also does not work at all
8432 on some architectures due to restrictions in the CSE pass.
8434 This optimization is enabled by default.
8436 @item -fvariable-expansion-in-unroller
8437 @opindex fvariable-expansion-in-unroller
8438 With this option, the compiler creates multiple copies of some
8439 local variables when unrolling a loop, which can result in superior code.
8441 @item -fpartial-inlining
8442 @opindex fpartial-inlining
8443 Inline parts of functions. This option has any effect only
8444 when inlining itself is turned on by the @option{-finline-functions}
8445 or @option{-finline-small-functions} options.
8447 Enabled at level @option{-O2}.
8449 @item -fpredictive-commoning
8450 @opindex fpredictive-commoning
8451 Perform predictive commoning optimization, i.e., reusing computations
8452 (especially memory loads and stores) performed in previous
8453 iterations of loops.
8455 This option is enabled at level @option{-O3}.
8457 @item -fprefetch-loop-arrays
8458 @opindex fprefetch-loop-arrays
8459 If supported by the target machine, generate instructions to prefetch
8460 memory to improve the performance of loops that access large arrays.
8462 This option may generate better or worse code; results are highly
8463 dependent on the structure of loops within the source code.
8465 Disabled at level @option{-Os}.
8467 @item -fno-printf-return-value
8468 @opindex fno-printf-return-value
8469 Do not substitute constants for known return value of formatted output
8470 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
8471 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
8472 transformation allows GCC to optimize or even eliminate branches based
8473 on the known return value of these functions called with arguments that
8474 are either constant, or whose values are known to be in a range that
8475 makes determining the exact return value possible. For example, when
8476 @option{-fprintf-return-value} is in effect, both the branch and the
8477 body of the @code{if} statement (but not the call to @code{snprint})
8478 can be optimized away when @code{i} is a 32-bit or smaller integer
8479 because the return value is guaranteed to be at most 8.
8483 if (snprintf (buf, "%08x", i) >= sizeof buf)
8487 The @option{-fprintf-return-value} option relies on other optimizations
8488 and yields best results with @option{-O2}. It works in tandem with the
8489 @option{-Wformat-overflow} and @option{-Wformat-truncation} options.
8490 The @option{-fprintf-return-value} option is enabled by default.
8493 @itemx -fno-peephole2
8494 @opindex fno-peephole
8495 @opindex fno-peephole2
8496 Disable any machine-specific peephole optimizations. The difference
8497 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8498 are implemented in the compiler; some targets use one, some use the
8499 other, a few use both.
8501 @option{-fpeephole} is enabled by default.
8502 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8504 @item -fno-guess-branch-probability
8505 @opindex fno-guess-branch-probability
8506 Do not guess branch probabilities using heuristics.
8508 GCC uses heuristics to guess branch probabilities if they are
8509 not provided by profiling feedback (@option{-fprofile-arcs}). These
8510 heuristics are based on the control flow graph. If some branch probabilities
8511 are specified by @code{__builtin_expect}, then the heuristics are
8512 used to guess branch probabilities for the rest of the control flow graph,
8513 taking the @code{__builtin_expect} info into account. The interactions
8514 between the heuristics and @code{__builtin_expect} can be complex, and in
8515 some cases, it may be useful to disable the heuristics so that the effects
8516 of @code{__builtin_expect} are easier to understand.
8518 The default is @option{-fguess-branch-probability} at levels
8519 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8521 @item -freorder-blocks
8522 @opindex freorder-blocks
8523 Reorder basic blocks in the compiled function in order to reduce number of
8524 taken branches and improve code locality.
8526 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8528 @item -freorder-blocks-algorithm=@var{algorithm}
8529 @opindex freorder-blocks-algorithm
8530 Use the specified algorithm for basic block reordering. The
8531 @var{algorithm} argument can be @samp{simple}, which does not increase
8532 code size (except sometimes due to secondary effects like alignment),
8533 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
8534 put all often executed code together, minimizing the number of branches
8535 executed by making extra copies of code.
8537 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
8538 @samp{stc} at levels @option{-O2}, @option{-O3}.
8540 @item -freorder-blocks-and-partition
8541 @opindex freorder-blocks-and-partition
8542 In addition to reordering basic blocks in the compiled function, in order
8543 to reduce number of taken branches, partitions hot and cold basic blocks
8544 into separate sections of the assembly and @file{.o} files, to improve
8545 paging and cache locality performance.
8547 This optimization is automatically turned off in the presence of
8548 exception handling, for linkonce sections, for functions with a user-defined
8549 section attribute and on any architecture that does not support named
8552 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8554 @item -freorder-functions
8555 @opindex freorder-functions
8556 Reorder functions in the object file in order to
8557 improve code locality. This is implemented by using special
8558 subsections @code{.text.hot} for most frequently executed functions and
8559 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8560 the linker so object file format must support named sections and linker must
8561 place them in a reasonable way.
8563 Also profile feedback must be available to make this option effective. See
8564 @option{-fprofile-arcs} for details.
8566 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8568 @item -fstrict-aliasing
8569 @opindex fstrict-aliasing
8570 Allow the compiler to assume the strictest aliasing rules applicable to
8571 the language being compiled. For C (and C++), this activates
8572 optimizations based on the type of expressions. In particular, an
8573 object of one type is assumed never to reside at the same address as an
8574 object of a different type, unless the types are almost the same. For
8575 example, an @code{unsigned int} can alias an @code{int}, but not a
8576 @code{void*} or a @code{double}. A character type may alias any other
8579 @anchor{Type-punning}Pay special attention to code like this:
8592 The practice of reading from a different union member than the one most
8593 recently written to (called ``type-punning'') is common. Even with
8594 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8595 is accessed through the union type. So, the code above works as
8596 expected. @xref{Structures unions enumerations and bit-fields
8597 implementation}. However, this code might not:
8608 Similarly, access by taking the address, casting the resulting pointer
8609 and dereferencing the result has undefined behavior, even if the cast
8610 uses a union type, e.g.:
8614 return ((union a_union *) &d)->i;
8618 The @option{-fstrict-aliasing} option is enabled at levels
8619 @option{-O2}, @option{-O3}, @option{-Os}.
8621 @item -fstrict-overflow
8622 @opindex fstrict-overflow
8623 Allow the compiler to assume strict signed overflow rules, depending
8624 on the language being compiled. For C (and C++) this means that
8625 overflow when doing arithmetic with signed numbers is undefined, which
8626 means that the compiler may assume that it does not happen. This
8627 permits various optimizations. For example, the compiler assumes
8628 that an expression like @code{i + 10 > i} is always true for
8629 signed @code{i}. This assumption is only valid if signed overflow is
8630 undefined, as the expression is false if @code{i + 10} overflows when
8631 using twos complement arithmetic. When this option is in effect any
8632 attempt to determine whether an operation on signed numbers
8633 overflows must be written carefully to not actually involve overflow.
8635 This option also allows the compiler to assume strict pointer
8636 semantics: given a pointer to an object, if adding an offset to that
8637 pointer does not produce a pointer to the same object, the addition is
8638 undefined. This permits the compiler to conclude that @code{p + u >
8639 p} is always true for a pointer @code{p} and unsigned integer
8640 @code{u}. This assumption is only valid because pointer wraparound is
8641 undefined, as the expression is false if @code{p + u} overflows using
8642 twos complement arithmetic.
8644 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8645 that integer signed overflow is fully defined: it wraps. When
8646 @option{-fwrapv} is used, there is no difference between
8647 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8648 integers. With @option{-fwrapv} certain types of overflow are
8649 permitted. For example, if the compiler gets an overflow when doing
8650 arithmetic on constants, the overflowed value can still be used with
8651 @option{-fwrapv}, but not otherwise.
8653 The @option{-fstrict-overflow} option is enabled at levels
8654 @option{-O2}, @option{-O3}, @option{-Os}.
8656 @item -falign-functions
8657 @itemx -falign-functions=@var{n}
8658 @opindex falign-functions
8659 Align the start of functions to the next power-of-two greater than
8660 @var{n}, skipping up to @var{n} bytes. For instance,
8661 @option{-falign-functions=32} aligns functions to the next 32-byte
8662 boundary, but @option{-falign-functions=24} aligns to the next
8663 32-byte boundary only if this can be done by skipping 23 bytes or less.
8665 @option{-fno-align-functions} and @option{-falign-functions=1} are
8666 equivalent and mean that functions are not aligned.
8668 Some assemblers only support this flag when @var{n} is a power of two;
8669 in that case, it is rounded up.
8671 If @var{n} is not specified or is zero, use a machine-dependent default.
8673 Enabled at levels @option{-O2}, @option{-O3}.
8675 @item -flimit-function-alignment
8676 If this option is enabled, the compiler tries to avoid unnecessarily
8677 overaligning functions. It attempts to instruct the assembler to align
8678 by the amount specified by @option{-falign-functions}, but not to
8679 skip more bytes than the size of the function.
8681 @item -falign-labels
8682 @itemx -falign-labels=@var{n}
8683 @opindex falign-labels
8684 Align all branch targets to a power-of-two boundary, skipping up to
8685 @var{n} bytes like @option{-falign-functions}. This option can easily
8686 make code slower, because it must insert dummy operations for when the
8687 branch target is reached in the usual flow of the code.
8689 @option{-fno-align-labels} and @option{-falign-labels=1} are
8690 equivalent and mean that labels are not aligned.
8692 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8693 are greater than this value, then their values are used instead.
8695 If @var{n} is not specified or is zero, use a machine-dependent default
8696 which is very likely to be @samp{1}, meaning no alignment.
8698 Enabled at levels @option{-O2}, @option{-O3}.
8701 @itemx -falign-loops=@var{n}
8702 @opindex falign-loops
8703 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8704 like @option{-falign-functions}. If the loops are
8705 executed many times, this makes up for any execution of the dummy
8708 @option{-fno-align-loops} and @option{-falign-loops=1} are
8709 equivalent and mean that loops are not aligned.
8711 If @var{n} is not specified or is zero, use a machine-dependent default.
8713 Enabled at levels @option{-O2}, @option{-O3}.
8716 @itemx -falign-jumps=@var{n}
8717 @opindex falign-jumps
8718 Align branch targets to a power-of-two boundary, for branch targets
8719 where the targets can only be reached by jumping, skipping up to @var{n}
8720 bytes like @option{-falign-functions}. In this case, no dummy operations
8723 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8724 equivalent and mean that loops are not aligned.
8726 If @var{n} is not specified or is zero, use a machine-dependent default.
8728 Enabled at levels @option{-O2}, @option{-O3}.
8730 @item -funit-at-a-time
8731 @opindex funit-at-a-time
8732 This option is left for compatibility reasons. @option{-funit-at-a-time}
8733 has no effect, while @option{-fno-unit-at-a-time} implies
8734 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8738 @item -fno-toplevel-reorder
8739 @opindex fno-toplevel-reorder
8740 Do not reorder top-level functions, variables, and @code{asm}
8741 statements. Output them in the same order that they appear in the
8742 input file. When this option is used, unreferenced static variables
8743 are not removed. This option is intended to support existing code
8744 that relies on a particular ordering. For new code, it is better to
8745 use attributes when possible.
8747 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8748 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8753 Constructs webs as commonly used for register allocation purposes and assign
8754 each web individual pseudo register. This allows the register allocation pass
8755 to operate on pseudos directly, but also strengthens several other optimization
8756 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8757 however, make debugging impossible, since variables no longer stay in a
8760 Enabled by default with @option{-funroll-loops}.
8762 @item -fwhole-program
8763 @opindex fwhole-program
8764 Assume that the current compilation unit represents the whole program being
8765 compiled. All public functions and variables with the exception of @code{main}
8766 and those merged by attribute @code{externally_visible} become static functions
8767 and in effect are optimized more aggressively by interprocedural optimizers.
8769 This option should not be used in combination with @option{-flto}.
8770 Instead relying on a linker plugin should provide safer and more precise
8773 @item -flto[=@var{n}]
8775 This option runs the standard link-time optimizer. When invoked
8776 with source code, it generates GIMPLE (one of GCC's internal
8777 representations) and writes it to special ELF sections in the object
8778 file. When the object files are linked together, all the function
8779 bodies are read from these ELF sections and instantiated as if they
8780 had been part of the same translation unit.
8782 To use the link-time optimizer, @option{-flto} and optimization
8783 options should be specified at compile time and during the final link.
8784 It is recommended that you compile all the files participating in the
8785 same link with the same options and also specify those options at
8790 gcc -c -O2 -flto foo.c
8791 gcc -c -O2 -flto bar.c
8792 gcc -o myprog -flto -O2 foo.o bar.o
8795 The first two invocations to GCC save a bytecode representation
8796 of GIMPLE into special ELF sections inside @file{foo.o} and
8797 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8798 @file{foo.o} and @file{bar.o}, merges the two files into a single
8799 internal image, and compiles the result as usual. Since both
8800 @file{foo.o} and @file{bar.o} are merged into a single image, this
8801 causes all the interprocedural analyses and optimizations in GCC to
8802 work across the two files as if they were a single one. This means,
8803 for example, that the inliner is able to inline functions in
8804 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8806 Another (simpler) way to enable link-time optimization is:
8809 gcc -o myprog -flto -O2 foo.c bar.c
8812 The above generates bytecode for @file{foo.c} and @file{bar.c},
8813 merges them together into a single GIMPLE representation and optimizes
8814 them as usual to produce @file{myprog}.
8816 The only important thing to keep in mind is that to enable link-time
8817 optimizations you need to use the GCC driver to perform the link step.
8818 GCC then automatically performs link-time optimization if any of the
8819 objects involved were compiled with the @option{-flto} command-line option.
8821 should specify the optimization options to be used for link-time
8822 optimization though GCC tries to be clever at guessing an
8823 optimization level to use from the options used at compile time
8824 if you fail to specify one at link time. You can always override
8825 the automatic decision to do link-time optimization at link time
8826 by passing @option{-fno-lto} to the link command.
8828 To make whole program optimization effective, it is necessary to make
8829 certain whole program assumptions. The compiler needs to know
8830 what functions and variables can be accessed by libraries and runtime
8831 outside of the link-time optimized unit. When supported by the linker,
8832 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8833 to the compiler about used and externally visible symbols. When
8834 the linker plugin is not available, @option{-fwhole-program} should be
8835 used to allow the compiler to make these assumptions, which leads
8836 to more aggressive optimization decisions.
8838 When @option{-fuse-linker-plugin} is not enabled, when a file is
8839 compiled with @option{-flto}, the generated object file is larger than
8840 a regular object file because it contains GIMPLE bytecodes and the usual
8841 final code (see @option{-ffat-lto-objects}. This means that
8842 object files with LTO information can be linked as normal object
8843 files; if @option{-fno-lto} is passed to the linker, no
8844 interprocedural optimizations are applied. Note that when
8845 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8846 but you cannot perform a regular, non-LTO link on them.
8848 Additionally, the optimization flags used to compile individual files
8849 are not necessarily related to those used at link time. For instance,
8852 gcc -c -O0 -ffat-lto-objects -flto foo.c
8853 gcc -c -O0 -ffat-lto-objects -flto bar.c
8854 gcc -o myprog -O3 foo.o bar.o
8857 This produces individual object files with unoptimized assembler
8858 code, but the resulting binary @file{myprog} is optimized at
8859 @option{-O3}. If, instead, the final binary is generated with
8860 @option{-fno-lto}, then @file{myprog} is not optimized.
8862 When producing the final binary, GCC only
8863 applies link-time optimizations to those files that contain bytecode.
8864 Therefore, you can mix and match object files and libraries with
8865 GIMPLE bytecodes and final object code. GCC automatically selects
8866 which files to optimize in LTO mode and which files to link without
8869 There are some code generation flags preserved by GCC when
8870 generating bytecodes, as they need to be used during the final link
8871 stage. Generally options specified at link time override those
8872 specified at compile time.
8874 If you do not specify an optimization level option @option{-O} at
8875 link time, then GCC uses the highest optimization level
8876 used when compiling the object files.
8878 Currently, the following options and their settings are taken from
8879 the first object file that explicitly specifies them:
8880 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8881 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8882 and all the @option{-m} target flags.
8884 Certain ABI-changing flags are required to match in all compilation units,
8885 and trying to override this at link time with a conflicting value
8886 is ignored. This includes options such as @option{-freg-struct-return}
8887 and @option{-fpcc-struct-return}.
8889 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8890 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8891 are passed through to the link stage and merged conservatively for
8892 conflicting translation units. Specifically
8893 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8894 precedence; and for example @option{-ffp-contract=off} takes precedence
8895 over @option{-ffp-contract=fast}. You can override them at link time.
8897 If LTO encounters objects with C linkage declared with incompatible
8898 types in separate translation units to be linked together (undefined
8899 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8900 issued. The behavior is still undefined at run time. Similar
8901 diagnostics may be raised for other languages.
8903 Another feature of LTO is that it is possible to apply interprocedural
8904 optimizations on files written in different languages:
8909 gfortran -c -flto baz.f90
8910 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8913 Notice that the final link is done with @command{g++} to get the C++
8914 runtime libraries and @option{-lgfortran} is added to get the Fortran
8915 runtime libraries. In general, when mixing languages in LTO mode, you
8916 should use the same link command options as when mixing languages in a
8917 regular (non-LTO) compilation.
8919 If object files containing GIMPLE bytecode are stored in a library archive, say
8920 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8921 are using a linker with plugin support. To create static libraries suitable
8922 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8923 and @command{ranlib};
8924 to show the symbols of object files with GIMPLE bytecode, use
8925 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8926 and @command{nm} have been compiled with plugin support. At link time, use the the
8927 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8928 the LTO optimization process:
8931 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8934 With the linker plugin enabled, the linker extracts the needed
8935 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8936 to make them part of the aggregated GIMPLE image to be optimized.
8938 If you are not using a linker with plugin support and/or do not
8939 enable the linker plugin, then the objects inside @file{libfoo.a}
8940 are extracted and linked as usual, but they do not participate
8941 in the LTO optimization process. In order to make a static library suitable
8942 for both LTO optimization and usual linkage, compile its object files with
8943 @option{-flto} @option{-ffat-lto-objects}.
8945 Link-time optimizations do not require the presence of the whole program to
8946 operate. If the program does not require any symbols to be exported, it is
8947 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8948 the interprocedural optimizers to use more aggressive assumptions which may
8949 lead to improved optimization opportunities.
8950 Use of @option{-fwhole-program} is not needed when linker plugin is
8951 active (see @option{-fuse-linker-plugin}).
8953 The current implementation of LTO makes no
8954 attempt to generate bytecode that is portable between different
8955 types of hosts. The bytecode files are versioned and there is a
8956 strict version check, so bytecode files generated in one version of
8957 GCC do not work with an older or newer version of GCC.
8959 Link-time optimization does not work well with generation of debugging
8960 information. Combining @option{-flto} with
8961 @option{-g} is currently experimental and expected to produce unexpected
8964 If you specify the optional @var{n}, the optimization and code
8965 generation done at link time is executed in parallel using @var{n}
8966 parallel jobs by utilizing an installed @command{make} program. The
8967 environment variable @env{MAKE} may be used to override the program
8968 used. The default value for @var{n} is 1.
8970 You can also specify @option{-flto=jobserver} to use GNU make's
8971 job server mode to determine the number of parallel jobs. This
8972 is useful when the Makefile calling GCC is already executing in parallel.
8973 You must prepend a @samp{+} to the command recipe in the parent Makefile
8974 for this to work. This option likely only works if @env{MAKE} is
8977 @item -flto-partition=@var{alg}
8978 @opindex flto-partition
8979 Specify the partitioning algorithm used by the link-time optimizer.
8980 The value is either @samp{1to1} to specify a partitioning mirroring
8981 the original source files or @samp{balanced} to specify partitioning
8982 into equally sized chunks (whenever possible) or @samp{max} to create
8983 new partition for every symbol where possible. Specifying @samp{none}
8984 as an algorithm disables partitioning and streaming completely.
8985 The default value is @samp{balanced}. While @samp{1to1} can be used
8986 as an workaround for various code ordering issues, the @samp{max}
8987 partitioning is intended for internal testing only.
8988 The value @samp{one} specifies that exactly one partition should be
8989 used while the value @samp{none} bypasses partitioning and executes
8990 the link-time optimization step directly from the WPA phase.
8992 @item -flto-odr-type-merging
8993 @opindex flto-odr-type-merging
8994 Enable streaming of mangled types names of C++ types and their unification
8995 at link time. This increases size of LTO object files, but enables
8996 diagnostics about One Definition Rule violations.
8998 @item -flto-compression-level=@var{n}
8999 @opindex flto-compression-level
9000 This option specifies the level of compression used for intermediate
9001 language written to LTO object files, and is only meaningful in
9002 conjunction with LTO mode (@option{-flto}). Valid
9003 values are 0 (no compression) to 9 (maximum compression). Values
9004 outside this range are clamped to either 0 or 9. If the option is not
9005 given, a default balanced compression setting is used.
9007 @item -fuse-linker-plugin
9008 @opindex fuse-linker-plugin
9009 Enables the use of a linker plugin during link-time optimization. This
9010 option relies on plugin support in the linker, which is available in gold
9011 or in GNU ld 2.21 or newer.
9013 This option enables the extraction of object files with GIMPLE bytecode out
9014 of library archives. This improves the quality of optimization by exposing
9015 more code to the link-time optimizer. This information specifies what
9016 symbols can be accessed externally (by non-LTO object or during dynamic
9017 linking). Resulting code quality improvements on binaries (and shared
9018 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9019 See @option{-flto} for a description of the effect of this flag and how to
9022 This option is enabled by default when LTO support in GCC is enabled
9023 and GCC was configured for use with
9024 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9026 @item -ffat-lto-objects
9027 @opindex ffat-lto-objects
9028 Fat LTO objects are object files that contain both the intermediate language
9029 and the object code. This makes them usable for both LTO linking and normal
9030 linking. This option is effective only when compiling with @option{-flto}
9031 and is ignored at link time.
9033 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9034 requires the complete toolchain to be aware of LTO. It requires a linker with
9035 linker plugin support for basic functionality. Additionally,
9036 @command{nm}, @command{ar} and @command{ranlib}
9037 need to support linker plugins to allow a full-featured build environment
9038 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9039 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9040 to these tools. With non fat LTO makefiles need to be modified to use them.
9042 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9045 @item -fcompare-elim
9046 @opindex fcompare-elim
9047 After register allocation and post-register allocation instruction splitting,
9048 identify arithmetic instructions that compute processor flags similar to a
9049 comparison operation based on that arithmetic. If possible, eliminate the
9050 explicit comparison operation.
9052 This pass only applies to certain targets that cannot explicitly represent
9053 the comparison operation before register allocation is complete.
9055 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9057 @item -fcprop-registers
9058 @opindex fcprop-registers
9059 After register allocation and post-register allocation instruction splitting,
9060 perform a copy-propagation pass to try to reduce scheduling dependencies
9061 and occasionally eliminate the copy.
9063 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9065 @item -fprofile-correction
9066 @opindex fprofile-correction
9067 Profiles collected using an instrumented binary for multi-threaded programs may
9068 be inconsistent due to missed counter updates. When this option is specified,
9069 GCC uses heuristics to correct or smooth out such inconsistencies. By
9070 default, GCC emits an error message when an inconsistent profile is detected.
9073 @itemx -fprofile-use=@var{path}
9074 @opindex fprofile-use
9075 Enable profile feedback-directed optimizations,
9076 and the following optimizations
9077 which are generally profitable only with profile feedback available:
9078 @option{-fbranch-probabilities}, @option{-fvpt},
9079 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9080 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9082 Before you can use this option, you must first generate profiling information.
9083 @xref{Instrumentation Options}, for information about the
9084 @option{-fprofile-generate} option.
9086 By default, GCC emits an error message if the feedback profiles do not
9087 match the source code. This error can be turned into a warning by using
9088 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9091 If @var{path} is specified, GCC looks at the @var{path} to find
9092 the profile feedback data files. See @option{-fprofile-dir}.
9094 @item -fauto-profile
9095 @itemx -fauto-profile=@var{path}
9096 @opindex fauto-profile
9097 Enable sampling-based feedback-directed optimizations,
9098 and the following optimizations
9099 which are generally profitable only with profile feedback available:
9100 @option{-fbranch-probabilities}, @option{-fvpt},
9101 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9102 @option{-ftree-vectorize},
9103 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9104 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9105 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9107 @var{path} is the name of a file containing AutoFDO profile information.
9108 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9110 Producing an AutoFDO profile data file requires running your program
9111 with the @command{perf} utility on a supported GNU/Linux target system.
9112 For more information, see @uref{https://perf.wiki.kernel.org/}.
9116 perf record -e br_inst_retired:near_taken -b -o perf.data \
9120 Then use the @command{create_gcov} tool to convert the raw profile data
9121 to a format that can be used by GCC.@ You must also supply the
9122 unstripped binary for your program to this tool.
9123 See @uref{https://github.com/google/autofdo}.
9127 create_gcov --binary=your_program.unstripped --profile=perf.data \
9132 The following options control compiler behavior regarding floating-point
9133 arithmetic. These options trade off between speed and
9134 correctness. All must be specifically enabled.
9138 @opindex ffloat-store
9139 Do not store floating-point variables in registers, and inhibit other
9140 options that might change whether a floating-point value is taken from a
9143 @cindex floating-point precision
9144 This option prevents undesirable excess precision on machines such as
9145 the 68000 where the floating registers (of the 68881) keep more
9146 precision than a @code{double} is supposed to have. Similarly for the
9147 x86 architecture. For most programs, the excess precision does only
9148 good, but a few programs rely on the precise definition of IEEE floating
9149 point. Use @option{-ffloat-store} for such programs, after modifying
9150 them to store all pertinent intermediate computations into variables.
9152 @item -fexcess-precision=@var{style}
9153 @opindex fexcess-precision
9154 This option allows further control over excess precision on machines
9155 where floating-point operations occur in a format with more precision or
9156 range than the IEEE standard and interchange floating-point types. By
9157 default, @option{-fexcess-precision=fast} is in effect; this means that
9158 operations may be carried out in a wider precision than the types specified
9159 in the source if that would result in faster code, and it is unpredictable
9160 when rounding to the types specified in the source code takes place.
9161 When compiling C, if @option{-fexcess-precision=standard} is specified then
9162 excess precision follows the rules specified in ISO C99; in particular,
9163 both casts and assignments cause values to be rounded to their
9164 semantic types (whereas @option{-ffloat-store} only affects
9165 assignments). This option is enabled by default for C if a strict
9166 conformance option such as @option{-std=c99} is used.
9167 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
9168 regardless of whether a strict conformance option is used.
9171 @option{-fexcess-precision=standard} is not implemented for languages
9172 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
9173 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9174 semantics apply without excess precision, and in the latter, rounding
9179 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9180 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9181 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
9182 @option{-fexcess-precision=fast}.
9184 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9186 This option is not turned on by any @option{-O} option besides
9187 @option{-Ofast} since it can result in incorrect output for programs
9188 that depend on an exact implementation of IEEE or ISO rules/specifications
9189 for math functions. It may, however, yield faster code for programs
9190 that do not require the guarantees of these specifications.
9192 @item -fno-math-errno
9193 @opindex fno-math-errno
9194 Do not set @code{errno} after calling math functions that are executed
9195 with a single instruction, e.g., @code{sqrt}. A program that relies on
9196 IEEE exceptions for math error handling may want to use this flag
9197 for speed while maintaining IEEE arithmetic compatibility.
9199 This option is not turned on by any @option{-O} option since
9200 it can result in incorrect output for programs that depend on
9201 an exact implementation of IEEE or ISO rules/specifications for
9202 math functions. It may, however, yield faster code for programs
9203 that do not require the guarantees of these specifications.
9205 The default is @option{-fmath-errno}.
9207 On Darwin systems, the math library never sets @code{errno}. There is
9208 therefore no reason for the compiler to consider the possibility that
9209 it might, and @option{-fno-math-errno} is the default.
9211 @item -funsafe-math-optimizations
9212 @opindex funsafe-math-optimizations
9214 Allow optimizations for floating-point arithmetic that (a) assume
9215 that arguments and results are valid and (b) may violate IEEE or
9216 ANSI standards. When used at link time, it may include libraries
9217 or startup files that change the default FPU control word or other
9218 similar optimizations.
9220 This option is not turned on by any @option{-O} option since
9221 it can result in incorrect output for programs that depend on
9222 an exact implementation of IEEE or ISO rules/specifications for
9223 math functions. It may, however, yield faster code for programs
9224 that do not require the guarantees of these specifications.
9225 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9226 @option{-fassociative-math} and @option{-freciprocal-math}.
9228 The default is @option{-fno-unsafe-math-optimizations}.
9230 @item -fassociative-math
9231 @opindex fassociative-math
9233 Allow re-association of operands in series of floating-point operations.
9234 This violates the ISO C and C++ language standard by possibly changing
9235 computation result. NOTE: re-ordering may change the sign of zero as
9236 well as ignore NaNs and inhibit or create underflow or overflow (and
9237 thus cannot be used on code that relies on rounding behavior like
9238 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9239 and thus may not be used when ordered comparisons are required.
9240 This option requires that both @option{-fno-signed-zeros} and
9241 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9242 much sense with @option{-frounding-math}. For Fortran the option
9243 is automatically enabled when both @option{-fno-signed-zeros} and
9244 @option{-fno-trapping-math} are in effect.
9246 The default is @option{-fno-associative-math}.
9248 @item -freciprocal-math
9249 @opindex freciprocal-math
9251 Allow the reciprocal of a value to be used instead of dividing by
9252 the value if this enables optimizations. For example @code{x / y}
9253 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9254 is subject to common subexpression elimination. Note that this loses
9255 precision and increases the number of flops operating on the value.
9257 The default is @option{-fno-reciprocal-math}.
9259 @item -ffinite-math-only
9260 @opindex ffinite-math-only
9261 Allow optimizations for floating-point arithmetic that assume
9262 that arguments and results are not NaNs or +-Infs.
9264 This option is not turned on by any @option{-O} option since
9265 it can result in incorrect output for programs that depend on
9266 an exact implementation of IEEE or ISO rules/specifications for
9267 math functions. It may, however, yield faster code for programs
9268 that do not require the guarantees of these specifications.
9270 The default is @option{-fno-finite-math-only}.
9272 @item -fno-signed-zeros
9273 @opindex fno-signed-zeros
9274 Allow optimizations for floating-point arithmetic that ignore the
9275 signedness of zero. IEEE arithmetic specifies the behavior of
9276 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9277 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9278 This option implies that the sign of a zero result isn't significant.
9280 The default is @option{-fsigned-zeros}.
9282 @item -fno-trapping-math
9283 @opindex fno-trapping-math
9284 Compile code assuming that floating-point operations cannot generate
9285 user-visible traps. These traps include division by zero, overflow,
9286 underflow, inexact result and invalid operation. This option requires
9287 that @option{-fno-signaling-nans} be in effect. Setting this option may
9288 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9290 This option should never be turned on by any @option{-O} option since
9291 it can result in incorrect output for programs that depend on
9292 an exact implementation of IEEE or ISO rules/specifications for
9295 The default is @option{-ftrapping-math}.
9297 @item -frounding-math
9298 @opindex frounding-math
9299 Disable transformations and optimizations that assume default floating-point
9300 rounding behavior. This is round-to-zero for all floating point
9301 to integer conversions, and round-to-nearest for all other arithmetic
9302 truncations. This option should be specified for programs that change
9303 the FP rounding mode dynamically, or that may be executed with a
9304 non-default rounding mode. This option disables constant folding of
9305 floating-point expressions at compile time (which may be affected by
9306 rounding mode) and arithmetic transformations that are unsafe in the
9307 presence of sign-dependent rounding modes.
9309 The default is @option{-fno-rounding-math}.
9311 This option is experimental and does not currently guarantee to
9312 disable all GCC optimizations that are affected by rounding mode.
9313 Future versions of GCC may provide finer control of this setting
9314 using C99's @code{FENV_ACCESS} pragma. This command-line option
9315 will be used to specify the default state for @code{FENV_ACCESS}.
9317 @item -fsignaling-nans
9318 @opindex fsignaling-nans
9319 Compile code assuming that IEEE signaling NaNs may generate user-visible
9320 traps during floating-point operations. Setting this option disables
9321 optimizations that may change the number of exceptions visible with
9322 signaling NaNs. This option implies @option{-ftrapping-math}.
9324 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9327 The default is @option{-fno-signaling-nans}.
9329 This option is experimental and does not currently guarantee to
9330 disable all GCC optimizations that affect signaling NaN behavior.
9332 @item -fno-fp-int-builtin-inexact
9333 @opindex fno-fp-int-builtin-inexact
9334 Do not allow the built-in functions @code{ceil}, @code{floor},
9335 @code{round} and @code{trunc}, and their @code{float} and @code{long
9336 double} variants, to generate code that raises the ``inexact''
9337 floating-point exception for noninteger arguments. ISO C99 and C11
9338 allow these functions to raise the ``inexact'' exception, but ISO/IEC
9339 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
9342 The default is @option{-ffp-int-builtin-inexact}, allowing the
9343 exception to be raised. This option does nothing unless
9344 @option{-ftrapping-math} is in effect.
9346 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
9347 generate a call to a library function then the ``inexact'' exception
9348 may be raised if the library implementation does not follow TS 18661.
9350 @item -fsingle-precision-constant
9351 @opindex fsingle-precision-constant
9352 Treat floating-point constants as single precision instead of
9353 implicitly converting them to double-precision constants.
9355 @item -fcx-limited-range
9356 @opindex fcx-limited-range
9357 When enabled, this option states that a range reduction step is not
9358 needed when performing complex division. Also, there is no checking
9359 whether the result of a complex multiplication or division is @code{NaN
9360 + I*NaN}, with an attempt to rescue the situation in that case. The
9361 default is @option{-fno-cx-limited-range}, but is enabled by
9362 @option{-ffast-math}.
9364 This option controls the default setting of the ISO C99
9365 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9368 @item -fcx-fortran-rules
9369 @opindex fcx-fortran-rules
9370 Complex multiplication and division follow Fortran rules. Range
9371 reduction is done as part of complex division, but there is no checking
9372 whether the result of a complex multiplication or division is @code{NaN
9373 + I*NaN}, with an attempt to rescue the situation in that case.
9375 The default is @option{-fno-cx-fortran-rules}.
9379 The following options control optimizations that may improve
9380 performance, but are not enabled by any @option{-O} options. This
9381 section includes experimental options that may produce broken code.
9384 @item -fbranch-probabilities
9385 @opindex fbranch-probabilities
9386 After running a program compiled with @option{-fprofile-arcs}
9387 (@pxref{Instrumentation Options}),
9388 you can compile it a second time using
9389 @option{-fbranch-probabilities}, to improve optimizations based on
9390 the number of times each branch was taken. When a program
9391 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9392 counts to a file called @file{@var{sourcename}.gcda} for each source
9393 file. The information in this data file is very dependent on the
9394 structure of the generated code, so you must use the same source code
9395 and the same optimization options for both compilations.
9397 With @option{-fbranch-probabilities}, GCC puts a
9398 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9399 These can be used to improve optimization. Currently, they are only
9400 used in one place: in @file{reorg.c}, instead of guessing which path a
9401 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9402 exactly determine which path is taken more often.
9404 @item -fprofile-values
9405 @opindex fprofile-values
9406 If combined with @option{-fprofile-arcs}, it adds code so that some
9407 data about values of expressions in the program is gathered.
9409 With @option{-fbranch-probabilities}, it reads back the data gathered
9410 from profiling values of expressions for usage in optimizations.
9412 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9414 @item -fprofile-reorder-functions
9415 @opindex fprofile-reorder-functions
9416 Function reordering based on profile instrumentation collects
9417 first time of execution of a function and orders these functions
9420 Enabled with @option{-fprofile-use}.
9424 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9425 to add code to gather information about values of expressions.
9427 With @option{-fbranch-probabilities}, it reads back the data gathered
9428 and actually performs the optimizations based on them.
9429 Currently the optimizations include specialization of division operations
9430 using the knowledge about the value of the denominator.
9432 @item -frename-registers
9433 @opindex frename-registers
9434 Attempt to avoid false dependencies in scheduled code by making use
9435 of registers left over after register allocation. This optimization
9436 most benefits processors with lots of registers. Depending on the
9437 debug information format adopted by the target, however, it can
9438 make debugging impossible, since variables no longer stay in
9439 a ``home register''.
9441 Enabled by default with @option{-funroll-loops}.
9443 @item -fschedule-fusion
9444 @opindex fschedule-fusion
9445 Performs a target dependent pass over the instruction stream to schedule
9446 instructions of same type together because target machine can execute them
9447 more efficiently if they are adjacent to each other in the instruction flow.
9449 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9453 Perform tail duplication to enlarge superblock size. This transformation
9454 simplifies the control flow of the function allowing other optimizations to do
9457 Enabled with @option{-fprofile-use}.
9459 @item -funroll-loops
9460 @opindex funroll-loops
9461 Unroll loops whose number of iterations can be determined at compile time or
9462 upon entry to the loop. @option{-funroll-loops} implies
9463 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9464 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9465 a small constant number of iterations). This option makes code larger, and may
9466 or may not make it run faster.
9468 Enabled with @option{-fprofile-use}.
9470 @item -funroll-all-loops
9471 @opindex funroll-all-loops
9472 Unroll all loops, even if their number of iterations is uncertain when
9473 the loop is entered. This usually makes programs run more slowly.
9474 @option{-funroll-all-loops} implies the same options as
9475 @option{-funroll-loops}.
9478 @opindex fpeel-loops
9479 Peels loops for which there is enough information that they do not
9480 roll much (from profile feedback or static analysis). It also turns on
9481 complete loop peeling (i.e.@: complete removal of loops with small constant
9482 number of iterations).
9484 Enabled with @option{-O3} and/or @option{-fprofile-use}.
9486 @item -fmove-loop-invariants
9487 @opindex fmove-loop-invariants
9488 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9489 at level @option{-O1}
9492 @opindex fsplit-loops
9493 Split a loop into two if it contains a condition that's always true
9494 for one side of the iteration space and false for the other.
9496 @item -funswitch-loops
9497 @opindex funswitch-loops
9498 Move branches with loop invariant conditions out of the loop, with duplicates
9499 of the loop on both branches (modified according to result of the condition).
9501 @item -ffunction-sections
9502 @itemx -fdata-sections
9503 @opindex ffunction-sections
9504 @opindex fdata-sections
9505 Place each function or data item into its own section in the output
9506 file if the target supports arbitrary sections. The name of the
9507 function or the name of the data item determines the section's name
9510 Use these options on systems where the linker can perform optimizations
9511 to improve locality of reference in the instruction space. Most systems
9512 using the ELF object format and SPARC processors running Solaris 2 have
9513 linkers with such optimizations. AIX may have these optimizations in
9516 Only use these options when there are significant benefits from doing
9517 so. When you specify these options, the assembler and linker
9518 create larger object and executable files and are also slower.
9519 You cannot use @command{gprof} on all systems if you
9520 specify this option, and you may have problems with debugging if
9521 you specify both this option and @option{-g}.
9523 @item -fbranch-target-load-optimize
9524 @opindex fbranch-target-load-optimize
9525 Perform branch target register load optimization before prologue / epilogue
9527 The use of target registers can typically be exposed only during reload,
9528 thus hoisting loads out of loops and doing inter-block scheduling needs
9529 a separate optimization pass.
9531 @item -fbranch-target-load-optimize2
9532 @opindex fbranch-target-load-optimize2
9533 Perform branch target register load optimization after prologue / epilogue
9536 @item -fbtr-bb-exclusive
9537 @opindex fbtr-bb-exclusive
9538 When performing branch target register load optimization, don't reuse
9539 branch target registers within any basic block.
9542 @opindex fstdarg-opt
9543 Optimize the prologue of variadic argument functions with respect to usage of
9546 @item -fsection-anchors
9547 @opindex fsection-anchors
9548 Try to reduce the number of symbolic address calculations by using
9549 shared ``anchor'' symbols to address nearby objects. This transformation
9550 can help to reduce the number of GOT entries and GOT accesses on some
9553 For example, the implementation of the following function @code{foo}:
9557 int foo (void) @{ return a + b + c; @}
9561 usually calculates the addresses of all three variables, but if you
9562 compile it with @option{-fsection-anchors}, it accesses the variables
9563 from a common anchor point instead. The effect is similar to the
9564 following pseudocode (which isn't valid C):
9569 register int *xr = &x;
9570 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9574 Not all targets support this option.
9576 @item --param @var{name}=@var{value}
9578 In some places, GCC uses various constants to control the amount of
9579 optimization that is done. For example, GCC does not inline functions
9580 that contain more than a certain number of instructions. You can
9581 control some of these constants on the command line using the
9582 @option{--param} option.
9584 The names of specific parameters, and the meaning of the values, are
9585 tied to the internals of the compiler, and are subject to change
9586 without notice in future releases.
9588 In each case, the @var{value} is an integer. The allowable choices for
9592 @item predictable-branch-outcome
9593 When branch is predicted to be taken with probability lower than this threshold
9594 (in percent), then it is considered well predictable. The default is 10.
9596 @item max-rtl-if-conversion-insns
9597 RTL if-conversion tries to remove conditional branches around a block and
9598 replace them with conditionally executed instructions. This parameter
9599 gives the maximum number of instructions in a block which should be
9600 considered for if-conversion. The default is 10, though the compiler will
9601 also use other heuristics to decide whether if-conversion is likely to be
9604 @item max-rtl-if-conversion-predictable-cost
9605 @item max-rtl-if-conversion-unpredictable-cost
9606 RTL if-conversion will try to remove conditional branches around a block
9607 and replace them with conditionally executed instructions. These parameters
9608 give the maximum permissible cost for the sequence that would be generated
9609 by if-conversion depending on whether the branch is statically determined
9610 to be predictable or not. The units for this parameter are the same as
9611 those for the GCC internal seq_cost metric. The compiler will try to
9612 provide a reasonable default for this parameter using the BRANCH_COST
9615 @item max-crossjump-edges
9616 The maximum number of incoming edges to consider for cross-jumping.
9617 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9618 the number of edges incoming to each block. Increasing values mean
9619 more aggressive optimization, making the compilation time increase with
9620 probably small improvement in executable size.
9622 @item min-crossjump-insns
9623 The minimum number of instructions that must be matched at the end
9624 of two blocks before cross-jumping is performed on them. This
9625 value is ignored in the case where all instructions in the block being
9626 cross-jumped from are matched. The default value is 5.
9628 @item max-grow-copy-bb-insns
9629 The maximum code size expansion factor when copying basic blocks
9630 instead of jumping. The expansion is relative to a jump instruction.
9631 The default value is 8.
9633 @item max-goto-duplication-insns
9634 The maximum number of instructions to duplicate to a block that jumps
9635 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9636 passes, GCC factors computed gotos early in the compilation process,
9637 and unfactors them as late as possible. Only computed jumps at the
9638 end of a basic blocks with no more than max-goto-duplication-insns are
9639 unfactored. The default value is 8.
9641 @item max-delay-slot-insn-search
9642 The maximum number of instructions to consider when looking for an
9643 instruction to fill a delay slot. If more than this arbitrary number of
9644 instructions are searched, the time savings from filling the delay slot
9645 are minimal, so stop searching. Increasing values mean more
9646 aggressive optimization, making the compilation time increase with probably
9647 small improvement in execution time.
9649 @item max-delay-slot-live-search
9650 When trying to fill delay slots, the maximum number of instructions to
9651 consider when searching for a block with valid live register
9652 information. Increasing this arbitrarily chosen value means more
9653 aggressive optimization, increasing the compilation time. This parameter
9654 should be removed when the delay slot code is rewritten to maintain the
9657 @item max-gcse-memory
9658 The approximate maximum amount of memory that can be allocated in
9659 order to perform the global common subexpression elimination
9660 optimization. If more memory than specified is required, the
9661 optimization is not done.
9663 @item max-gcse-insertion-ratio
9664 If the ratio of expression insertions to deletions is larger than this value
9665 for any expression, then RTL PRE inserts or removes the expression and thus
9666 leaves partially redundant computations in the instruction stream. The default value is 20.
9668 @item max-pending-list-length
9669 The maximum number of pending dependencies scheduling allows
9670 before flushing the current state and starting over. Large functions
9671 with few branches or calls can create excessively large lists which
9672 needlessly consume memory and resources.
9674 @item max-modulo-backtrack-attempts
9675 The maximum number of backtrack attempts the scheduler should make
9676 when modulo scheduling a loop. Larger values can exponentially increase
9679 @item max-inline-insns-single
9680 Several parameters control the tree inliner used in GCC@.
9681 This number sets the maximum number of instructions (counted in GCC's
9682 internal representation) in a single function that the tree inliner
9683 considers for inlining. This only affects functions declared
9684 inline and methods implemented in a class declaration (C++).
9685 The default value is 400.
9687 @item max-inline-insns-auto
9688 When you use @option{-finline-functions} (included in @option{-O3}),
9689 a lot of functions that would otherwise not be considered for inlining
9690 by the compiler are investigated. To those functions, a different
9691 (more restrictive) limit compared to functions declared inline can
9693 The default value is 40.
9695 @item inline-min-speedup
9696 When estimated performance improvement of caller + callee runtime exceeds this
9697 threshold (in percent), the function can be inlined regardless of the limit on
9698 @option{--param max-inline-insns-single} and @option{--param
9699 max-inline-insns-auto}.
9701 @item large-function-insns
9702 The limit specifying really large functions. For functions larger than this
9703 limit after inlining, inlining is constrained by
9704 @option{--param large-function-growth}. This parameter is useful primarily
9705 to avoid extreme compilation time caused by non-linear algorithms used by the
9707 The default value is 2700.
9709 @item large-function-growth
9710 Specifies maximal growth of large function caused by inlining in percents.
9711 The default value is 100 which limits large function growth to 2.0 times
9714 @item large-unit-insns
9715 The limit specifying large translation unit. Growth caused by inlining of
9716 units larger than this limit is limited by @option{--param inline-unit-growth}.
9717 For small units this might be too tight.
9718 For example, consider a unit consisting of function A
9719 that is inline and B that just calls A three times. If B is small relative to
9720 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9721 large units consisting of small inlineable functions, however, the overall unit
9722 growth limit is needed to avoid exponential explosion of code size. Thus for
9723 smaller units, the size is increased to @option{--param large-unit-insns}
9724 before applying @option{--param inline-unit-growth}. The default is 10000.
9726 @item inline-unit-growth
9727 Specifies maximal overall growth of the compilation unit caused by inlining.
9728 The default value is 20 which limits unit growth to 1.2 times the original
9729 size. Cold functions (either marked cold via an attribute or by profile
9730 feedback) are not accounted into the unit size.
9732 @item ipcp-unit-growth
9733 Specifies maximal overall growth of the compilation unit caused by
9734 interprocedural constant propagation. The default value is 10 which limits
9735 unit growth to 1.1 times the original size.
9737 @item large-stack-frame
9738 The limit specifying large stack frames. While inlining the algorithm is trying
9739 to not grow past this limit too much. The default value is 256 bytes.
9741 @item large-stack-frame-growth
9742 Specifies maximal growth of large stack frames caused by inlining in percents.
9743 The default value is 1000 which limits large stack frame growth to 11 times
9746 @item max-inline-insns-recursive
9747 @itemx max-inline-insns-recursive-auto
9748 Specifies the maximum number of instructions an out-of-line copy of a
9749 self-recursive inline
9750 function can grow into by performing recursive inlining.
9752 @option{--param max-inline-insns-recursive} applies to functions
9754 For functions not declared inline, recursive inlining
9755 happens only when @option{-finline-functions} (included in @option{-O3}) is
9756 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
9757 default value is 450.
9759 @item max-inline-recursive-depth
9760 @itemx max-inline-recursive-depth-auto
9761 Specifies the maximum recursion depth used for recursive inlining.
9763 @option{--param max-inline-recursive-depth} applies to functions
9764 declared inline. For functions not declared inline, recursive inlining
9765 happens only when @option{-finline-functions} (included in @option{-O3}) is
9766 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9769 @item min-inline-recursive-probability
9770 Recursive inlining is profitable only for function having deep recursion
9771 in average and can hurt for function having little recursion depth by
9772 increasing the prologue size or complexity of function body to other
9775 When profile feedback is available (see @option{-fprofile-generate}) the actual
9776 recursion depth can be guessed from the probability that function recurses
9777 via a given call expression. This parameter limits inlining only to call
9778 expressions whose probability exceeds the given threshold (in percents).
9779 The default value is 10.
9781 @item early-inlining-insns
9782 Specify growth that the early inliner can make. In effect it increases
9783 the amount of inlining for code having a large abstraction penalty.
9784 The default value is 14.
9786 @item max-early-inliner-iterations
9787 Limit of iterations of the early inliner. This basically bounds
9788 the number of nested indirect calls the early inliner can resolve.
9789 Deeper chains are still handled by late inlining.
9791 @item comdat-sharing-probability
9792 Probability (in percent) that C++ inline function with comdat visibility
9793 are shared across multiple compilation units. The default value is 20.
9795 @item profile-func-internal-id
9796 A parameter to control whether to use function internal id in profile
9797 database lookup. If the value is 0, the compiler uses an id that
9798 is based on function assembler name and filename, which makes old profile
9799 data more tolerant to source changes such as function reordering etc.
9800 The default value is 0.
9802 @item min-vect-loop-bound
9803 The minimum number of iterations under which loops are not vectorized
9804 when @option{-ftree-vectorize} is used. The number of iterations after
9805 vectorization needs to be greater than the value specified by this option
9806 to allow vectorization. The default value is 0.
9808 @item gcse-cost-distance-ratio
9809 Scaling factor in calculation of maximum distance an expression
9810 can be moved by GCSE optimizations. This is currently supported only in the
9811 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9812 is with simple expressions, i.e., the expressions that have cost
9813 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9814 hoisting of simple expressions. The default value is 10.
9816 @item gcse-unrestricted-cost
9817 Cost, roughly measured as the cost of a single typical machine
9818 instruction, at which GCSE optimizations do not constrain
9819 the distance an expression can travel. This is currently
9820 supported only in the code hoisting pass. The lesser the cost,
9821 the more aggressive code hoisting is. Specifying 0
9822 allows all expressions to travel unrestricted distances.
9823 The default value is 3.
9825 @item max-hoist-depth
9826 The depth of search in the dominator tree for expressions to hoist.
9827 This is used to avoid quadratic behavior in hoisting algorithm.
9828 The value of 0 does not limit on the search, but may slow down compilation
9829 of huge functions. The default value is 30.
9831 @item max-tail-merge-comparisons
9832 The maximum amount of similar bbs to compare a bb with. This is used to
9833 avoid quadratic behavior in tree tail merging. The default value is 10.
9835 @item max-tail-merge-iterations
9836 The maximum amount of iterations of the pass over the function. This is used to
9837 limit compilation time in tree tail merging. The default value is 2.
9839 @item store-merging-allow-unaligned
9840 Allow the store merging pass to introduce unaligned stores if it is legal to
9841 do so. The default value is 1.
9843 @item max-stores-to-merge
9844 The maximum number of stores to attempt to merge into wider stores in the store
9845 merging pass. The minimum value is 2 and the default is 64.
9847 @item max-unrolled-insns
9848 The maximum number of instructions that a loop may have to be unrolled.
9849 If a loop is unrolled, this parameter also determines how many times
9850 the loop code is unrolled.
9852 @item max-average-unrolled-insns
9853 The maximum number of instructions biased by probabilities of their execution
9854 that a loop may have to be unrolled. If a loop is unrolled,
9855 this parameter also determines how many times the loop code is unrolled.
9857 @item max-unroll-times
9858 The maximum number of unrollings of a single loop.
9860 @item max-peeled-insns
9861 The maximum number of instructions that a loop may have to be peeled.
9862 If a loop is peeled, this parameter also determines how many times
9863 the loop code is peeled.
9865 @item max-peel-times
9866 The maximum number of peelings of a single loop.
9868 @item max-peel-branches
9869 The maximum number of branches on the hot path through the peeled sequence.
9871 @item max-completely-peeled-insns
9872 The maximum number of insns of a completely peeled loop.
9874 @item max-completely-peel-times
9875 The maximum number of iterations of a loop to be suitable for complete peeling.
9877 @item max-completely-peel-loop-nest-depth
9878 The maximum depth of a loop nest suitable for complete peeling.
9880 @item max-unswitch-insns
9881 The maximum number of insns of an unswitched loop.
9883 @item max-unswitch-level
9884 The maximum number of branches unswitched in a single loop.
9886 @item max-loop-headers-insns
9887 The maximum number of insns in loop header duplicated by he copy loop headers
9891 The minimum cost of an expensive expression in the loop invariant motion.
9893 @item iv-consider-all-candidates-bound
9894 Bound on number of candidates for induction variables, below which
9895 all candidates are considered for each use in induction variable
9896 optimizations. If there are more candidates than this,
9897 only the most relevant ones are considered to avoid quadratic time complexity.
9899 @item iv-max-considered-uses
9900 The induction variable optimizations give up on loops that contain more
9901 induction variable uses.
9903 @item iv-always-prune-cand-set-bound
9904 If the number of candidates in the set is smaller than this value,
9905 always try to remove unnecessary ivs from the set
9906 when adding a new one.
9908 @item avg-loop-niter
9909 Average number of iterations of a loop.
9911 @item dse-max-object-size
9912 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
9913 Larger values may result in larger compilation times.
9915 @item scev-max-expr-size
9916 Bound on size of expressions used in the scalar evolutions analyzer.
9917 Large expressions slow the analyzer.
9919 @item scev-max-expr-complexity
9920 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9921 Complex expressions slow the analyzer.
9923 @item max-tree-if-conversion-phi-args
9924 Maximum number of arguments in a PHI supported by TREE if conversion
9925 unless the loop is marked with simd pragma.
9927 @item vect-max-version-for-alignment-checks
9928 The maximum number of run-time checks that can be performed when
9929 doing loop versioning for alignment in the vectorizer.
9931 @item vect-max-version-for-alias-checks
9932 The maximum number of run-time checks that can be performed when
9933 doing loop versioning for alias in the vectorizer.
9935 @item vect-max-peeling-for-alignment
9936 The maximum number of loop peels to enhance access alignment
9937 for vectorizer. Value -1 means no limit.
9939 @item max-iterations-to-track
9940 The maximum number of iterations of a loop the brute-force algorithm
9941 for analysis of the number of iterations of the loop tries to evaluate.
9943 @item hot-bb-count-ws-permille
9944 A basic block profile count is considered hot if it contributes to
9945 the given permillage (i.e. 0...1000) of the entire profiled execution.
9947 @item hot-bb-frequency-fraction
9948 Select fraction of the entry block frequency of executions of basic block in
9949 function given basic block needs to have to be considered hot.
9951 @item max-predicted-iterations
9952 The maximum number of loop iterations we predict statically. This is useful
9953 in cases where a function contains a single loop with known bound and
9954 another loop with unknown bound.
9955 The known number of iterations is predicted correctly, while
9956 the unknown number of iterations average to roughly 10. This means that the
9957 loop without bounds appears artificially cold relative to the other one.
9959 @item builtin-expect-probability
9960 Control the probability of the expression having the specified value. This
9961 parameter takes a percentage (i.e. 0 ... 100) as input.
9962 The default probability of 90 is obtained empirically.
9964 @item align-threshold
9966 Select fraction of the maximal frequency of executions of a basic block in
9967 a function to align the basic block.
9969 @item align-loop-iterations
9971 A loop expected to iterate at least the selected number of iterations is
9974 @item tracer-dynamic-coverage
9975 @itemx tracer-dynamic-coverage-feedback
9977 This value is used to limit superblock formation once the given percentage of
9978 executed instructions is covered. This limits unnecessary code size
9981 The @option{tracer-dynamic-coverage-feedback} parameter
9982 is used only when profile
9983 feedback is available. The real profiles (as opposed to statically estimated
9984 ones) are much less balanced allowing the threshold to be larger value.
9986 @item tracer-max-code-growth
9987 Stop tail duplication once code growth has reached given percentage. This is
9988 a rather artificial limit, as most of the duplicates are eliminated later in
9989 cross jumping, so it may be set to much higher values than is the desired code
9992 @item tracer-min-branch-ratio
9994 Stop reverse growth when the reverse probability of best edge is less than this
9995 threshold (in percent).
9997 @item tracer-min-branch-probability
9998 @itemx tracer-min-branch-probability-feedback
10000 Stop forward growth if the best edge has probability lower than this
10003 Similarly to @option{tracer-dynamic-coverage} two parameters are
10004 provided. @option{tracer-min-branch-probability-feedback} is used for
10005 compilation with profile feedback and @option{tracer-min-branch-probability}
10006 compilation without. The value for compilation with profile feedback
10007 needs to be more conservative (higher) in order to make tracer
10010 @item max-cse-path-length
10012 The maximum number of basic blocks on path that CSE considers.
10015 @item max-cse-insns
10016 The maximum number of instructions CSE processes before flushing.
10017 The default is 1000.
10019 @item ggc-min-expand
10021 GCC uses a garbage collector to manage its own memory allocation. This
10022 parameter specifies the minimum percentage by which the garbage
10023 collector's heap should be allowed to expand between collections.
10024 Tuning this may improve compilation speed; it has no effect on code
10027 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10028 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10029 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10030 GCC is not able to calculate RAM on a particular platform, the lower
10031 bound of 30% is used. Setting this parameter and
10032 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10033 every opportunity. This is extremely slow, but can be useful for
10036 @item ggc-min-heapsize
10038 Minimum size of the garbage collector's heap before it begins bothering
10039 to collect garbage. The first collection occurs after the heap expands
10040 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10041 tuning this may improve compilation speed, and has no effect on code
10044 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10045 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10046 with a lower bound of 4096 (four megabytes) and an upper bound of
10047 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10048 particular platform, the lower bound is used. Setting this parameter
10049 very large effectively disables garbage collection. Setting this
10050 parameter and @option{ggc-min-expand} to zero causes a full collection
10051 to occur at every opportunity.
10053 @item max-reload-search-insns
10054 The maximum number of instruction reload should look backward for equivalent
10055 register. Increasing values mean more aggressive optimization, making the
10056 compilation time increase with probably slightly better performance.
10057 The default value is 100.
10059 @item max-cselib-memory-locations
10060 The maximum number of memory locations cselib should take into account.
10061 Increasing values mean more aggressive optimization, making the compilation time
10062 increase with probably slightly better performance. The default value is 500.
10064 @item max-sched-ready-insns
10065 The maximum number of instructions ready to be issued the scheduler should
10066 consider at any given time during the first scheduling pass. Increasing
10067 values mean more thorough searches, making the compilation time increase
10068 with probably little benefit. The default value is 100.
10070 @item max-sched-region-blocks
10071 The maximum number of blocks in a region to be considered for
10072 interblock scheduling. The default value is 10.
10074 @item max-pipeline-region-blocks
10075 The maximum number of blocks in a region to be considered for
10076 pipelining in the selective scheduler. The default value is 15.
10078 @item max-sched-region-insns
10079 The maximum number of insns in a region to be considered for
10080 interblock scheduling. The default value is 100.
10082 @item max-pipeline-region-insns
10083 The maximum number of insns in a region to be considered for
10084 pipelining in the selective scheduler. The default value is 200.
10086 @item min-spec-prob
10087 The minimum probability (in percents) of reaching a source block
10088 for interblock speculative scheduling. The default value is 40.
10090 @item max-sched-extend-regions-iters
10091 The maximum number of iterations through CFG to extend regions.
10092 A value of 0 (the default) disables region extensions.
10094 @item max-sched-insn-conflict-delay
10095 The maximum conflict delay for an insn to be considered for speculative motion.
10096 The default value is 3.
10098 @item sched-spec-prob-cutoff
10099 The minimal probability of speculation success (in percents), so that
10100 speculative insns are scheduled.
10101 The default value is 40.
10103 @item sched-state-edge-prob-cutoff
10104 The minimum probability an edge must have for the scheduler to save its
10106 The default value is 10.
10108 @item sched-mem-true-dep-cost
10109 Minimal distance (in CPU cycles) between store and load targeting same
10110 memory locations. The default value is 1.
10112 @item selsched-max-lookahead
10113 The maximum size of the lookahead window of selective scheduling. It is a
10114 depth of search for available instructions.
10115 The default value is 50.
10117 @item selsched-max-sched-times
10118 The maximum number of times that an instruction is scheduled during
10119 selective scheduling. This is the limit on the number of iterations
10120 through which the instruction may be pipelined. The default value is 2.
10122 @item selsched-insns-to-rename
10123 The maximum number of best instructions in the ready list that are considered
10124 for renaming in the selective scheduler. The default value is 2.
10127 The minimum value of stage count that swing modulo scheduler
10128 generates. The default value is 2.
10130 @item max-last-value-rtl
10131 The maximum size measured as number of RTLs that can be recorded in an expression
10132 in combiner for a pseudo register as last known value of that register. The default
10135 @item max-combine-insns
10136 The maximum number of instructions the RTL combiner tries to combine.
10137 The default value is 2 at @option{-Og} and 4 otherwise.
10139 @item integer-share-limit
10140 Small integer constants can use a shared data structure, reducing the
10141 compiler's memory usage and increasing its speed. This sets the maximum
10142 value of a shared integer constant. The default value is 256.
10144 @item ssp-buffer-size
10145 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10146 protection when @option{-fstack-protection} is used.
10148 @item min-size-for-stack-sharing
10149 The minimum size of variables taking part in stack slot sharing when not
10150 optimizing. The default value is 32.
10152 @item max-jump-thread-duplication-stmts
10153 Maximum number of statements allowed in a block that needs to be
10154 duplicated when threading jumps.
10156 @item max-fields-for-field-sensitive
10157 Maximum number of fields in a structure treated in
10158 a field sensitive manner during pointer analysis. The default is zero
10159 for @option{-O0} and @option{-O1},
10160 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10162 @item prefetch-latency
10163 Estimate on average number of instructions that are executed before
10164 prefetch finishes. The distance prefetched ahead is proportional
10165 to this constant. Increasing this number may also lead to less
10166 streams being prefetched (see @option{simultaneous-prefetches}).
10168 @item simultaneous-prefetches
10169 Maximum number of prefetches that can run at the same time.
10171 @item l1-cache-line-size
10172 The size of cache line in L1 cache, in bytes.
10174 @item l1-cache-size
10175 The size of L1 cache, in kilobytes.
10177 @item l2-cache-size
10178 The size of L2 cache, in kilobytes.
10180 @item min-insn-to-prefetch-ratio
10181 The minimum ratio between the number of instructions and the
10182 number of prefetches to enable prefetching in a loop.
10184 @item prefetch-min-insn-to-mem-ratio
10185 The minimum ratio between the number of instructions and the
10186 number of memory references to enable prefetching in a loop.
10188 @item use-canonical-types
10189 Whether the compiler should use the ``canonical'' type system. By
10190 default, this should always be 1, which uses a more efficient internal
10191 mechanism for comparing types in C++ and Objective-C++. However, if
10192 bugs in the canonical type system are causing compilation failures,
10193 set this value to 0 to disable canonical types.
10195 @item switch-conversion-max-branch-ratio
10196 Switch initialization conversion refuses to create arrays that are
10197 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10198 branches in the switch.
10200 @item max-partial-antic-length
10201 Maximum length of the partial antic set computed during the tree
10202 partial redundancy elimination optimization (@option{-ftree-pre}) when
10203 optimizing at @option{-O3} and above. For some sorts of source code
10204 the enhanced partial redundancy elimination optimization can run away,
10205 consuming all of the memory available on the host machine. This
10206 parameter sets a limit on the length of the sets that are computed,
10207 which prevents the runaway behavior. Setting a value of 0 for
10208 this parameter allows an unlimited set length.
10210 @item sccvn-max-scc-size
10211 Maximum size of a strongly connected component (SCC) during SCCVN
10212 processing. If this limit is hit, SCCVN processing for the whole
10213 function is not done and optimizations depending on it are
10214 disabled. The default maximum SCC size is 10000.
10216 @item sccvn-max-alias-queries-per-access
10217 Maximum number of alias-oracle queries we perform when looking for
10218 redundancies for loads and stores. If this limit is hit the search
10219 is aborted and the load or store is not considered redundant. The
10220 number of queries is algorithmically limited to the number of
10221 stores on all paths from the load to the function entry.
10222 The default maximum number of queries is 1000.
10224 @item ira-max-loops-num
10225 IRA uses regional register allocation by default. If a function
10226 contains more loops than the number given by this parameter, only at most
10227 the given number of the most frequently-executed loops form regions
10228 for regional register allocation. The default value of the
10231 @item ira-max-conflict-table-size
10232 Although IRA uses a sophisticated algorithm to compress the conflict
10233 table, the table can still require excessive amounts of memory for
10234 huge functions. If the conflict table for a function could be more
10235 than the size in MB given by this parameter, the register allocator
10236 instead uses a faster, simpler, and lower-quality
10237 algorithm that does not require building a pseudo-register conflict table.
10238 The default value of the parameter is 2000.
10240 @item ira-loop-reserved-regs
10241 IRA can be used to evaluate more accurate register pressure in loops
10242 for decisions to move loop invariants (see @option{-O3}). The number
10243 of available registers reserved for some other purposes is given
10244 by this parameter. The default value of the parameter is 2, which is
10245 the minimal number of registers needed by typical instructions.
10246 This value is the best found from numerous experiments.
10248 @item lra-inheritance-ebb-probability-cutoff
10249 LRA tries to reuse values reloaded in registers in subsequent insns.
10250 This optimization is called inheritance. EBB is used as a region to
10251 do this optimization. The parameter defines a minimal fall-through
10252 edge probability in percentage used to add BB to inheritance EBB in
10253 LRA. The default value of the parameter is 40. The value was chosen
10254 from numerous runs of SPEC2000 on x86-64.
10256 @item loop-invariant-max-bbs-in-loop
10257 Loop invariant motion can be very expensive, both in compilation time and
10258 in amount of needed compile-time memory, with very large loops. Loops
10259 with more basic blocks than this parameter won't have loop invariant
10260 motion optimization performed on them. The default value of the
10261 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10263 @item loop-max-datarefs-for-datadeps
10264 Building data dependencies is expensive for very large loops. This
10265 parameter limits the number of data references in loops that are
10266 considered for data dependence analysis. These large loops are no
10267 handled by the optimizations using loop data dependencies.
10268 The default value is 1000.
10270 @item max-vartrack-size
10271 Sets a maximum number of hash table slots to use during variable
10272 tracking dataflow analysis of any function. If this limit is exceeded
10273 with variable tracking at assignments enabled, analysis for that
10274 function is retried without it, after removing all debug insns from
10275 the function. If the limit is exceeded even without debug insns, var
10276 tracking analysis is completely disabled for the function. Setting
10277 the parameter to zero makes it unlimited.
10279 @item max-vartrack-expr-depth
10280 Sets a maximum number of recursion levels when attempting to map
10281 variable names or debug temporaries to value expressions. This trades
10282 compilation time for more complete debug information. If this is set too
10283 low, value expressions that are available and could be represented in
10284 debug information may end up not being used; setting this higher may
10285 enable the compiler to find more complex debug expressions, but compile
10286 time and memory use may grow. The default is 12.
10288 @item min-nondebug-insn-uid
10289 Use uids starting at this parameter for nondebug insns. The range below
10290 the parameter is reserved exclusively for debug insns created by
10291 @option{-fvar-tracking-assignments}, but debug insns may get
10292 (non-overlapping) uids above it if the reserved range is exhausted.
10294 @item ipa-sra-ptr-growth-factor
10295 IPA-SRA replaces a pointer to an aggregate with one or more new
10296 parameters only when their cumulative size is less or equal to
10297 @option{ipa-sra-ptr-growth-factor} times the size of the original
10300 @item sra-max-scalarization-size-Ospeed
10301 @item sra-max-scalarization-size-Osize
10302 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10303 replace scalar parts of aggregates with uses of independent scalar
10304 variables. These parameters control the maximum size, in storage units,
10305 of aggregate which is considered for replacement when compiling for
10307 (@option{sra-max-scalarization-size-Ospeed}) or size
10308 (@option{sra-max-scalarization-size-Osize}) respectively.
10310 @item tm-max-aggregate-size
10311 When making copies of thread-local variables in a transaction, this
10312 parameter specifies the size in bytes after which variables are
10313 saved with the logging functions as opposed to save/restore code
10314 sequence pairs. This option only applies when using
10317 @item graphite-max-nb-scop-params
10318 To avoid exponential effects in the Graphite loop transforms, the
10319 number of parameters in a Static Control Part (SCoP) is bounded. The
10320 default value is 10 parameters. A variable whose value is unknown at
10321 compilation time and defined outside a SCoP is a parameter of the SCoP.
10323 @item graphite-max-bbs-per-function
10324 To avoid exponential effects in the detection of SCoPs, the size of
10325 the functions analyzed by Graphite is bounded. The default value is
10328 @item loop-block-tile-size
10329 Loop blocking or strip mining transforms, enabled with
10330 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10331 loop in the loop nest by a given number of iterations. The strip
10332 length can be changed using the @option{loop-block-tile-size}
10333 parameter. The default value is 51 iterations.
10335 @item loop-unroll-jam-size
10336 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10337 default value is 4.
10339 @item loop-unroll-jam-depth
10340 Specify the dimension to be unrolled (counting from the most inner loop)
10341 for the @option{-floop-unroll-and-jam}. The default value is 2.
10343 @item ipa-cp-value-list-size
10344 IPA-CP attempts to track all possible values and types passed to a function's
10345 parameter in order to propagate them and perform devirtualization.
10346 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10347 stores per one formal parameter of a function.
10349 @item ipa-cp-eval-threshold
10350 IPA-CP calculates its own score of cloning profitability heuristics
10351 and performs those cloning opportunities with scores that exceed
10352 @option{ipa-cp-eval-threshold}.
10354 @item ipa-cp-recursion-penalty
10355 Percentage penalty the recursive functions will receive when they
10356 are evaluated for cloning.
10358 @item ipa-cp-single-call-penalty
10359 Percentage penalty functions containing a single call to another
10360 function will receive when they are evaluated for cloning.
10363 @item ipa-max-agg-items
10364 IPA-CP is also capable to propagate a number of scalar values passed
10365 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10366 number of such values per one parameter.
10368 @item ipa-cp-loop-hint-bonus
10369 When IPA-CP determines that a cloning candidate would make the number
10370 of iterations of a loop known, it adds a bonus of
10371 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10374 @item ipa-cp-array-index-hint-bonus
10375 When IPA-CP determines that a cloning candidate would make the index of
10376 an array access known, it adds a bonus of
10377 @option{ipa-cp-array-index-hint-bonus} to the profitability
10378 score of the candidate.
10380 @item ipa-max-aa-steps
10381 During its analysis of function bodies, IPA-CP employs alias analysis
10382 in order to track values pointed to by function parameters. In order
10383 not spend too much time analyzing huge functions, it gives up and
10384 consider all memory clobbered after examining
10385 @option{ipa-max-aa-steps} statements modifying memory.
10387 @item lto-partitions
10388 Specify desired number of partitions produced during WHOPR compilation.
10389 The number of partitions should exceed the number of CPUs used for compilation.
10390 The default value is 32.
10392 @item lto-min-partition
10393 Size of minimal partition for WHOPR (in estimated instructions).
10394 This prevents expenses of splitting very small programs into too many
10397 @item lto-max-partition
10398 Size of max partition for WHOPR (in estimated instructions).
10399 to provide an upper bound for individual size of partition.
10400 Meant to be used only with balanced partitioning.
10402 @item cxx-max-namespaces-for-diagnostic-help
10403 The maximum number of namespaces to consult for suggestions when C++
10404 name lookup fails for an identifier. The default is 1000.
10406 @item sink-frequency-threshold
10407 The maximum relative execution frequency (in percents) of the target block
10408 relative to a statement's original block to allow statement sinking of a
10409 statement. Larger numbers result in more aggressive statement sinking.
10410 The default value is 75. A small positive adjustment is applied for
10411 statements with memory operands as those are even more profitable so sink.
10413 @item max-stores-to-sink
10414 The maximum number of conditional store pairs that can be sunk. Set to 0
10415 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10416 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10418 @item allow-store-data-races
10419 Allow optimizers to introduce new data races on stores.
10420 Set to 1 to allow, otherwise to 0. This option is enabled by default
10421 at optimization level @option{-Ofast}.
10423 @item case-values-threshold
10424 The smallest number of different values for which it is best to use a
10425 jump-table instead of a tree of conditional branches. If the value is
10426 0, use the default for the machine. The default is 0.
10428 @item tree-reassoc-width
10429 Set the maximum number of instructions executed in parallel in
10430 reassociated tree. This parameter overrides target dependent
10431 heuristics used by default if has non zero value.
10433 @item sched-pressure-algorithm
10434 Choose between the two available implementations of
10435 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10436 and is the more likely to prevent instructions from being reordered.
10437 Algorithm 2 was designed to be a compromise between the relatively
10438 conservative approach taken by algorithm 1 and the rather aggressive
10439 approach taken by the default scheduler. It relies more heavily on
10440 having a regular register file and accurate register pressure classes.
10441 See @file{haifa-sched.c} in the GCC sources for more details.
10443 The default choice depends on the target.
10445 @item max-slsr-cand-scan
10446 Set the maximum number of existing candidates that are considered when
10447 seeking a basis for a new straight-line strength reduction candidate.
10450 Enable buffer overflow detection for global objects. This kind
10451 of protection is enabled by default if you are using
10452 @option{-fsanitize=address} option.
10453 To disable global objects protection use @option{--param asan-globals=0}.
10456 Enable buffer overflow detection for stack objects. This kind of
10457 protection is enabled by default when using @option{-fsanitize=address}.
10458 To disable stack protection use @option{--param asan-stack=0} option.
10460 @item asan-instrument-reads
10461 Enable buffer overflow detection for memory reads. This kind of
10462 protection is enabled by default when using @option{-fsanitize=address}.
10463 To disable memory reads protection use
10464 @option{--param asan-instrument-reads=0}.
10466 @item asan-instrument-writes
10467 Enable buffer overflow detection for memory writes. This kind of
10468 protection is enabled by default when using @option{-fsanitize=address}.
10469 To disable memory writes protection use
10470 @option{--param asan-instrument-writes=0} option.
10472 @item asan-memintrin
10473 Enable detection for built-in functions. This kind of protection
10474 is enabled by default when using @option{-fsanitize=address}.
10475 To disable built-in functions protection use
10476 @option{--param asan-memintrin=0}.
10478 @item asan-use-after-return
10479 Enable detection of use-after-return. This kind of protection
10480 is enabled by default when using @option{-fsanitize=address} option.
10481 To disable use-after-return detection use
10482 @option{--param asan-use-after-return=0}.
10484 Note: The check is disabled by default at runtime. To enable the check,
10485 you should set environment variable @env{ASAN_OPTIONS} to
10486 @code{detect_stack_use_after_return=1}.
10488 @item asan-instrumentation-with-call-threshold
10489 If number of memory accesses in function being instrumented
10490 is greater or equal to this number, use callbacks instead of inline checks.
10491 E.g. to disable inline code use
10492 @option{--param asan-instrumentation-with-call-threshold=0}.
10494 @item use-after-scope-direct-emission-threshold
10495 If size of a local variable in bytes is smaller or equal to this number,
10496 direct instruction emission is utilized to poison and unpoison local variables.
10497 Default value in 256.
10499 @item chkp-max-ctor-size
10500 Static constructors generated by Pointer Bounds Checker may become very
10501 large and significantly increase compile time at optimization level
10502 @option{-O1} and higher. This parameter is a maximum number of statements
10503 in a single generated constructor. Default value is 5000.
10505 @item max-fsm-thread-path-insns
10506 Maximum number of instructions to copy when duplicating blocks on a
10507 finite state automaton jump thread path. The default is 100.
10509 @item max-fsm-thread-length
10510 Maximum number of basic blocks on a finite state automaton jump thread
10511 path. The default is 10.
10513 @item max-fsm-thread-paths
10514 Maximum number of new jump thread paths to create for a finite state
10515 automaton. The default is 50.
10517 @item parloops-chunk-size
10518 Chunk size of omp schedule for loops parallelized by parloops. The default
10521 @item parloops-schedule
10522 Schedule type of omp schedule for loops parallelized by parloops (static,
10523 dynamic, guided, auto, runtime). The default is static.
10525 @item max-ssa-name-query-depth
10526 Maximum depth of recursion when querying properties of SSA names in things
10527 like fold routines. One level of recursion corresponds to following a
10530 @item hsa-gen-debug-stores
10531 Enable emission of special debug stores within HSA kernels which are
10532 then read and reported by libgomp plugin. Generation of these stores
10533 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
10536 @item max-speculative-devirt-maydefs
10537 The maximum number of may-defs we analyze when looking for a must-def
10538 specifying the dynamic type of an object that invokes a virtual call
10539 we may be able to devirtualize speculatively.
10541 @item max-vrp-switch-assertions
10542 The maximum number of assertions to add along the default edge of a switch
10543 statement during VRP. The default is 10.
10547 @node Instrumentation Options
10548 @section Program Instrumentation Options
10549 @cindex instrumentation options
10550 @cindex program instrumentation options
10551 @cindex run-time error checking options
10552 @cindex profiling options
10553 @cindex options, program instrumentation
10554 @cindex options, run-time error checking
10555 @cindex options, profiling
10557 GCC supports a number of command-line options that control adding
10558 run-time instrumentation to the code it normally generates.
10559 For example, one purpose of instrumentation is collect profiling
10560 statistics for use in finding program hot spots, code coverage
10561 analysis, or profile-guided optimizations.
10562 Another class of program instrumentation is adding run-time checking
10563 to detect programming errors like invalid pointer
10564 dereferences or out-of-bounds array accesses, as well as deliberately
10565 hostile attacks such as stack smashing or C++ vtable hijacking.
10566 There is also a general hook which can be used to implement other
10567 forms of tracing or function-level instrumentation for debug or
10568 program analysis purposes.
10571 @cindex @command{prof}
10574 Generate extra code to write profile information suitable for the
10575 analysis program @command{prof}. You must use this option when compiling
10576 the source files you want data about, and you must also use it when
10579 @cindex @command{gprof}
10582 Generate extra code to write profile information suitable for the
10583 analysis program @command{gprof}. You must use this option when compiling
10584 the source files you want data about, and you must also use it when
10587 @item -fprofile-arcs
10588 @opindex fprofile-arcs
10589 Add code so that program flow @dfn{arcs} are instrumented. During
10590 execution the program records how many times each branch and call is
10591 executed and how many times it is taken or returns. On targets that support
10592 constructors with priority support, profiling properly handles constructors,
10593 destructors and C++ constructors (and destructors) of classes which are used
10594 as a type of a global variable.
10597 program exits it saves this data to a file called
10598 @file{@var{auxname}.gcda} for each source file. The data may be used for
10599 profile-directed optimizations (@option{-fbranch-probabilities}), or for
10600 test coverage analysis (@option{-ftest-coverage}). Each object file's
10601 @var{auxname} is generated from the name of the output file, if
10602 explicitly specified and it is not the final executable, otherwise it is
10603 the basename of the source file. In both cases any suffix is removed
10604 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
10605 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
10606 @xref{Cross-profiling}.
10608 @cindex @command{gcov}
10612 This option is used to compile and link code instrumented for coverage
10613 analysis. The option is a synonym for @option{-fprofile-arcs}
10614 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
10615 linking). See the documentation for those options for more details.
10620 Compile the source files with @option{-fprofile-arcs} plus optimization
10621 and code generation options. For test coverage analysis, use the
10622 additional @option{-ftest-coverage} option. You do not need to profile
10623 every source file in a program.
10626 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
10627 (the latter implies the former).
10630 Run the program on a representative workload to generate the arc profile
10631 information. This may be repeated any number of times. You can run
10632 concurrent instances of your program, and provided that the file system
10633 supports locking, the data files will be correctly updated. Also
10634 @code{fork} calls are detected and correctly handled (double counting
10638 For profile-directed optimizations, compile the source files again with
10639 the same optimization and code generation options plus
10640 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
10641 Control Optimization}).
10644 For test coverage analysis, use @command{gcov} to produce human readable
10645 information from the @file{.gcno} and @file{.gcda} files. Refer to the
10646 @command{gcov} documentation for further information.
10650 With @option{-fprofile-arcs}, for each function of your program GCC
10651 creates a program flow graph, then finds a spanning tree for the graph.
10652 Only arcs that are not on the spanning tree have to be instrumented: the
10653 compiler adds code to count the number of times that these arcs are
10654 executed. When an arc is the only exit or only entrance to a block, the
10655 instrumentation code can be added to the block; otherwise, a new basic
10656 block must be created to hold the instrumentation code.
10659 @item -ftest-coverage
10660 @opindex ftest-coverage
10661 Produce a notes file that the @command{gcov} code-coverage utility
10662 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
10663 show program coverage. Each source file's note file is called
10664 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
10665 above for a description of @var{auxname} and instructions on how to
10666 generate test coverage data. Coverage data matches the source files
10667 more closely if you do not optimize.
10669 @item -fprofile-dir=@var{path}
10670 @opindex fprofile-dir
10672 Set the directory to search for the profile data files in to @var{path}.
10673 This option affects only the profile data generated by
10674 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
10675 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
10676 and its related options. Both absolute and relative paths can be used.
10677 By default, GCC uses the current directory as @var{path}, thus the
10678 profile data file appears in the same directory as the object file.
10680 @item -fprofile-generate
10681 @itemx -fprofile-generate=@var{path}
10682 @opindex fprofile-generate
10684 Enable options usually used for instrumenting application to produce
10685 profile useful for later recompilation with profile feedback based
10686 optimization. You must use @option{-fprofile-generate} both when
10687 compiling and when linking your program.
10689 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
10691 If @var{path} is specified, GCC looks at the @var{path} to find
10692 the profile feedback data files. See @option{-fprofile-dir}.
10694 To optimize the program based on the collected profile information, use
10695 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
10697 @item -fprofile-update=@var{method}
10698 @opindex fprofile-update
10700 Alter the update method for an application instrumented for profile
10701 feedback based optimization. The @var{method} argument should be one of
10702 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
10703 The first one is useful for single-threaded applications,
10704 while the second one prevents profile corruption by emitting thread-safe code.
10706 @strong{Warning:} When an application does not properly join all threads
10707 (or creates an detached thread), a profile file can be still corrupted.
10709 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
10710 when supported by a target, or to @samp{single} otherwise. The GCC driver
10711 automatically selects @samp{prefer-atomic} when @option{-pthread}
10712 is present in the command line.
10714 @item -fsanitize=address
10715 @opindex fsanitize=address
10716 Enable AddressSanitizer, a fast memory error detector.
10717 Memory access instructions are instrumented to detect
10718 out-of-bounds and use-after-free bugs.
10719 The option enables @option{-fsanitize-address-use-after-scope}.
10720 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
10721 more details. The run-time behavior can be influenced using the
10722 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
10723 the available options are shown at startup of the instrumented program. See
10724 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
10725 for a list of supported options.
10726 The option can't be combined with @option{-fsanitize=thread}.
10728 @item -fsanitize=kernel-address
10729 @opindex fsanitize=kernel-address
10730 Enable AddressSanitizer for Linux kernel.
10731 See @uref{https://github.com/google/kasan/wiki} for more details.
10733 @item -fsanitize=thread
10734 @opindex fsanitize=thread
10735 Enable ThreadSanitizer, a fast data race detector.
10736 Memory access instructions are instrumented to detect
10737 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
10738 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
10739 environment variable; see
10740 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
10742 The option can't be combined with @option{-fsanitize=address}
10743 and/or @option{-fsanitize=leak}.
10745 @item -fsanitize=leak
10746 @opindex fsanitize=leak
10747 Enable LeakSanitizer, a memory leak detector.
10748 This option only matters for linking of executables and
10749 the executable is linked against a library that overrides @code{malloc}
10750 and other allocator functions. See
10751 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
10752 details. The run-time behavior can be influenced using the
10753 @env{LSAN_OPTIONS} environment variable.
10754 The option can't be combined with @option{-fsanitize=thread}.
10756 @item -fsanitize=undefined
10757 @opindex fsanitize=undefined
10758 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
10759 Various computations are instrumented to detect undefined behavior
10760 at runtime. Current suboptions are:
10764 @item -fsanitize=shift
10765 @opindex fsanitize=shift
10766 This option enables checking that the result of a shift operation is
10767 not undefined. Note that what exactly is considered undefined differs
10768 slightly between C and C++, as well as between ISO C90 and C99, etc.
10769 This option has two suboptions, @option{-fsanitize=shift-base} and
10770 @option{-fsanitize=shift-exponent}.
10772 @item -fsanitize=shift-exponent
10773 @opindex fsanitize=shift-exponent
10774 This option enables checking that the second argument of a shift operation
10775 is not negative and is smaller than the precision of the promoted first
10778 @item -fsanitize=shift-base
10779 @opindex fsanitize=shift-base
10780 If the second argument of a shift operation is within range, check that the
10781 result of a shift operation is not undefined. Note that what exactly is
10782 considered undefined differs slightly between C and C++, as well as between
10783 ISO C90 and C99, etc.
10785 @item -fsanitize=integer-divide-by-zero
10786 @opindex fsanitize=integer-divide-by-zero
10787 Detect integer division by zero as well as @code{INT_MIN / -1} division.
10789 @item -fsanitize=unreachable
10790 @opindex fsanitize=unreachable
10791 With this option, the compiler turns the @code{__builtin_unreachable}
10792 call into a diagnostics message call instead. When reaching the
10793 @code{__builtin_unreachable} call, the behavior is undefined.
10795 @item -fsanitize=vla-bound
10796 @opindex fsanitize=vla-bound
10797 This option instructs the compiler to check that the size of a variable
10798 length array is positive.
10800 @item -fsanitize=null
10801 @opindex fsanitize=null
10802 This option enables pointer checking. Particularly, the application
10803 built with this option turned on will issue an error message when it
10804 tries to dereference a NULL pointer, or if a reference (possibly an
10805 rvalue reference) is bound to a NULL pointer, or if a method is invoked
10806 on an object pointed by a NULL pointer.
10808 @item -fsanitize=return
10809 @opindex fsanitize=return
10810 This option enables return statement checking. Programs
10811 built with this option turned on will issue an error message
10812 when the end of a non-void function is reached without actually
10813 returning a value. This option works in C++ only.
10815 @item -fsanitize=signed-integer-overflow
10816 @opindex fsanitize=signed-integer-overflow
10817 This option enables signed integer overflow checking. We check that
10818 the result of @code{+}, @code{*}, and both unary and binary @code{-}
10819 does not overflow in the signed arithmetics. Note, integer promotion
10820 rules must be taken into account. That is, the following is not an
10823 signed char a = SCHAR_MAX;
10827 @item -fsanitize=bounds
10828 @opindex fsanitize=bounds
10829 This option enables instrumentation of array bounds. Various out of bounds
10830 accesses are detected. Flexible array members, flexible array member-like
10831 arrays, and initializers of variables with static storage are not instrumented.
10833 @item -fsanitize=bounds-strict
10834 @opindex fsanitize=bounds-strict
10835 This option enables strict instrumentation of array bounds. Most out of bounds
10836 accesses are detected, including flexible array members and flexible array
10837 member-like arrays. Initializers of variables with static storage are not
10840 @item -fsanitize=alignment
10841 @opindex fsanitize=alignment
10843 This option enables checking of alignment of pointers when they are
10844 dereferenced, or when a reference is bound to insufficiently aligned target,
10845 or when a method or constructor is invoked on insufficiently aligned object.
10847 @item -fsanitize=object-size
10848 @opindex fsanitize=object-size
10849 This option enables instrumentation of memory references using the
10850 @code{__builtin_object_size} function. Various out of bounds pointer
10851 accesses are detected.
10853 @item -fsanitize=float-divide-by-zero
10854 @opindex fsanitize=float-divide-by-zero
10855 Detect floating-point division by zero. Unlike other similar options,
10856 @option{-fsanitize=float-divide-by-zero} is not enabled by
10857 @option{-fsanitize=undefined}, since floating-point division by zero can
10858 be a legitimate way of obtaining infinities and NaNs.
10860 @item -fsanitize=float-cast-overflow
10861 @opindex fsanitize=float-cast-overflow
10862 This option enables floating-point type to integer conversion checking.
10863 We check that the result of the conversion does not overflow.
10864 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10865 not enabled by @option{-fsanitize=undefined}.
10866 This option does not work well with @code{FE_INVALID} exceptions enabled.
10868 @item -fsanitize=nonnull-attribute
10869 @opindex fsanitize=nonnull-attribute
10871 This option enables instrumentation of calls, checking whether null values
10872 are not passed to arguments marked as requiring a non-null value by the
10873 @code{nonnull} function attribute.
10875 @item -fsanitize=returns-nonnull-attribute
10876 @opindex fsanitize=returns-nonnull-attribute
10878 This option enables instrumentation of return statements in functions
10879 marked with @code{returns_nonnull} function attribute, to detect returning
10880 of null values from such functions.
10882 @item -fsanitize=bool
10883 @opindex fsanitize=bool
10885 This option enables instrumentation of loads from bool. If a value other
10886 than 0/1 is loaded, a run-time error is issued.
10888 @item -fsanitize=enum
10889 @opindex fsanitize=enum
10891 This option enables instrumentation of loads from an enum type. If
10892 a value outside the range of values for the enum type is loaded,
10893 a run-time error is issued.
10895 @item -fsanitize=vptr
10896 @opindex fsanitize=vptr
10898 This option enables instrumentation of C++ member function calls, member
10899 accesses and some conversions between pointers to base and derived classes,
10900 to verify the referenced object has the correct dynamic type.
10904 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10905 @option{-fsanitize=undefined} gives a diagnostic message.
10906 This currently works only for the C family of languages.
10908 @item -fno-sanitize=all
10909 @opindex fno-sanitize=all
10911 This option disables all previously enabled sanitizers.
10912 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10915 @item -fasan-shadow-offset=@var{number}
10916 @opindex fasan-shadow-offset
10917 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10918 It is useful for experimenting with different shadow memory layouts in
10919 Kernel AddressSanitizer.
10921 @item -fsanitize-sections=@var{s1},@var{s2},...
10922 @opindex fsanitize-sections
10923 Sanitize global variables in selected user-defined sections. @var{si} may
10926 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10927 @opindex fsanitize-recover
10928 @opindex fno-sanitize-recover
10929 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10930 mentioned in comma-separated list of @var{opts}. Enabling this option
10931 for a sanitizer component causes it to attempt to continue
10932 running the program as if no error happened. This means multiple
10933 runtime errors can be reported in a single program run, and the exit
10934 code of the program may indicate success even when errors
10935 have been reported. The @option{-fno-sanitize-recover=} option
10936 can be used to alter
10937 this behavior: only the first detected error is reported
10938 and program then exits with a non-zero exit code.
10940 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10941 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10942 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10943 @option{-fsanitize=bounds-strict},
10944 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10945 For these sanitizers error recovery is turned on by default,
10946 except @option{-fsanitize=address}, for which this feature is experimental.
10947 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10948 accepted, the former enables recovery for all sanitizers that support it,
10949 the latter disables recovery for all sanitizers that support it.
10951 Even if a recovery mode is turned on the compiler side, it needs to be also
10952 enabled on the runtime library side, otherwise the failures are still fatal.
10953 The runtime library defaults to @code{halt_on_error=0} for
10954 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10955 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10956 setting the @code{halt_on_error} flag in the corresponding environment variable.
10958 Syntax without an explicit @var{opts} parameter is deprecated. It is
10959 equivalent to specifying an @var{opts} list of:
10962 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
10965 @item -fsanitize-address-use-after-scope
10966 @opindex fsanitize-address-use-after-scope
10967 Enable sanitization of local variables to detect use-after-scope bugs.
10968 The option sets @option{-fstack-reuse} to @samp{none}.
10970 @item -fsanitize-undefined-trap-on-error
10971 @opindex fsanitize-undefined-trap-on-error
10972 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10973 report undefined behavior using @code{__builtin_trap} rather than
10974 a @code{libubsan} library routine. The advantage of this is that the
10975 @code{libubsan} library is not needed and is not linked in, so this
10976 is usable even in freestanding environments.
10978 @item -fsanitize-coverage=trace-pc
10979 @opindex fsanitize-coverage=trace-pc
10980 Enable coverage-guided fuzzing code instrumentation.
10981 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10983 @item -fbounds-check
10984 @opindex fbounds-check
10985 For front ends that support it, generate additional code to check that
10986 indices used to access arrays are within the declared range. This is
10987 currently only supported by the Fortran front end, where this option
10990 @item -fcheck-pointer-bounds
10991 @opindex fcheck-pointer-bounds
10992 @opindex fno-check-pointer-bounds
10993 @cindex Pointer Bounds Checker options
10994 Enable Pointer Bounds Checker instrumentation. Each memory reference
10995 is instrumented with checks of the pointer used for memory access against
10996 bounds associated with that pointer.
10999 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
11000 and @option{-mmpx} are required to enable this feature.
11001 MPX-based instrumentation requires
11002 a runtime library to enable MPX in hardware and handle bounds
11003 violation signals. By default when @option{-fcheck-pointer-bounds}
11004 and @option{-mmpx} options are used to link a program, the GCC driver
11005 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
11006 Bounds checking on calls to dynamic libraries requires a linker
11007 with @option{-z bndplt} support; if GCC was configured with a linker
11008 without support for this option (including the Gold linker and older
11009 versions of ld), a warning is given if you link with @option{-mmpx}
11010 without also specifying @option{-static}, since the overall effectiveness
11011 of the bounds checking protection is reduced.
11012 See also @option{-static-libmpxwrappers}.
11014 MPX-based instrumentation
11015 may be used for debugging and also may be included in production code
11016 to increase program security. Depending on usage, you may
11017 have different requirements for the runtime library. The current version
11018 of the MPX runtime library is more oriented for use as a debugging
11019 tool. MPX runtime library usage implies @option{-lpthread}. See
11020 also @option{-static-libmpx}. The runtime library behavior can be
11021 influenced using various @env{CHKP_RT_*} environment variables. See
11022 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
11025 Generated instrumentation may be controlled by various
11026 @option{-fchkp-*} options and by the @code{bnd_variable_size}
11027 structure field attribute (@pxref{Type Attributes}) and
11028 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
11029 (@pxref{Function Attributes}). GCC also provides a number of built-in
11030 functions for controlling the Pointer Bounds Checker. @xref{Pointer
11031 Bounds Checker builtins}, for more information.
11033 @item -fchkp-check-incomplete-type
11034 @opindex fchkp-check-incomplete-type
11035 @opindex fno-chkp-check-incomplete-type
11036 Generate pointer bounds checks for variables with incomplete type.
11037 Enabled by default.
11039 @item -fchkp-narrow-bounds
11040 @opindex fchkp-narrow-bounds
11041 @opindex fno-chkp-narrow-bounds
11042 Controls bounds used by Pointer Bounds Checker for pointers to object
11043 fields. If narrowing is enabled then field bounds are used. Otherwise
11044 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
11045 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
11047 @item -fchkp-first-field-has-own-bounds
11048 @opindex fchkp-first-field-has-own-bounds
11049 @opindex fno-chkp-first-field-has-own-bounds
11050 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
11051 first field in the structure. By default a pointer to the first field has
11052 the same bounds as a pointer to the whole structure.
11054 @item -fchkp-flexible-struct-trailing-arrays
11055 @opindex fchkp-flexible-struct-trailing-arrays
11056 @opindex fno-chkp-flexible-struct-trailing-arrays
11057 Forces Pointer Bounds Checker to treat all trailing arrays in structures as
11058 possibly flexible. By default only array fields with zero length or that are
11059 marked with attribute bnd_variable_size are treated as flexible.
11061 @item -fchkp-narrow-to-innermost-array
11062 @opindex fchkp-narrow-to-innermost-array
11063 @opindex fno-chkp-narrow-to-innermost-array
11064 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
11065 case of nested static array access. By default this option is disabled and
11066 bounds of the outermost array are used.
11068 @item -fchkp-optimize
11069 @opindex fchkp-optimize
11070 @opindex fno-chkp-optimize
11071 Enables Pointer Bounds Checker optimizations. Enabled by default at
11072 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
11074 @item -fchkp-use-fast-string-functions
11075 @opindex fchkp-use-fast-string-functions
11076 @opindex fno-chkp-use-fast-string-functions
11077 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
11078 by Pointer Bounds Checker. Disabled by default.
11080 @item -fchkp-use-nochk-string-functions
11081 @opindex fchkp-use-nochk-string-functions
11082 @opindex fno-chkp-use-nochk-string-functions
11083 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
11084 by Pointer Bounds Checker. Disabled by default.
11086 @item -fchkp-use-static-bounds
11087 @opindex fchkp-use-static-bounds
11088 @opindex fno-chkp-use-static-bounds
11089 Allow Pointer Bounds Checker to generate static bounds holding
11090 bounds of static variables. Enabled by default.
11092 @item -fchkp-use-static-const-bounds
11093 @opindex fchkp-use-static-const-bounds
11094 @opindex fno-chkp-use-static-const-bounds
11095 Use statically-initialized bounds for constant bounds instead of
11096 generating them each time they are required. By default enabled when
11097 @option{-fchkp-use-static-bounds} is enabled.
11099 @item -fchkp-treat-zero-dynamic-size-as-infinite
11100 @opindex fchkp-treat-zero-dynamic-size-as-infinite
11101 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
11102 With this option, objects with incomplete type whose
11103 dynamically-obtained size is zero are treated as having infinite size
11104 instead by Pointer Bounds
11105 Checker. This option may be helpful if a program is linked with a library
11106 missing size information for some symbols. Disabled by default.
11108 @item -fchkp-check-read
11109 @opindex fchkp-check-read
11110 @opindex fno-chkp-check-read
11111 Instructs Pointer Bounds Checker to generate checks for all read
11112 accesses to memory. Enabled by default.
11114 @item -fchkp-check-write
11115 @opindex fchkp-check-write
11116 @opindex fno-chkp-check-write
11117 Instructs Pointer Bounds Checker to generate checks for all write
11118 accesses to memory. Enabled by default.
11120 @item -fchkp-store-bounds
11121 @opindex fchkp-store-bounds
11122 @opindex fno-chkp-store-bounds
11123 Instructs Pointer Bounds Checker to generate bounds stores for
11124 pointer writes. Enabled by default.
11126 @item -fchkp-instrument-calls
11127 @opindex fchkp-instrument-calls
11128 @opindex fno-chkp-instrument-calls
11129 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
11130 Enabled by default.
11132 @item -fchkp-instrument-marked-only
11133 @opindex fchkp-instrument-marked-only
11134 @opindex fno-chkp-instrument-marked-only
11135 Instructs Pointer Bounds Checker to instrument only functions
11136 marked with the @code{bnd_instrument} attribute
11137 (@pxref{Function Attributes}). Disabled by default.
11139 @item -fchkp-use-wrappers
11140 @opindex fchkp-use-wrappers
11141 @opindex fno-chkp-use-wrappers
11142 Allows Pointer Bounds Checker to replace calls to built-in functions
11143 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
11144 is used to link a program, the GCC driver automatically links
11145 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
11146 Enabled by default.
11148 @item -fstack-protector
11149 @opindex fstack-protector
11150 Emit extra code to check for buffer overflows, such as stack smashing
11151 attacks. This is done by adding a guard variable to functions with
11152 vulnerable objects. This includes functions that call @code{alloca}, and
11153 functions with buffers larger than 8 bytes. The guards are initialized
11154 when a function is entered and then checked when the function exits.
11155 If a guard check fails, an error message is printed and the program exits.
11157 @item -fstack-protector-all
11158 @opindex fstack-protector-all
11159 Like @option{-fstack-protector} except that all functions are protected.
11161 @item -fstack-protector-strong
11162 @opindex fstack-protector-strong
11163 Like @option{-fstack-protector} but includes additional functions to
11164 be protected --- those that have local array definitions, or have
11165 references to local frame addresses.
11167 @item -fstack-protector-explicit
11168 @opindex fstack-protector-explicit
11169 Like @option{-fstack-protector} but only protects those functions which
11170 have the @code{stack_protect} attribute.
11172 @item -fstack-check
11173 @opindex fstack-check
11174 Generate code to verify that you do not go beyond the boundary of the
11175 stack. You should specify this flag if you are running in an
11176 environment with multiple threads, but you only rarely need to specify it in
11177 a single-threaded environment since stack overflow is automatically
11178 detected on nearly all systems if there is only one stack.
11180 Note that this switch does not actually cause checking to be done; the
11181 operating system or the language runtime must do that. The switch causes
11182 generation of code to ensure that they see the stack being extended.
11184 You can additionally specify a string parameter: @samp{no} means no
11185 checking, @samp{generic} means force the use of old-style checking,
11186 @samp{specific} means use the best checking method and is equivalent
11187 to bare @option{-fstack-check}.
11189 Old-style checking is a generic mechanism that requires no specific
11190 target support in the compiler but comes with the following drawbacks:
11194 Modified allocation strategy for large objects: they are always
11195 allocated dynamically if their size exceeds a fixed threshold.
11198 Fixed limit on the size of the static frame of functions: when it is
11199 topped by a particular function, stack checking is not reliable and
11200 a warning is issued by the compiler.
11203 Inefficiency: because of both the modified allocation strategy and the
11204 generic implementation, code performance is hampered.
11207 Note that old-style stack checking is also the fallback method for
11208 @samp{specific} if no target support has been added in the compiler.
11210 @item -fstack-limit-register=@var{reg}
11211 @itemx -fstack-limit-symbol=@var{sym}
11212 @itemx -fno-stack-limit
11213 @opindex fstack-limit-register
11214 @opindex fstack-limit-symbol
11215 @opindex fno-stack-limit
11216 Generate code to ensure that the stack does not grow beyond a certain value,
11217 either the value of a register or the address of a symbol. If a larger
11218 stack is required, a signal is raised at run time. For most targets,
11219 the signal is raised before the stack overruns the boundary, so
11220 it is possible to catch the signal without taking special precautions.
11222 For instance, if the stack starts at absolute address @samp{0x80000000}
11223 and grows downwards, you can use the flags
11224 @option{-fstack-limit-symbol=__stack_limit} and
11225 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11226 of 128KB@. Note that this may only work with the GNU linker.
11228 You can locally override stack limit checking by using the
11229 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
11231 @item -fsplit-stack
11232 @opindex fsplit-stack
11233 Generate code to automatically split the stack before it overflows.
11234 The resulting program has a discontiguous stack which can only
11235 overflow if the program is unable to allocate any more memory. This
11236 is most useful when running threaded programs, as it is no longer
11237 necessary to calculate a good stack size to use for each thread. This
11238 is currently only implemented for the x86 targets running
11241 When code compiled with @option{-fsplit-stack} calls code compiled
11242 without @option{-fsplit-stack}, there may not be much stack space
11243 available for the latter code to run. If compiling all code,
11244 including library code, with @option{-fsplit-stack} is not an option,
11245 then the linker can fix up these calls so that the code compiled
11246 without @option{-fsplit-stack} always has a large stack. Support for
11247 this is implemented in the gold linker in GNU binutils release 2.21
11250 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
11251 @opindex fvtable-verify
11252 This option is only available when compiling C++ code.
11253 It turns on (or off, if using @option{-fvtable-verify=none}) the security
11254 feature that verifies at run time, for every virtual call, that
11255 the vtable pointer through which the call is made is valid for the type of
11256 the object, and has not been corrupted or overwritten. If an invalid vtable
11257 pointer is detected at run time, an error is reported and execution of the
11258 program is immediately halted.
11260 This option causes run-time data structures to be built at program startup,
11261 which are used for verifying the vtable pointers.
11262 The options @samp{std} and @samp{preinit}
11263 control the timing of when these data structures are built. In both cases the
11264 data structures are built before execution reaches @code{main}. Using
11265 @option{-fvtable-verify=std} causes the data structures to be built after
11266 shared libraries have been loaded and initialized.
11267 @option{-fvtable-verify=preinit} causes them to be built before shared
11268 libraries have been loaded and initialized.
11270 If this option appears multiple times in the command line with different
11271 values specified, @samp{none} takes highest priority over both @samp{std} and
11272 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
11275 @opindex fvtv-debug
11276 When used in conjunction with @option{-fvtable-verify=std} or
11277 @option{-fvtable-verify=preinit}, causes debug versions of the
11278 runtime functions for the vtable verification feature to be called.
11279 This flag also causes the compiler to log information about which
11280 vtable pointers it finds for each class.
11281 This information is written to a file named @file{vtv_set_ptr_data.log}
11282 in the directory named by the environment variable @env{VTV_LOGS_DIR}
11283 if that is defined or the current working directory otherwise.
11285 Note: This feature @emph{appends} data to the log file. If you want a fresh log
11286 file, be sure to delete any existing one.
11289 @opindex fvtv-counts
11290 This is a debugging flag. When used in conjunction with
11291 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
11292 causes the compiler to keep track of the total number of virtual calls
11293 it encounters and the number of verifications it inserts. It also
11294 counts the number of calls to certain run-time library functions
11295 that it inserts and logs this information for each compilation unit.
11296 The compiler writes this information to a file named
11297 @file{vtv_count_data.log} in the directory named by the environment
11298 variable @env{VTV_LOGS_DIR} if that is defined or the current working
11299 directory otherwise. It also counts the size of the vtable pointer sets
11300 for each class, and writes this information to @file{vtv_class_set_sizes.log}
11301 in the same directory.
11303 Note: This feature @emph{appends} data to the log files. To get fresh log
11304 files, be sure to delete any existing ones.
11306 @item -finstrument-functions
11307 @opindex finstrument-functions
11308 Generate instrumentation calls for entry and exit to functions. Just
11309 after function entry and just before function exit, the following
11310 profiling functions are called with the address of the current
11311 function and its call site. (On some platforms,
11312 @code{__builtin_return_address} does not work beyond the current
11313 function, so the call site information may not be available to the
11314 profiling functions otherwise.)
11317 void __cyg_profile_func_enter (void *this_fn,
11319 void __cyg_profile_func_exit (void *this_fn,
11323 The first argument is the address of the start of the current function,
11324 which may be looked up exactly in the symbol table.
11326 This instrumentation is also done for functions expanded inline in other
11327 functions. The profiling calls indicate where, conceptually, the
11328 inline function is entered and exited. This means that addressable
11329 versions of such functions must be available. If all your uses of a
11330 function are expanded inline, this may mean an additional expansion of
11331 code size. If you use @code{extern inline} in your C code, an
11332 addressable version of such functions must be provided. (This is
11333 normally the case anyway, but if you get lucky and the optimizer always
11334 expands the functions inline, you might have gotten away without
11335 providing static copies.)
11337 A function may be given the attribute @code{no_instrument_function}, in
11338 which case this instrumentation is not done. This can be used, for
11339 example, for the profiling functions listed above, high-priority
11340 interrupt routines, and any functions from which the profiling functions
11341 cannot safely be called (perhaps signal handlers, if the profiling
11342 routines generate output or allocate memory).
11344 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
11345 @opindex finstrument-functions-exclude-file-list
11347 Set the list of functions that are excluded from instrumentation (see
11348 the description of @option{-finstrument-functions}). If the file that
11349 contains a function definition matches with one of @var{file}, then
11350 that function is not instrumented. The match is done on substrings:
11351 if the @var{file} parameter is a substring of the file name, it is
11352 considered to be a match.
11357 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
11361 excludes any inline function defined in files whose pathnames
11362 contain @file{/bits/stl} or @file{include/sys}.
11364 If, for some reason, you want to include letter @samp{,} in one of
11365 @var{sym}, write @samp{\,}. For example,
11366 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
11367 (note the single quote surrounding the option).
11369 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
11370 @opindex finstrument-functions-exclude-function-list
11372 This is similar to @option{-finstrument-functions-exclude-file-list},
11373 but this option sets the list of function names to be excluded from
11374 instrumentation. The function name to be matched is its user-visible
11375 name, such as @code{vector<int> blah(const vector<int> &)}, not the
11376 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
11377 match is done on substrings: if the @var{sym} parameter is a substring
11378 of the function name, it is considered to be a match. For C99 and C++
11379 extended identifiers, the function name must be given in UTF-8, not
11380 using universal character names.
11385 @node Preprocessor Options
11386 @section Options Controlling the Preprocessor
11387 @cindex preprocessor options
11388 @cindex options, preprocessor
11390 These options control the C preprocessor, which is run on each C source
11391 file before actual compilation.
11393 If you use the @option{-E} option, nothing is done except preprocessing.
11394 Some of these options make sense only together with @option{-E} because
11395 they cause the preprocessor output to be unsuitable for actual
11398 In addition to the options listed here, there are a number of options
11399 to control search paths for include files documented in
11400 @ref{Directory Options}.
11401 Options to control preprocessor diagnostics are listed in
11402 @ref{Warning Options}.
11405 @include cppopts.texi
11407 @item -Wp,@var{option}
11409 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11410 and pass @var{option} directly through to the preprocessor. If
11411 @var{option} contains commas, it is split into multiple options at the
11412 commas. However, many options are modified, translated or interpreted
11413 by the compiler driver before being passed to the preprocessor, and
11414 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11415 interface is undocumented and subject to change, so whenever possible
11416 you should avoid using @option{-Wp} and let the driver handle the
11419 @item -Xpreprocessor @var{option}
11420 @opindex Xpreprocessor
11421 Pass @var{option} as an option to the preprocessor. You can use this to
11422 supply system-specific preprocessor options that GCC does not
11425 If you want to pass an option that takes an argument, you must use
11426 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11428 @item -no-integrated-cpp
11429 @opindex no-integrated-cpp
11430 Perform preprocessing as a separate pass before compilation.
11431 By default, GCC performs preprocessing as an integrated part of
11432 input tokenization and parsing.
11433 If this option is provided, the appropriate language front end
11434 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11435 and Objective-C, respectively) is instead invoked twice,
11436 once for preprocessing only and once for actual compilation
11437 of the preprocessed input.
11438 This option may be useful in conjunction with the @option{-B} or
11439 @option{-wrapper} options to specify an alternate preprocessor or
11440 perform additional processing of the program source between
11441 normal preprocessing and compilation.
11445 @node Assembler Options
11446 @section Passing Options to the Assembler
11448 @c prevent bad page break with this line
11449 You can pass options to the assembler.
11452 @item -Wa,@var{option}
11454 Pass @var{option} as an option to the assembler. If @var{option}
11455 contains commas, it is split into multiple options at the commas.
11457 @item -Xassembler @var{option}
11458 @opindex Xassembler
11459 Pass @var{option} as an option to the assembler. You can use this to
11460 supply system-specific assembler options that GCC does not
11463 If you want to pass an option that takes an argument, you must use
11464 @option{-Xassembler} twice, once for the option and once for the argument.
11469 @section Options for Linking
11470 @cindex link options
11471 @cindex options, linking
11473 These options come into play when the compiler links object files into
11474 an executable output file. They are meaningless if the compiler is
11475 not doing a link step.
11479 @item @var{object-file-name}
11480 A file name that does not end in a special recognized suffix is
11481 considered to name an object file or library. (Object files are
11482 distinguished from libraries by the linker according to the file
11483 contents.) If linking is done, these object files are used as input
11492 If any of these options is used, then the linker is not run, and
11493 object file names should not be used as arguments. @xref{Overall
11497 @opindex fuse-ld=bfd
11498 Use the @command{bfd} linker instead of the default linker.
11500 @item -fuse-ld=gold
11501 @opindex fuse-ld=gold
11502 Use the @command{gold} linker instead of the default linker.
11505 @item -l@var{library}
11506 @itemx -l @var{library}
11508 Search the library named @var{library} when linking. (The second
11509 alternative with the library as a separate argument is only for
11510 POSIX compliance and is not recommended.)
11512 It makes a difference where in the command you write this option; the
11513 linker searches and processes libraries and object files in the order they
11514 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11515 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11516 to functions in @samp{z}, those functions may not be loaded.
11518 The linker searches a standard list of directories for the library,
11519 which is actually a file named @file{lib@var{library}.a}. The linker
11520 then uses this file as if it had been specified precisely by name.
11522 The directories searched include several standard system directories
11523 plus any that you specify with @option{-L}.
11525 Normally the files found this way are library files---archive files
11526 whose members are object files. The linker handles an archive file by
11527 scanning through it for members which define symbols that have so far
11528 been referenced but not defined. But if the file that is found is an
11529 ordinary object file, it is linked in the usual fashion. The only
11530 difference between using an @option{-l} option and specifying a file name
11531 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11532 and searches several directories.
11536 You need this special case of the @option{-l} option in order to
11537 link an Objective-C or Objective-C++ program.
11539 @item -nostartfiles
11540 @opindex nostartfiles
11541 Do not use the standard system startup files when linking.
11542 The standard system libraries are used normally, unless @option{-nostdlib}
11543 or @option{-nodefaultlibs} is used.
11545 @item -nodefaultlibs
11546 @opindex nodefaultlibs
11547 Do not use the standard system libraries when linking.
11548 Only the libraries you specify are passed to the linker, and options
11549 specifying linkage of the system libraries, such as @option{-static-libgcc}
11550 or @option{-shared-libgcc}, are ignored.
11551 The standard startup files are used normally, unless @option{-nostartfiles}
11554 The compiler may generate calls to @code{memcmp},
11555 @code{memset}, @code{memcpy} and @code{memmove}.
11556 These entries are usually resolved by entries in
11557 libc. These entry points should be supplied through some other
11558 mechanism when this option is specified.
11562 Do not use the standard system startup files or libraries when linking.
11563 No startup files and only the libraries you specify are passed to
11564 the linker, and options specifying linkage of the system libraries, such as
11565 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11567 The compiler may generate calls to @code{memcmp}, @code{memset},
11568 @code{memcpy} and @code{memmove}.
11569 These entries are usually resolved by entries in
11570 libc. These entry points should be supplied through some other
11571 mechanism when this option is specified.
11573 @cindex @option{-lgcc}, use with @option{-nostdlib}
11574 @cindex @option{-nostdlib} and unresolved references
11575 @cindex unresolved references and @option{-nostdlib}
11576 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11577 @cindex @option{-nodefaultlibs} and unresolved references
11578 @cindex unresolved references and @option{-nodefaultlibs}
11579 One of the standard libraries bypassed by @option{-nostdlib} and
11580 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11581 which GCC uses to overcome shortcomings of particular machines, or special
11582 needs for some languages.
11583 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11584 Collection (GCC) Internals},
11585 for more discussion of @file{libgcc.a}.)
11586 In most cases, you need @file{libgcc.a} even when you want to avoid
11587 other standard libraries. In other words, when you specify @option{-nostdlib}
11588 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11589 This ensures that you have no unresolved references to internal GCC
11590 library subroutines.
11591 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11592 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11593 GNU Compiler Collection (GCC) Internals}.)
11597 Produce a position independent executable on targets that support it.
11598 For predictable results, you must also specify the same set of options
11599 used for compilation (@option{-fpie}, @option{-fPIE},
11600 or model suboptions) when you specify this linker option.
11604 Don't produce a position independent executable.
11608 Link with the POSIX threads library. This option is supported on
11609 GNU/Linux targets, most other Unix derivatives, and also on
11610 x86 Cygwin and MinGW targets. On some targets this option also sets
11611 flags for the preprocessor, so it should be used consistently for both
11612 compilation and linking.
11616 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11617 that support it. This instructs the linker to add all symbols, not
11618 only used ones, to the dynamic symbol table. This option is needed
11619 for some uses of @code{dlopen} or to allow obtaining backtraces
11620 from within a program.
11624 Remove all symbol table and relocation information from the executable.
11628 On systems that support dynamic linking, this prevents linking with the shared
11629 libraries. On other systems, this option has no effect.
11633 Produce a shared object which can then be linked with other objects to
11634 form an executable. Not all systems support this option. For predictable
11635 results, you must also specify the same set of options used for compilation
11636 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11637 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11638 needs to build supplementary stub code for constructors to work. On
11639 multi-libbed systems, @samp{gcc -shared} must select the correct support
11640 libraries to link against. Failing to supply the correct flags may lead
11641 to subtle defects. Supplying them in cases where they are not necessary
11644 @item -shared-libgcc
11645 @itemx -static-libgcc
11646 @opindex shared-libgcc
11647 @opindex static-libgcc
11648 On systems that provide @file{libgcc} as a shared library, these options
11649 force the use of either the shared or static version, respectively.
11650 If no shared version of @file{libgcc} was built when the compiler was
11651 configured, these options have no effect.
11653 There are several situations in which an application should use the
11654 shared @file{libgcc} instead of the static version. The most common
11655 of these is when the application wishes to throw and catch exceptions
11656 across different shared libraries. In that case, each of the libraries
11657 as well as the application itself should use the shared @file{libgcc}.
11659 Therefore, the G++ and driver automatically adds @option{-shared-libgcc}
11660 whenever you build a shared library or a main executable, because C++
11661 programs typically use exceptions, so this is the right thing to do.
11663 If, instead, you use the GCC driver to create shared libraries, you may
11664 find that they are not always linked with the shared @file{libgcc}.
11665 If GCC finds, at its configuration time, that you have a non-GNU linker
11666 or a GNU linker that does not support option @option{--eh-frame-hdr},
11667 it links the shared version of @file{libgcc} into shared libraries
11668 by default. Otherwise, it takes advantage of the linker and optimizes
11669 away the linking with the shared version of @file{libgcc}, linking with
11670 the static version of libgcc by default. This allows exceptions to
11671 propagate through such shared libraries, without incurring relocation
11672 costs at library load time.
11674 However, if a library or main executable is supposed to throw or catch
11675 exceptions, you must link it using the G++ driver, as appropriate
11676 for the languages used in the program, or using the option
11677 @option{-shared-libgcc}, such that it is linked with the shared
11680 @item -static-libasan
11681 @opindex static-libasan
11682 When the @option{-fsanitize=address} option is used to link a program,
11683 the GCC driver automatically links against @option{libasan}. If
11684 @file{libasan} is available as a shared library, and the @option{-static}
11685 option is not used, then this links against the shared version of
11686 @file{libasan}. The @option{-static-libasan} option directs the GCC
11687 driver to link @file{libasan} statically, without necessarily linking
11688 other libraries statically.
11690 @item -static-libtsan
11691 @opindex static-libtsan
11692 When the @option{-fsanitize=thread} option is used to link a program,
11693 the GCC driver automatically links against @option{libtsan}. If
11694 @file{libtsan} is available as a shared library, and the @option{-static}
11695 option is not used, then this links against the shared version of
11696 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11697 driver to link @file{libtsan} statically, without necessarily linking
11698 other libraries statically.
11700 @item -static-liblsan
11701 @opindex static-liblsan
11702 When the @option{-fsanitize=leak} option is used to link a program,
11703 the GCC driver automatically links against @option{liblsan}. If
11704 @file{liblsan} is available as a shared library, and the @option{-static}
11705 option is not used, then this links against the shared version of
11706 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11707 driver to link @file{liblsan} statically, without necessarily linking
11708 other libraries statically.
11710 @item -static-libubsan
11711 @opindex static-libubsan
11712 When the @option{-fsanitize=undefined} option is used to link a program,
11713 the GCC driver automatically links against @option{libubsan}. If
11714 @file{libubsan} is available as a shared library, and the @option{-static}
11715 option is not used, then this links against the shared version of
11716 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11717 driver to link @file{libubsan} statically, without necessarily linking
11718 other libraries statically.
11720 @item -static-libmpx
11721 @opindex static-libmpx
11722 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11723 used to link a program, the GCC driver automatically links against
11724 @file{libmpx}. If @file{libmpx} is available as a shared library,
11725 and the @option{-static} option is not used, then this links against
11726 the shared version of @file{libmpx}. The @option{-static-libmpx}
11727 option directs the GCC driver to link @file{libmpx} statically,
11728 without necessarily linking other libraries statically.
11730 @item -static-libmpxwrappers
11731 @opindex static-libmpxwrappers
11732 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11733 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11734 GCC driver automatically links against @file{libmpxwrappers}. If
11735 @file{libmpxwrappers} is available as a shared library, and the
11736 @option{-static} option is not used, then this links against the shared
11737 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11738 option directs the GCC driver to link @file{libmpxwrappers} statically,
11739 without necessarily linking other libraries statically.
11741 @item -static-libstdc++
11742 @opindex static-libstdc++
11743 When the @command{g++} program is used to link a C++ program, it
11744 normally automatically links against @option{libstdc++}. If
11745 @file{libstdc++} is available as a shared library, and the
11746 @option{-static} option is not used, then this links against the
11747 shared version of @file{libstdc++}. That is normally fine. However, it
11748 is sometimes useful to freeze the version of @file{libstdc++} used by
11749 the program without going all the way to a fully static link. The
11750 @option{-static-libstdc++} option directs the @command{g++} driver to
11751 link @file{libstdc++} statically, without necessarily linking other
11752 libraries statically.
11756 Bind references to global symbols when building a shared object. Warn
11757 about any unresolved references (unless overridden by the link editor
11758 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11761 @item -T @var{script}
11763 @cindex linker script
11764 Use @var{script} as the linker script. This option is supported by most
11765 systems using the GNU linker. On some targets, such as bare-board
11766 targets without an operating system, the @option{-T} option may be required
11767 when linking to avoid references to undefined symbols.
11769 @item -Xlinker @var{option}
11771 Pass @var{option} as an option to the linker. You can use this to
11772 supply system-specific linker options that GCC does not recognize.
11774 If you want to pass an option that takes a separate argument, you must use
11775 @option{-Xlinker} twice, once for the option and once for the argument.
11776 For example, to pass @option{-assert definitions}, you must write
11777 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11778 @option{-Xlinker "-assert definitions"}, because this passes the entire
11779 string as a single argument, which is not what the linker expects.
11781 When using the GNU linker, it is usually more convenient to pass
11782 arguments to linker options using the @option{@var{option}=@var{value}}
11783 syntax than as separate arguments. For example, you can specify
11784 @option{-Xlinker -Map=output.map} rather than
11785 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11786 this syntax for command-line options.
11788 @item -Wl,@var{option}
11790 Pass @var{option} as an option to the linker. If @var{option} contains
11791 commas, it is split into multiple options at the commas. You can use this
11792 syntax to pass an argument to the option.
11793 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11794 linker. When using the GNU linker, you can also get the same effect with
11795 @option{-Wl,-Map=output.map}.
11797 @item -u @var{symbol}
11799 Pretend the symbol @var{symbol} is undefined, to force linking of
11800 library modules to define it. You can use @option{-u} multiple times with
11801 different symbols to force loading of additional library modules.
11803 @item -z @var{keyword}
11805 @option{-z} is passed directly on to the linker along with the keyword
11806 @var{keyword}. See the section in the documentation of your linker for
11807 permitted values and their meanings.
11810 @node Directory Options
11811 @section Options for Directory Search
11812 @cindex directory options
11813 @cindex options, directory search
11814 @cindex search path
11816 These options specify directories to search for header files, for
11817 libraries and for parts of the compiler:
11820 @include cppdiropts.texi
11822 @item -iplugindir=@var{dir}
11823 @opindex iplugindir=
11824 Set the directory to search for plugins that are passed
11825 by @option{-fplugin=@var{name}} instead of
11826 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11827 to be used by the user, but only passed by the driver.
11831 Add directory @var{dir} to the list of directories to be searched
11834 @item -B@var{prefix}
11836 This option specifies where to find the executables, libraries,
11837 include files, and data files of the compiler itself.
11839 The compiler driver program runs one or more of the subprograms
11840 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11841 @var{prefix} as a prefix for each program it tries to run, both with and
11842 without @samp{@var{machine}/@var{version}/} for the corresponding target
11843 machine and compiler version.
11845 For each subprogram to be run, the compiler driver first tries the
11846 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11847 is not specified, the driver tries two standard prefixes,
11848 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11849 those results in a file name that is found, the unmodified program
11850 name is searched for using the directories specified in your
11851 @env{PATH} environment variable.
11853 The compiler checks to see if the path provided by @option{-B}
11854 refers to a directory, and if necessary it adds a directory
11855 separator character at the end of the path.
11857 @option{-B} prefixes that effectively specify directory names also apply
11858 to libraries in the linker, because the compiler translates these
11859 options into @option{-L} options for the linker. They also apply to
11860 include files in the preprocessor, because the compiler translates these
11861 options into @option{-isystem} options for the preprocessor. In this case,
11862 the compiler appends @samp{include} to the prefix.
11864 The runtime support file @file{libgcc.a} can also be searched for using
11865 the @option{-B} prefix, if needed. If it is not found there, the two
11866 standard prefixes above are tried, and that is all. The file is left
11867 out of the link if it is not found by those means.
11869 Another way to specify a prefix much like the @option{-B} prefix is to use
11870 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11873 As a special kludge, if the path provided by @option{-B} is
11874 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11875 9, then it is replaced by @file{[dir/]include}. This is to help
11876 with boot-strapping the compiler.
11878 @item -no-canonical-prefixes
11879 @opindex no-canonical-prefixes
11880 Do not expand any symbolic links, resolve references to @samp{/../}
11881 or @samp{/./}, or make the path absolute when generating a relative
11884 @item --sysroot=@var{dir}
11886 Use @var{dir} as the logical root directory for headers and libraries.
11887 For example, if the compiler normally searches for headers in
11888 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11889 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11891 If you use both this option and the @option{-isysroot} option, then
11892 the @option{--sysroot} option applies to libraries, but the
11893 @option{-isysroot} option applies to header files.
11895 The GNU linker (beginning with version 2.16) has the necessary support
11896 for this option. If your linker does not support this option, the
11897 header file aspect of @option{--sysroot} still works, but the
11898 library aspect does not.
11900 @item --no-sysroot-suffix
11901 @opindex no-sysroot-suffix
11902 For some targets, a suffix is added to the root directory specified
11903 with @option{--sysroot}, depending on the other options used, so that
11904 headers may for example be found in
11905 @file{@var{dir}/@var{suffix}/usr/include} instead of
11906 @file{@var{dir}/usr/include}. This option disables the addition of
11911 @node Code Gen Options
11912 @section Options for Code Generation Conventions
11913 @cindex code generation conventions
11914 @cindex options, code generation
11915 @cindex run-time options
11917 These machine-independent options control the interface conventions
11918 used in code generation.
11920 Most of them have both positive and negative forms; the negative form
11921 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11922 one of the forms is listed---the one that is not the default. You
11923 can figure out the other form by either removing @samp{no-} or adding
11927 @item -fstack-reuse=@var{reuse-level}
11928 @opindex fstack_reuse
11929 This option controls stack space reuse for user declared local/auto variables
11930 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11931 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11932 local variables and temporaries, @samp{named_vars} enables the reuse only for
11933 user defined local variables with names, and @samp{none} disables stack reuse
11934 completely. The default value is @samp{all}. The option is needed when the
11935 program extends the lifetime of a scoped local variable or a compiler generated
11936 temporary beyond the end point defined by the language. When a lifetime of
11937 a variable ends, and if the variable lives in memory, the optimizing compiler
11938 has the freedom to reuse its stack space with other temporaries or scoped
11939 local variables whose live range does not overlap with it. Legacy code extending
11940 local lifetime is likely to break with the stack reuse optimization.
11959 if (*p == 10) // out of scope use of local1
11970 A(int k) : i(k), j(k) @{ @}
11977 void foo(const A& ar)
11984 foo(A(10)); // temp object's lifetime ends when foo returns
11990 ap->i+= 10; // ap references out of scope temp whose space
11991 // is reused with a. What is the value of ap->i?
11996 The lifetime of a compiler generated temporary is well defined by the C++
11997 standard. When a lifetime of a temporary ends, and if the temporary lives
11998 in memory, the optimizing compiler has the freedom to reuse its stack
11999 space with other temporaries or scoped local variables whose live range
12000 does not overlap with it. However some of the legacy code relies on
12001 the behavior of older compilers in which temporaries' stack space is
12002 not reused, the aggressive stack reuse can lead to runtime errors. This
12003 option is used to control the temporary stack reuse optimization.
12007 This option generates traps for signed overflow on addition, subtraction,
12008 multiplication operations.
12009 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12010 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12011 @option{-fwrapv} being effective. Note that only active options override, so
12012 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12013 results in @option{-ftrapv} being effective.
12017 This option instructs the compiler to assume that signed arithmetic
12018 overflow of addition, subtraction and multiplication wraps around
12019 using twos-complement representation. This flag enables some optimizations
12020 and disables others.
12021 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
12022 @option{-ftrapv} @option{-fwrapv} on the command-line results in
12023 @option{-fwrapv} being effective. Note that only active options override, so
12024 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
12025 results in @option{-ftrapv} being effective.
12028 @opindex fexceptions
12029 Enable exception handling. Generates extra code needed to propagate
12030 exceptions. For some targets, this implies GCC generates frame
12031 unwind information for all functions, which can produce significant data
12032 size overhead, although it does not affect execution. If you do not
12033 specify this option, GCC enables it by default for languages like
12034 C++ that normally require exception handling, and disables it for
12035 languages like C that do not normally require it. However, you may need
12036 to enable this option when compiling C code that needs to interoperate
12037 properly with exception handlers written in C++. You may also wish to
12038 disable this option if you are compiling older C++ programs that don't
12039 use exception handling.
12041 @item -fnon-call-exceptions
12042 @opindex fnon-call-exceptions
12043 Generate code that allows trapping instructions to throw exceptions.
12044 Note that this requires platform-specific runtime support that does
12045 not exist everywhere. Moreover, it only allows @emph{trapping}
12046 instructions to throw exceptions, i.e.@: memory references or floating-point
12047 instructions. It does not allow exceptions to be thrown from
12048 arbitrary signal handlers such as @code{SIGALRM}.
12050 @item -fdelete-dead-exceptions
12051 @opindex fdelete-dead-exceptions
12052 Consider that instructions that may throw exceptions but don't otherwise
12053 contribute to the execution of the program can be optimized away.
12054 This option is enabled by default for the Ada front end, as permitted by
12055 the Ada language specification.
12056 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
12058 @item -funwind-tables
12059 @opindex funwind-tables
12060 Similar to @option{-fexceptions}, except that it just generates any needed
12061 static data, but does not affect the generated code in any other way.
12062 You normally do not need to enable this option; instead, a language processor
12063 that needs this handling enables it on your behalf.
12065 @item -fasynchronous-unwind-tables
12066 @opindex fasynchronous-unwind-tables
12067 Generate unwind table in DWARF format, if supported by target machine. The
12068 table is exact at each instruction boundary, so it can be used for stack
12069 unwinding from asynchronous events (such as debugger or garbage collector).
12071 @item -fno-gnu-unique
12072 @opindex fno-gnu-unique
12073 On systems with recent GNU assembler and C library, the C++ compiler
12074 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
12075 of template static data members and static local variables in inline
12076 functions are unique even in the presence of @code{RTLD_LOCAL}; this
12077 is necessary to avoid problems with a library used by two different
12078 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
12079 therefore disagreeing with the other one about the binding of the
12080 symbol. But this causes @code{dlclose} to be ignored for affected
12081 DSOs; if your program relies on reinitialization of a DSO via
12082 @code{dlclose} and @code{dlopen}, you can use
12083 @option{-fno-gnu-unique}.
12085 @item -fpcc-struct-return
12086 @opindex fpcc-struct-return
12087 Return ``short'' @code{struct} and @code{union} values in memory like
12088 longer ones, rather than in registers. This convention is less
12089 efficient, but it has the advantage of allowing intercallability between
12090 GCC-compiled files and files compiled with other compilers, particularly
12091 the Portable C Compiler (pcc).
12093 The precise convention for returning structures in memory depends
12094 on the target configuration macros.
12096 Short structures and unions are those whose size and alignment match
12097 that of some integer type.
12099 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12100 switch is not binary compatible with code compiled with the
12101 @option{-freg-struct-return} switch.
12102 Use it to conform to a non-default application binary interface.
12104 @item -freg-struct-return
12105 @opindex freg-struct-return
12106 Return @code{struct} and @code{union} values in registers when possible.
12107 This is more efficient for small structures than
12108 @option{-fpcc-struct-return}.
12110 If you specify neither @option{-fpcc-struct-return} nor
12111 @option{-freg-struct-return}, GCC defaults to whichever convention is
12112 standard for the target. If there is no standard convention, GCC
12113 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12114 the principal compiler. In those cases, we can choose the standard, and
12115 we chose the more efficient register return alternative.
12117 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12118 switch is not binary compatible with code compiled with the
12119 @option{-fpcc-struct-return} switch.
12120 Use it to conform to a non-default application binary interface.
12122 @item -fshort-enums
12123 @opindex fshort-enums
12124 Allocate to an @code{enum} type only as many bytes as it needs for the
12125 declared range of possible values. Specifically, the @code{enum} type
12126 is equivalent to the smallest integer type that has enough room.
12128 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12129 code that is not binary compatible with code generated without that switch.
12130 Use it to conform to a non-default application binary interface.
12132 @item -fshort-wchar
12133 @opindex fshort-wchar
12134 Override the underlying type for @code{wchar_t} to be @code{short
12135 unsigned int} instead of the default for the target. This option is
12136 useful for building programs to run under WINE@.
12138 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12139 code that is not binary compatible with code generated without that switch.
12140 Use it to conform to a non-default application binary interface.
12143 @opindex fno-common
12144 @cindex tentative definitions
12145 In C code, this option controls the placement of global variables
12146 defined without an initializer, known as @dfn{tentative definitions}
12147 in the C standard. Tentative definitions are distinct from declarations
12148 of a variable with the @code{extern} keyword, which do not allocate storage.
12150 Unix C compilers have traditionally allocated storage for
12151 uninitialized global variables in a common block. This allows the
12152 linker to resolve all tentative definitions of the same variable
12153 in different compilation units to the same object, or to a non-tentative
12155 This is the behavior specified by @option{-fcommon}, and is the default for
12156 GCC on most targets.
12157 On the other hand, this behavior is not required by ISO
12158 C, and on some targets may carry a speed or code size penalty on
12159 variable references.
12161 The @option{-fno-common} option specifies that the compiler should instead
12162 place uninitialized global variables in the data section of the object file.
12163 This inhibits the merging of tentative definitions by the linker so
12164 you get a multiple-definition error if the same
12165 variable is defined in more than one compilation unit.
12166 Compiling with @option{-fno-common} is useful on targets for which
12167 it provides better performance, or if you wish to verify that the
12168 program will work on other systems that always treat uninitialized
12169 variable definitions this way.
12173 Ignore the @code{#ident} directive.
12175 @item -finhibit-size-directive
12176 @opindex finhibit-size-directive
12177 Don't output a @code{.size} assembler directive, or anything else that
12178 would cause trouble if the function is split in the middle, and the
12179 two halves are placed at locations far apart in memory. This option is
12180 used when compiling @file{crtstuff.c}; you should not need to use it
12183 @item -fverbose-asm
12184 @opindex fverbose-asm
12185 Put extra commentary information in the generated assembly code to
12186 make it more readable. This option is generally only of use to those
12187 who actually need to read the generated assembly code (perhaps while
12188 debugging the compiler itself).
12190 @option{-fno-verbose-asm}, the default, causes the
12191 extra information to be omitted and is useful when comparing two assembler
12194 The added comments include:
12199 information on the compiler version and command-line options,
12202 the source code lines associated with the assembly instructions,
12203 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
12206 hints on which high-level expressions correspond to
12207 the various assembly instruction operands.
12211 For example, given this C source file:
12219 for (i = 0; i < n; i++)
12226 compiling to (x86_64) assembly via @option{-S} and emitting the result
12227 direct to stdout via @option{-o} @option{-}
12230 gcc -S test.c -fverbose-asm -Os -o -
12233 gives output similar to this:
12237 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
12244 .type test, @@function
12248 # test.c:4: int total = 0;
12249 xorl %eax, %eax # <retval>
12250 # test.c:6: for (i = 0; i < n; i++)
12251 xorl %edx, %edx # i
12253 # test.c:6: for (i = 0; i < n; i++)
12254 cmpl %edi, %edx # n, i
12256 # test.c:7: total += i * i;
12257 movl %edx, %ecx # i, tmp92
12258 imull %edx, %ecx # i, tmp92
12259 # test.c:6: for (i = 0; i < n; i++)
12261 # test.c:7: total += i * i;
12262 addl %ecx, %eax # tmp92, <retval>
12270 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
12271 .section .note.GNU-stack,"",@@progbits
12274 The comments are intended for humans rather than machines and hence the
12275 precise format of the comments is subject to change.
12277 @item -frecord-gcc-switches
12278 @opindex frecord-gcc-switches
12279 This switch causes the command line used to invoke the
12280 compiler to be recorded into the object file that is being created.
12281 This switch is only implemented on some targets and the exact format
12282 of the recording is target and binary file format dependent, but it
12283 usually takes the form of a section containing ASCII text. This
12284 switch is related to the @option{-fverbose-asm} switch, but that
12285 switch only records information in the assembler output file as
12286 comments, so it never reaches the object file.
12287 See also @option{-grecord-gcc-switches} for another
12288 way of storing compiler options into the object file.
12292 @cindex global offset table
12294 Generate position-independent code (PIC) suitable for use in a shared
12295 library, if supported for the target machine. Such code accesses all
12296 constant addresses through a global offset table (GOT)@. The dynamic
12297 loader resolves the GOT entries when the program starts (the dynamic
12298 loader is not part of GCC; it is part of the operating system). If
12299 the GOT size for the linked executable exceeds a machine-specific
12300 maximum size, you get an error message from the linker indicating that
12301 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12302 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
12303 on the m68k and RS/6000. The x86 has no such limit.)
12305 Position-independent code requires special support, and therefore works
12306 only on certain machines. For the x86, GCC supports PIC for System V
12307 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12308 position-independent.
12310 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12315 If supported for the target machine, emit position-independent code,
12316 suitable for dynamic linking and avoiding any limit on the size of the
12317 global offset table. This option makes a difference on AArch64, m68k,
12318 PowerPC and SPARC@.
12320 Position-independent code requires special support, and therefore works
12321 only on certain machines.
12323 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12330 These options are similar to @option{-fpic} and @option{-fPIC}, but
12331 generated position independent code can be only linked into executables.
12332 Usually these options are used when @option{-pie} GCC option is
12333 used during linking.
12335 @option{-fpie} and @option{-fPIE} both define the macros
12336 @code{__pie__} and @code{__PIE__}. The macros have the value 1
12337 for @option{-fpie} and 2 for @option{-fPIE}.
12341 Do not use the PLT for external function calls in position-independent code.
12342 Instead, load the callee address at call sites from the GOT and branch to it.
12343 This leads to more efficient code by eliminating PLT stubs and exposing
12344 GOT loads to optimizations. On architectures such as 32-bit x86 where
12345 PLT stubs expect the GOT pointer in a specific register, this gives more
12346 register allocation freedom to the compiler.
12347 Lazy binding requires use of the PLT;
12348 with @option{-fno-plt} all external symbols are resolved at load time.
12350 Alternatively, the function attribute @code{noplt} can be used to avoid calls
12351 through the PLT for specific external functions.
12353 In position-dependent code, a few targets also convert calls to
12354 functions that are marked to not use the PLT to use the GOT instead.
12356 @item -fno-jump-tables
12357 @opindex fno-jump-tables
12358 Do not use jump tables for switch statements even where it would be
12359 more efficient than other code generation strategies. This option is
12360 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12361 building code that forms part of a dynamic linker and cannot
12362 reference the address of a jump table. On some targets, jump tables
12363 do not require a GOT and this option is not needed.
12365 @item -ffixed-@var{reg}
12367 Treat the register named @var{reg} as a fixed register; generated code
12368 should never refer to it (except perhaps as a stack pointer, frame
12369 pointer or in some other fixed role).
12371 @var{reg} must be the name of a register. The register names accepted
12372 are machine-specific and are defined in the @code{REGISTER_NAMES}
12373 macro in the machine description macro file.
12375 This flag does not have a negative form, because it specifies a
12378 @item -fcall-used-@var{reg}
12379 @opindex fcall-used
12380 Treat the register named @var{reg} as an allocable register that is
12381 clobbered by function calls. It may be allocated for temporaries or
12382 variables that do not live across a call. Functions compiled this way
12383 do not save and restore the register @var{reg}.
12385 It is an error to use this flag with the frame pointer or stack pointer.
12386 Use of this flag for other registers that have fixed pervasive roles in
12387 the machine's execution model produces disastrous results.
12389 This flag does not have a negative form, because it specifies a
12392 @item -fcall-saved-@var{reg}
12393 @opindex fcall-saved
12394 Treat the register named @var{reg} as an allocable register saved by
12395 functions. It may be allocated even for temporaries or variables that
12396 live across a call. Functions compiled this way save and restore
12397 the register @var{reg} if they use it.
12399 It is an error to use this flag with the frame pointer or stack pointer.
12400 Use of this flag for other registers that have fixed pervasive roles in
12401 the machine's execution model produces disastrous results.
12403 A different sort of disaster results from the use of this flag for
12404 a register in which function values may be returned.
12406 This flag does not have a negative form, because it specifies a
12409 @item -fpack-struct[=@var{n}]
12410 @opindex fpack-struct
12411 Without a value specified, pack all structure members together without
12412 holes. When a value is specified (which must be a small power of two), pack
12413 structure members according to this value, representing the maximum
12414 alignment (that is, objects with default alignment requirements larger than
12415 this are output potentially unaligned at the next fitting location.
12417 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12418 code that is not binary compatible with code generated without that switch.
12419 Additionally, it makes the code suboptimal.
12420 Use it to conform to a non-default application binary interface.
12422 @item -fleading-underscore
12423 @opindex fleading-underscore
12424 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12425 change the way C symbols are represented in the object file. One use
12426 is to help link with legacy assembly code.
12428 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12429 generate code that is not binary compatible with code generated without that
12430 switch. Use it to conform to a non-default application binary interface.
12431 Not all targets provide complete support for this switch.
12433 @item -ftls-model=@var{model}
12434 @opindex ftls-model
12435 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12436 The @var{model} argument should be one of @samp{global-dynamic},
12437 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
12438 Note that the choice is subject to optimization: the compiler may use
12439 a more efficient model for symbols not visible outside of the translation
12440 unit, or if @option{-fpic} is not given on the command line.
12442 The default without @option{-fpic} is @samp{initial-exec}; with
12443 @option{-fpic} the default is @samp{global-dynamic}.
12445 @item -ftrampolines
12446 @opindex ftrampolines
12447 For targets that normally need trampolines for nested functions, always
12448 generate them instead of using descriptors. Otherwise, for targets that
12449 do not need them, like for example HP-PA or IA-64, do nothing.
12451 A trampoline is a small piece of code that is created at run time on the
12452 stack when the address of a nested function is taken, and is used to call
12453 the nested function indirectly. Therefore, it requires the stack to be
12454 made executable in order for the program to work properly.
12456 @option{-fno-trampolines} is enabled by default on a language by language
12457 basis to let the compiler avoid generating them, if it computes that this
12458 is safe, and replace them with descriptors. Descriptors are made up of data
12459 only, but the generated code must be prepared to deal with them. As of this
12460 writing, @option{-fno-trampolines} is enabled by default only for Ada.
12462 Moreover, code compiled with @option{-ftrampolines} and code compiled with
12463 @option{-fno-trampolines} are not binary compatible if nested functions are
12464 present. This option must therefore be used on a program-wide basis and be
12465 manipulated with extreme care.
12467 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
12468 @opindex fvisibility
12469 Set the default ELF image symbol visibility to the specified option---all
12470 symbols are marked with this unless overridden within the code.
12471 Using this feature can very substantially improve linking and
12472 load times of shared object libraries, produce more optimized
12473 code, provide near-perfect API export and prevent symbol clashes.
12474 It is @strong{strongly} recommended that you use this in any shared objects
12477 Despite the nomenclature, @samp{default} always means public; i.e.,
12478 available to be linked against from outside the shared object.
12479 @samp{protected} and @samp{internal} are pretty useless in real-world
12480 usage so the only other commonly used option is @samp{hidden}.
12481 The default if @option{-fvisibility} isn't specified is
12482 @samp{default}, i.e., make every symbol public.
12484 A good explanation of the benefits offered by ensuring ELF
12485 symbols have the correct visibility is given by ``How To Write
12486 Shared Libraries'' by Ulrich Drepper (which can be found at
12487 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
12488 solution made possible by this option to marking things hidden when
12489 the default is public is to make the default hidden and mark things
12490 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
12491 and @code{__attribute__ ((visibility("default")))} instead of
12492 @code{__declspec(dllexport)} you get almost identical semantics with
12493 identical syntax. This is a great boon to those working with
12494 cross-platform projects.
12496 For those adding visibility support to existing code, you may find
12497 @code{#pragma GCC visibility} of use. This works by you enclosing
12498 the declarations you wish to set visibility for with (for example)
12499 @code{#pragma GCC visibility push(hidden)} and
12500 @code{#pragma GCC visibility pop}.
12501 Bear in mind that symbol visibility should be viewed @strong{as
12502 part of the API interface contract} and thus all new code should
12503 always specify visibility when it is not the default; i.e., declarations
12504 only for use within the local DSO should @strong{always} be marked explicitly
12505 as hidden as so to avoid PLT indirection overheads---making this
12506 abundantly clear also aids readability and self-documentation of the code.
12507 Note that due to ISO C++ specification requirements, @code{operator new} and
12508 @code{operator delete} must always be of default visibility.
12510 Be aware that headers from outside your project, in particular system
12511 headers and headers from any other library you use, may not be
12512 expecting to be compiled with visibility other than the default. You
12513 may need to explicitly say @code{#pragma GCC visibility push(default)}
12514 before including any such headers.
12516 @code{extern} declarations are not affected by @option{-fvisibility}, so
12517 a lot of code can be recompiled with @option{-fvisibility=hidden} with
12518 no modifications. However, this means that calls to @code{extern}
12519 functions with no explicit visibility use the PLT, so it is more
12520 effective to use @code{__attribute ((visibility))} and/or
12521 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
12522 declarations should be treated as hidden.
12524 Note that @option{-fvisibility} does affect C++ vague linkage
12525 entities. This means that, for instance, an exception class that is
12526 be thrown between DSOs must be explicitly marked with default
12527 visibility so that the @samp{type_info} nodes are unified between
12530 An overview of these techniques, their benefits and how to use them
12531 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
12533 @item -fstrict-volatile-bitfields
12534 @opindex fstrict-volatile-bitfields
12535 This option should be used if accesses to volatile bit-fields (or other
12536 structure fields, although the compiler usually honors those types
12537 anyway) should use a single access of the width of the
12538 field's type, aligned to a natural alignment if possible. For
12539 example, targets with memory-mapped peripheral registers might require
12540 all such accesses to be 16 bits wide; with this flag you can
12541 declare all peripheral bit-fields as @code{unsigned short} (assuming short
12542 is 16 bits on these targets) to force GCC to use 16-bit accesses
12543 instead of, perhaps, a more efficient 32-bit access.
12545 If this option is disabled, the compiler uses the most efficient
12546 instruction. In the previous example, that might be a 32-bit load
12547 instruction, even though that accesses bytes that do not contain
12548 any portion of the bit-field, or memory-mapped registers unrelated to
12549 the one being updated.
12551 In some cases, such as when the @code{packed} attribute is applied to a
12552 structure field, it may not be possible to access the field with a single
12553 read or write that is correctly aligned for the target machine. In this
12554 case GCC falls back to generating multiple accesses rather than code that
12555 will fault or truncate the result at run time.
12557 Note: Due to restrictions of the C/C++11 memory model, write accesses are
12558 not allowed to touch non bit-field members. It is therefore recommended
12559 to define all bits of the field's type as bit-field members.
12561 The default value of this option is determined by the application binary
12562 interface for the target processor.
12564 @item -fsync-libcalls
12565 @opindex fsync-libcalls
12566 This option controls whether any out-of-line instance of the @code{__sync}
12567 family of functions may be used to implement the C++11 @code{__atomic}
12568 family of functions.
12570 The default value of this option is enabled, thus the only useful form
12571 of the option is @option{-fno-sync-libcalls}. This option is used in
12572 the implementation of the @file{libatomic} runtime library.
12576 @node Developer Options
12577 @section GCC Developer Options
12578 @cindex developer options
12579 @cindex debugging GCC
12580 @cindex debug dump options
12581 @cindex dump options
12582 @cindex compilation statistics
12584 This section describes command-line options that are primarily of
12585 interest to GCC developers, including options to support compiler
12586 testing and investigation of compiler bugs and compile-time
12587 performance problems. This includes options that produce debug dumps
12588 at various points in the compilation; that print statistics such as
12589 memory use and execution time; and that print information about GCC's
12590 configuration, such as where it searches for libraries. You should
12591 rarely need to use any of these options for ordinary compilation and
12596 @item -d@var{letters}
12597 @itemx -fdump-rtl-@var{pass}
12598 @itemx -fdump-rtl-@var{pass}=@var{filename}
12600 @opindex fdump-rtl-@var{pass}
12601 Says to make debugging dumps during compilation at times specified by
12602 @var{letters}. This is used for debugging the RTL-based passes of the
12603 compiler. The file names for most of the dumps are made by appending
12604 a pass number and a word to the @var{dumpname}, and the files are
12605 created in the directory of the output file. In case of
12606 @option{=@var{filename}} option, the dump is output on the given file
12607 instead of the pass numbered dump files. Note that the pass number is
12608 assigned as passes are registered into the pass manager. Most passes
12609 are registered in the order that they will execute and for these passes
12610 the number corresponds to the pass execution order. However, passes
12611 registered by plugins, passes specific to compilation targets, or
12612 passes that are otherwise registered after all the other passes are
12613 numbered higher than a pass named "final", even if they are executed
12614 earlier. @var{dumpname} is generated from the name of the output
12615 file if explicitly specified and not an executable, otherwise it is
12616 the basename of the source file.
12618 Some @option{-d@var{letters}} switches have different meaning when
12619 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
12620 for information about preprocessor-specific dump options.
12622 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
12623 @option{-d} option @var{letters}. Here are the possible
12624 letters for use in @var{pass} and @var{letters}, and their meanings:
12628 @item -fdump-rtl-alignments
12629 @opindex fdump-rtl-alignments
12630 Dump after branch alignments have been computed.
12632 @item -fdump-rtl-asmcons
12633 @opindex fdump-rtl-asmcons
12634 Dump after fixing rtl statements that have unsatisfied in/out constraints.
12636 @item -fdump-rtl-auto_inc_dec
12637 @opindex fdump-rtl-auto_inc_dec
12638 Dump after auto-inc-dec discovery. This pass is only run on
12639 architectures that have auto inc or auto dec instructions.
12641 @item -fdump-rtl-barriers
12642 @opindex fdump-rtl-barriers
12643 Dump after cleaning up the barrier instructions.
12645 @item -fdump-rtl-bbpart
12646 @opindex fdump-rtl-bbpart
12647 Dump after partitioning hot and cold basic blocks.
12649 @item -fdump-rtl-bbro
12650 @opindex fdump-rtl-bbro
12651 Dump after block reordering.
12653 @item -fdump-rtl-btl1
12654 @itemx -fdump-rtl-btl2
12655 @opindex fdump-rtl-btl2
12656 @opindex fdump-rtl-btl2
12657 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
12658 after the two branch
12659 target load optimization passes.
12661 @item -fdump-rtl-bypass
12662 @opindex fdump-rtl-bypass
12663 Dump after jump bypassing and control flow optimizations.
12665 @item -fdump-rtl-combine
12666 @opindex fdump-rtl-combine
12667 Dump after the RTL instruction combination pass.
12669 @item -fdump-rtl-compgotos
12670 @opindex fdump-rtl-compgotos
12671 Dump after duplicating the computed gotos.
12673 @item -fdump-rtl-ce1
12674 @itemx -fdump-rtl-ce2
12675 @itemx -fdump-rtl-ce3
12676 @opindex fdump-rtl-ce1
12677 @opindex fdump-rtl-ce2
12678 @opindex fdump-rtl-ce3
12679 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
12680 @option{-fdump-rtl-ce3} enable dumping after the three
12681 if conversion passes.
12683 @item -fdump-rtl-cprop_hardreg
12684 @opindex fdump-rtl-cprop_hardreg
12685 Dump after hard register copy propagation.
12687 @item -fdump-rtl-csa
12688 @opindex fdump-rtl-csa
12689 Dump after combining stack adjustments.
12691 @item -fdump-rtl-cse1
12692 @itemx -fdump-rtl-cse2
12693 @opindex fdump-rtl-cse1
12694 @opindex fdump-rtl-cse2
12695 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
12696 the two common subexpression elimination passes.
12698 @item -fdump-rtl-dce
12699 @opindex fdump-rtl-dce
12700 Dump after the standalone dead code elimination passes.
12702 @item -fdump-rtl-dbr
12703 @opindex fdump-rtl-dbr
12704 Dump after delayed branch scheduling.
12706 @item -fdump-rtl-dce1
12707 @itemx -fdump-rtl-dce2
12708 @opindex fdump-rtl-dce1
12709 @opindex fdump-rtl-dce2
12710 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
12711 the two dead store elimination passes.
12713 @item -fdump-rtl-eh
12714 @opindex fdump-rtl-eh
12715 Dump after finalization of EH handling code.
12717 @item -fdump-rtl-eh_ranges
12718 @opindex fdump-rtl-eh_ranges
12719 Dump after conversion of EH handling range regions.
12721 @item -fdump-rtl-expand
12722 @opindex fdump-rtl-expand
12723 Dump after RTL generation.
12725 @item -fdump-rtl-fwprop1
12726 @itemx -fdump-rtl-fwprop2
12727 @opindex fdump-rtl-fwprop1
12728 @opindex fdump-rtl-fwprop2
12729 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
12730 dumping after the two forward propagation passes.
12732 @item -fdump-rtl-gcse1
12733 @itemx -fdump-rtl-gcse2
12734 @opindex fdump-rtl-gcse1
12735 @opindex fdump-rtl-gcse2
12736 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
12737 after global common subexpression elimination.
12739 @item -fdump-rtl-init-regs
12740 @opindex fdump-rtl-init-regs
12741 Dump after the initialization of the registers.
12743 @item -fdump-rtl-initvals
12744 @opindex fdump-rtl-initvals
12745 Dump after the computation of the initial value sets.
12747 @item -fdump-rtl-into_cfglayout
12748 @opindex fdump-rtl-into_cfglayout
12749 Dump after converting to cfglayout mode.
12751 @item -fdump-rtl-ira
12752 @opindex fdump-rtl-ira
12753 Dump after iterated register allocation.
12755 @item -fdump-rtl-jump
12756 @opindex fdump-rtl-jump
12757 Dump after the second jump optimization.
12759 @item -fdump-rtl-loop2
12760 @opindex fdump-rtl-loop2
12761 @option{-fdump-rtl-loop2} enables dumping after the rtl
12762 loop optimization passes.
12764 @item -fdump-rtl-mach
12765 @opindex fdump-rtl-mach
12766 Dump after performing the machine dependent reorganization pass, if that
12769 @item -fdump-rtl-mode_sw
12770 @opindex fdump-rtl-mode_sw
12771 Dump after removing redundant mode switches.
12773 @item -fdump-rtl-rnreg
12774 @opindex fdump-rtl-rnreg
12775 Dump after register renumbering.
12777 @item -fdump-rtl-outof_cfglayout
12778 @opindex fdump-rtl-outof_cfglayout
12779 Dump after converting from cfglayout mode.
12781 @item -fdump-rtl-peephole2
12782 @opindex fdump-rtl-peephole2
12783 Dump after the peephole pass.
12785 @item -fdump-rtl-postreload
12786 @opindex fdump-rtl-postreload
12787 Dump after post-reload optimizations.
12789 @item -fdump-rtl-pro_and_epilogue
12790 @opindex fdump-rtl-pro_and_epilogue
12791 Dump after generating the function prologues and epilogues.
12793 @item -fdump-rtl-sched1
12794 @itemx -fdump-rtl-sched2
12795 @opindex fdump-rtl-sched1
12796 @opindex fdump-rtl-sched2
12797 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
12798 after the basic block scheduling passes.
12800 @item -fdump-rtl-ree
12801 @opindex fdump-rtl-ree
12802 Dump after sign/zero extension elimination.
12804 @item -fdump-rtl-seqabstr
12805 @opindex fdump-rtl-seqabstr
12806 Dump after common sequence discovery.
12808 @item -fdump-rtl-shorten
12809 @opindex fdump-rtl-shorten
12810 Dump after shortening branches.
12812 @item -fdump-rtl-sibling
12813 @opindex fdump-rtl-sibling
12814 Dump after sibling call optimizations.
12816 @item -fdump-rtl-split1
12817 @itemx -fdump-rtl-split2
12818 @itemx -fdump-rtl-split3
12819 @itemx -fdump-rtl-split4
12820 @itemx -fdump-rtl-split5
12821 @opindex fdump-rtl-split1
12822 @opindex fdump-rtl-split2
12823 @opindex fdump-rtl-split3
12824 @opindex fdump-rtl-split4
12825 @opindex fdump-rtl-split5
12826 These options enable dumping after five rounds of
12827 instruction splitting.
12829 @item -fdump-rtl-sms
12830 @opindex fdump-rtl-sms
12831 Dump after modulo scheduling. This pass is only run on some
12834 @item -fdump-rtl-stack
12835 @opindex fdump-rtl-stack
12836 Dump after conversion from GCC's ``flat register file'' registers to the
12837 x87's stack-like registers. This pass is only run on x86 variants.
12839 @item -fdump-rtl-subreg1
12840 @itemx -fdump-rtl-subreg2
12841 @opindex fdump-rtl-subreg1
12842 @opindex fdump-rtl-subreg2
12843 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
12844 the two subreg expansion passes.
12846 @item -fdump-rtl-unshare
12847 @opindex fdump-rtl-unshare
12848 Dump after all rtl has been unshared.
12850 @item -fdump-rtl-vartrack
12851 @opindex fdump-rtl-vartrack
12852 Dump after variable tracking.
12854 @item -fdump-rtl-vregs
12855 @opindex fdump-rtl-vregs
12856 Dump after converting virtual registers to hard registers.
12858 @item -fdump-rtl-web
12859 @opindex fdump-rtl-web
12860 Dump after live range splitting.
12862 @item -fdump-rtl-regclass
12863 @itemx -fdump-rtl-subregs_of_mode_init
12864 @itemx -fdump-rtl-subregs_of_mode_finish
12865 @itemx -fdump-rtl-dfinit
12866 @itemx -fdump-rtl-dfinish
12867 @opindex fdump-rtl-regclass
12868 @opindex fdump-rtl-subregs_of_mode_init
12869 @opindex fdump-rtl-subregs_of_mode_finish
12870 @opindex fdump-rtl-dfinit
12871 @opindex fdump-rtl-dfinish
12872 These dumps are defined but always produce empty files.
12875 @itemx -fdump-rtl-all
12877 @opindex fdump-rtl-all
12878 Produce all the dumps listed above.
12882 Annotate the assembler output with miscellaneous debugging information.
12886 Dump all macro definitions, at the end of preprocessing, in addition to
12891 Produce a core dump whenever an error occurs.
12895 Annotate the assembler output with a comment indicating which
12896 pattern and alternative is used. The length of each instruction is
12901 Dump the RTL in the assembler output as a comment before each instruction.
12902 Also turns on @option{-dp} annotation.
12906 Just generate RTL for a function instead of compiling it. Usually used
12907 with @option{-fdump-rtl-expand}.
12910 @item -fdump-noaddr
12911 @opindex fdump-noaddr
12912 When doing debugging dumps, suppress address output. This makes it more
12913 feasible to use diff on debugging dumps for compiler invocations with
12914 different compiler binaries and/or different
12915 text / bss / data / heap / stack / dso start locations.
12918 @opindex freport-bug
12919 Collect and dump debug information into a temporary file if an
12920 internal compiler error (ICE) occurs.
12922 @item -fdump-unnumbered
12923 @opindex fdump-unnumbered
12924 When doing debugging dumps, suppress instruction numbers and address output.
12925 This makes it more feasible to use diff on debugging dumps for compiler
12926 invocations with different options, in particular with and without
12929 @item -fdump-unnumbered-links
12930 @opindex fdump-unnumbered-links
12931 When doing debugging dumps (see @option{-d} option above), suppress
12932 instruction numbers for the links to the previous and next instructions
12935 @item -fdump-translation-unit @r{(C++ only)}
12936 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12937 @opindex fdump-translation-unit
12938 Dump a representation of the tree structure for the entire translation
12939 unit to a file. The file name is made by appending @file{.tu} to the
12940 source file name, and the file is created in the same directory as the
12941 output file. If the @samp{-@var{options}} form is used, @var{options}
12942 controls the details of the dump as described for the
12943 @option{-fdump-tree} options.
12945 @item -fdump-class-hierarchy @r{(C++ only)}
12946 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12947 @opindex fdump-class-hierarchy
12948 Dump a representation of each class's hierarchy and virtual function
12949 table layout to a file. The file name is made by appending
12950 @file{.class} to the source file name, and the file is created in the
12951 same directory as the output file. If the @samp{-@var{options}} form
12952 is used, @var{options} controls the details of the dump as described
12953 for the @option{-fdump-tree} options.
12955 @item -fdump-ipa-@var{switch}
12957 Control the dumping at various stages of inter-procedural analysis
12958 language tree to a file. The file name is generated by appending a
12959 switch specific suffix to the source file name, and the file is created
12960 in the same directory as the output file. The following dumps are
12965 Enables all inter-procedural analysis dumps.
12968 Dumps information about call-graph optimization, unused function removal,
12969 and inlining decisions.
12972 Dump after function inlining.
12976 @item -fdump-passes
12977 @opindex fdump-passes
12978 Print on @file{stderr} the list of optimization passes that are turned
12979 on and off by the current command-line options.
12981 @item -fdump-statistics-@var{option}
12982 @opindex fdump-statistics
12983 Enable and control dumping of pass statistics in a separate file. The
12984 file name is generated by appending a suffix ending in
12985 @samp{.statistics} to the source file name, and the file is created in
12986 the same directory as the output file. If the @samp{-@var{option}}
12987 form is used, @samp{-stats} causes counters to be summed over the
12988 whole compilation unit while @samp{-details} dumps every event as
12989 the passes generate them. The default with no option is to sum
12990 counters for each function compiled.
12992 @item -fdump-tree-all
12993 @itemx -fdump-tree-@var{switch}
12994 @itemx -fdump-tree-@var{switch}-@var{options}
12995 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
12996 @opindex fdump-tree-all
12997 @opindex fdump-tree
12998 Control the dumping at various stages of processing the intermediate
12999 language tree to a file. The file name is generated by appending a
13000 switch-specific suffix to the source file name, and the file is
13001 created in the same directory as the output file. In case of
13002 @option{=@var{filename}} option, the dump is output on the given file
13003 instead of the auto named dump files. If the @samp{-@var{options}}
13004 form is used, @var{options} is a list of @samp{-} separated options
13005 which control the details of the dump. Not all options are applicable
13006 to all dumps; those that are not meaningful are ignored. The
13007 following options are available
13011 Print the address of each node. Usually this is not meaningful as it
13012 changes according to the environment and source file. Its primary use
13013 is for tying up a dump file with a debug environment.
13015 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
13016 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
13017 use working backward from mangled names in the assembly file.
13019 When dumping front-end intermediate representations, inhibit dumping
13020 of members of a scope or body of a function merely because that scope
13021 has been reached. Only dump such items when they are directly reachable
13022 by some other path.
13024 When dumping pretty-printed trees, this option inhibits dumping the
13025 bodies of control structures.
13027 When dumping RTL, print the RTL in slim (condensed) form instead of
13028 the default LISP-like representation.
13030 Print a raw representation of the tree. By default, trees are
13031 pretty-printed into a C-like representation.
13033 Enable more detailed dumps (not honored by every dump option). Also
13034 include information from the optimization passes.
13036 Enable dumping various statistics about the pass (not honored by every dump
13039 Enable showing basic block boundaries (disabled in raw dumps).
13041 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
13042 dump a representation of the control flow graph suitable for viewing with
13043 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
13044 the file is pretty-printed as a subgraph, so that GraphViz can render them
13045 all in a single plot.
13047 This option currently only works for RTL dumps, and the RTL is always
13048 dumped in slim form.
13050 Enable showing virtual operands for every statement.
13052 Enable showing line numbers for statements.
13054 Enable showing the unique ID (@code{DECL_UID}) for each variable.
13056 Enable showing the tree dump for each statement.
13058 Enable showing the EH region number holding each statement.
13060 Enable showing scalar evolution analysis details.
13062 Enable showing optimization information (only available in certain
13065 Enable showing missed optimization information (only available in certain
13068 Enable other detailed optimization information (only available in
13070 @item =@var{filename}
13071 Instead of an auto named dump file, output into the given file
13072 name. The file names @file{stdout} and @file{stderr} are treated
13073 specially and are considered already open standard streams. For
13077 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
13078 -fdump-tree-pre=/dev/stderr file.c
13081 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
13082 output on to @file{stderr}. If two conflicting dump filenames are
13083 given for the same pass, then the latter option overrides the earlier
13087 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
13088 and @option{lineno}.
13091 Turn on all optimization options, i.e., @option{optimized},
13092 @option{missed}, and @option{note}.
13095 To determine what tree dumps are available or find the dump for a pass
13096 of interest follow the steps below.
13100 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
13101 look for a code that corresponds to the pass you are interested in.
13102 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
13103 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
13104 The number at the end distinguishes distinct invocations of the same pass.
13106 To enable the creation of the dump file, append the pass code to
13107 the @option{-fdump-} option prefix and invoke GCC with it. For example,
13108 to enable the dump from the Early Value Range Propagation pass, invoke
13109 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
13110 specify the name of the dump file. If you don't specify one, GCC
13111 creates as described below.
13113 Find the pass dump in a file whose name is composed of three components
13114 separated by a period: the name of the source file GCC was invoked to
13115 compile, a numeric suffix indicating the pass number followed by the
13116 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
13117 and finally the pass code. For example, the Early VRP pass dump might
13118 be in a file named @file{myfile.c.038t.evrp} in the current working
13119 directory. Note that the numeric codes are not stable and may change
13120 from one version of GCC to another.
13124 @itemx -fopt-info-@var{options}
13125 @itemx -fopt-info-@var{options}=@var{filename}
13127 Controls optimization dumps from various optimization passes. If the
13128 @samp{-@var{options}} form is used, @var{options} is a list of
13129 @samp{-} separated option keywords to select the dump details and
13132 The @var{options} can be divided into two groups: options describing the
13133 verbosity of the dump, and options describing which optimizations
13134 should be included. The options from both the groups can be freely
13135 mixed as they are non-overlapping. However, in case of any conflicts,
13136 the later options override the earlier options on the command
13139 The following options control the dump verbosity:
13143 Print information when an optimization is successfully applied. It is
13144 up to a pass to decide which information is relevant. For example, the
13145 vectorizer passes print the source location of loops which are
13146 successfully vectorized.
13148 Print information about missed optimizations. Individual passes
13149 control which information to include in the output.
13151 Print verbose information about optimizations, such as certain
13152 transformations, more detailed messages about decisions etc.
13154 Print detailed optimization information. This includes
13155 @samp{optimized}, @samp{missed}, and @samp{note}.
13158 One or more of the following option keywords can be used to describe a
13159 group of optimizations:
13163 Enable dumps from all interprocedural optimizations.
13165 Enable dumps from all loop optimizations.
13167 Enable dumps from all inlining optimizations.
13169 Enable dumps from all vectorization optimizations.
13171 Enable dumps from all optimizations. This is a superset of
13172 the optimization groups listed above.
13175 If @var{options} is
13176 omitted, it defaults to @samp{optimized-optall}, which means to dump all
13177 info about successful optimizations from all the passes.
13179 If the @var{filename} is provided, then the dumps from all the
13180 applicable optimizations are concatenated into the @var{filename}.
13181 Otherwise the dump is output onto @file{stderr}. Though multiple
13182 @option{-fopt-info} options are accepted, only one of them can include
13183 a @var{filename}. If other filenames are provided then all but the
13184 first such option are ignored.
13186 Note that the output @var{filename} is overwritten
13187 in case of multiple translation units. If a combined output from
13188 multiple translation units is desired, @file{stderr} should be used
13191 In the following example, the optimization info is output to
13200 gcc -O3 -fopt-info-missed=missed.all
13204 outputs missed optimization report from all the passes into
13205 @file{missed.all}, and this one:
13208 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
13212 prints information about missed optimization opportunities from
13213 vectorization passes on @file{stderr}.
13214 Note that @option{-fopt-info-vec-missed} is equivalent to
13215 @option{-fopt-info-missed-vec}.
13217 As another example,
13219 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
13223 outputs information about missed optimizations as well as
13224 optimized locations from all the inlining passes into
13230 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
13234 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
13235 in conflict since only one output file is allowed. In this case, only
13236 the first option takes effect and the subsequent options are
13237 ignored. Thus only @file{vec.miss} is produced which contains
13238 dumps from the vectorizer about missed opportunities.
13240 @item -fsched-verbose=@var{n}
13241 @opindex fsched-verbose
13242 On targets that use instruction scheduling, this option controls the
13243 amount of debugging output the scheduler prints to the dump files.
13245 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
13246 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
13247 For @var{n} greater than one, it also output basic block probabilities,
13248 detailed ready list information and unit/insn info. For @var{n} greater
13249 than two, it includes RTL at abort point, control-flow and regions info.
13250 And for @var{n} over four, @option{-fsched-verbose} also includes
13255 @item -fenable-@var{kind}-@var{pass}
13256 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
13260 This is a set of options that are used to explicitly disable/enable
13261 optimization passes. These options are intended for use for debugging GCC.
13262 Compiler users should use regular options for enabling/disabling
13267 @item -fdisable-ipa-@var{pass}
13268 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13269 statically invoked in the compiler multiple times, the pass name should be
13270 appended with a sequential number starting from 1.
13272 @item -fdisable-rtl-@var{pass}
13273 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
13274 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
13275 statically invoked in the compiler multiple times, the pass name should be
13276 appended with a sequential number starting from 1. @var{range-list} is a
13277 comma-separated list of function ranges or assembler names. Each range is a number
13278 pair separated by a colon. The range is inclusive in both ends. If the range
13279 is trivial, the number pair can be simplified as a single number. If the
13280 function's call graph node's @var{uid} falls within one of the specified ranges,
13281 the @var{pass} is disabled for that function. The @var{uid} is shown in the
13282 function header of a dump file, and the pass names can be dumped by using
13283 option @option{-fdump-passes}.
13285 @item -fdisable-tree-@var{pass}
13286 @itemx -fdisable-tree-@var{pass}=@var{range-list}
13287 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
13290 @item -fenable-ipa-@var{pass}
13291 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
13292 statically invoked in the compiler multiple times, the pass name should be
13293 appended with a sequential number starting from 1.
13295 @item -fenable-rtl-@var{pass}
13296 @itemx -fenable-rtl-@var{pass}=@var{range-list}
13297 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
13298 description and examples.
13300 @item -fenable-tree-@var{pass}
13301 @itemx -fenable-tree-@var{pass}=@var{range-list}
13302 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
13303 of option arguments.
13307 Here are some examples showing uses of these options.
13311 # disable ccp1 for all functions
13312 -fdisable-tree-ccp1
13313 # disable complete unroll for function whose cgraph node uid is 1
13314 -fenable-tree-cunroll=1
13315 # disable gcse2 for functions at the following ranges [1,1],
13316 # [300,400], and [400,1000]
13317 # disable gcse2 for functions foo and foo2
13318 -fdisable-rtl-gcse2=foo,foo2
13319 # disable early inlining
13320 -fdisable-tree-einline
13321 # disable ipa inlining
13322 -fdisable-ipa-inline
13323 # enable tree full unroll
13324 -fenable-tree-unroll
13329 @itemx -fchecking=@var{n}
13331 @opindex fno-checking
13332 Enable internal consistency checking. The default depends on
13333 the compiler configuration. @option{-fchecking=2} enables further
13334 internal consistency checking that might affect code generation.
13336 @item -frandom-seed=@var{string}
13337 @opindex frandom-seed
13338 This option provides a seed that GCC uses in place of
13339 random numbers in generating certain symbol names
13340 that have to be different in every compiled file. It is also used to
13341 place unique stamps in coverage data files and the object files that
13342 produce them. You can use the @option{-frandom-seed} option to produce
13343 reproducibly identical object files.
13345 The @var{string} can either be a number (decimal, octal or hex) or an
13346 arbitrary string (in which case it's converted to a number by
13349 The @var{string} should be different for every file you compile.
13352 @itemx -save-temps=cwd
13353 @opindex save-temps
13354 Store the usual ``temporary'' intermediate files permanently; place them
13355 in the current directory and name them based on the source file. Thus,
13356 compiling @file{foo.c} with @option{-c -save-temps} produces files
13357 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
13358 preprocessed @file{foo.i} output file even though the compiler now
13359 normally uses an integrated preprocessor.
13361 When used in combination with the @option{-x} command-line option,
13362 @option{-save-temps} is sensible enough to avoid over writing an
13363 input source file with the same extension as an intermediate file.
13364 The corresponding intermediate file may be obtained by renaming the
13365 source file before using @option{-save-temps}.
13367 If you invoke GCC in parallel, compiling several different source
13368 files that share a common base name in different subdirectories or the
13369 same source file compiled for multiple output destinations, it is
13370 likely that the different parallel compilers will interfere with each
13371 other, and overwrite the temporary files. For instance:
13374 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
13375 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
13378 may result in @file{foo.i} and @file{foo.o} being written to
13379 simultaneously by both compilers.
13381 @item -save-temps=obj
13382 @opindex save-temps=obj
13383 Store the usual ``temporary'' intermediate files permanently. If the
13384 @option{-o} option is used, the temporary files are based on the
13385 object file. If the @option{-o} option is not used, the
13386 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
13391 gcc -save-temps=obj -c foo.c
13392 gcc -save-temps=obj -c bar.c -o dir/xbar.o
13393 gcc -save-temps=obj foobar.c -o dir2/yfoobar
13397 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
13398 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
13399 @file{dir2/yfoobar.o}.
13401 @item -time@r{[}=@var{file}@r{]}
13403 Report the CPU time taken by each subprocess in the compilation
13404 sequence. For C source files, this is the compiler proper and assembler
13405 (plus the linker if linking is done).
13407 Without the specification of an output file, the output looks like this:
13414 The first number on each line is the ``user time'', that is time spent
13415 executing the program itself. The second number is ``system time'',
13416 time spent executing operating system routines on behalf of the program.
13417 Both numbers are in seconds.
13419 With the specification of an output file, the output is appended to the
13420 named file, and it looks like this:
13423 0.12 0.01 cc1 @var{options}
13424 0.00 0.01 as @var{options}
13427 The ``user time'' and the ``system time'' are moved before the program
13428 name, and the options passed to the program are displayed, so that one
13429 can later tell what file was being compiled, and with which options.
13431 @item -fdump-final-insns@r{[}=@var{file}@r{]}
13432 @opindex fdump-final-insns
13433 Dump the final internal representation (RTL) to @var{file}. If the
13434 optional argument is omitted (or if @var{file} is @code{.}), the name
13435 of the dump file is determined by appending @code{.gkd} to the
13436 compilation output file name.
13438 @item -fcompare-debug@r{[}=@var{opts}@r{]}
13439 @opindex fcompare-debug
13440 @opindex fno-compare-debug
13441 If no error occurs during compilation, run the compiler a second time,
13442 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
13443 passed to the second compilation. Dump the final internal
13444 representation in both compilations, and print an error if they differ.
13446 If the equal sign is omitted, the default @option{-gtoggle} is used.
13448 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
13449 and nonzero, implicitly enables @option{-fcompare-debug}. If
13450 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
13451 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
13454 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
13455 is equivalent to @option{-fno-compare-debug}, which disables the dumping
13456 of the final representation and the second compilation, preventing even
13457 @env{GCC_COMPARE_DEBUG} from taking effect.
13459 To verify full coverage during @option{-fcompare-debug} testing, set
13460 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
13461 which GCC rejects as an invalid option in any actual compilation
13462 (rather than preprocessing, assembly or linking). To get just a
13463 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
13464 not overridden} will do.
13466 @item -fcompare-debug-second
13467 @opindex fcompare-debug-second
13468 This option is implicitly passed to the compiler for the second
13469 compilation requested by @option{-fcompare-debug}, along with options to
13470 silence warnings, and omitting other options that would cause
13471 side-effect compiler outputs to files or to the standard output. Dump
13472 files and preserved temporary files are renamed so as to contain the
13473 @code{.gk} additional extension during the second compilation, to avoid
13474 overwriting those generated by the first.
13476 When this option is passed to the compiler driver, it causes the
13477 @emph{first} compilation to be skipped, which makes it useful for little
13478 other than debugging the compiler proper.
13482 Turn off generation of debug info, if leaving out this option
13483 generates it, or turn it on at level 2 otherwise. The position of this
13484 argument in the command line does not matter; it takes effect after all
13485 other options are processed, and it does so only once, no matter how
13486 many times it is given. This is mainly intended to be used with
13487 @option{-fcompare-debug}.
13489 @item -fvar-tracking-assignments-toggle
13490 @opindex fvar-tracking-assignments-toggle
13491 @opindex fno-var-tracking-assignments-toggle
13492 Toggle @option{-fvar-tracking-assignments}, in the same way that
13493 @option{-gtoggle} toggles @option{-g}.
13497 Makes the compiler print out each function name as it is compiled, and
13498 print some statistics about each pass when it finishes.
13500 @item -ftime-report
13501 @opindex ftime-report
13502 Makes the compiler print some statistics about the time consumed by each
13503 pass when it finishes.
13505 @item -ftime-report-details
13506 @opindex ftime-report-details
13507 Record the time consumed by infrastructure parts separately for each pass.
13509 @item -fira-verbose=@var{n}
13510 @opindex fira-verbose
13511 Control the verbosity of the dump file for the integrated register allocator.
13512 The default value is 5. If the value @var{n} is greater or equal to 10,
13513 the dump output is sent to stderr using the same format as @var{n} minus 10.
13516 @opindex flto-report
13517 Prints a report with internal details on the workings of the link-time
13518 optimizer. The contents of this report vary from version to version.
13519 It is meant to be useful to GCC developers when processing object
13520 files in LTO mode (via @option{-flto}).
13522 Disabled by default.
13524 @item -flto-report-wpa
13525 @opindex flto-report-wpa
13526 Like @option{-flto-report}, but only print for the WPA phase of Link
13530 @opindex fmem-report
13531 Makes the compiler print some statistics about permanent memory
13532 allocation when it finishes.
13534 @item -fmem-report-wpa
13535 @opindex fmem-report-wpa
13536 Makes the compiler print some statistics about permanent memory
13537 allocation for the WPA phase only.
13539 @item -fpre-ipa-mem-report
13540 @opindex fpre-ipa-mem-report
13541 @item -fpost-ipa-mem-report
13542 @opindex fpost-ipa-mem-report
13543 Makes the compiler print some statistics about permanent memory
13544 allocation before or after interprocedural optimization.
13546 @item -fprofile-report
13547 @opindex fprofile-report
13548 Makes the compiler print some statistics about consistency of the
13549 (estimated) profile and effect of individual passes.
13551 @item -fstack-usage
13552 @opindex fstack-usage
13553 Makes the compiler output stack usage information for the program, on a
13554 per-function basis. The filename for the dump is made by appending
13555 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
13556 the output file, if explicitly specified and it is not an executable,
13557 otherwise it is the basename of the source file. An entry is made up
13562 The name of the function.
13566 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
13569 The qualifier @code{static} means that the function manipulates the stack
13570 statically: a fixed number of bytes are allocated for the frame on function
13571 entry and released on function exit; no stack adjustments are otherwise made
13572 in the function. The second field is this fixed number of bytes.
13574 The qualifier @code{dynamic} means that the function manipulates the stack
13575 dynamically: in addition to the static allocation described above, stack
13576 adjustments are made in the body of the function, for example to push/pop
13577 arguments around function calls. If the qualifier @code{bounded} is also
13578 present, the amount of these adjustments is bounded at compile time and
13579 the second field is an upper bound of the total amount of stack used by
13580 the function. If it is not present, the amount of these adjustments is
13581 not bounded at compile time and the second field only represents the
13586 Emit statistics about front-end processing at the end of the compilation.
13587 This option is supported only by the C++ front end, and
13588 the information is generally only useful to the G++ development team.
13590 @item -fdbg-cnt-list
13591 @opindex fdbg-cnt-list
13592 Print the name and the counter upper bound for all debug counters.
13595 @item -fdbg-cnt=@var{counter-value-list}
13597 Set the internal debug counter upper bound. @var{counter-value-list}
13598 is a comma-separated list of @var{name}:@var{value} pairs
13599 which sets the upper bound of each debug counter @var{name} to @var{value}.
13600 All debug counters have the initial upper bound of @code{UINT_MAX};
13601 thus @code{dbg_cnt} returns true always unless the upper bound
13602 is set by this option.
13603 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
13604 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
13606 @item -print-file-name=@var{library}
13607 @opindex print-file-name
13608 Print the full absolute name of the library file @var{library} that
13609 would be used when linking---and don't do anything else. With this
13610 option, GCC does not compile or link anything; it just prints the
13613 @item -print-multi-directory
13614 @opindex print-multi-directory
13615 Print the directory name corresponding to the multilib selected by any
13616 other switches present in the command line. This directory is supposed
13617 to exist in @env{GCC_EXEC_PREFIX}.
13619 @item -print-multi-lib
13620 @opindex print-multi-lib
13621 Print the mapping from multilib directory names to compiler switches
13622 that enable them. The directory name is separated from the switches by
13623 @samp{;}, and each switch starts with an @samp{@@} instead of the
13624 @samp{-}, without spaces between multiple switches. This is supposed to
13625 ease shell processing.
13627 @item -print-multi-os-directory
13628 @opindex print-multi-os-directory
13629 Print the path to OS libraries for the selected
13630 multilib, relative to some @file{lib} subdirectory. If OS libraries are
13631 present in the @file{lib} subdirectory and no multilibs are used, this is
13632 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
13633 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
13634 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
13635 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
13637 @item -print-multiarch
13638 @opindex print-multiarch
13639 Print the path to OS libraries for the selected multiarch,
13640 relative to some @file{lib} subdirectory.
13642 @item -print-prog-name=@var{program}
13643 @opindex print-prog-name
13644 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
13646 @item -print-libgcc-file-name
13647 @opindex print-libgcc-file-name
13648 Same as @option{-print-file-name=libgcc.a}.
13650 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
13651 but you do want to link with @file{libgcc.a}. You can do:
13654 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
13657 @item -print-search-dirs
13658 @opindex print-search-dirs
13659 Print the name of the configured installation directory and a list of
13660 program and library directories @command{gcc} searches---and don't do anything else.
13662 This is useful when @command{gcc} prints the error message
13663 @samp{installation problem, cannot exec cpp0: No such file or directory}.
13664 To resolve this you either need to put @file{cpp0} and the other compiler
13665 components where @command{gcc} expects to find them, or you can set the environment
13666 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
13667 Don't forget the trailing @samp{/}.
13668 @xref{Environment Variables}.
13670 @item -print-sysroot
13671 @opindex print-sysroot
13672 Print the target sysroot directory that is used during
13673 compilation. This is the target sysroot specified either at configure
13674 time or using the @option{--sysroot} option, possibly with an extra
13675 suffix that depends on compilation options. If no target sysroot is
13676 specified, the option prints nothing.
13678 @item -print-sysroot-headers-suffix
13679 @opindex print-sysroot-headers-suffix
13680 Print the suffix added to the target sysroot when searching for
13681 headers, or give an error if the compiler is not configured with such
13682 a suffix---and don't do anything else.
13685 @opindex dumpmachine
13686 Print the compiler's target machine (for example,
13687 @samp{i686-pc-linux-gnu})---and don't do anything else.
13690 @opindex dumpversion
13691 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
13692 anything else. This is the compiler version used in filesystem paths,
13693 specs, can be depending on how the compiler has been configured just
13694 a single number (major version), two numbers separated by dot (major and
13695 minor version) or three numbers separated by dots (major, minor and patchlevel
13698 @item -dumpfullversion
13699 @opindex dumpfullversion
13700 Print the full compiler version, always 3 numbers separated by dots,
13701 major, minor and patchlevel version.
13705 Print the compiler's built-in specs---and don't do anything else. (This
13706 is used when GCC itself is being built.) @xref{Spec Files}.
13709 @node Submodel Options
13710 @section Machine-Dependent Options
13711 @cindex submodel options
13712 @cindex specifying hardware config
13713 @cindex hardware models and configurations, specifying
13714 @cindex target-dependent options
13715 @cindex machine-dependent options
13717 Each target machine supported by GCC can have its own options---for
13718 example, to allow you to compile for a particular processor variant or
13719 ABI, or to control optimizations specific to that machine. By
13720 convention, the names of machine-specific options start with
13723 Some configurations of the compiler also support additional target-specific
13724 options, usually for compatibility with other compilers on the same
13727 @c This list is ordered alphanumerically by subsection name.
13728 @c It should be the same order and spelling as these options are listed
13729 @c in Machine Dependent Options
13732 * AArch64 Options::
13733 * Adapteva Epiphany Options::
13737 * Blackfin Options::
13742 * DEC Alpha Options::
13746 * GNU/Linux Options::
13756 * MicroBlaze Options::
13759 * MN10300 Options::
13763 * Nios II Options::
13764 * Nvidia PTX Options::
13766 * picoChip Options::
13767 * PowerPC Options::
13770 * RS/6000 and PowerPC Options::
13772 * S/390 and zSeries Options::
13775 * Solaris 2 Options::
13778 * System V Options::
13779 * TILE-Gx Options::
13780 * TILEPro Options::
13785 * VxWorks Options::
13787 * x86 Windows Options::
13788 * Xstormy16 Options::
13790 * zSeries Options::
13793 @node AArch64 Options
13794 @subsection AArch64 Options
13795 @cindex AArch64 Options
13797 These options are defined for AArch64 implementations:
13801 @item -mabi=@var{name}
13803 Generate code for the specified data model. Permissible values
13804 are @samp{ilp32} for SysV-like data model where int, long int and pointers
13805 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
13806 but long int and pointers are 64 bits.
13808 The default depends on the specific target configuration. Note that
13809 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
13810 entire program with the same ABI, and link with a compatible set of libraries.
13813 @opindex mbig-endian
13814 Generate big-endian code. This is the default when GCC is configured for an
13815 @samp{aarch64_be-*-*} target.
13817 @item -mgeneral-regs-only
13818 @opindex mgeneral-regs-only
13819 Generate code which uses only the general-purpose registers. This will prevent
13820 the compiler from using floating-point and Advanced SIMD registers but will not
13821 impose any restrictions on the assembler.
13823 @item -mlittle-endian
13824 @opindex mlittle-endian
13825 Generate little-endian code. This is the default when GCC is configured for an
13826 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13828 @item -mcmodel=tiny
13829 @opindex mcmodel=tiny
13830 Generate code for the tiny code model. The program and its statically defined
13831 symbols must be within 1MB of each other. Programs can be statically or
13832 dynamically linked.
13834 @item -mcmodel=small
13835 @opindex mcmodel=small
13836 Generate code for the small code model. The program and its statically defined
13837 symbols must be within 4GB of each other. Programs can be statically or
13838 dynamically linked. This is the default code model.
13840 @item -mcmodel=large
13841 @opindex mcmodel=large
13842 Generate code for the large code model. This makes no assumptions about
13843 addresses and sizes of sections. Programs can be statically linked only.
13845 @item -mstrict-align
13846 @opindex mstrict-align
13847 Avoid generating memory accesses that may not be aligned on a natural object
13848 boundary as described in the architecture specification.
13850 @item -momit-leaf-frame-pointer
13851 @itemx -mno-omit-leaf-frame-pointer
13852 @opindex momit-leaf-frame-pointer
13853 @opindex mno-omit-leaf-frame-pointer
13854 Omit or keep the frame pointer in leaf functions. The former behavior is the
13857 @item -mtls-dialect=desc
13858 @opindex mtls-dialect=desc
13859 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13860 of TLS variables. This is the default.
13862 @item -mtls-dialect=traditional
13863 @opindex mtls-dialect=traditional
13864 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13867 @item -mtls-size=@var{size}
13869 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13870 This option requires binutils 2.26 or newer.
13872 @item -mfix-cortex-a53-835769
13873 @itemx -mno-fix-cortex-a53-835769
13874 @opindex mfix-cortex-a53-835769
13875 @opindex mno-fix-cortex-a53-835769
13876 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13877 This involves inserting a NOP instruction between memory instructions and
13878 64-bit integer multiply-accumulate instructions.
13880 @item -mfix-cortex-a53-843419
13881 @itemx -mno-fix-cortex-a53-843419
13882 @opindex mfix-cortex-a53-843419
13883 @opindex mno-fix-cortex-a53-843419
13884 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13885 This erratum workaround is made at link time and this will only pass the
13886 corresponding flag to the linker.
13888 @item -mlow-precision-recip-sqrt
13889 @item -mno-low-precision-recip-sqrt
13890 @opindex mlow-precision-recip-sqrt
13891 @opindex mno-low-precision-recip-sqrt
13892 Enable or disable the reciprocal square root approximation.
13893 This option only has an effect if @option{-ffast-math} or
13894 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13895 precision of reciprocal square root results to about 16 bits for
13896 single precision and to 32 bits for double precision.
13898 @item -mlow-precision-sqrt
13899 @item -mno-low-precision-sqrt
13900 @opindex -mlow-precision-sqrt
13901 @opindex -mno-low-precision-sqrt
13902 Enable or disable the square root approximation.
13903 This option only has an effect if @option{-ffast-math} or
13904 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13905 precision of square root results to about 16 bits for
13906 single precision and to 32 bits for double precision.
13907 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13909 @item -mlow-precision-div
13910 @item -mno-low-precision-div
13911 @opindex -mlow-precision-div
13912 @opindex -mno-low-precision-div
13913 Enable or disable the division approximation.
13914 This option only has an effect if @option{-ffast-math} or
13915 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13916 precision of division results to about 16 bits for
13917 single precision and to 32 bits for double precision.
13919 @item -march=@var{name}
13921 Specify the name of the target architecture and, optionally, one or
13922 more feature modifiers. This option has the form
13923 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13925 The permissible values for @var{arch} are @samp{armv8-a},
13926 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @var{native}.
13928 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
13929 support for the ARMv8.3-A architecture extensions.
13931 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
13932 support for the ARMv8.2-A architecture extensions.
13934 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
13935 support for the ARMv8.1-A architecture extension. In particular, it
13936 enables the @samp{+crc} and @samp{+lse} features.
13938 The value @samp{native} is available on native AArch64 GNU/Linux and
13939 causes the compiler to pick the architecture of the host system. This
13940 option has no effect if the compiler is unable to recognize the
13941 architecture of the host system,
13943 The permissible values for @var{feature} are listed in the sub-section
13944 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13945 Feature Modifiers}. Where conflicting feature modifiers are
13946 specified, the right-most feature is used.
13948 GCC uses @var{name} to determine what kind of instructions it can emit
13949 when generating assembly code. If @option{-march} is specified
13950 without either of @option{-mtune} or @option{-mcpu} also being
13951 specified, the code is tuned to perform well across a range of target
13952 processors implementing the target architecture.
13954 @item -mtune=@var{name}
13956 Specify the name of the target processor for which GCC should tune the
13957 performance of the code. Permissible values for this option are:
13958 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
13959 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{falkor},
13960 @samp{qdf24xx}, @samp{xgene1}, @samp{vulcan}, @samp{thunderx},
13961 @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
13962 @samp{thunderxt83}, @samp{thunderx2t99}, @samp{cortex-a57.cortex-a53},
13963 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
13964 @samp{cortex-a73.cortex-a53}, @samp{native}.
13966 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
13967 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
13968 specify that GCC should tune for a big.LITTLE system.
13970 Additionally on native AArch64 GNU/Linux systems the value
13971 @samp{native} tunes performance to the host system. This option has no effect
13972 if the compiler is unable to recognize the processor of the host system.
13974 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13975 are specified, the code is tuned to perform well across a range
13976 of target processors.
13978 This option cannot be suffixed by feature modifiers.
13980 @item -mcpu=@var{name}
13982 Specify the name of the target processor, optionally suffixed by one
13983 or more feature modifiers. This option has the form
13984 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13985 the permissible values for @var{cpu} are the same as those available
13986 for @option{-mtune}. The permissible values for @var{feature} are
13987 documented in the sub-section on
13988 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13989 Feature Modifiers}. Where conflicting feature modifiers are
13990 specified, the right-most feature is used.
13992 GCC uses @var{name} to determine what kind of instructions it can emit when
13993 generating assembly code (as if by @option{-march}) and to determine
13994 the target processor for which to tune for performance (as if
13995 by @option{-mtune}). Where this option is used in conjunction
13996 with @option{-march} or @option{-mtune}, those options take precedence
13997 over the appropriate part of this option.
13999 @item -moverride=@var{string}
14001 Override tuning decisions made by the back-end in response to a
14002 @option{-mtune=} switch. The syntax, semantics, and accepted values
14003 for @var{string} in this option are not guaranteed to be consistent
14006 This option is only intended to be useful when developing GCC.
14008 @item -mpc-relative-literal-loads
14009 @opindex mpc-relative-literal-loads
14010 Enable PC-relative literal loads. With this option literal pools are
14011 accessed using a single instruction and emitted after each function. This
14012 limits the maximum size of functions to 1MB. This is enabled by default for
14013 @option{-mcmodel=tiny}.
14015 @item -msign-return-address=@var{scope}
14016 @opindex msign-return-address
14017 Select the function scope on which return address signing will be applied.
14018 Permissible values are @samp{none}, which disables return address signing,
14019 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
14020 functions, and @samp{all}, which enables pointer signing for all functions. The
14021 default value is @samp{none}.
14025 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
14026 @anchor{aarch64-feature-modifiers}
14027 @cindex @option{-march} feature modifiers
14028 @cindex @option{-mcpu} feature modifiers
14029 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
14030 the following and their inverses @option{no@var{feature}}:
14034 Enable CRC extension. This is on by default for
14035 @option{-march=armv8.1-a}.
14037 Enable Crypto extension. This also enables Advanced SIMD and floating-point
14040 Enable floating-point instructions. This is on by default for all possible
14041 values for options @option{-march} and @option{-mcpu}.
14043 Enable Advanced SIMD instructions. This also enables floating-point
14044 instructions. This is on by default for all possible values for options
14045 @option{-march} and @option{-mcpu}.
14047 Enable Large System Extension instructions. This is on by default for
14048 @option{-march=armv8.1-a}.
14050 Enable FP16 extension. This also enables floating-point instructions.
14054 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
14055 Conversely, @option{nofp} implies @option{nosimd}, which implies
14058 @node Adapteva Epiphany Options
14059 @subsection Adapteva Epiphany Options
14061 These @samp{-m} options are defined for Adapteva Epiphany:
14064 @item -mhalf-reg-file
14065 @opindex mhalf-reg-file
14066 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
14067 That allows code to run on hardware variants that lack these registers.
14069 @item -mprefer-short-insn-regs
14070 @opindex mprefer-short-insn-regs
14071 Preferentially allocate registers that allow short instruction generation.
14072 This can result in increased instruction count, so this may either reduce or
14073 increase overall code size.
14075 @item -mbranch-cost=@var{num}
14076 @opindex mbranch-cost
14077 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14078 This cost is only a heuristic and is not guaranteed to produce
14079 consistent results across releases.
14083 Enable the generation of conditional moves.
14085 @item -mnops=@var{num}
14087 Emit @var{num} NOPs before every other generated instruction.
14089 @item -mno-soft-cmpsf
14090 @opindex mno-soft-cmpsf
14091 For single-precision floating-point comparisons, emit an @code{fsub} instruction
14092 and test the flags. This is faster than a software comparison, but can
14093 get incorrect results in the presence of NaNs, or when two different small
14094 numbers are compared such that their difference is calculated as zero.
14095 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
14096 software comparisons.
14098 @item -mstack-offset=@var{num}
14099 @opindex mstack-offset
14100 Set the offset between the top of the stack and the stack pointer.
14101 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
14102 can be used by leaf functions without stack allocation.
14103 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
14104 Note also that this option changes the ABI; compiling a program with a
14105 different stack offset than the libraries have been compiled with
14106 generally does not work.
14107 This option can be useful if you want to evaluate if a different stack
14108 offset would give you better code, but to actually use a different stack
14109 offset to build working programs, it is recommended to configure the
14110 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
14112 @item -mno-round-nearest
14113 @opindex mno-round-nearest
14114 Make the scheduler assume that the rounding mode has been set to
14115 truncating. The default is @option{-mround-nearest}.
14118 @opindex mlong-calls
14119 If not otherwise specified by an attribute, assume all calls might be beyond
14120 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
14121 function address into a register before performing a (otherwise direct) call.
14122 This is the default.
14124 @item -mshort-calls
14125 @opindex short-calls
14126 If not otherwise specified by an attribute, assume all direct calls are
14127 in the range of the @code{b} / @code{bl} instructions, so use these instructions
14128 for direct calls. The default is @option{-mlong-calls}.
14132 Assume addresses can be loaded as 16-bit unsigned values. This does not
14133 apply to function addresses for which @option{-mlong-calls} semantics
14136 @item -mfp-mode=@var{mode}
14138 Set the prevailing mode of the floating-point unit.
14139 This determines the floating-point mode that is provided and expected
14140 at function call and return time. Making this mode match the mode you
14141 predominantly need at function start can make your programs smaller and
14142 faster by avoiding unnecessary mode switches.
14144 @var{mode} can be set to one the following values:
14148 Any mode at function entry is valid, and retained or restored when
14149 the function returns, and when it calls other functions.
14150 This mode is useful for compiling libraries or other compilation units
14151 you might want to incorporate into different programs with different
14152 prevailing FPU modes, and the convenience of being able to use a single
14153 object file outweighs the size and speed overhead for any extra
14154 mode switching that might be needed, compared with what would be needed
14155 with a more specific choice of prevailing FPU mode.
14158 This is the mode used for floating-point calculations with
14159 truncating (i.e.@: round towards zero) rounding mode. That includes
14160 conversion from floating point to integer.
14162 @item round-nearest
14163 This is the mode used for floating-point calculations with
14164 round-to-nearest-or-even rounding mode.
14167 This is the mode used to perform integer calculations in the FPU, e.g.@:
14168 integer multiply, or integer multiply-and-accumulate.
14171 The default is @option{-mfp-mode=caller}
14173 @item -mnosplit-lohi
14174 @itemx -mno-postinc
14175 @itemx -mno-postmodify
14176 @opindex mnosplit-lohi
14177 @opindex mno-postinc
14178 @opindex mno-postmodify
14179 Code generation tweaks that disable, respectively, splitting of 32-bit
14180 loads, generation of post-increment addresses, and generation of
14181 post-modify addresses. The defaults are @option{msplit-lohi},
14182 @option{-mpost-inc}, and @option{-mpost-modify}.
14184 @item -mnovect-double
14185 @opindex mno-vect-double
14186 Change the preferred SIMD mode to SImode. The default is
14187 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
14189 @item -max-vect-align=@var{num}
14190 @opindex max-vect-align
14191 The maximum alignment for SIMD vector mode types.
14192 @var{num} may be 4 or 8. The default is 8.
14193 Note that this is an ABI change, even though many library function
14194 interfaces are unaffected if they don't use SIMD vector modes
14195 in places that affect size and/or alignment of relevant types.
14197 @item -msplit-vecmove-early
14198 @opindex msplit-vecmove-early
14199 Split vector moves into single word moves before reload. In theory this
14200 can give better register allocation, but so far the reverse seems to be
14201 generally the case.
14203 @item -m1reg-@var{reg}
14205 Specify a register to hold the constant @minus{}1, which makes loading small negative
14206 constants and certain bitmasks faster.
14207 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
14208 which specify use of that register as a fixed register,
14209 and @samp{none}, which means that no register is used for this
14210 purpose. The default is @option{-m1reg-none}.
14215 @subsection ARC Options
14216 @cindex ARC options
14218 The following options control the architecture variant for which code
14221 @c architecture variants
14224 @item -mbarrel-shifter
14225 @opindex mbarrel-shifter
14226 Generate instructions supported by barrel shifter. This is the default
14227 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
14229 @item -mcpu=@var{cpu}
14231 Set architecture type, register usage, and instruction scheduling
14232 parameters for @var{cpu}. There are also shortcut alias options
14233 available for backward compatibility and convenience. Supported
14234 values for @var{cpu} are
14240 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
14244 Compile for ARC601. Alias: @option{-mARC601}.
14249 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
14250 This is the default when configured with @option{--with-cpu=arc700}@.
14253 Compile for ARC EM.
14256 Compile for ARC HS.
14260 Compile for ARC EM cpu with no hardware extension.
14264 Compile for ARC EM4 cpu.
14268 Compile for ARC EM4 DMIPS cpu.
14272 Compile for ARC EM4 DMIPS cpu with single precision floating point
14277 Compile for ARC EM4 DMIPS cpu with single precision floating point and
14278 double assists instructions.
14282 Compile for ARC HS cpu with no hardware extension, except the atomic
14287 Compile for ARC HS34 cpu.
14291 Compile for ARC HS38 cpu.
14294 @opindex hs38_linux
14295 Compile for ARC HS38 cpu with all hardware extensions on.
14298 @opindex arc600_norm
14299 Compile for ARC 600 cpu with norm instruction enabled.
14301 @item arc600_mul32x16
14302 @opindex arc600_mul32x16
14303 Compile for ARC 600 cpu with norm and mul32x16 instructions enabled.
14306 @opindex arc600_mul64
14307 Compile for ARC 600 cpu with norm and mul64 instructions enabled.
14310 @opindex arc601_norm
14311 Compile for ARC 601 cpu with norm instruction enabled.
14313 @item arc601_mul32x16
14314 @opindex arc601_mul32x16
14315 Compile for ARC 601 cpu with norm and mul32x16 instructions enabled.
14318 @opindex arc601_mul64
14319 Compile for ARC 601 cpu with norm and mul64 instructions enabled.
14323 Compile for ARC 700 on NPS400 chip.
14329 @itemx -mdpfp-compact
14330 @opindex mdpfp-compact
14331 FPX: Generate Double Precision FPX instructions, tuned for the compact
14335 @opindex mdpfp-fast
14336 FPX: Generate Double Precision FPX instructions, tuned for the fast
14339 @item -mno-dpfp-lrsr
14340 @opindex mno-dpfp-lrsr
14341 Disable LR and SR instructions from using FPX extension aux registers.
14345 Generate Extended arithmetic instructions. Currently only
14346 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
14347 supported. This is always enabled for @option{-mcpu=ARC700}.
14351 Do not generate mpy instructions for ARC700. This instruction is
14356 Generate 32x16 bit multiply and mac instructions.
14360 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
14364 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
14369 @itemx -mspfp-compact
14370 @opindex mspfp-compact
14371 FPX: Generate Single Precision FPX instructions, tuned for the compact
14375 @opindex mspfp-fast
14376 FPX: Generate Single Precision FPX instructions, tuned for the fast
14381 Enable generation of ARC SIMD instructions via target-specific
14382 builtins. Only valid for @option{-mcpu=ARC700}.
14385 @opindex msoft-float
14386 This option ignored; it is provided for compatibility purposes only.
14387 Software floating point code is emitted by default, and this default
14388 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
14389 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
14390 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
14394 Generate swap instructions.
14398 This enables Locked Load/Store Conditional extension to implement
14399 atomic memopry built-in functions. Not available for ARC 6xx or ARC
14404 Enable DIV/REM instructions for ARCv2 cores.
14406 @item -mcode-density
14407 @opindex mcode-density
14408 Enable code density instructions for ARC EM, default on for ARC HS.
14412 Enable double load/store operations for ARC HS cores.
14414 @item -mtp-regno=@var{regno}
14416 Specify thread pointer register number.
14418 @item -mmpy-option=@var{multo}
14419 @opindex mmpy-option
14420 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
14421 the default value. The recognized values for @var{multo} are:
14425 No multiplier available.
14429 The multiply option is set to w: 16x16 multiplier, fully pipelined.
14430 The following instructions are enabled: MPYW, and MPYUW.
14434 The multiply option is set to wlh1: 32x32 multiplier, fully
14435 pipelined (1 stage). The following instructions are additionally
14436 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
14440 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
14441 (2 stages). The following instructions are additionally enabled: MPY,
14442 MPYU, MPYM, MPYMU, and MPY_S.
14446 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
14447 sequential. The following instructions are additionally enabled: MPY,
14448 MPYU, MPYM, MPYMU, and MPY_S.
14452 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
14453 sequential. The following instructions are additionally enabled: MPY,
14454 MPYU, MPYM, MPYMU, and MPY_S.
14458 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
14459 sequential. The following instructions are additionally enabled: MPY,
14460 MPYU, MPYM, MPYMU, and MPY_S.
14464 This option is only available for ARCv2 cores@.
14466 @item -mfpu=@var{fpu}
14468 Enables specific floating-point hardware extension for ARCv2
14469 core. Supported values for @var{fpu} are:
14475 Enables support for single precision floating point hardware
14480 Enables support for double precision floating point hardware
14481 extensions. The single precision floating point extension is also
14482 enabled. Not available for ARC EM@.
14486 Enables support for double precision floating point hardware
14487 extensions using double precision assist instructions. The single
14488 precision floating point extension is also enabled. This option is
14489 only available for ARC EM@.
14493 Enables support for double precision floating point hardware
14494 extensions using double precision assist instructions, and simple
14495 precision square-root and divide hardware extensions. The single
14496 precision floating point extension is also enabled. This option is
14497 only available for ARC EM@.
14501 Enables support for double precision floating point hardware
14502 extensions using double precision assist instructions, and simple
14503 precision fused multiple and add hardware extension. The single
14504 precision floating point extension is also enabled. This option is
14505 only available for ARC EM@.
14509 Enables support for double precision floating point hardware
14510 extensions using double precision assist instructions, and all simple
14511 precision hardware extensions. The single precision floating point
14512 extension is also enabled. This option is only available for ARC EM@.
14516 Enables support for single precision floating point, and single
14517 precision square-root and divide hardware extensions@.
14521 Enables support for double precision floating point, and double
14522 precision square-root and divide hardware extensions. This option
14523 includes option @samp{fpus_div}. Not available for ARC EM@.
14527 Enables support for single precision floating point, and single
14528 precision fused multiple and add hardware extensions@.
14532 Enables support for double precision floating point, and double
14533 precision fused multiple and add hardware extensions. This option
14534 includes option @samp{fpus_fma}. Not available for ARC EM@.
14538 Enables support for all single precision floating point hardware
14543 Enables support for all single and double precision floating point
14544 hardware extensions. Not available for ARC EM@.
14550 The following options are passed through to the assembler, and also
14551 define preprocessor macro symbols.
14553 @c Flags used by the assembler, but for which we define preprocessor
14554 @c macro symbols as well.
14557 @opindex mdsp-packa
14558 Passed down to the assembler to enable the DSP Pack A extensions.
14559 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
14564 Passed down to the assembler to enable the dual viterbi butterfly
14565 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
14566 option is deprecated.
14568 @c ARC700 4.10 extension instruction
14571 Passed down to the assembler to enable the Locked Load/Store
14572 Conditional extension. Also sets the preprocessor symbol
14577 Passed down to the assembler. Also sets the preprocessor symbol
14578 @code{__Xxmac_d16}. This option is deprecated.
14582 Passed down to the assembler. Also sets the preprocessor symbol
14583 @code{__Xxmac_24}. This option is deprecated.
14585 @c ARC700 4.10 extension instruction
14588 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
14589 extension instruction. Also sets the preprocessor symbol
14590 @code{__Xrtsc}. This option is deprecated.
14592 @c ARC700 4.10 extension instruction
14595 Passed down to the assembler to enable the swap byte ordering
14596 extension instruction. Also sets the preprocessor symbol
14600 @opindex mtelephony
14601 Passed down to the assembler to enable dual and single operand
14602 instructions for telephony. Also sets the preprocessor symbol
14603 @code{__Xtelephony}. This option is deprecated.
14607 Passed down to the assembler to enable the XY Memory extension. Also
14608 sets the preprocessor symbol @code{__Xxy}.
14612 The following options control how the assembly code is annotated:
14614 @c Assembly annotation options
14618 Annotate assembler instructions with estimated addresses.
14620 @item -mannotate-align
14621 @opindex mannotate-align
14622 Explain what alignment considerations lead to the decision to make an
14623 instruction short or long.
14627 The following options are passed through to the linker:
14629 @c options passed through to the linker
14633 Passed through to the linker, to specify use of the @code{arclinux} emulation.
14634 This option is enabled by default in tool chains built for
14635 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
14636 when profiling is not requested.
14638 @item -marclinux_prof
14639 @opindex marclinux_prof
14640 Passed through to the linker, to specify use of the
14641 @code{arclinux_prof} emulation. This option is enabled by default in
14642 tool chains built for @w{@code{arc-linux-uclibc}} and
14643 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
14647 The following options control the semantics of generated code:
14649 @c semantically relevant code generation options
14652 @opindex mlong-calls
14653 Generate call insns as register indirect calls, thus providing access
14654 to the full 32-bit address range.
14656 @item -mmedium-calls
14657 @opindex mmedium-calls
14658 Don't use less than 25 bit addressing range for calls, which is the
14659 offset available for an unconditional branch-and-link
14660 instruction. Conditional execution of function calls is suppressed, to
14661 allow use of the 25-bit range, rather than the 21-bit range with
14662 conditional branch-and-link. This is the default for tool chains built
14663 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
14667 Do not generate sdata references. This is the default for tool chains
14668 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
14671 @item -mvolatile-cache
14672 @opindex mvolatile-cache
14673 Use ordinarily cached memory accesses for volatile references. This is the
14676 @item -mno-volatile-cache
14677 @opindex mno-volatile-cache
14678 Enable cache bypass for volatile references.
14682 The following options fine tune code generation:
14683 @c code generation tuning options
14686 @opindex malign-call
14687 Do alignment optimizations for call instructions.
14689 @item -mauto-modify-reg
14690 @opindex mauto-modify-reg
14691 Enable the use of pre/post modify with register displacement.
14693 @item -mbbit-peephole
14694 @opindex mbbit-peephole
14695 Enable bbit peephole2.
14699 This option disables a target-specific pass in @file{arc_reorg} to
14700 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
14701 generation driven by the combiner pass.
14703 @item -mcase-vector-pcrel
14704 @opindex mcase-vector-pcrel
14705 Use pc-relative switch case tables - this enables case table shortening.
14706 This is the default for @option{-Os}.
14708 @item -mcompact-casesi
14709 @opindex mcompact-casesi
14710 Enable compact casesi pattern. This is the default for @option{-Os},
14711 and only available for ARCv1 cores.
14713 @item -mno-cond-exec
14714 @opindex mno-cond-exec
14715 Disable ARCompact specific pass to generate conditional execution instructions.
14716 Due to delay slot scheduling and interactions between operand numbers,
14717 literal sizes, instruction lengths, and the support for conditional execution,
14718 the target-independent pass to generate conditional execution is often lacking,
14719 so the ARC port has kept a special pass around that tries to find more
14720 conditional execution generating opportunities after register allocation,
14721 branch shortening, and delay slot scheduling have been done. This pass
14722 generally, but not always, improves performance and code size, at the cost of
14723 extra compilation time, which is why there is an option to switch it off.
14724 If you have a problem with call instructions exceeding their allowable
14725 offset range because they are conditionalized, you should consider using
14726 @option{-mmedium-calls} instead.
14728 @item -mearly-cbranchsi
14729 @opindex mearly-cbranchsi
14730 Enable pre-reload use of the cbranchsi pattern.
14732 @item -mexpand-adddi
14733 @opindex mexpand-adddi
14734 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
14735 @code{add.f}, @code{adc} etc.
14737 @item -mindexed-loads
14738 @opindex mindexed-loads
14739 Enable the use of indexed loads. This can be problematic because some
14740 optimizers then assume that indexed stores exist, which is not
14744 Enable Local Register Allocation. This is still experimental for ARC,
14745 so by default the compiler uses standard reload
14746 (i.e. @option{-mno-lra}).
14748 @item -mlra-priority-none
14749 @opindex mlra-priority-none
14750 Don't indicate any priority for target registers.
14752 @item -mlra-priority-compact
14753 @opindex mlra-priority-compact
14754 Indicate target register priority for r0..r3 / r12..r15.
14756 @item -mlra-priority-noncompact
14757 @opindex mlra-priority-noncompact
14758 Reduce target register priority for r0..r3 / r12..r15.
14760 @item -mno-millicode
14761 @opindex mno-millicode
14762 When optimizing for size (using @option{-Os}), prologues and epilogues
14763 that have to save or restore a large number of registers are often
14764 shortened by using call to a special function in libgcc; this is
14765 referred to as a @emph{millicode} call. As these calls can pose
14766 performance issues, and/or cause linking issues when linking in a
14767 nonstandard way, this option is provided to turn off millicode call
14771 @opindex mmixed-code
14772 Tweak register allocation to help 16-bit instruction generation.
14773 This generally has the effect of decreasing the average instruction size
14774 while increasing the instruction count.
14778 Enable 'q' instruction alternatives.
14779 This is the default for @option{-Os}.
14783 Enable Rcq constraint handling - most short code generation depends on this.
14784 This is the default.
14788 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
14789 This is the default.
14791 @item -msize-level=@var{level}
14792 @opindex msize-level
14793 Fine-tune size optimization with regards to instruction lengths and alignment.
14794 The recognized values for @var{level} are:
14797 No size optimization. This level is deprecated and treated like @samp{1}.
14800 Short instructions are used opportunistically.
14803 In addition, alignment of loops and of code after barriers are dropped.
14806 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
14810 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
14811 the behavior when this is not set is equivalent to level @samp{1}.
14813 @item -mtune=@var{cpu}
14815 Set instruction scheduling parameters for @var{cpu}, overriding any implied
14816 by @option{-mcpu=}.
14818 Supported values for @var{cpu} are
14822 Tune for ARC600 cpu.
14825 Tune for ARC601 cpu.
14828 Tune for ARC700 cpu with standard multiplier block.
14831 Tune for ARC700 cpu with XMAC block.
14834 Tune for ARC725D cpu.
14837 Tune for ARC750D cpu.
14841 @item -mmultcost=@var{num}
14843 Cost to assume for a multiply instruction, with @samp{4} being equal to a
14844 normal instruction.
14846 @item -munalign-prob-threshold=@var{probability}
14847 @opindex munalign-prob-threshold
14848 Set probability threshold for unaligning branches.
14849 When tuning for @samp{ARC700} and optimizing for speed, branches without
14850 filled delay slot are preferably emitted unaligned and long, unless
14851 profiling indicates that the probability for the branch to be taken
14852 is below @var{probability}. @xref{Cross-profiling}.
14853 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
14857 The following options are maintained for backward compatibility, but
14858 are now deprecated and will be removed in a future release:
14860 @c Deprecated options
14868 @opindex mbig-endian
14871 Compile code for big endian targets. Use of these options is now
14872 deprecated. Users wanting big-endian code, should use the
14873 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
14874 building the tool chain, for which big-endian is the default.
14876 @item -mlittle-endian
14877 @opindex mlittle-endian
14880 Compile code for little endian targets. Use of these options is now
14881 deprecated. Users wanting little-endian code should use the
14882 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
14883 building the tool chain, for which little-endian is the default.
14885 @item -mbarrel_shifter
14886 @opindex mbarrel_shifter
14887 Replaced by @option{-mbarrel-shifter}.
14889 @item -mdpfp_compact
14890 @opindex mdpfp_compact
14891 Replaced by @option{-mdpfp-compact}.
14894 @opindex mdpfp_fast
14895 Replaced by @option{-mdpfp-fast}.
14898 @opindex mdsp_packa
14899 Replaced by @option{-mdsp-packa}.
14903 Replaced by @option{-mea}.
14907 Replaced by @option{-mmac-24}.
14911 Replaced by @option{-mmac-d16}.
14913 @item -mspfp_compact
14914 @opindex mspfp_compact
14915 Replaced by @option{-mspfp-compact}.
14918 @opindex mspfp_fast
14919 Replaced by @option{-mspfp-fast}.
14921 @item -mtune=@var{cpu}
14923 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14924 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14925 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
14927 @item -multcost=@var{num}
14929 Replaced by @option{-mmultcost}.
14934 @subsection ARM Options
14935 @cindex ARM options
14937 These @samp{-m} options are defined for the ARM port:
14940 @item -mabi=@var{name}
14942 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
14943 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
14946 @opindex mapcs-frame
14947 Generate a stack frame that is compliant with the ARM Procedure Call
14948 Standard for all functions, even if this is not strictly necessary for
14949 correct execution of the code. Specifying @option{-fomit-frame-pointer}
14950 with this option causes the stack frames not to be generated for
14951 leaf functions. The default is @option{-mno-apcs-frame}.
14952 This option is deprecated.
14956 This is a synonym for @option{-mapcs-frame} and is deprecated.
14959 @c not currently implemented
14960 @item -mapcs-stack-check
14961 @opindex mapcs-stack-check
14962 Generate code to check the amount of stack space available upon entry to
14963 every function (that actually uses some stack space). If there is
14964 insufficient space available then either the function
14965 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
14966 called, depending upon the amount of stack space required. The runtime
14967 system is required to provide these functions. The default is
14968 @option{-mno-apcs-stack-check}, since this produces smaller code.
14970 @c not currently implemented
14971 @item -mapcs-reentrant
14972 @opindex mapcs-reentrant
14973 Generate reentrant, position-independent code. The default is
14974 @option{-mno-apcs-reentrant}.
14977 @item -mthumb-interwork
14978 @opindex mthumb-interwork
14979 Generate code that supports calling between the ARM and Thumb
14980 instruction sets. Without this option, on pre-v5 architectures, the
14981 two instruction sets cannot be reliably used inside one program. The
14982 default is @option{-mno-thumb-interwork}, since slightly larger code
14983 is generated when @option{-mthumb-interwork} is specified. In AAPCS
14984 configurations this option is meaningless.
14986 @item -mno-sched-prolog
14987 @opindex mno-sched-prolog
14988 Prevent the reordering of instructions in the function prologue, or the
14989 merging of those instruction with the instructions in the function's
14990 body. This means that all functions start with a recognizable set
14991 of instructions (or in fact one of a choice from a small set of
14992 different function prologues), and this information can be used to
14993 locate the start of functions inside an executable piece of code. The
14994 default is @option{-msched-prolog}.
14996 @item -mfloat-abi=@var{name}
14997 @opindex mfloat-abi
14998 Specifies which floating-point ABI to use. Permissible values
14999 are: @samp{soft}, @samp{softfp} and @samp{hard}.
15001 Specifying @samp{soft} causes GCC to generate output containing
15002 library calls for floating-point operations.
15003 @samp{softfp} allows the generation of code using hardware floating-point
15004 instructions, but still uses the soft-float calling conventions.
15005 @samp{hard} allows generation of floating-point instructions
15006 and uses FPU-specific calling conventions.
15008 The default depends on the specific target configuration. Note that
15009 the hard-float and soft-float ABIs are not link-compatible; you must
15010 compile your entire program with the same ABI, and link with a
15011 compatible set of libraries.
15013 @item -mlittle-endian
15014 @opindex mlittle-endian
15015 Generate code for a processor running in little-endian mode. This is
15016 the default for all standard configurations.
15019 @opindex mbig-endian
15020 Generate code for a processor running in big-endian mode; the default is
15021 to compile code for a little-endian processor.
15023 @item -march=@var{name}
15025 This specifies the name of the target ARM architecture. GCC uses this
15026 name to determine what kind of instructions it can emit when generating
15027 assembly code. This option can be used in conjunction with or instead
15028 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
15029 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
15030 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
15031 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
15032 @samp{armv6kz}, @samp{armv6s-m},
15033 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
15034 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
15035 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
15036 @samp{armv8.1-a+crc}, @samp{armv8-m.base}, @samp{armv8-m.main},
15037 @samp{armv8-m.main+dsp}, @samp{iwmmxt}, @samp{iwmmxt2}.
15039 Architecture revisions older than @samp{armv4t} are deprecated.
15041 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
15042 the (now mandatory) SVC instruction.
15044 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
15047 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
15050 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
15051 architecture together with the optional CRC32 extensions.
15053 @option{-march=armv8.1-a} enables compiler support for the ARMv8.1-A
15054 architecture. This also enables the features provided by
15055 @option{-march=armv8-a+crc}.
15057 @option{-march=armv8.2-a} enables compiler support for the ARMv8.2-A
15058 architecture. This also enables the features provided by
15059 @option{-march=armv8.1-a}.
15061 @option{-march=armv8.2-a+fp16} enables compiler support for the
15062 ARMv8.2-A architecture with the optional FP16 instructions extension.
15063 This also enables the features provided by @option{-march=armv8.1-a}
15064 and implies @option{-mfp16-format=ieee}.
15066 @option{-march=native} causes the compiler to auto-detect the architecture
15067 of the build computer. At present, this feature is only supported on
15068 GNU/Linux, and not all architectures are recognized. If the auto-detect
15069 is unsuccessful the option has no effect.
15071 @item -mtune=@var{name}
15073 This option specifies the name of the target ARM processor for
15074 which GCC should tune the performance of the code.
15075 For some ARM implementations better performance can be obtained by using
15077 Permissible names are: @samp{arm2}, @samp{arm250},
15078 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
15079 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
15080 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
15081 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
15083 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
15084 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
15085 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
15086 @samp{strongarm1110},
15087 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
15088 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
15089 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
15090 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
15091 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
15092 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
15093 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
15094 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
15095 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
15096 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
15097 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
15098 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
15106 @samp{cortex-m0plus},
15107 @samp{cortex-m1.small-multiply},
15108 @samp{cortex-m0.small-multiply},
15109 @samp{cortex-m0plus.small-multiply},
15113 @samp{marvell-pj4},
15114 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
15115 @samp{fa526}, @samp{fa626},
15116 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
15119 Additionally, this option can specify that GCC should tune the performance
15120 of the code for a big.LITTLE system. Permissible names are:
15121 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
15122 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15123 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
15125 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
15126 performance for a blend of processors within architecture @var{arch}.
15127 The aim is to generate code that run well on the current most popular
15128 processors, balancing between optimizations that benefit some CPUs in the
15129 range, and avoiding performance pitfalls of other CPUs. The effects of
15130 this option may change in future GCC versions as CPU models come and go.
15132 @option{-mtune=native} causes the compiler to auto-detect the CPU
15133 of the build computer. At present, this feature is only supported on
15134 GNU/Linux, and not all architectures are recognized. If the auto-detect is
15135 unsuccessful the option has no effect.
15137 @item -mcpu=@var{name}
15139 This specifies the name of the target ARM processor. GCC uses this name
15140 to derive the name of the target ARM architecture (as if specified
15141 by @option{-march}) and the ARM processor type for which to tune for
15142 performance (as if specified by @option{-mtune}). Where this option
15143 is used in conjunction with @option{-march} or @option{-mtune},
15144 those options take precedence over the appropriate part of this option.
15146 Permissible names for this option are the same as those for
15149 @option{-mcpu=generic-@var{arch}} is also permissible, and is
15150 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
15151 See @option{-mtune} for more information.
15153 @option{-mcpu=native} causes the compiler to auto-detect the CPU
15154 of the build computer. At present, this feature is only supported on
15155 GNU/Linux, and not all architectures are recognized. If the auto-detect
15156 is unsuccessful the option has no effect.
15158 @item -mfpu=@var{name}
15160 This specifies what floating-point hardware (or hardware emulation) is
15161 available on the target. Permissible names are: @samp{vfpv2}, @samp{vfpv3},
15162 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
15163 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
15164 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
15165 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
15166 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
15167 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
15168 is an alias for @samp{vfpv2}.
15170 If @option{-msoft-float} is specified this specifies the format of
15171 floating-point values.
15173 If the selected floating-point hardware includes the NEON extension
15174 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
15175 operations are not generated by GCC's auto-vectorization pass unless
15176 @option{-funsafe-math-optimizations} is also specified. This is
15177 because NEON hardware does not fully implement the IEEE 754 standard for
15178 floating-point arithmetic (in particular denormal values are treated as
15179 zero), so the use of NEON instructions may lead to a loss of precision.
15181 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}).
15183 @item -mfp16-format=@var{name}
15184 @opindex mfp16-format
15185 Specify the format of the @code{__fp16} half-precision floating-point type.
15186 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
15187 the default is @samp{none}, in which case the @code{__fp16} type is not
15188 defined. @xref{Half-Precision}, for more information.
15190 @item -mstructure-size-boundary=@var{n}
15191 @opindex mstructure-size-boundary
15192 The sizes of all structures and unions are rounded up to a multiple
15193 of the number of bits set by this option. Permissible values are 8, 32
15194 and 64. The default value varies for different toolchains. For the COFF
15195 targeted toolchain the default value is 8. A value of 64 is only allowed
15196 if the underlying ABI supports it.
15198 Specifying a larger number can produce faster, more efficient code, but
15199 can also increase the size of the program. Different values are potentially
15200 incompatible. Code compiled with one value cannot necessarily expect to
15201 work with code or libraries compiled with another value, if they exchange
15202 information using structures or unions.
15204 @item -mabort-on-noreturn
15205 @opindex mabort-on-noreturn
15206 Generate a call to the function @code{abort} at the end of a
15207 @code{noreturn} function. It is executed if the function tries to
15211 @itemx -mno-long-calls
15212 @opindex mlong-calls
15213 @opindex mno-long-calls
15214 Tells the compiler to perform function calls by first loading the
15215 address of the function into a register and then performing a subroutine
15216 call on this register. This switch is needed if the target function
15217 lies outside of the 64-megabyte addressing range of the offset-based
15218 version of subroutine call instruction.
15220 Even if this switch is enabled, not all function calls are turned
15221 into long calls. The heuristic is that static functions, functions
15222 that have the @code{short_call} attribute, functions that are inside
15223 the scope of a @code{#pragma no_long_calls} directive, and functions whose
15224 definitions have already been compiled within the current compilation
15225 unit are not turned into long calls. The exceptions to this rule are
15226 that weak function definitions, functions with the @code{long_call}
15227 attribute or the @code{section} attribute, and functions that are within
15228 the scope of a @code{#pragma long_calls} directive are always
15229 turned into long calls.
15231 This feature is not enabled by default. Specifying
15232 @option{-mno-long-calls} restores the default behavior, as does
15233 placing the function calls within the scope of a @code{#pragma
15234 long_calls_off} directive. Note these switches have no effect on how
15235 the compiler generates code to handle function calls via function
15238 @item -msingle-pic-base
15239 @opindex msingle-pic-base
15240 Treat the register used for PIC addressing as read-only, rather than
15241 loading it in the prologue for each function. The runtime system is
15242 responsible for initializing this register with an appropriate value
15243 before execution begins.
15245 @item -mpic-register=@var{reg}
15246 @opindex mpic-register
15247 Specify the register to be used for PIC addressing.
15248 For standard PIC base case, the default is any suitable register
15249 determined by compiler. For single PIC base case, the default is
15250 @samp{R9} if target is EABI based or stack-checking is enabled,
15251 otherwise the default is @samp{R10}.
15253 @item -mpic-data-is-text-relative
15254 @opindex mpic-data-is-text-relative
15255 Assume that the displacement between the text and data segments is fixed
15256 at static link time. This permits using PC-relative addressing
15257 operations to access data known to be in the data segment. For
15258 non-VxWorks RTP targets, this option is enabled by default. When
15259 disabled on such targets, it will enable @option{-msingle-pic-base} by
15262 @item -mpoke-function-name
15263 @opindex mpoke-function-name
15264 Write the name of each function into the text section, directly
15265 preceding the function prologue. The generated code is similar to this:
15269 .ascii "arm_poke_function_name", 0
15272 .word 0xff000000 + (t1 - t0)
15273 arm_poke_function_name
15275 stmfd sp!, @{fp, ip, lr, pc@}
15279 When performing a stack backtrace, code can inspect the value of
15280 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
15281 location @code{pc - 12} and the top 8 bits are set, then we know that
15282 there is a function name embedded immediately preceding this location
15283 and has length @code{((pc[-3]) & 0xff000000)}.
15290 Select between generating code that executes in ARM and Thumb
15291 states. The default for most configurations is to generate code
15292 that executes in ARM state, but the default can be changed by
15293 configuring GCC with the @option{--with-mode=}@var{state}
15296 You can also override the ARM and Thumb mode for each function
15297 by using the @code{target("thumb")} and @code{target("arm")} function attributes
15298 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
15301 @opindex mtpcs-frame
15302 Generate a stack frame that is compliant with the Thumb Procedure Call
15303 Standard for all non-leaf functions. (A leaf function is one that does
15304 not call any other functions.) The default is @option{-mno-tpcs-frame}.
15306 @item -mtpcs-leaf-frame
15307 @opindex mtpcs-leaf-frame
15308 Generate a stack frame that is compliant with the Thumb Procedure Call
15309 Standard for all leaf functions. (A leaf function is one that does
15310 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
15312 @item -mcallee-super-interworking
15313 @opindex mcallee-super-interworking
15314 Gives all externally visible functions in the file being compiled an ARM
15315 instruction set header which switches to Thumb mode before executing the
15316 rest of the function. This allows these functions to be called from
15317 non-interworking code. This option is not valid in AAPCS configurations
15318 because interworking is enabled by default.
15320 @item -mcaller-super-interworking
15321 @opindex mcaller-super-interworking
15322 Allows calls via function pointers (including virtual functions) to
15323 execute correctly regardless of whether the target code has been
15324 compiled for interworking or not. There is a small overhead in the cost
15325 of executing a function pointer if this option is enabled. This option
15326 is not valid in AAPCS configurations because interworking is enabled
15329 @item -mtp=@var{name}
15331 Specify the access model for the thread local storage pointer. The valid
15332 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
15333 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
15334 (supported in the arm6k architecture), and @samp{auto}, which uses the
15335 best available method for the selected processor. The default setting is
15338 @item -mtls-dialect=@var{dialect}
15339 @opindex mtls-dialect
15340 Specify the dialect to use for accessing thread local storage. Two
15341 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
15342 @samp{gnu} dialect selects the original GNU scheme for supporting
15343 local and global dynamic TLS models. The @samp{gnu2} dialect
15344 selects the GNU descriptor scheme, which provides better performance
15345 for shared libraries. The GNU descriptor scheme is compatible with
15346 the original scheme, but does require new assembler, linker and
15347 library support. Initial and local exec TLS models are unaffected by
15348 this option and always use the original scheme.
15350 @item -mword-relocations
15351 @opindex mword-relocations
15352 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
15353 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
15354 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
15357 @item -mfix-cortex-m3-ldrd
15358 @opindex mfix-cortex-m3-ldrd
15359 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
15360 with overlapping destination and base registers are used. This option avoids
15361 generating these instructions. This option is enabled by default when
15362 @option{-mcpu=cortex-m3} is specified.
15364 @item -munaligned-access
15365 @itemx -mno-unaligned-access
15366 @opindex munaligned-access
15367 @opindex mno-unaligned-access
15368 Enables (or disables) reading and writing of 16- and 32- bit values
15369 from addresses that are not 16- or 32- bit aligned. By default
15370 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
15371 ARMv8-M Baseline architectures, and enabled for all other
15372 architectures. If unaligned access is not enabled then words in packed
15373 data structures are accessed a byte at a time.
15375 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
15376 generated object file to either true or false, depending upon the
15377 setting of this option. If unaligned access is enabled then the
15378 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
15381 @item -mneon-for-64bits
15382 @opindex mneon-for-64bits
15383 Enables using Neon to handle scalar 64-bits operations. This is
15384 disabled by default since the cost of moving data from core registers
15387 @item -mslow-flash-data
15388 @opindex mslow-flash-data
15389 Assume loading data from flash is slower than fetching instruction.
15390 Therefore literal load is minimized for better performance.
15391 This option is only supported when compiling for ARMv7 M-profile and
15394 @item -masm-syntax-unified
15395 @opindex masm-syntax-unified
15396 Assume inline assembler is using unified asm syntax. The default is
15397 currently off which implies divided syntax. This option has no impact
15398 on Thumb2. However, this may change in future releases of GCC.
15399 Divided syntax should be considered deprecated.
15401 @item -mrestrict-it
15402 @opindex mrestrict-it
15403 Restricts generation of IT blocks to conform to the rules of ARMv8.
15404 IT blocks can only contain a single 16-bit instruction from a select
15405 set of instructions. This option is on by default for ARMv8 Thumb mode.
15407 @item -mprint-tune-info
15408 @opindex mprint-tune-info
15409 Print CPU tuning information as comment in assembler file. This is
15410 an option used only for regression testing of the compiler and not
15411 intended for ordinary use in compiling code. This option is disabled
15415 @opindex mpure-code
15416 Do not allow constant data to be placed in code sections.
15417 Additionally, when compiling for ELF object format give all text sections the
15418 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
15419 is only available when generating non-pic code for ARMv7-M targets.
15423 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
15424 Development Tools Engineering Specification", which can be found on
15425 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
15429 @subsection AVR Options
15430 @cindex AVR Options
15432 These options are defined for AVR implementations:
15435 @item -mmcu=@var{mcu}
15437 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
15439 The default for this option is@tie{}@samp{avr2}.
15441 GCC supports the following AVR devices and ISAs:
15443 @include avr-mmcu.texi
15448 Assume that all data in static storage can be accessed by LDS / STS
15449 instructions. This option has only an effect on reduced Tiny devices like
15450 ATtiny40. See also the @code{absdata}
15451 @ref{AVR Variable Attributes,variable attribute}.
15453 @item -maccumulate-args
15454 @opindex maccumulate-args
15455 Accumulate outgoing function arguments and acquire/release the needed
15456 stack space for outgoing function arguments once in function
15457 prologue/epilogue. Without this option, outgoing arguments are pushed
15458 before calling a function and popped afterwards.
15460 Popping the arguments after the function call can be expensive on
15461 AVR so that accumulating the stack space might lead to smaller
15462 executables because arguments need not be removed from the
15463 stack after such a function call.
15465 This option can lead to reduced code size for functions that perform
15466 several calls to functions that get their arguments on the stack like
15467 calls to printf-like functions.
15469 @item -mbranch-cost=@var{cost}
15470 @opindex mbranch-cost
15471 Set the branch costs for conditional branch instructions to
15472 @var{cost}. Reasonable values for @var{cost} are small, non-negative
15473 integers. The default branch cost is 0.
15475 @item -mcall-prologues
15476 @opindex mcall-prologues
15477 Functions prologues/epilogues are expanded as calls to appropriate
15478 subroutines. Code size is smaller.
15482 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
15483 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
15484 and @code{long long} is 4 bytes. Please note that this option does not
15485 conform to the C standards, but it results in smaller code
15488 @item -mn-flash=@var{num}
15490 Assume that the flash memory has a size of
15491 @var{num} times 64@tie{}KiB.
15493 @item -mno-interrupts
15494 @opindex mno-interrupts
15495 Generated code is not compatible with hardware interrupts.
15496 Code size is smaller.
15500 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
15501 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
15502 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
15503 the assembler's command line and the @option{--relax} option to the
15504 linker's command line.
15506 Jump relaxing is performed by the linker because jump offsets are not
15507 known before code is located. Therefore, the assembler code generated by the
15508 compiler is the same, but the instructions in the executable may
15509 differ from instructions in the assembler code.
15511 Relaxing must be turned on if linker stubs are needed, see the
15512 section on @code{EIND} and linker stubs below.
15516 Assume that the device supports the Read-Modify-Write
15517 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
15521 Treat the stack pointer register as an 8-bit register,
15522 i.e.@: assume the high byte of the stack pointer is zero.
15523 In general, you don't need to set this option by hand.
15525 This option is used internally by the compiler to select and
15526 build multilibs for architectures @code{avr2} and @code{avr25}.
15527 These architectures mix devices with and without @code{SPH}.
15528 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
15529 the compiler driver adds or removes this option from the compiler
15530 proper's command line, because the compiler then knows if the device
15531 or architecture has an 8-bit stack pointer and thus no @code{SPH}
15536 Use address register @code{X} in a way proposed by the hardware. This means
15537 that @code{X} is only used in indirect, post-increment or
15538 pre-decrement addressing.
15540 Without this option, the @code{X} register may be used in the same way
15541 as @code{Y} or @code{Z} which then is emulated by additional
15543 For example, loading a value with @code{X+const} addressing with a
15544 small non-negative @code{const < 64} to a register @var{Rn} is
15548 adiw r26, const ; X += const
15549 ld @var{Rn}, X ; @var{Rn} = *X
15550 sbiw r26, const ; X -= const
15554 @opindex mtiny-stack
15555 Only change the lower 8@tie{}bits of the stack pointer.
15557 @item -mfract-convert-truncate
15558 @opindex mfract-convert-truncate
15559 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
15562 @opindex nodevicelib
15563 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
15565 @item -Waddr-space-convert
15566 @opindex Waddr-space-convert
15567 Warn about conversions between address spaces in the case where the
15568 resulting address space is not contained in the incoming address space.
15570 @item -Wmisspelled-isr
15571 @opindex Wmisspelled-isr
15572 Warn if the ISR is misspelled, i.e. without __vector prefix.
15573 Enabled by default.
15576 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
15577 @cindex @code{EIND}
15578 Pointers in the implementation are 16@tie{}bits wide.
15579 The address of a function or label is represented as word address so
15580 that indirect jumps and calls can target any code address in the
15581 range of 64@tie{}Ki words.
15583 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
15584 bytes of program memory space, there is a special function register called
15585 @code{EIND} that serves as most significant part of the target address
15586 when @code{EICALL} or @code{EIJMP} instructions are used.
15588 Indirect jumps and calls on these devices are handled as follows by
15589 the compiler and are subject to some limitations:
15594 The compiler never sets @code{EIND}.
15597 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
15598 instructions or might read @code{EIND} directly in order to emulate an
15599 indirect call/jump by means of a @code{RET} instruction.
15602 The compiler assumes that @code{EIND} never changes during the startup
15603 code or during the application. In particular, @code{EIND} is not
15604 saved/restored in function or interrupt service routine
15608 For indirect calls to functions and computed goto, the linker
15609 generates @emph{stubs}. Stubs are jump pads sometimes also called
15610 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
15611 The stub contains a direct jump to the desired address.
15614 Linker relaxation must be turned on so that the linker generates
15615 the stubs correctly in all situations. See the compiler option
15616 @option{-mrelax} and the linker option @option{--relax}.
15617 There are corner cases where the linker is supposed to generate stubs
15618 but aborts without relaxation and without a helpful error message.
15621 The default linker script is arranged for code with @code{EIND = 0}.
15622 If code is supposed to work for a setup with @code{EIND != 0}, a custom
15623 linker script has to be used in order to place the sections whose
15624 name start with @code{.trampolines} into the segment where @code{EIND}
15628 The startup code from libgcc never sets @code{EIND}.
15629 Notice that startup code is a blend of code from libgcc and AVR-LibC.
15630 For the impact of AVR-LibC on @code{EIND}, see the
15631 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
15634 It is legitimate for user-specific startup code to set up @code{EIND}
15635 early, for example by means of initialization code located in
15636 section @code{.init3}. Such code runs prior to general startup code
15637 that initializes RAM and calls constructors, but after the bit
15638 of startup code from AVR-LibC that sets @code{EIND} to the segment
15639 where the vector table is located.
15641 #include <avr/io.h>
15644 __attribute__((section(".init3"),naked,used,no_instrument_function))
15645 init3_set_eind (void)
15647 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
15648 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
15653 The @code{__trampolines_start} symbol is defined in the linker script.
15656 Stubs are generated automatically by the linker if
15657 the following two conditions are met:
15660 @item The address of a label is taken by means of the @code{gs} modifier
15661 (short for @emph{generate stubs}) like so:
15663 LDI r24, lo8(gs(@var{func}))
15664 LDI r25, hi8(gs(@var{func}))
15666 @item The final location of that label is in a code segment
15667 @emph{outside} the segment where the stubs are located.
15671 The compiler emits such @code{gs} modifiers for code labels in the
15672 following situations:
15674 @item Taking address of a function or code label.
15675 @item Computed goto.
15676 @item If prologue-save function is used, see @option{-mcall-prologues}
15677 command-line option.
15678 @item Switch/case dispatch tables. If you do not want such dispatch
15679 tables you can specify the @option{-fno-jump-tables} command-line option.
15680 @item C and C++ constructors/destructors called during startup/shutdown.
15681 @item If the tools hit a @code{gs()} modifier explained above.
15685 Jumping to non-symbolic addresses like so is @emph{not} supported:
15690 /* Call function at word address 0x2 */
15691 return ((int(*)(void)) 0x2)();
15695 Instead, a stub has to be set up, i.e.@: the function has to be called
15696 through a symbol (@code{func_4} in the example):
15701 extern int func_4 (void);
15703 /* Call function at byte address 0x4 */
15708 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
15709 Alternatively, @code{func_4} can be defined in the linker script.
15712 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
15713 @cindex @code{RAMPD}
15714 @cindex @code{RAMPX}
15715 @cindex @code{RAMPY}
15716 @cindex @code{RAMPZ}
15717 Some AVR devices support memories larger than the 64@tie{}KiB range
15718 that can be accessed with 16-bit pointers. To access memory locations
15719 outside this 64@tie{}KiB range, the content of a @code{RAMP}
15720 register is used as high part of the address:
15721 The @code{X}, @code{Y}, @code{Z} address register is concatenated
15722 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
15723 register, respectively, to get a wide address. Similarly,
15724 @code{RAMPD} is used together with direct addressing.
15728 The startup code initializes the @code{RAMP} special function
15729 registers with zero.
15732 If a @ref{AVR Named Address Spaces,named address space} other than
15733 generic or @code{__flash} is used, then @code{RAMPZ} is set
15734 as needed before the operation.
15737 If the device supports RAM larger than 64@tie{}KiB and the compiler
15738 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
15739 is reset to zero after the operation.
15742 If the device comes with a specific @code{RAMP} register, the ISR
15743 prologue/epilogue saves/restores that SFR and initializes it with
15744 zero in case the ISR code might (implicitly) use it.
15747 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
15748 If you use inline assembler to read from locations outside the
15749 16-bit address range and change one of the @code{RAMP} registers,
15750 you must reset it to zero after the access.
15754 @subsubsection AVR Built-in Macros
15756 GCC defines several built-in macros so that the user code can test
15757 for the presence or absence of features. Almost any of the following
15758 built-in macros are deduced from device capabilities and thus
15759 triggered by the @option{-mmcu=} command-line option.
15761 For even more AVR-specific built-in macros see
15762 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
15767 Build-in macro that resolves to a decimal number that identifies the
15768 architecture and depends on the @option{-mmcu=@var{mcu}} option.
15769 Possible values are:
15771 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
15772 @code{4}, @code{5}, @code{51}, @code{6}
15774 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
15775 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
15779 @code{100}, @code{102}, @code{104},
15780 @code{105}, @code{106}, @code{107}
15782 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
15783 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
15784 If @var{mcu} specifies a device, this built-in macro is set
15785 accordingly. For example, with @option{-mmcu=atmega8} the macro is
15786 defined to @code{4}.
15788 @item __AVR_@var{Device}__
15789 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
15790 the device's name. For example, @option{-mmcu=atmega8} defines the
15791 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
15792 @code{__AVR_ATtiny261A__}, etc.
15794 The built-in macros' names follow
15795 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
15796 the device name as from the AVR user manual. The difference between
15797 @var{Device} in the built-in macro and @var{device} in
15798 @option{-mmcu=@var{device}} is that the latter is always lowercase.
15800 If @var{device} is not a device but only a core architecture like
15801 @samp{avr51}, this macro is not defined.
15803 @item __AVR_DEVICE_NAME__
15804 Setting @option{-mmcu=@var{device}} defines this built-in macro to
15805 the device's name. For example, with @option{-mmcu=atmega8} the macro
15806 is defined to @code{atmega8}.
15808 If @var{device} is not a device but only a core architecture like
15809 @samp{avr51}, this macro is not defined.
15811 @item __AVR_XMEGA__
15812 The device / architecture belongs to the XMEGA family of devices.
15814 @item __AVR_HAVE_ELPM__
15815 The device has the @code{ELPM} instruction.
15817 @item __AVR_HAVE_ELPMX__
15818 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
15819 R@var{n},Z+} instructions.
15821 @item __AVR_HAVE_MOVW__
15822 The device has the @code{MOVW} instruction to perform 16-bit
15823 register-register moves.
15825 @item __AVR_HAVE_LPMX__
15826 The device has the @code{LPM R@var{n},Z} and
15827 @code{LPM R@var{n},Z+} instructions.
15829 @item __AVR_HAVE_MUL__
15830 The device has a hardware multiplier.
15832 @item __AVR_HAVE_JMP_CALL__
15833 The device has the @code{JMP} and @code{CALL} instructions.
15834 This is the case for devices with at least 16@tie{}KiB of program
15837 @item __AVR_HAVE_EIJMP_EICALL__
15838 @itemx __AVR_3_BYTE_PC__
15839 The device has the @code{EIJMP} and @code{EICALL} instructions.
15840 This is the case for devices with more than 128@tie{}KiB of program memory.
15841 This also means that the program counter
15842 (PC) is 3@tie{}bytes wide.
15844 @item __AVR_2_BYTE_PC__
15845 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
15846 with up to 128@tie{}KiB of program memory.
15848 @item __AVR_HAVE_8BIT_SP__
15849 @itemx __AVR_HAVE_16BIT_SP__
15850 The stack pointer (SP) register is treated as 8-bit respectively
15851 16-bit register by the compiler.
15852 The definition of these macros is affected by @option{-mtiny-stack}.
15854 @item __AVR_HAVE_SPH__
15856 The device has the SPH (high part of stack pointer) special function
15857 register or has an 8-bit stack pointer, respectively.
15858 The definition of these macros is affected by @option{-mmcu=} and
15859 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
15862 @item __AVR_HAVE_RAMPD__
15863 @itemx __AVR_HAVE_RAMPX__
15864 @itemx __AVR_HAVE_RAMPY__
15865 @itemx __AVR_HAVE_RAMPZ__
15866 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
15867 @code{RAMPZ} special function register, respectively.
15869 @item __NO_INTERRUPTS__
15870 This macro reflects the @option{-mno-interrupts} command-line option.
15872 @item __AVR_ERRATA_SKIP__
15873 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
15874 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
15875 instructions because of a hardware erratum. Skip instructions are
15876 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
15877 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
15880 @item __AVR_ISA_RMW__
15881 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
15883 @item __AVR_SFR_OFFSET__=@var{offset}
15884 Instructions that can address I/O special function registers directly
15885 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
15886 address as if addressed by an instruction to access RAM like @code{LD}
15887 or @code{STS}. This offset depends on the device architecture and has
15888 to be subtracted from the RAM address in order to get the
15889 respective I/O@tie{}address.
15891 @item __WITH_AVRLIBC__
15892 The compiler is configured to be used together with AVR-Libc.
15893 See the @option{--with-avrlibc} configure option.
15897 @node Blackfin Options
15898 @subsection Blackfin Options
15899 @cindex Blackfin Options
15902 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
15904 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
15905 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
15906 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
15907 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
15908 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
15909 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
15910 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
15911 @samp{bf561}, @samp{bf592}.
15913 The optional @var{sirevision} specifies the silicon revision of the target
15914 Blackfin processor. Any workarounds available for the targeted silicon revision
15915 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
15916 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
15917 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
15918 hexadecimal digits representing the major and minor numbers in the silicon
15919 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
15920 is not defined. If @var{sirevision} is @samp{any}, the
15921 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
15922 If this optional @var{sirevision} is not used, GCC assumes the latest known
15923 silicon revision of the targeted Blackfin processor.
15925 GCC defines a preprocessor macro for the specified @var{cpu}.
15926 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
15927 provided by libgloss to be linked in if @option{-msim} is not given.
15929 Without this option, @samp{bf532} is used as the processor by default.
15931 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
15932 only the preprocessor macro is defined.
15936 Specifies that the program will be run on the simulator. This causes
15937 the simulator BSP provided by libgloss to be linked in. This option
15938 has effect only for @samp{bfin-elf} toolchain.
15939 Certain other options, such as @option{-mid-shared-library} and
15940 @option{-mfdpic}, imply @option{-msim}.
15942 @item -momit-leaf-frame-pointer
15943 @opindex momit-leaf-frame-pointer
15944 Don't keep the frame pointer in a register for leaf functions. This
15945 avoids the instructions to save, set up and restore frame pointers and
15946 makes an extra register available in leaf functions. The option
15947 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
15948 which might make debugging harder.
15950 @item -mspecld-anomaly
15951 @opindex mspecld-anomaly
15952 When enabled, the compiler ensures that the generated code does not
15953 contain speculative loads after jump instructions. If this option is used,
15954 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
15956 @item -mno-specld-anomaly
15957 @opindex mno-specld-anomaly
15958 Don't generate extra code to prevent speculative loads from occurring.
15960 @item -mcsync-anomaly
15961 @opindex mcsync-anomaly
15962 When enabled, the compiler ensures that the generated code does not
15963 contain CSYNC or SSYNC instructions too soon after conditional branches.
15964 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
15966 @item -mno-csync-anomaly
15967 @opindex mno-csync-anomaly
15968 Don't generate extra code to prevent CSYNC or SSYNC instructions from
15969 occurring too soon after a conditional branch.
15973 When enabled, the compiler is free to take advantage of the knowledge that
15974 the entire program fits into the low 64k of memory.
15977 @opindex mno-low-64k
15978 Assume that the program is arbitrarily large. This is the default.
15980 @item -mstack-check-l1
15981 @opindex mstack-check-l1
15982 Do stack checking using information placed into L1 scratchpad memory by the
15985 @item -mid-shared-library
15986 @opindex mid-shared-library
15987 Generate code that supports shared libraries via the library ID method.
15988 This allows for execute in place and shared libraries in an environment
15989 without virtual memory management. This option implies @option{-fPIC}.
15990 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15992 @item -mno-id-shared-library
15993 @opindex mno-id-shared-library
15994 Generate code that doesn't assume ID-based shared libraries are being used.
15995 This is the default.
15997 @item -mleaf-id-shared-library
15998 @opindex mleaf-id-shared-library
15999 Generate code that supports shared libraries via the library ID method,
16000 but assumes that this library or executable won't link against any other
16001 ID shared libraries. That allows the compiler to use faster code for jumps
16004 @item -mno-leaf-id-shared-library
16005 @opindex mno-leaf-id-shared-library
16006 Do not assume that the code being compiled won't link against any ID shared
16007 libraries. Slower code is generated for jump and call insns.
16009 @item -mshared-library-id=n
16010 @opindex mshared-library-id
16011 Specifies the identification number of the ID-based shared library being
16012 compiled. Specifying a value of 0 generates more compact code; specifying
16013 other values forces the allocation of that number to the current
16014 library but is no more space- or time-efficient than omitting this option.
16018 Generate code that allows the data segment to be located in a different
16019 area of memory from the text segment. This allows for execute in place in
16020 an environment without virtual memory management by eliminating relocations
16021 against the text section.
16023 @item -mno-sep-data
16024 @opindex mno-sep-data
16025 Generate code that assumes that the data segment follows the text segment.
16026 This is the default.
16029 @itemx -mno-long-calls
16030 @opindex mlong-calls
16031 @opindex mno-long-calls
16032 Tells the compiler to perform function calls by first loading the
16033 address of the function into a register and then performing a subroutine
16034 call on this register. This switch is needed if the target function
16035 lies outside of the 24-bit addressing range of the offset-based
16036 version of subroutine call instruction.
16038 This feature is not enabled by default. Specifying
16039 @option{-mno-long-calls} restores the default behavior. Note these
16040 switches have no effect on how the compiler generates code to handle
16041 function calls via function pointers.
16045 Link with the fast floating-point library. This library relaxes some of
16046 the IEEE floating-point standard's rules for checking inputs against
16047 Not-a-Number (NAN), in the interest of performance.
16050 @opindex minline-plt
16051 Enable inlining of PLT entries in function calls to functions that are
16052 not known to bind locally. It has no effect without @option{-mfdpic}.
16055 @opindex mmulticore
16056 Build a standalone application for multicore Blackfin processors.
16057 This option causes proper start files and link scripts supporting
16058 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
16059 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
16061 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
16062 selects the one-application-per-core programming model. Without
16063 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
16064 programming model is used. In this model, the main function of Core B
16065 should be named as @code{coreb_main}.
16067 If this option is not used, the single-core application programming
16072 Build a standalone application for Core A of BF561 when using
16073 the one-application-per-core programming model. Proper start files
16074 and link scripts are used to support Core A, and the macro
16075 @code{__BFIN_COREA} is defined.
16076 This option can only be used in conjunction with @option{-mmulticore}.
16080 Build a standalone application for Core B of BF561 when using
16081 the one-application-per-core programming model. Proper start files
16082 and link scripts are used to support Core B, and the macro
16083 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
16084 should be used instead of @code{main}.
16085 This option can only be used in conjunction with @option{-mmulticore}.
16089 Build a standalone application for SDRAM. Proper start files and
16090 link scripts are used to put the application into SDRAM, and the macro
16091 @code{__BFIN_SDRAM} is defined.
16092 The loader should initialize SDRAM before loading the application.
16096 Assume that ICPLBs are enabled at run time. This has an effect on certain
16097 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
16098 are enabled; for standalone applications the default is off.
16102 @subsection C6X Options
16103 @cindex C6X Options
16106 @item -march=@var{name}
16108 This specifies the name of the target architecture. GCC uses this
16109 name to determine what kind of instructions it can emit when generating
16110 assembly code. Permissible names are: @samp{c62x},
16111 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
16114 @opindex mbig-endian
16115 Generate code for a big-endian target.
16117 @item -mlittle-endian
16118 @opindex mlittle-endian
16119 Generate code for a little-endian target. This is the default.
16123 Choose startup files and linker script suitable for the simulator.
16125 @item -msdata=default
16126 @opindex msdata=default
16127 Put small global and static data in the @code{.neardata} section,
16128 which is pointed to by register @code{B14}. Put small uninitialized
16129 global and static data in the @code{.bss} section, which is adjacent
16130 to the @code{.neardata} section. Put small read-only data into the
16131 @code{.rodata} section. The corresponding sections used for large
16132 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
16135 @opindex msdata=all
16136 Put all data, not just small objects, into the sections reserved for
16137 small data, and use addressing relative to the @code{B14} register to
16141 @opindex msdata=none
16142 Make no use of the sections reserved for small data, and use absolute
16143 addresses to access all data. Put all initialized global and static
16144 data in the @code{.fardata} section, and all uninitialized data in the
16145 @code{.far} section. Put all constant data into the @code{.const}
16150 @subsection CRIS Options
16151 @cindex CRIS Options
16153 These options are defined specifically for the CRIS ports.
16156 @item -march=@var{architecture-type}
16157 @itemx -mcpu=@var{architecture-type}
16160 Generate code for the specified architecture. The choices for
16161 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
16162 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
16163 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
16166 @item -mtune=@var{architecture-type}
16168 Tune to @var{architecture-type} everything applicable about the generated
16169 code, except for the ABI and the set of available instructions. The
16170 choices for @var{architecture-type} are the same as for
16171 @option{-march=@var{architecture-type}}.
16173 @item -mmax-stack-frame=@var{n}
16174 @opindex mmax-stack-frame
16175 Warn when the stack frame of a function exceeds @var{n} bytes.
16181 The options @option{-metrax4} and @option{-metrax100} are synonyms for
16182 @option{-march=v3} and @option{-march=v8} respectively.
16184 @item -mmul-bug-workaround
16185 @itemx -mno-mul-bug-workaround
16186 @opindex mmul-bug-workaround
16187 @opindex mno-mul-bug-workaround
16188 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
16189 models where it applies. This option is active by default.
16193 Enable CRIS-specific verbose debug-related information in the assembly
16194 code. This option also has the effect of turning off the @samp{#NO_APP}
16195 formatted-code indicator to the assembler at the beginning of the
16200 Do not use condition-code results from previous instruction; always emit
16201 compare and test instructions before use of condition codes.
16203 @item -mno-side-effects
16204 @opindex mno-side-effects
16205 Do not emit instructions with side effects in addressing modes other than
16208 @item -mstack-align
16209 @itemx -mno-stack-align
16210 @itemx -mdata-align
16211 @itemx -mno-data-align
16212 @itemx -mconst-align
16213 @itemx -mno-const-align
16214 @opindex mstack-align
16215 @opindex mno-stack-align
16216 @opindex mdata-align
16217 @opindex mno-data-align
16218 @opindex mconst-align
16219 @opindex mno-const-align
16220 These options (@samp{no-} options) arrange (eliminate arrangements) for the
16221 stack frame, individual data and constants to be aligned for the maximum
16222 single data access size for the chosen CPU model. The default is to
16223 arrange for 32-bit alignment. ABI details such as structure layout are
16224 not affected by these options.
16232 Similar to the stack- data- and const-align options above, these options
16233 arrange for stack frame, writable data and constants to all be 32-bit,
16234 16-bit or 8-bit aligned. The default is 32-bit alignment.
16236 @item -mno-prologue-epilogue
16237 @itemx -mprologue-epilogue
16238 @opindex mno-prologue-epilogue
16239 @opindex mprologue-epilogue
16240 With @option{-mno-prologue-epilogue}, the normal function prologue and
16241 epilogue which set up the stack frame are omitted and no return
16242 instructions or return sequences are generated in the code. Use this
16243 option only together with visual inspection of the compiled code: no
16244 warnings or errors are generated when call-saved registers must be saved,
16245 or storage for local variables needs to be allocated.
16249 @opindex mno-gotplt
16251 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
16252 instruction sequences that load addresses for functions from the PLT part
16253 of the GOT rather than (traditional on other architectures) calls to the
16254 PLT@. The default is @option{-mgotplt}.
16258 Legacy no-op option only recognized with the cris-axis-elf and
16259 cris-axis-linux-gnu targets.
16263 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
16267 This option, recognized for the cris-axis-elf, arranges
16268 to link with input-output functions from a simulator library. Code,
16269 initialized data and zero-initialized data are allocated consecutively.
16273 Like @option{-sim}, but pass linker options to locate initialized data at
16274 0x40000000 and zero-initialized data at 0x80000000.
16278 @subsection CR16 Options
16279 @cindex CR16 Options
16281 These options are defined specifically for the CR16 ports.
16287 Enable the use of multiply-accumulate instructions. Disabled by default.
16291 @opindex mcr16cplus
16293 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
16298 Links the library libsim.a which is in compatible with simulator. Applicable
16299 to ELF compiler only.
16303 Choose integer type as 32-bit wide.
16307 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
16309 @item -mdata-model=@var{model}
16310 @opindex mdata-model
16311 Choose a data model. The choices for @var{model} are @samp{near},
16312 @samp{far} or @samp{medium}. @samp{medium} is default.
16313 However, @samp{far} is not valid with @option{-mcr16c}, as the
16314 CR16C architecture does not support the far data model.
16317 @node Darwin Options
16318 @subsection Darwin Options
16319 @cindex Darwin options
16321 These options are defined for all architectures running the Darwin operating
16324 FSF GCC on Darwin does not create ``fat'' object files; it creates
16325 an object file for the single architecture that GCC was built to
16326 target. Apple's GCC on Darwin does create ``fat'' files if multiple
16327 @option{-arch} options are used; it does so by running the compiler or
16328 linker multiple times and joining the results together with
16331 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
16332 @samp{i686}) is determined by the flags that specify the ISA
16333 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
16334 @option{-force_cpusubtype_ALL} option can be used to override this.
16336 The Darwin tools vary in their behavior when presented with an ISA
16337 mismatch. The assembler, @file{as}, only permits instructions to
16338 be used that are valid for the subtype of the file it is generating,
16339 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
16340 The linker for shared libraries, @file{/usr/bin/libtool}, fails
16341 and prints an error if asked to create a shared library with a less
16342 restrictive subtype than its input files (for instance, trying to put
16343 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
16344 for executables, @command{ld}, quietly gives the executable the most
16345 restrictive subtype of any of its input files.
16350 Add the framework directory @var{dir} to the head of the list of
16351 directories to be searched for header files. These directories are
16352 interleaved with those specified by @option{-I} options and are
16353 scanned in a left-to-right order.
16355 A framework directory is a directory with frameworks in it. A
16356 framework is a directory with a @file{Headers} and/or
16357 @file{PrivateHeaders} directory contained directly in it that ends
16358 in @file{.framework}. The name of a framework is the name of this
16359 directory excluding the @file{.framework}. Headers associated with
16360 the framework are found in one of those two directories, with
16361 @file{Headers} being searched first. A subframework is a framework
16362 directory that is in a framework's @file{Frameworks} directory.
16363 Includes of subframework headers can only appear in a header of a
16364 framework that contains the subframework, or in a sibling subframework
16365 header. Two subframeworks are siblings if they occur in the same
16366 framework. A subframework should not have the same name as a
16367 framework; a warning is issued if this is violated. Currently a
16368 subframework cannot have subframeworks; in the future, the mechanism
16369 may be extended to support this. The standard frameworks can be found
16370 in @file{/System/Library/Frameworks} and
16371 @file{/Library/Frameworks}. An example include looks like
16372 @code{#include <Framework/header.h>}, where @file{Framework} denotes
16373 the name of the framework and @file{header.h} is found in the
16374 @file{PrivateHeaders} or @file{Headers} directory.
16376 @item -iframework@var{dir}
16377 @opindex iframework
16378 Like @option{-F} except the directory is a treated as a system
16379 directory. The main difference between this @option{-iframework} and
16380 @option{-F} is that with @option{-iframework} the compiler does not
16381 warn about constructs contained within header files found via
16382 @var{dir}. This option is valid only for the C family of languages.
16386 Emit debugging information for symbols that are used. For stabs
16387 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
16388 This is by default ON@.
16392 Emit debugging information for all symbols and types.
16394 @item -mmacosx-version-min=@var{version}
16395 The earliest version of MacOS X that this executable will run on
16396 is @var{version}. Typical values of @var{version} include @code{10.1},
16397 @code{10.2}, and @code{10.3.9}.
16399 If the compiler was built to use the system's headers by default,
16400 then the default for this option is the system version on which the
16401 compiler is running, otherwise the default is to make choices that
16402 are compatible with as many systems and code bases as possible.
16406 Enable kernel development mode. The @option{-mkernel} option sets
16407 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
16408 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
16409 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
16410 applicable. This mode also sets @option{-mno-altivec},
16411 @option{-msoft-float}, @option{-fno-builtin} and
16412 @option{-mlong-branch} for PowerPC targets.
16414 @item -mone-byte-bool
16415 @opindex mone-byte-bool
16416 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
16417 By default @code{sizeof(bool)} is @code{4} when compiling for
16418 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
16419 option has no effect on x86.
16421 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
16422 to generate code that is not binary compatible with code generated
16423 without that switch. Using this switch may require recompiling all
16424 other modules in a program, including system libraries. Use this
16425 switch to conform to a non-default data model.
16427 @item -mfix-and-continue
16428 @itemx -ffix-and-continue
16429 @itemx -findirect-data
16430 @opindex mfix-and-continue
16431 @opindex ffix-and-continue
16432 @opindex findirect-data
16433 Generate code suitable for fast turnaround development, such as to
16434 allow GDB to dynamically load @file{.o} files into already-running
16435 programs. @option{-findirect-data} and @option{-ffix-and-continue}
16436 are provided for backwards compatibility.
16440 Loads all members of static archive libraries.
16441 See man ld(1) for more information.
16443 @item -arch_errors_fatal
16444 @opindex arch_errors_fatal
16445 Cause the errors having to do with files that have the wrong architecture
16448 @item -bind_at_load
16449 @opindex bind_at_load
16450 Causes the output file to be marked such that the dynamic linker will
16451 bind all undefined references when the file is loaded or launched.
16455 Produce a Mach-o bundle format file.
16456 See man ld(1) for more information.
16458 @item -bundle_loader @var{executable}
16459 @opindex bundle_loader
16460 This option specifies the @var{executable} that will load the build
16461 output file being linked. See man ld(1) for more information.
16464 @opindex dynamiclib
16465 When passed this option, GCC produces a dynamic library instead of
16466 an executable when linking, using the Darwin @file{libtool} command.
16468 @item -force_cpusubtype_ALL
16469 @opindex force_cpusubtype_ALL
16470 This causes GCC's output file to have the @samp{ALL} subtype, instead of
16471 one controlled by the @option{-mcpu} or @option{-march} option.
16473 @item -allowable_client @var{client_name}
16474 @itemx -client_name
16475 @itemx -compatibility_version
16476 @itemx -current_version
16478 @itemx -dependency-file
16480 @itemx -dylinker_install_name
16482 @itemx -exported_symbols_list
16485 @itemx -flat_namespace
16486 @itemx -force_flat_namespace
16487 @itemx -headerpad_max_install_names
16490 @itemx -install_name
16491 @itemx -keep_private_externs
16492 @itemx -multi_module
16493 @itemx -multiply_defined
16494 @itemx -multiply_defined_unused
16497 @itemx -no_dead_strip_inits_and_terms
16498 @itemx -nofixprebinding
16499 @itemx -nomultidefs
16501 @itemx -noseglinkedit
16502 @itemx -pagezero_size
16504 @itemx -prebind_all_twolevel_modules
16505 @itemx -private_bundle
16507 @itemx -read_only_relocs
16509 @itemx -sectobjectsymbols
16513 @itemx -sectobjectsymbols
16516 @itemx -segs_read_only_addr
16518 @itemx -segs_read_write_addr
16519 @itemx -seg_addr_table
16520 @itemx -seg_addr_table_filename
16521 @itemx -seglinkedit
16523 @itemx -segs_read_only_addr
16524 @itemx -segs_read_write_addr
16525 @itemx -single_module
16527 @itemx -sub_library
16529 @itemx -sub_umbrella
16530 @itemx -twolevel_namespace
16533 @itemx -unexported_symbols_list
16534 @itemx -weak_reference_mismatches
16535 @itemx -whatsloaded
16536 @opindex allowable_client
16537 @opindex client_name
16538 @opindex compatibility_version
16539 @opindex current_version
16540 @opindex dead_strip
16541 @opindex dependency-file
16542 @opindex dylib_file
16543 @opindex dylinker_install_name
16545 @opindex exported_symbols_list
16547 @opindex flat_namespace
16548 @opindex force_flat_namespace
16549 @opindex headerpad_max_install_names
16550 @opindex image_base
16552 @opindex install_name
16553 @opindex keep_private_externs
16554 @opindex multi_module
16555 @opindex multiply_defined
16556 @opindex multiply_defined_unused
16557 @opindex noall_load
16558 @opindex no_dead_strip_inits_and_terms
16559 @opindex nofixprebinding
16560 @opindex nomultidefs
16562 @opindex noseglinkedit
16563 @opindex pagezero_size
16565 @opindex prebind_all_twolevel_modules
16566 @opindex private_bundle
16567 @opindex read_only_relocs
16569 @opindex sectobjectsymbols
16572 @opindex sectcreate
16573 @opindex sectobjectsymbols
16576 @opindex segs_read_only_addr
16577 @opindex segs_read_write_addr
16578 @opindex seg_addr_table
16579 @opindex seg_addr_table_filename
16580 @opindex seglinkedit
16582 @opindex segs_read_only_addr
16583 @opindex segs_read_write_addr
16584 @opindex single_module
16586 @opindex sub_library
16587 @opindex sub_umbrella
16588 @opindex twolevel_namespace
16591 @opindex unexported_symbols_list
16592 @opindex weak_reference_mismatches
16593 @opindex whatsloaded
16594 These options are passed to the Darwin linker. The Darwin linker man page
16595 describes them in detail.
16598 @node DEC Alpha Options
16599 @subsection DEC Alpha Options
16601 These @samp{-m} options are defined for the DEC Alpha implementations:
16604 @item -mno-soft-float
16605 @itemx -msoft-float
16606 @opindex mno-soft-float
16607 @opindex msoft-float
16608 Use (do not use) the hardware floating-point instructions for
16609 floating-point operations. When @option{-msoft-float} is specified,
16610 functions in @file{libgcc.a} are used to perform floating-point
16611 operations. Unless they are replaced by routines that emulate the
16612 floating-point operations, or compiled in such a way as to call such
16613 emulations routines, these routines issue floating-point
16614 operations. If you are compiling for an Alpha without floating-point
16615 operations, you must ensure that the library is built so as not to call
16618 Note that Alpha implementations without floating-point operations are
16619 required to have floating-point registers.
16622 @itemx -mno-fp-regs
16624 @opindex mno-fp-regs
16625 Generate code that uses (does not use) the floating-point register set.
16626 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
16627 register set is not used, floating-point operands are passed in integer
16628 registers as if they were integers and floating-point results are passed
16629 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
16630 so any function with a floating-point argument or return value called by code
16631 compiled with @option{-mno-fp-regs} must also be compiled with that
16634 A typical use of this option is building a kernel that does not use,
16635 and hence need not save and restore, any floating-point registers.
16639 The Alpha architecture implements floating-point hardware optimized for
16640 maximum performance. It is mostly compliant with the IEEE floating-point
16641 standard. However, for full compliance, software assistance is
16642 required. This option generates code fully IEEE-compliant code
16643 @emph{except} that the @var{inexact-flag} is not maintained (see below).
16644 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
16645 defined during compilation. The resulting code is less efficient but is
16646 able to correctly support denormalized numbers and exceptional IEEE
16647 values such as not-a-number and plus/minus infinity. Other Alpha
16648 compilers call this option @option{-ieee_with_no_inexact}.
16650 @item -mieee-with-inexact
16651 @opindex mieee-with-inexact
16652 This is like @option{-mieee} except the generated code also maintains
16653 the IEEE @var{inexact-flag}. Turning on this option causes the
16654 generated code to implement fully-compliant IEEE math. In addition to
16655 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
16656 macro. On some Alpha implementations the resulting code may execute
16657 significantly slower than the code generated by default. Since there is
16658 very little code that depends on the @var{inexact-flag}, you should
16659 normally not specify this option. Other Alpha compilers call this
16660 option @option{-ieee_with_inexact}.
16662 @item -mfp-trap-mode=@var{trap-mode}
16663 @opindex mfp-trap-mode
16664 This option controls what floating-point related traps are enabled.
16665 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
16666 The trap mode can be set to one of four values:
16670 This is the default (normal) setting. The only traps that are enabled
16671 are the ones that cannot be disabled in software (e.g., division by zero
16675 In addition to the traps enabled by @samp{n}, underflow traps are enabled
16679 Like @samp{u}, but the instructions are marked to be safe for software
16680 completion (see Alpha architecture manual for details).
16683 Like @samp{su}, but inexact traps are enabled as well.
16686 @item -mfp-rounding-mode=@var{rounding-mode}
16687 @opindex mfp-rounding-mode
16688 Selects the IEEE rounding mode. Other Alpha compilers call this option
16689 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
16694 Normal IEEE rounding mode. Floating-point numbers are rounded towards
16695 the nearest machine number or towards the even machine number in case
16699 Round towards minus infinity.
16702 Chopped rounding mode. Floating-point numbers are rounded towards zero.
16705 Dynamic rounding mode. A field in the floating-point control register
16706 (@var{fpcr}, see Alpha architecture reference manual) controls the
16707 rounding mode in effect. The C library initializes this register for
16708 rounding towards plus infinity. Thus, unless your program modifies the
16709 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
16712 @item -mtrap-precision=@var{trap-precision}
16713 @opindex mtrap-precision
16714 In the Alpha architecture, floating-point traps are imprecise. This
16715 means without software assistance it is impossible to recover from a
16716 floating trap and program execution normally needs to be terminated.
16717 GCC can generate code that can assist operating system trap handlers
16718 in determining the exact location that caused a floating-point trap.
16719 Depending on the requirements of an application, different levels of
16720 precisions can be selected:
16724 Program precision. This option is the default and means a trap handler
16725 can only identify which program caused a floating-point exception.
16728 Function precision. The trap handler can determine the function that
16729 caused a floating-point exception.
16732 Instruction precision. The trap handler can determine the exact
16733 instruction that caused a floating-point exception.
16736 Other Alpha compilers provide the equivalent options called
16737 @option{-scope_safe} and @option{-resumption_safe}.
16739 @item -mieee-conformant
16740 @opindex mieee-conformant
16741 This option marks the generated code as IEEE conformant. You must not
16742 use this option unless you also specify @option{-mtrap-precision=i} and either
16743 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
16744 is to emit the line @samp{.eflag 48} in the function prologue of the
16745 generated assembly file.
16747 @item -mbuild-constants
16748 @opindex mbuild-constants
16749 Normally GCC examines a 32- or 64-bit integer constant to
16750 see if it can construct it from smaller constants in two or three
16751 instructions. If it cannot, it outputs the constant as a literal and
16752 generates code to load it from the data segment at run time.
16754 Use this option to require GCC to construct @emph{all} integer constants
16755 using code, even if it takes more instructions (the maximum is six).
16757 You typically use this option to build a shared library dynamic
16758 loader. Itself a shared library, it must relocate itself in memory
16759 before it can find the variables and constants in its own data segment.
16777 Indicate whether GCC should generate code to use the optional BWX,
16778 CIX, FIX and MAX instruction sets. The default is to use the instruction
16779 sets supported by the CPU type specified via @option{-mcpu=} option or that
16780 of the CPU on which GCC was built if none is specified.
16783 @itemx -mfloat-ieee
16784 @opindex mfloat-vax
16785 @opindex mfloat-ieee
16786 Generate code that uses (does not use) VAX F and G floating-point
16787 arithmetic instead of IEEE single and double precision.
16789 @item -mexplicit-relocs
16790 @itemx -mno-explicit-relocs
16791 @opindex mexplicit-relocs
16792 @opindex mno-explicit-relocs
16793 Older Alpha assemblers provided no way to generate symbol relocations
16794 except via assembler macros. Use of these macros does not allow
16795 optimal instruction scheduling. GNU binutils as of version 2.12
16796 supports a new syntax that allows the compiler to explicitly mark
16797 which relocations should apply to which instructions. This option
16798 is mostly useful for debugging, as GCC detects the capabilities of
16799 the assembler when it is built and sets the default accordingly.
16802 @itemx -mlarge-data
16803 @opindex msmall-data
16804 @opindex mlarge-data
16805 When @option{-mexplicit-relocs} is in effect, static data is
16806 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
16807 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
16808 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
16809 16-bit relocations off of the @code{$gp} register. This limits the
16810 size of the small data area to 64KB, but allows the variables to be
16811 directly accessed via a single instruction.
16813 The default is @option{-mlarge-data}. With this option the data area
16814 is limited to just below 2GB@. Programs that require more than 2GB of
16815 data must use @code{malloc} or @code{mmap} to allocate the data in the
16816 heap instead of in the program's data segment.
16818 When generating code for shared libraries, @option{-fpic} implies
16819 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
16822 @itemx -mlarge-text
16823 @opindex msmall-text
16824 @opindex mlarge-text
16825 When @option{-msmall-text} is used, the compiler assumes that the
16826 code of the entire program (or shared library) fits in 4MB, and is
16827 thus reachable with a branch instruction. When @option{-msmall-data}
16828 is used, the compiler can assume that all local symbols share the
16829 same @code{$gp} value, and thus reduce the number of instructions
16830 required for a function call from 4 to 1.
16832 The default is @option{-mlarge-text}.
16834 @item -mcpu=@var{cpu_type}
16836 Set the instruction set and instruction scheduling parameters for
16837 machine type @var{cpu_type}. You can specify either the @samp{EV}
16838 style name or the corresponding chip number. GCC supports scheduling
16839 parameters for the EV4, EV5 and EV6 family of processors and
16840 chooses the default values for the instruction set from the processor
16841 you specify. If you do not specify a processor type, GCC defaults
16842 to the processor on which the compiler was built.
16844 Supported values for @var{cpu_type} are
16850 Schedules as an EV4 and has no instruction set extensions.
16854 Schedules as an EV5 and has no instruction set extensions.
16858 Schedules as an EV5 and supports the BWX extension.
16863 Schedules as an EV5 and supports the BWX and MAX extensions.
16867 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
16871 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
16874 Native toolchains also support the value @samp{native},
16875 which selects the best architecture option for the host processor.
16876 @option{-mcpu=native} has no effect if GCC does not recognize
16879 @item -mtune=@var{cpu_type}
16881 Set only the instruction scheduling parameters for machine type
16882 @var{cpu_type}. The instruction set is not changed.
16884 Native toolchains also support the value @samp{native},
16885 which selects the best architecture option for the host processor.
16886 @option{-mtune=native} has no effect if GCC does not recognize
16889 @item -mmemory-latency=@var{time}
16890 @opindex mmemory-latency
16891 Sets the latency the scheduler should assume for typical memory
16892 references as seen by the application. This number is highly
16893 dependent on the memory access patterns used by the application
16894 and the size of the external cache on the machine.
16896 Valid options for @var{time} are
16900 A decimal number representing clock cycles.
16906 The compiler contains estimates of the number of clock cycles for
16907 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
16908 (also called Dcache, Scache, and Bcache), as well as to main memory.
16909 Note that L3 is only valid for EV5.
16915 @subsection FR30 Options
16916 @cindex FR30 Options
16918 These options are defined specifically for the FR30 port.
16922 @item -msmall-model
16923 @opindex msmall-model
16924 Use the small address space model. This can produce smaller code, but
16925 it does assume that all symbolic values and addresses fit into a
16930 Assume that runtime support has been provided and so there is no need
16931 to include the simulator library (@file{libsim.a}) on the linker
16937 @subsection FT32 Options
16938 @cindex FT32 Options
16940 These options are defined specifically for the FT32 port.
16946 Specifies that the program will be run on the simulator. This causes
16947 an alternate runtime startup and library to be linked.
16948 You must not use this option when generating programs that will run on
16949 real hardware; you must provide your own runtime library for whatever
16950 I/O functions are needed.
16954 Enable Local Register Allocation. This is still experimental for FT32,
16955 so by default the compiler uses standard reload.
16959 Do not use div and mod instructions.
16964 @subsection FRV Options
16965 @cindex FRV Options
16971 Only use the first 32 general-purpose registers.
16976 Use all 64 general-purpose registers.
16981 Use only the first 32 floating-point registers.
16986 Use all 64 floating-point registers.
16989 @opindex mhard-float
16991 Use hardware instructions for floating-point operations.
16994 @opindex msoft-float
16996 Use library routines for floating-point operations.
17001 Dynamically allocate condition code registers.
17006 Do not try to dynamically allocate condition code registers, only
17007 use @code{icc0} and @code{fcc0}.
17012 Change ABI to use double word insns.
17017 Do not use double word instructions.
17022 Use floating-point double instructions.
17025 @opindex mno-double
17027 Do not use floating-point double instructions.
17032 Use media instructions.
17037 Do not use media instructions.
17042 Use multiply and add/subtract instructions.
17045 @opindex mno-muladd
17047 Do not use multiply and add/subtract instructions.
17052 Select the FDPIC ABI, which uses function descriptors to represent
17053 pointers to functions. Without any PIC/PIE-related options, it
17054 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
17055 assumes GOT entries and small data are within a 12-bit range from the
17056 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
17057 are computed with 32 bits.
17058 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17061 @opindex minline-plt
17063 Enable inlining of PLT entries in function calls to functions that are
17064 not known to bind locally. It has no effect without @option{-mfdpic}.
17065 It's enabled by default if optimizing for speed and compiling for
17066 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
17067 optimization option such as @option{-O3} or above is present in the
17073 Assume a large TLS segment when generating thread-local code.
17078 Do not assume a large TLS segment when generating thread-local code.
17083 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
17084 that is known to be in read-only sections. It's enabled by default,
17085 except for @option{-fpic} or @option{-fpie}: even though it may help
17086 make the global offset table smaller, it trades 1 instruction for 4.
17087 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
17088 one of which may be shared by multiple symbols, and it avoids the need
17089 for a GOT entry for the referenced symbol, so it's more likely to be a
17090 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
17092 @item -multilib-library-pic
17093 @opindex multilib-library-pic
17095 Link with the (library, not FD) pic libraries. It's implied by
17096 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
17097 @option{-fpic} without @option{-mfdpic}. You should never have to use
17101 @opindex mlinked-fp
17103 Follow the EABI requirement of always creating a frame pointer whenever
17104 a stack frame is allocated. This option is enabled by default and can
17105 be disabled with @option{-mno-linked-fp}.
17108 @opindex mlong-calls
17110 Use indirect addressing to call functions outside the current
17111 compilation unit. This allows the functions to be placed anywhere
17112 within the 32-bit address space.
17114 @item -malign-labels
17115 @opindex malign-labels
17117 Try to align labels to an 8-byte boundary by inserting NOPs into the
17118 previous packet. This option only has an effect when VLIW packing
17119 is enabled. It doesn't create new packets; it merely adds NOPs to
17122 @item -mlibrary-pic
17123 @opindex mlibrary-pic
17125 Generate position-independent EABI code.
17130 Use only the first four media accumulator registers.
17135 Use all eight media accumulator registers.
17140 Pack VLIW instructions.
17145 Do not pack VLIW instructions.
17148 @opindex mno-eflags
17150 Do not mark ABI switches in e_flags.
17153 @opindex mcond-move
17155 Enable the use of conditional-move instructions (default).
17157 This switch is mainly for debugging the compiler and will likely be removed
17158 in a future version.
17160 @item -mno-cond-move
17161 @opindex mno-cond-move
17163 Disable the use of conditional-move instructions.
17165 This switch is mainly for debugging the compiler and will likely be removed
17166 in a future version.
17171 Enable the use of conditional set instructions (default).
17173 This switch is mainly for debugging the compiler and will likely be removed
17174 in a future version.
17179 Disable the use of conditional set instructions.
17181 This switch is mainly for debugging the compiler and will likely be removed
17182 in a future version.
17185 @opindex mcond-exec
17187 Enable the use of conditional execution (default).
17189 This switch is mainly for debugging the compiler and will likely be removed
17190 in a future version.
17192 @item -mno-cond-exec
17193 @opindex mno-cond-exec
17195 Disable the use of conditional execution.
17197 This switch is mainly for debugging the compiler and will likely be removed
17198 in a future version.
17200 @item -mvliw-branch
17201 @opindex mvliw-branch
17203 Run a pass to pack branches into VLIW instructions (default).
17205 This switch is mainly for debugging the compiler and will likely be removed
17206 in a future version.
17208 @item -mno-vliw-branch
17209 @opindex mno-vliw-branch
17211 Do not run a pass to pack branches into VLIW instructions.
17213 This switch is mainly for debugging the compiler and will likely be removed
17214 in a future version.
17216 @item -mmulti-cond-exec
17217 @opindex mmulti-cond-exec
17219 Enable optimization of @code{&&} and @code{||} in conditional execution
17222 This switch is mainly for debugging the compiler and will likely be removed
17223 in a future version.
17225 @item -mno-multi-cond-exec
17226 @opindex mno-multi-cond-exec
17228 Disable optimization of @code{&&} and @code{||} in conditional execution.
17230 This switch is mainly for debugging the compiler and will likely be removed
17231 in a future version.
17233 @item -mnested-cond-exec
17234 @opindex mnested-cond-exec
17236 Enable nested conditional execution optimizations (default).
17238 This switch is mainly for debugging the compiler and will likely be removed
17239 in a future version.
17241 @item -mno-nested-cond-exec
17242 @opindex mno-nested-cond-exec
17244 Disable nested conditional execution optimizations.
17246 This switch is mainly for debugging the compiler and will likely be removed
17247 in a future version.
17249 @item -moptimize-membar
17250 @opindex moptimize-membar
17252 This switch removes redundant @code{membar} instructions from the
17253 compiler-generated code. It is enabled by default.
17255 @item -mno-optimize-membar
17256 @opindex mno-optimize-membar
17258 This switch disables the automatic removal of redundant @code{membar}
17259 instructions from the generated code.
17261 @item -mtomcat-stats
17262 @opindex mtomcat-stats
17264 Cause gas to print out tomcat statistics.
17266 @item -mcpu=@var{cpu}
17269 Select the processor type for which to generate code. Possible values are
17270 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
17271 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
17275 @node GNU/Linux Options
17276 @subsection GNU/Linux Options
17278 These @samp{-m} options are defined for GNU/Linux targets:
17283 Use the GNU C library. This is the default except
17284 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
17285 @samp{*-*-linux-*android*} targets.
17289 Use uClibc C library. This is the default on
17290 @samp{*-*-linux-*uclibc*} targets.
17294 Use the musl C library. This is the default on
17295 @samp{*-*-linux-*musl*} targets.
17299 Use Bionic C library. This is the default on
17300 @samp{*-*-linux-*android*} targets.
17304 Compile code compatible with Android platform. This is the default on
17305 @samp{*-*-linux-*android*} targets.
17307 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
17308 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
17309 this option makes the GCC driver pass Android-specific options to the linker.
17310 Finally, this option causes the preprocessor macro @code{__ANDROID__}
17313 @item -tno-android-cc
17314 @opindex tno-android-cc
17315 Disable compilation effects of @option{-mandroid}, i.e., do not enable
17316 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
17317 @option{-fno-rtti} by default.
17319 @item -tno-android-ld
17320 @opindex tno-android-ld
17321 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
17322 linking options to the linker.
17326 @node H8/300 Options
17327 @subsection H8/300 Options
17329 These @samp{-m} options are defined for the H8/300 implementations:
17334 Shorten some address references at link time, when possible; uses the
17335 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
17336 ld, Using ld}, for a fuller description.
17340 Generate code for the H8/300H@.
17344 Generate code for the H8S@.
17348 Generate code for the H8S and H8/300H in the normal mode. This switch
17349 must be used either with @option{-mh} or @option{-ms}.
17353 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
17357 Extended registers are stored on stack before execution of function
17358 with monitor attribute. Default option is @option{-mexr}.
17359 This option is valid only for H8S targets.
17363 Extended registers are not stored on stack before execution of function
17364 with monitor attribute. Default option is @option{-mno-exr}.
17365 This option is valid only for H8S targets.
17369 Make @code{int} data 32 bits by default.
17372 @opindex malign-300
17373 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
17374 The default for the H8/300H and H8S is to align longs and floats on
17376 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
17377 This option has no effect on the H8/300.
17381 @subsection HPPA Options
17382 @cindex HPPA Options
17384 These @samp{-m} options are defined for the HPPA family of computers:
17387 @item -march=@var{architecture-type}
17389 Generate code for the specified architecture. The choices for
17390 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
17391 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
17392 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
17393 architecture option for your machine. Code compiled for lower numbered
17394 architectures runs on higher numbered architectures, but not the
17397 @item -mpa-risc-1-0
17398 @itemx -mpa-risc-1-1
17399 @itemx -mpa-risc-2-0
17400 @opindex mpa-risc-1-0
17401 @opindex mpa-risc-1-1
17402 @opindex mpa-risc-2-0
17403 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
17405 @item -mcaller-copies
17406 @opindex mcaller-copies
17407 The caller copies function arguments passed by hidden reference. This
17408 option should be used with care as it is not compatible with the default
17409 32-bit runtime. However, only aggregates larger than eight bytes are
17410 passed by hidden reference and the option provides better compatibility
17413 @item -mjump-in-delay
17414 @opindex mjump-in-delay
17415 This option is ignored and provided for compatibility purposes only.
17417 @item -mdisable-fpregs
17418 @opindex mdisable-fpregs
17419 Prevent floating-point registers from being used in any manner. This is
17420 necessary for compiling kernels that perform lazy context switching of
17421 floating-point registers. If you use this option and attempt to perform
17422 floating-point operations, the compiler aborts.
17424 @item -mdisable-indexing
17425 @opindex mdisable-indexing
17426 Prevent the compiler from using indexing address modes. This avoids some
17427 rather obscure problems when compiling MIG generated code under MACH@.
17429 @item -mno-space-regs
17430 @opindex mno-space-regs
17431 Generate code that assumes the target has no space registers. This allows
17432 GCC to generate faster indirect calls and use unscaled index address modes.
17434 Such code is suitable for level 0 PA systems and kernels.
17436 @item -mfast-indirect-calls
17437 @opindex mfast-indirect-calls
17438 Generate code that assumes calls never cross space boundaries. This
17439 allows GCC to emit code that performs faster indirect calls.
17441 This option does not work in the presence of shared libraries or nested
17444 @item -mfixed-range=@var{register-range}
17445 @opindex mfixed-range
17446 Generate code treating the given register range as fixed registers.
17447 A fixed register is one that the register allocator cannot use. This is
17448 useful when compiling kernel code. A register range is specified as
17449 two registers separated by a dash. Multiple register ranges can be
17450 specified separated by a comma.
17452 @item -mlong-load-store
17453 @opindex mlong-load-store
17454 Generate 3-instruction load and store sequences as sometimes required by
17455 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
17458 @item -mportable-runtime
17459 @opindex mportable-runtime
17460 Use the portable calling conventions proposed by HP for ELF systems.
17464 Enable the use of assembler directives only GAS understands.
17466 @item -mschedule=@var{cpu-type}
17468 Schedule code according to the constraints for the machine type
17469 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
17470 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
17471 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
17472 proper scheduling option for your machine. The default scheduling is
17476 @opindex mlinker-opt
17477 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
17478 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
17479 linkers in which they give bogus error messages when linking some programs.
17482 @opindex msoft-float
17483 Generate output containing library calls for floating point.
17484 @strong{Warning:} the requisite libraries are not available for all HPPA
17485 targets. Normally the facilities of the machine's usual C compiler are
17486 used, but this cannot be done directly in cross-compilation. You must make
17487 your own arrangements to provide suitable library functions for
17490 @option{-msoft-float} changes the calling convention in the output file;
17491 therefore, it is only useful if you compile @emph{all} of a program with
17492 this option. In particular, you need to compile @file{libgcc.a}, the
17493 library that comes with GCC, with @option{-msoft-float} in order for
17498 Generate the predefine, @code{_SIO}, for server IO@. The default is
17499 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
17500 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
17501 options are available under HP-UX and HI-UX@.
17505 Use options specific to GNU @command{ld}.
17506 This passes @option{-shared} to @command{ld} when
17507 building a shared library. It is the default when GCC is configured,
17508 explicitly or implicitly, with the GNU linker. This option does not
17509 affect which @command{ld} is called; it only changes what parameters
17510 are passed to that @command{ld}.
17511 The @command{ld} that is called is determined by the
17512 @option{--with-ld} configure option, GCC's program search path, and
17513 finally by the user's @env{PATH}. The linker used by GCC can be printed
17514 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
17515 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17519 Use options specific to HP @command{ld}.
17520 This passes @option{-b} to @command{ld} when building
17521 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
17522 links. It is the default when GCC is configured, explicitly or
17523 implicitly, with the HP linker. This option does not affect
17524 which @command{ld} is called; it only changes what parameters are passed to that
17526 The @command{ld} that is called is determined by the @option{--with-ld}
17527 configure option, GCC's program search path, and finally by the user's
17528 @env{PATH}. The linker used by GCC can be printed using @samp{which
17529 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
17530 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
17533 @opindex mno-long-calls
17534 Generate code that uses long call sequences. This ensures that a call
17535 is always able to reach linker generated stubs. The default is to generate
17536 long calls only when the distance from the call site to the beginning
17537 of the function or translation unit, as the case may be, exceeds a
17538 predefined limit set by the branch type being used. The limits for
17539 normal calls are 7,600,000 and 240,000 bytes, respectively for the
17540 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
17543 Distances are measured from the beginning of functions when using the
17544 @option{-ffunction-sections} option, or when using the @option{-mgas}
17545 and @option{-mno-portable-runtime} options together under HP-UX with
17548 It is normally not desirable to use this option as it degrades
17549 performance. However, it may be useful in large applications,
17550 particularly when partial linking is used to build the application.
17552 The types of long calls used depends on the capabilities of the
17553 assembler and linker, and the type of code being generated. The
17554 impact on systems that support long absolute calls, and long pic
17555 symbol-difference or pc-relative calls should be relatively small.
17556 However, an indirect call is used on 32-bit ELF systems in pic code
17557 and it is quite long.
17559 @item -munix=@var{unix-std}
17561 Generate compiler predefines and select a startfile for the specified
17562 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
17563 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
17564 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
17565 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
17566 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
17569 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
17570 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
17571 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
17572 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
17573 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
17574 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
17576 It is @emph{important} to note that this option changes the interfaces
17577 for various library routines. It also affects the operational behavior
17578 of the C library. Thus, @emph{extreme} care is needed in using this
17581 Library code that is intended to operate with more than one UNIX
17582 standard must test, set and restore the variable @code{__xpg4_extended_mask}
17583 as appropriate. Most GNU software doesn't provide this capability.
17587 Suppress the generation of link options to search libdld.sl when the
17588 @option{-static} option is specified on HP-UX 10 and later.
17592 The HP-UX implementation of setlocale in libc has a dependency on
17593 libdld.sl. There isn't an archive version of libdld.sl. Thus,
17594 when the @option{-static} option is specified, special link options
17595 are needed to resolve this dependency.
17597 On HP-UX 10 and later, the GCC driver adds the necessary options to
17598 link with libdld.sl when the @option{-static} option is specified.
17599 This causes the resulting binary to be dynamic. On the 64-bit port,
17600 the linkers generate dynamic binaries by default in any case. The
17601 @option{-nolibdld} option can be used to prevent the GCC driver from
17602 adding these link options.
17606 Add support for multithreading with the @dfn{dce thread} library
17607 under HP-UX@. This option sets flags for both the preprocessor and
17611 @node IA-64 Options
17612 @subsection IA-64 Options
17613 @cindex IA-64 Options
17615 These are the @samp{-m} options defined for the Intel IA-64 architecture.
17619 @opindex mbig-endian
17620 Generate code for a big-endian target. This is the default for HP-UX@.
17622 @item -mlittle-endian
17623 @opindex mlittle-endian
17624 Generate code for a little-endian target. This is the default for AIX5
17630 @opindex mno-gnu-as
17631 Generate (or don't) code for the GNU assembler. This is the default.
17632 @c Also, this is the default if the configure option @option{--with-gnu-as}
17638 @opindex mno-gnu-ld
17639 Generate (or don't) code for the GNU linker. This is the default.
17640 @c Also, this is the default if the configure option @option{--with-gnu-ld}
17645 Generate code that does not use a global pointer register. The result
17646 is not position independent code, and violates the IA-64 ABI@.
17648 @item -mvolatile-asm-stop
17649 @itemx -mno-volatile-asm-stop
17650 @opindex mvolatile-asm-stop
17651 @opindex mno-volatile-asm-stop
17652 Generate (or don't) a stop bit immediately before and after volatile asm
17655 @item -mregister-names
17656 @itemx -mno-register-names
17657 @opindex mregister-names
17658 @opindex mno-register-names
17659 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
17660 the stacked registers. This may make assembler output more readable.
17666 Disable (or enable) optimizations that use the small data section. This may
17667 be useful for working around optimizer bugs.
17669 @item -mconstant-gp
17670 @opindex mconstant-gp
17671 Generate code that uses a single constant global pointer value. This is
17672 useful when compiling kernel code.
17676 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
17677 This is useful when compiling firmware code.
17679 @item -minline-float-divide-min-latency
17680 @opindex minline-float-divide-min-latency
17681 Generate code for inline divides of floating-point values
17682 using the minimum latency algorithm.
17684 @item -minline-float-divide-max-throughput
17685 @opindex minline-float-divide-max-throughput
17686 Generate code for inline divides of floating-point values
17687 using the maximum throughput algorithm.
17689 @item -mno-inline-float-divide
17690 @opindex mno-inline-float-divide
17691 Do not generate inline code for divides of floating-point values.
17693 @item -minline-int-divide-min-latency
17694 @opindex minline-int-divide-min-latency
17695 Generate code for inline divides of integer values
17696 using the minimum latency algorithm.
17698 @item -minline-int-divide-max-throughput
17699 @opindex minline-int-divide-max-throughput
17700 Generate code for inline divides of integer values
17701 using the maximum throughput algorithm.
17703 @item -mno-inline-int-divide
17704 @opindex mno-inline-int-divide
17705 Do not generate inline code for divides of integer values.
17707 @item -minline-sqrt-min-latency
17708 @opindex minline-sqrt-min-latency
17709 Generate code for inline square roots
17710 using the minimum latency algorithm.
17712 @item -minline-sqrt-max-throughput
17713 @opindex minline-sqrt-max-throughput
17714 Generate code for inline square roots
17715 using the maximum throughput algorithm.
17717 @item -mno-inline-sqrt
17718 @opindex mno-inline-sqrt
17719 Do not generate inline code for @code{sqrt}.
17722 @itemx -mno-fused-madd
17723 @opindex mfused-madd
17724 @opindex mno-fused-madd
17725 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
17726 instructions. The default is to use these instructions.
17728 @item -mno-dwarf2-asm
17729 @itemx -mdwarf2-asm
17730 @opindex mno-dwarf2-asm
17731 @opindex mdwarf2-asm
17732 Don't (or do) generate assembler code for the DWARF line number debugging
17733 info. This may be useful when not using the GNU assembler.
17735 @item -mearly-stop-bits
17736 @itemx -mno-early-stop-bits
17737 @opindex mearly-stop-bits
17738 @opindex mno-early-stop-bits
17739 Allow stop bits to be placed earlier than immediately preceding the
17740 instruction that triggered the stop bit. This can improve instruction
17741 scheduling, but does not always do so.
17743 @item -mfixed-range=@var{register-range}
17744 @opindex mfixed-range
17745 Generate code treating the given register range as fixed registers.
17746 A fixed register is one that the register allocator cannot use. This is
17747 useful when compiling kernel code. A register range is specified as
17748 two registers separated by a dash. Multiple register ranges can be
17749 specified separated by a comma.
17751 @item -mtls-size=@var{tls-size}
17753 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
17756 @item -mtune=@var{cpu-type}
17758 Tune the instruction scheduling for a particular CPU, Valid values are
17759 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
17760 and @samp{mckinley}.
17766 Generate code for a 32-bit or 64-bit environment.
17767 The 32-bit environment sets int, long and pointer to 32 bits.
17768 The 64-bit environment sets int to 32 bits and long and pointer
17769 to 64 bits. These are HP-UX specific flags.
17771 @item -mno-sched-br-data-spec
17772 @itemx -msched-br-data-spec
17773 @opindex mno-sched-br-data-spec
17774 @opindex msched-br-data-spec
17775 (Dis/En)able data speculative scheduling before reload.
17776 This results in generation of @code{ld.a} instructions and
17777 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17778 The default setting is disabled.
17780 @item -msched-ar-data-spec
17781 @itemx -mno-sched-ar-data-spec
17782 @opindex msched-ar-data-spec
17783 @opindex mno-sched-ar-data-spec
17784 (En/Dis)able data speculative scheduling after reload.
17785 This results in generation of @code{ld.a} instructions and
17786 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
17787 The default setting is enabled.
17789 @item -mno-sched-control-spec
17790 @itemx -msched-control-spec
17791 @opindex mno-sched-control-spec
17792 @opindex msched-control-spec
17793 (Dis/En)able control speculative scheduling. This feature is
17794 available only during region scheduling (i.e.@: before reload).
17795 This results in generation of the @code{ld.s} instructions and
17796 the corresponding check instructions @code{chk.s}.
17797 The default setting is disabled.
17799 @item -msched-br-in-data-spec
17800 @itemx -mno-sched-br-in-data-spec
17801 @opindex msched-br-in-data-spec
17802 @opindex mno-sched-br-in-data-spec
17803 (En/Dis)able speculative scheduling of the instructions that
17804 are dependent on the data speculative loads before reload.
17805 This is effective only with @option{-msched-br-data-spec} enabled.
17806 The default setting is enabled.
17808 @item -msched-ar-in-data-spec
17809 @itemx -mno-sched-ar-in-data-spec
17810 @opindex msched-ar-in-data-spec
17811 @opindex mno-sched-ar-in-data-spec
17812 (En/Dis)able speculative scheduling of the instructions that
17813 are dependent on the data speculative loads after reload.
17814 This is effective only with @option{-msched-ar-data-spec} enabled.
17815 The default setting is enabled.
17817 @item -msched-in-control-spec
17818 @itemx -mno-sched-in-control-spec
17819 @opindex msched-in-control-spec
17820 @opindex mno-sched-in-control-spec
17821 (En/Dis)able speculative scheduling of the instructions that
17822 are dependent on the control speculative loads.
17823 This is effective only with @option{-msched-control-spec} enabled.
17824 The default setting is enabled.
17826 @item -mno-sched-prefer-non-data-spec-insns
17827 @itemx -msched-prefer-non-data-spec-insns
17828 @opindex mno-sched-prefer-non-data-spec-insns
17829 @opindex msched-prefer-non-data-spec-insns
17830 If enabled, data-speculative instructions are chosen for schedule
17831 only if there are no other choices at the moment. This makes
17832 the use of the data speculation much more conservative.
17833 The default setting is disabled.
17835 @item -mno-sched-prefer-non-control-spec-insns
17836 @itemx -msched-prefer-non-control-spec-insns
17837 @opindex mno-sched-prefer-non-control-spec-insns
17838 @opindex msched-prefer-non-control-spec-insns
17839 If enabled, control-speculative instructions are chosen for schedule
17840 only if there are no other choices at the moment. This makes
17841 the use of the control speculation much more conservative.
17842 The default setting is disabled.
17844 @item -mno-sched-count-spec-in-critical-path
17845 @itemx -msched-count-spec-in-critical-path
17846 @opindex mno-sched-count-spec-in-critical-path
17847 @opindex msched-count-spec-in-critical-path
17848 If enabled, speculative dependencies are considered during
17849 computation of the instructions priorities. This makes the use of the
17850 speculation a bit more conservative.
17851 The default setting is disabled.
17853 @item -msched-spec-ldc
17854 @opindex msched-spec-ldc
17855 Use a simple data speculation check. This option is on by default.
17857 @item -msched-control-spec-ldc
17858 @opindex msched-spec-ldc
17859 Use a simple check for control speculation. This option is on by default.
17861 @item -msched-stop-bits-after-every-cycle
17862 @opindex msched-stop-bits-after-every-cycle
17863 Place a stop bit after every cycle when scheduling. This option is on
17866 @item -msched-fp-mem-deps-zero-cost
17867 @opindex msched-fp-mem-deps-zero-cost
17868 Assume that floating-point stores and loads are not likely to cause a conflict
17869 when placed into the same instruction group. This option is disabled by
17872 @item -msel-sched-dont-check-control-spec
17873 @opindex msel-sched-dont-check-control-spec
17874 Generate checks for control speculation in selective scheduling.
17875 This flag is disabled by default.
17877 @item -msched-max-memory-insns=@var{max-insns}
17878 @opindex msched-max-memory-insns
17879 Limit on the number of memory insns per instruction group, giving lower
17880 priority to subsequent memory insns attempting to schedule in the same
17881 instruction group. Frequently useful to prevent cache bank conflicts.
17882 The default value is 1.
17884 @item -msched-max-memory-insns-hard-limit
17885 @opindex msched-max-memory-insns-hard-limit
17886 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
17887 disallowing more than that number in an instruction group.
17888 Otherwise, the limit is ``soft'', meaning that non-memory operations
17889 are preferred when the limit is reached, but memory operations may still
17895 @subsection LM32 Options
17896 @cindex LM32 options
17898 These @option{-m} options are defined for the LatticeMico32 architecture:
17901 @item -mbarrel-shift-enabled
17902 @opindex mbarrel-shift-enabled
17903 Enable barrel-shift instructions.
17905 @item -mdivide-enabled
17906 @opindex mdivide-enabled
17907 Enable divide and modulus instructions.
17909 @item -mmultiply-enabled
17910 @opindex multiply-enabled
17911 Enable multiply instructions.
17913 @item -msign-extend-enabled
17914 @opindex msign-extend-enabled
17915 Enable sign extend instructions.
17917 @item -muser-enabled
17918 @opindex muser-enabled
17919 Enable user-defined instructions.
17924 @subsection M32C Options
17925 @cindex M32C options
17928 @item -mcpu=@var{name}
17930 Select the CPU for which code is generated. @var{name} may be one of
17931 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
17932 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
17933 the M32C/80 series.
17937 Specifies that the program will be run on the simulator. This causes
17938 an alternate runtime library to be linked in which supports, for
17939 example, file I/O@. You must not use this option when generating
17940 programs that will run on real hardware; you must provide your own
17941 runtime library for whatever I/O functions are needed.
17943 @item -memregs=@var{number}
17945 Specifies the number of memory-based pseudo-registers GCC uses
17946 during code generation. These pseudo-registers are used like real
17947 registers, so there is a tradeoff between GCC's ability to fit the
17948 code into available registers, and the performance penalty of using
17949 memory instead of registers. Note that all modules in a program must
17950 be compiled with the same value for this option. Because of that, you
17951 must not use this option with GCC's default runtime libraries.
17955 @node M32R/D Options
17956 @subsection M32R/D Options
17957 @cindex M32R/D options
17959 These @option{-m} options are defined for Renesas M32R/D architectures:
17964 Generate code for the M32R/2@.
17968 Generate code for the M32R/X@.
17972 Generate code for the M32R@. This is the default.
17974 @item -mmodel=small
17975 @opindex mmodel=small
17976 Assume all objects live in the lower 16MB of memory (so that their addresses
17977 can be loaded with the @code{ld24} instruction), and assume all subroutines
17978 are reachable with the @code{bl} instruction.
17979 This is the default.
17981 The addressability of a particular object can be set with the
17982 @code{model} attribute.
17984 @item -mmodel=medium
17985 @opindex mmodel=medium
17986 Assume objects may be anywhere in the 32-bit address space (the compiler
17987 generates @code{seth/add3} instructions to load their addresses), and
17988 assume all subroutines are reachable with the @code{bl} instruction.
17990 @item -mmodel=large
17991 @opindex mmodel=large
17992 Assume objects may be anywhere in the 32-bit address space (the compiler
17993 generates @code{seth/add3} instructions to load their addresses), and
17994 assume subroutines may not be reachable with the @code{bl} instruction
17995 (the compiler generates the much slower @code{seth/add3/jl}
17996 instruction sequence).
17999 @opindex msdata=none
18000 Disable use of the small data area. Variables are put into
18001 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
18002 @code{section} attribute has been specified).
18003 This is the default.
18005 The small data area consists of sections @code{.sdata} and @code{.sbss}.
18006 Objects may be explicitly put in the small data area with the
18007 @code{section} attribute using one of these sections.
18009 @item -msdata=sdata
18010 @opindex msdata=sdata
18011 Put small global and static data in the small data area, but do not
18012 generate special code to reference them.
18015 @opindex msdata=use
18016 Put small global and static data in the small data area, and generate
18017 special instructions to reference them.
18021 @cindex smaller data references
18022 Put global and static objects less than or equal to @var{num} bytes
18023 into the small data or BSS sections instead of the normal data or BSS
18024 sections. The default value of @var{num} is 8.
18025 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
18026 for this option to have any effect.
18028 All modules should be compiled with the same @option{-G @var{num}} value.
18029 Compiling with different values of @var{num} may or may not work; if it
18030 doesn't the linker gives an error message---incorrect code is not
18035 Makes the M32R-specific code in the compiler display some statistics
18036 that might help in debugging programs.
18038 @item -malign-loops
18039 @opindex malign-loops
18040 Align all loops to a 32-byte boundary.
18042 @item -mno-align-loops
18043 @opindex mno-align-loops
18044 Do not enforce a 32-byte alignment for loops. This is the default.
18046 @item -missue-rate=@var{number}
18047 @opindex missue-rate=@var{number}
18048 Issue @var{number} instructions per cycle. @var{number} can only be 1
18051 @item -mbranch-cost=@var{number}
18052 @opindex mbranch-cost=@var{number}
18053 @var{number} can only be 1 or 2. If it is 1 then branches are
18054 preferred over conditional code, if it is 2, then the opposite applies.
18056 @item -mflush-trap=@var{number}
18057 @opindex mflush-trap=@var{number}
18058 Specifies the trap number to use to flush the cache. The default is
18059 12. Valid numbers are between 0 and 15 inclusive.
18061 @item -mno-flush-trap
18062 @opindex mno-flush-trap
18063 Specifies that the cache cannot be flushed by using a trap.
18065 @item -mflush-func=@var{name}
18066 @opindex mflush-func=@var{name}
18067 Specifies the name of the operating system function to call to flush
18068 the cache. The default is @samp{_flush_cache}, but a function call
18069 is only used if a trap is not available.
18071 @item -mno-flush-func
18072 @opindex mno-flush-func
18073 Indicates that there is no OS function for flushing the cache.
18077 @node M680x0 Options
18078 @subsection M680x0 Options
18079 @cindex M680x0 options
18081 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
18082 The default settings depend on which architecture was selected when
18083 the compiler was configured; the defaults for the most common choices
18087 @item -march=@var{arch}
18089 Generate code for a specific M680x0 or ColdFire instruction set
18090 architecture. Permissible values of @var{arch} for M680x0
18091 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
18092 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
18093 architectures are selected according to Freescale's ISA classification
18094 and the permissible values are: @samp{isaa}, @samp{isaaplus},
18095 @samp{isab} and @samp{isac}.
18097 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
18098 code for a ColdFire target. The @var{arch} in this macro is one of the
18099 @option{-march} arguments given above.
18101 When used together, @option{-march} and @option{-mtune} select code
18102 that runs on a family of similar processors but that is optimized
18103 for a particular microarchitecture.
18105 @item -mcpu=@var{cpu}
18107 Generate code for a specific M680x0 or ColdFire processor.
18108 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
18109 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
18110 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
18111 below, which also classifies the CPUs into families:
18113 @multitable @columnfractions 0.20 0.80
18114 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
18115 @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}
18116 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
18117 @item @samp{5206e} @tab @samp{5206e}
18118 @item @samp{5208} @tab @samp{5207} @samp{5208}
18119 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
18120 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
18121 @item @samp{5216} @tab @samp{5214} @samp{5216}
18122 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
18123 @item @samp{5225} @tab @samp{5224} @samp{5225}
18124 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
18125 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
18126 @item @samp{5249} @tab @samp{5249}
18127 @item @samp{5250} @tab @samp{5250}
18128 @item @samp{5271} @tab @samp{5270} @samp{5271}
18129 @item @samp{5272} @tab @samp{5272}
18130 @item @samp{5275} @tab @samp{5274} @samp{5275}
18131 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
18132 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
18133 @item @samp{5307} @tab @samp{5307}
18134 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
18135 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
18136 @item @samp{5407} @tab @samp{5407}
18137 @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}
18140 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
18141 @var{arch} is compatible with @var{cpu}. Other combinations of
18142 @option{-mcpu} and @option{-march} are rejected.
18144 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
18145 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
18146 where the value of @var{family} is given by the table above.
18148 @item -mtune=@var{tune}
18150 Tune the code for a particular microarchitecture within the
18151 constraints set by @option{-march} and @option{-mcpu}.
18152 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
18153 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
18154 and @samp{cpu32}. The ColdFire microarchitectures
18155 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
18157 You can also use @option{-mtune=68020-40} for code that needs
18158 to run relatively well on 68020, 68030 and 68040 targets.
18159 @option{-mtune=68020-60} is similar but includes 68060 targets
18160 as well. These two options select the same tuning decisions as
18161 @option{-m68020-40} and @option{-m68020-60} respectively.
18163 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
18164 when tuning for 680x0 architecture @var{arch}. It also defines
18165 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
18166 option is used. If GCC is tuning for a range of architectures,
18167 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
18168 it defines the macros for every architecture in the range.
18170 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
18171 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
18172 of the arguments given above.
18178 Generate output for a 68000. This is the default
18179 when the compiler is configured for 68000-based systems.
18180 It is equivalent to @option{-march=68000}.
18182 Use this option for microcontrollers with a 68000 or EC000 core,
18183 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
18187 Generate output for a 68010. This is the default
18188 when the compiler is configured for 68010-based systems.
18189 It is equivalent to @option{-march=68010}.
18195 Generate output for a 68020. This is the default
18196 when the compiler is configured for 68020-based systems.
18197 It is equivalent to @option{-march=68020}.
18201 Generate output for a 68030. This is the default when the compiler is
18202 configured for 68030-based systems. It is equivalent to
18203 @option{-march=68030}.
18207 Generate output for a 68040. This is the default when the compiler is
18208 configured for 68040-based systems. It is equivalent to
18209 @option{-march=68040}.
18211 This option inhibits the use of 68881/68882 instructions that have to be
18212 emulated by software on the 68040. Use this option if your 68040 does not
18213 have code to emulate those instructions.
18217 Generate output for a 68060. This is the default when the compiler is
18218 configured for 68060-based systems. It is equivalent to
18219 @option{-march=68060}.
18221 This option inhibits the use of 68020 and 68881/68882 instructions that
18222 have to be emulated by software on the 68060. Use this option if your 68060
18223 does not have code to emulate those instructions.
18227 Generate output for a CPU32. This is the default
18228 when the compiler is configured for CPU32-based systems.
18229 It is equivalent to @option{-march=cpu32}.
18231 Use this option for microcontrollers with a
18232 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
18233 68336, 68340, 68341, 68349 and 68360.
18237 Generate output for a 520X ColdFire CPU@. This is the default
18238 when the compiler is configured for 520X-based systems.
18239 It is equivalent to @option{-mcpu=5206}, and is now deprecated
18240 in favor of that option.
18242 Use this option for microcontroller with a 5200 core, including
18243 the MCF5202, MCF5203, MCF5204 and MCF5206.
18247 Generate output for a 5206e ColdFire CPU@. The option is now
18248 deprecated in favor of the equivalent @option{-mcpu=5206e}.
18252 Generate output for a member of the ColdFire 528X family.
18253 The option is now deprecated in favor of the equivalent
18254 @option{-mcpu=528x}.
18258 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
18259 in favor of the equivalent @option{-mcpu=5307}.
18263 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
18264 in favor of the equivalent @option{-mcpu=5407}.
18268 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
18269 This includes use of hardware floating-point instructions.
18270 The option is equivalent to @option{-mcpu=547x}, and is now
18271 deprecated in favor of that option.
18275 Generate output for a 68040, without using any of the new instructions.
18276 This results in code that can run relatively efficiently on either a
18277 68020/68881 or a 68030 or a 68040. The generated code does use the
18278 68881 instructions that are emulated on the 68040.
18280 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
18284 Generate output for a 68060, without using any of the new instructions.
18285 This results in code that can run relatively efficiently on either a
18286 68020/68881 or a 68030 or a 68040. The generated code does use the
18287 68881 instructions that are emulated on the 68060.
18289 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
18293 @opindex mhard-float
18295 Generate floating-point instructions. This is the default for 68020
18296 and above, and for ColdFire devices that have an FPU@. It defines the
18297 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
18298 on ColdFire targets.
18301 @opindex msoft-float
18302 Do not generate floating-point instructions; use library calls instead.
18303 This is the default for 68000, 68010, and 68832 targets. It is also
18304 the default for ColdFire devices that have no FPU.
18310 Generate (do not generate) ColdFire hardware divide and remainder
18311 instructions. If @option{-march} is used without @option{-mcpu},
18312 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
18313 architectures. Otherwise, the default is taken from the target CPU
18314 (either the default CPU, or the one specified by @option{-mcpu}). For
18315 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
18316 @option{-mcpu=5206e}.
18318 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
18322 Consider type @code{int} to be 16 bits wide, like @code{short int}.
18323 Additionally, parameters passed on the stack are also aligned to a
18324 16-bit boundary even on targets whose API mandates promotion to 32-bit.
18328 Do not consider type @code{int} to be 16 bits wide. This is the default.
18331 @itemx -mno-bitfield
18332 @opindex mnobitfield
18333 @opindex mno-bitfield
18334 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
18335 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
18339 Do use the bit-field instructions. The @option{-m68020} option implies
18340 @option{-mbitfield}. This is the default if you use a configuration
18341 designed for a 68020.
18345 Use a different function-calling convention, in which functions
18346 that take a fixed number of arguments return with the @code{rtd}
18347 instruction, which pops their arguments while returning. This
18348 saves one instruction in the caller since there is no need to pop
18349 the arguments there.
18351 This calling convention is incompatible with the one normally
18352 used on Unix, so you cannot use it if you need to call libraries
18353 compiled with the Unix compiler.
18355 Also, you must provide function prototypes for all functions that
18356 take variable numbers of arguments (including @code{printf});
18357 otherwise incorrect code is generated for calls to those
18360 In addition, seriously incorrect code results if you call a
18361 function with too many arguments. (Normally, extra arguments are
18362 harmlessly ignored.)
18364 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
18365 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
18369 Do not use the calling conventions selected by @option{-mrtd}.
18370 This is the default.
18373 @itemx -mno-align-int
18374 @opindex malign-int
18375 @opindex mno-align-int
18376 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
18377 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
18378 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
18379 Aligning variables on 32-bit boundaries produces code that runs somewhat
18380 faster on processors with 32-bit busses at the expense of more memory.
18382 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
18383 aligns structures containing the above types differently than
18384 most published application binary interface specifications for the m68k.
18388 Use the pc-relative addressing mode of the 68000 directly, instead of
18389 using a global offset table. At present, this option implies @option{-fpic},
18390 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
18391 not presently supported with @option{-mpcrel}, though this could be supported for
18392 68020 and higher processors.
18394 @item -mno-strict-align
18395 @itemx -mstrict-align
18396 @opindex mno-strict-align
18397 @opindex mstrict-align
18398 Do not (do) assume that unaligned memory references are handled by
18402 Generate code that allows the data segment to be located in a different
18403 area of memory from the text segment. This allows for execute-in-place in
18404 an environment without virtual memory management. This option implies
18407 @item -mno-sep-data
18408 Generate code that assumes that the data segment follows the text segment.
18409 This is the default.
18411 @item -mid-shared-library
18412 Generate code that supports shared libraries via the library ID method.
18413 This allows for execute-in-place and shared libraries in an environment
18414 without virtual memory management. This option implies @option{-fPIC}.
18416 @item -mno-id-shared-library
18417 Generate code that doesn't assume ID-based shared libraries are being used.
18418 This is the default.
18420 @item -mshared-library-id=n
18421 Specifies the identification number of the ID-based shared library being
18422 compiled. Specifying a value of 0 generates more compact code; specifying
18423 other values forces the allocation of that number to the current
18424 library, but is no more space- or time-efficient than omitting this option.
18430 When generating position-independent code for ColdFire, generate code
18431 that works if the GOT has more than 8192 entries. This code is
18432 larger and slower than code generated without this option. On M680x0
18433 processors, this option is not needed; @option{-fPIC} suffices.
18435 GCC normally uses a single instruction to load values from the GOT@.
18436 While this is relatively efficient, it only works if the GOT
18437 is smaller than about 64k. Anything larger causes the linker
18438 to report an error such as:
18440 @cindex relocation truncated to fit (ColdFire)
18442 relocation truncated to fit: R_68K_GOT16O foobar
18445 If this happens, you should recompile your code with @option{-mxgot}.
18446 It should then work with very large GOTs. However, code generated with
18447 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
18448 the value of a global symbol.
18450 Note that some linkers, including newer versions of the GNU linker,
18451 can create multiple GOTs and sort GOT entries. If you have such a linker,
18452 you should only need to use @option{-mxgot} when compiling a single
18453 object file that accesses more than 8192 GOT entries. Very few do.
18455 These options have no effect unless GCC is generating
18456 position-independent code.
18458 @item -mlong-jump-table-offsets
18459 @opindex mlong-jump-table-offsets
18460 Use 32-bit offsets in @code{switch} tables. The default is to use
18465 @node MCore Options
18466 @subsection MCore Options
18467 @cindex MCore options
18469 These are the @samp{-m} options defined for the Motorola M*Core
18475 @itemx -mno-hardlit
18477 @opindex mno-hardlit
18478 Inline constants into the code stream if it can be done in two
18479 instructions or less.
18485 Use the divide instruction. (Enabled by default).
18487 @item -mrelax-immediate
18488 @itemx -mno-relax-immediate
18489 @opindex mrelax-immediate
18490 @opindex mno-relax-immediate
18491 Allow arbitrary-sized immediates in bit operations.
18493 @item -mwide-bitfields
18494 @itemx -mno-wide-bitfields
18495 @opindex mwide-bitfields
18496 @opindex mno-wide-bitfields
18497 Always treat bit-fields as @code{int}-sized.
18499 @item -m4byte-functions
18500 @itemx -mno-4byte-functions
18501 @opindex m4byte-functions
18502 @opindex mno-4byte-functions
18503 Force all functions to be aligned to a 4-byte boundary.
18505 @item -mcallgraph-data
18506 @itemx -mno-callgraph-data
18507 @opindex mcallgraph-data
18508 @opindex mno-callgraph-data
18509 Emit callgraph information.
18512 @itemx -mno-slow-bytes
18513 @opindex mslow-bytes
18514 @opindex mno-slow-bytes
18515 Prefer word access when reading byte quantities.
18517 @item -mlittle-endian
18518 @itemx -mbig-endian
18519 @opindex mlittle-endian
18520 @opindex mbig-endian
18521 Generate code for a little-endian target.
18527 Generate code for the 210 processor.
18531 Assume that runtime support has been provided and so omit the
18532 simulator library (@file{libsim.a)} from the linker command line.
18534 @item -mstack-increment=@var{size}
18535 @opindex mstack-increment
18536 Set the maximum amount for a single stack increment operation. Large
18537 values can increase the speed of programs that contain functions
18538 that need a large amount of stack space, but they can also trigger a
18539 segmentation fault if the stack is extended too much. The default
18545 @subsection MeP Options
18546 @cindex MeP options
18552 Enables the @code{abs} instruction, which is the absolute difference
18553 between two registers.
18557 Enables all the optional instructions---average, multiply, divide, bit
18558 operations, leading zero, absolute difference, min/max, clip, and
18564 Enables the @code{ave} instruction, which computes the average of two
18567 @item -mbased=@var{n}
18569 Variables of size @var{n} bytes or smaller are placed in the
18570 @code{.based} section by default. Based variables use the @code{$tp}
18571 register as a base register, and there is a 128-byte limit to the
18572 @code{.based} section.
18576 Enables the bit operation instructions---bit test (@code{btstm}), set
18577 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
18578 test-and-set (@code{tas}).
18580 @item -mc=@var{name}
18582 Selects which section constant data is placed in. @var{name} may
18583 be @samp{tiny}, @samp{near}, or @samp{far}.
18587 Enables the @code{clip} instruction. Note that @option{-mclip} is not
18588 useful unless you also provide @option{-mminmax}.
18590 @item -mconfig=@var{name}
18592 Selects one of the built-in core configurations. Each MeP chip has
18593 one or more modules in it; each module has a core CPU and a variety of
18594 coprocessors, optional instructions, and peripherals. The
18595 @code{MeP-Integrator} tool, not part of GCC, provides these
18596 configurations through this option; using this option is the same as
18597 using all the corresponding command-line options. The default
18598 configuration is @samp{default}.
18602 Enables the coprocessor instructions. By default, this is a 32-bit
18603 coprocessor. Note that the coprocessor is normally enabled via the
18604 @option{-mconfig=} option.
18608 Enables the 32-bit coprocessor's instructions.
18612 Enables the 64-bit coprocessor's instructions.
18616 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
18620 Causes constant variables to be placed in the @code{.near} section.
18624 Enables the @code{div} and @code{divu} instructions.
18628 Generate big-endian code.
18632 Generate little-endian code.
18634 @item -mio-volatile
18635 @opindex mio-volatile
18636 Tells the compiler that any variable marked with the @code{io}
18637 attribute is to be considered volatile.
18641 Causes variables to be assigned to the @code{.far} section by default.
18645 Enables the @code{leadz} (leading zero) instruction.
18649 Causes variables to be assigned to the @code{.near} section by default.
18653 Enables the @code{min} and @code{max} instructions.
18657 Enables the multiplication and multiply-accumulate instructions.
18661 Disables all the optional instructions enabled by @option{-mall-opts}.
18665 Enables the @code{repeat} and @code{erepeat} instructions, used for
18666 low-overhead looping.
18670 Causes all variables to default to the @code{.tiny} section. Note
18671 that there is a 65536-byte limit to this section. Accesses to these
18672 variables use the @code{%gp} base register.
18676 Enables the saturation instructions. Note that the compiler does not
18677 currently generate these itself, but this option is included for
18678 compatibility with other tools, like @code{as}.
18682 Link the SDRAM-based runtime instead of the default ROM-based runtime.
18686 Link the simulator run-time libraries.
18690 Link the simulator runtime libraries, excluding built-in support
18691 for reset and exception vectors and tables.
18695 Causes all functions to default to the @code{.far} section. Without
18696 this option, functions default to the @code{.near} section.
18698 @item -mtiny=@var{n}
18700 Variables that are @var{n} bytes or smaller are allocated to the
18701 @code{.tiny} section. These variables use the @code{$gp} base
18702 register. The default for this option is 4, but note that there's a
18703 65536-byte limit to the @code{.tiny} section.
18707 @node MicroBlaze Options
18708 @subsection MicroBlaze Options
18709 @cindex MicroBlaze Options
18714 @opindex msoft-float
18715 Use software emulation for floating point (default).
18718 @opindex mhard-float
18719 Use hardware floating-point instructions.
18723 Do not optimize block moves, use @code{memcpy}.
18725 @item -mno-clearbss
18726 @opindex mno-clearbss
18727 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
18729 @item -mcpu=@var{cpu-type}
18731 Use features of, and schedule code for, the given CPU.
18732 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
18733 where @var{X} is a major version, @var{YY} is the minor version, and
18734 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
18735 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
18737 @item -mxl-soft-mul
18738 @opindex mxl-soft-mul
18739 Use software multiply emulation (default).
18741 @item -mxl-soft-div
18742 @opindex mxl-soft-div
18743 Use software emulation for divides (default).
18745 @item -mxl-barrel-shift
18746 @opindex mxl-barrel-shift
18747 Use the hardware barrel shifter.
18749 @item -mxl-pattern-compare
18750 @opindex mxl-pattern-compare
18751 Use pattern compare instructions.
18753 @item -msmall-divides
18754 @opindex msmall-divides
18755 Use table lookup optimization for small signed integer divisions.
18757 @item -mxl-stack-check
18758 @opindex mxl-stack-check
18759 This option is deprecated. Use @option{-fstack-check} instead.
18762 @opindex mxl-gp-opt
18763 Use GP-relative @code{.sdata}/@code{.sbss} sections.
18765 @item -mxl-multiply-high
18766 @opindex mxl-multiply-high
18767 Use multiply high instructions for high part of 32x32 multiply.
18769 @item -mxl-float-convert
18770 @opindex mxl-float-convert
18771 Use hardware floating-point conversion instructions.
18773 @item -mxl-float-sqrt
18774 @opindex mxl-float-sqrt
18775 Use hardware floating-point square root instruction.
18778 @opindex mbig-endian
18779 Generate code for a big-endian target.
18781 @item -mlittle-endian
18782 @opindex mlittle-endian
18783 Generate code for a little-endian target.
18786 @opindex mxl-reorder
18787 Use reorder instructions (swap and byte reversed load/store).
18789 @item -mxl-mode-@var{app-model}
18790 Select application model @var{app-model}. Valid models are
18793 normal executable (default), uses startup code @file{crt0.o}.
18796 for use with Xilinx Microprocessor Debugger (XMD) based
18797 software intrusive debug agent called xmdstub. This uses startup file
18798 @file{crt1.o} and sets the start address of the program to 0x800.
18801 for applications that are loaded using a bootloader.
18802 This model uses startup file @file{crt2.o} which does not contain a processor
18803 reset vector handler. This is suitable for transferring control on a
18804 processor reset to the bootloader rather than the application.
18807 for applications that do not require any of the
18808 MicroBlaze vectors. This option may be useful for applications running
18809 within a monitoring application. This model uses @file{crt3.o} as a startup file.
18812 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
18813 @option{-mxl-mode-@var{app-model}}.
18818 @subsection MIPS Options
18819 @cindex MIPS options
18825 Generate big-endian code.
18829 Generate little-endian code. This is the default for @samp{mips*el-*-*}
18832 @item -march=@var{arch}
18834 Generate code that runs on @var{arch}, which can be the name of a
18835 generic MIPS ISA, or the name of a particular processor.
18837 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
18838 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
18839 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
18840 @samp{mips64r5} and @samp{mips64r6}.
18841 The processor names are:
18842 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
18843 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
18844 @samp{5kc}, @samp{5kf},
18846 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
18847 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
18848 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
18849 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
18850 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
18853 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
18855 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
18856 @samp{m5100}, @samp{m5101},
18857 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
18860 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
18861 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
18862 @samp{rm7000}, @samp{rm9000},
18863 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
18866 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
18867 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
18868 @samp{xlr} and @samp{xlp}.
18869 The special value @samp{from-abi} selects the
18870 most compatible architecture for the selected ABI (that is,
18871 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
18873 The native Linux/GNU toolchain also supports the value @samp{native},
18874 which selects the best architecture option for the host processor.
18875 @option{-march=native} has no effect if GCC does not recognize
18878 In processor names, a final @samp{000} can be abbreviated as @samp{k}
18879 (for example, @option{-march=r2k}). Prefixes are optional, and
18880 @samp{vr} may be written @samp{r}.
18882 Names of the form @samp{@var{n}f2_1} refer to processors with
18883 FPUs clocked at half the rate of the core, names of the form
18884 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
18885 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
18886 processors with FPUs clocked a ratio of 3:2 with respect to the core.
18887 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
18888 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
18889 accepted as synonyms for @samp{@var{n}f1_1}.
18891 GCC defines two macros based on the value of this option. The first
18892 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
18893 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
18894 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
18895 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
18896 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
18898 Note that the @code{_MIPS_ARCH} macro uses the processor names given
18899 above. In other words, it has the full prefix and does not
18900 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
18901 the macro names the resolved architecture (either @code{"mips1"} or
18902 @code{"mips3"}). It names the default architecture when no
18903 @option{-march} option is given.
18905 @item -mtune=@var{arch}
18907 Optimize for @var{arch}. Among other things, this option controls
18908 the way instructions are scheduled, and the perceived cost of arithmetic
18909 operations. The list of @var{arch} values is the same as for
18912 When this option is not used, GCC optimizes for the processor
18913 specified by @option{-march}. By using @option{-march} and
18914 @option{-mtune} together, it is possible to generate code that
18915 runs on a family of processors, but optimize the code for one
18916 particular member of that family.
18918 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
18919 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
18920 @option{-march} ones described above.
18924 Equivalent to @option{-march=mips1}.
18928 Equivalent to @option{-march=mips2}.
18932 Equivalent to @option{-march=mips3}.
18936 Equivalent to @option{-march=mips4}.
18940 Equivalent to @option{-march=mips32}.
18944 Equivalent to @option{-march=mips32r3}.
18948 Equivalent to @option{-march=mips32r5}.
18952 Equivalent to @option{-march=mips32r6}.
18956 Equivalent to @option{-march=mips64}.
18960 Equivalent to @option{-march=mips64r2}.
18964 Equivalent to @option{-march=mips64r3}.
18968 Equivalent to @option{-march=mips64r5}.
18972 Equivalent to @option{-march=mips64r6}.
18977 @opindex mno-mips16
18978 Generate (do not generate) MIPS16 code. If GCC is targeting a
18979 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
18981 MIPS16 code generation can also be controlled on a per-function basis
18982 by means of @code{mips16} and @code{nomips16} attributes.
18983 @xref{Function Attributes}, for more information.
18985 @item -mflip-mips16
18986 @opindex mflip-mips16
18987 Generate MIPS16 code on alternating functions. This option is provided
18988 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
18989 not intended for ordinary use in compiling user code.
18991 @item -minterlink-compressed
18992 @item -mno-interlink-compressed
18993 @opindex minterlink-compressed
18994 @opindex mno-interlink-compressed
18995 Require (do not require) that code using the standard (uncompressed) MIPS ISA
18996 be link-compatible with MIPS16 and microMIPS code, and vice versa.
18998 For example, code using the standard ISA encoding cannot jump directly
18999 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
19000 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
19001 knows that the target of the jump is not compressed.
19003 @item -minterlink-mips16
19004 @itemx -mno-interlink-mips16
19005 @opindex minterlink-mips16
19006 @opindex mno-interlink-mips16
19007 Aliases of @option{-minterlink-compressed} and
19008 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
19009 and are retained for backwards compatibility.
19021 Generate code for the given ABI@.
19023 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
19024 generates 64-bit code when you select a 64-bit architecture, but you
19025 can use @option{-mgp32} to get 32-bit code instead.
19027 For information about the O64 ABI, see
19028 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
19030 GCC supports a variant of the o32 ABI in which floating-point registers
19031 are 64 rather than 32 bits wide. You can select this combination with
19032 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
19033 and @code{mfhc1} instructions and is therefore only supported for
19034 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
19036 The register assignments for arguments and return values remain the
19037 same, but each scalar value is passed in a single 64-bit register
19038 rather than a pair of 32-bit registers. For example, scalar
19039 floating-point values are returned in @samp{$f0} only, not a
19040 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
19041 remains the same in that the even-numbered double-precision registers
19044 Two additional variants of the o32 ABI are supported to enable
19045 a transition from 32-bit to 64-bit registers. These are FPXX
19046 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
19047 The FPXX extension mandates that all code must execute correctly
19048 when run using 32-bit or 64-bit registers. The code can be interlinked
19049 with either FP32 or FP64, but not both.
19050 The FP64A extension is similar to the FP64 extension but forbids the
19051 use of odd-numbered single-precision registers. This can be used
19052 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
19053 processors and allows both FP32 and FP64A code to interlink and
19054 run in the same process without changing FPU modes.
19057 @itemx -mno-abicalls
19059 @opindex mno-abicalls
19060 Generate (do not generate) code that is suitable for SVR4-style
19061 dynamic objects. @option{-mabicalls} is the default for SVR4-based
19066 Generate (do not generate) code that is fully position-independent,
19067 and that can therefore be linked into shared libraries. This option
19068 only affects @option{-mabicalls}.
19070 All @option{-mabicalls} code has traditionally been position-independent,
19071 regardless of options like @option{-fPIC} and @option{-fpic}. However,
19072 as an extension, the GNU toolchain allows executables to use absolute
19073 accesses for locally-binding symbols. It can also use shorter GP
19074 initialization sequences and generate direct calls to locally-defined
19075 functions. This mode is selected by @option{-mno-shared}.
19077 @option{-mno-shared} depends on binutils 2.16 or higher and generates
19078 objects that can only be linked by the GNU linker. However, the option
19079 does not affect the ABI of the final executable; it only affects the ABI
19080 of relocatable objects. Using @option{-mno-shared} generally makes
19081 executables both smaller and quicker.
19083 @option{-mshared} is the default.
19089 Assume (do not assume) that the static and dynamic linkers
19090 support PLTs and copy relocations. This option only affects
19091 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
19092 has no effect without @option{-msym32}.
19094 You can make @option{-mplt} the default by configuring
19095 GCC with @option{--with-mips-plt}. The default is
19096 @option{-mno-plt} otherwise.
19102 Lift (do not lift) the usual restrictions on the size of the global
19105 GCC normally uses a single instruction to load values from the GOT@.
19106 While this is relatively efficient, it only works if the GOT
19107 is smaller than about 64k. Anything larger causes the linker
19108 to report an error such as:
19110 @cindex relocation truncated to fit (MIPS)
19112 relocation truncated to fit: R_MIPS_GOT16 foobar
19115 If this happens, you should recompile your code with @option{-mxgot}.
19116 This works with very large GOTs, although the code is also
19117 less efficient, since it takes three instructions to fetch the
19118 value of a global symbol.
19120 Note that some linkers can create multiple GOTs. If you have such a
19121 linker, you should only need to use @option{-mxgot} when a single object
19122 file accesses more than 64k's worth of GOT entries. Very few do.
19124 These options have no effect unless GCC is generating position
19129 Assume that general-purpose registers are 32 bits wide.
19133 Assume that general-purpose registers are 64 bits wide.
19137 Assume that floating-point registers are 32 bits wide.
19141 Assume that floating-point registers are 64 bits wide.
19145 Do not assume the width of floating-point registers.
19148 @opindex mhard-float
19149 Use floating-point coprocessor instructions.
19152 @opindex msoft-float
19153 Do not use floating-point coprocessor instructions. Implement
19154 floating-point calculations using library calls instead.
19158 Equivalent to @option{-msoft-float}, but additionally asserts that the
19159 program being compiled does not perform any floating-point operations.
19160 This option is presently supported only by some bare-metal MIPS
19161 configurations, where it may select a special set of libraries
19162 that lack all floating-point support (including, for example, the
19163 floating-point @code{printf} formats).
19164 If code compiled with @option{-mno-float} accidentally contains
19165 floating-point operations, it is likely to suffer a link-time
19166 or run-time failure.
19168 @item -msingle-float
19169 @opindex msingle-float
19170 Assume that the floating-point coprocessor only supports single-precision
19173 @item -mdouble-float
19174 @opindex mdouble-float
19175 Assume that the floating-point coprocessor supports double-precision
19176 operations. This is the default.
19179 @itemx -mno-odd-spreg
19180 @opindex modd-spreg
19181 @opindex mno-odd-spreg
19182 Enable the use of odd-numbered single-precision floating-point registers
19183 for the o32 ABI. This is the default for processors that are known to
19184 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
19188 @itemx -mabs=legacy
19190 @opindex mabs=legacy
19191 These options control the treatment of the special not-a-number (NaN)
19192 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
19193 @code{neg.@i{fmt}} machine instructions.
19195 By default or when @option{-mabs=legacy} is used the legacy
19196 treatment is selected. In this case these instructions are considered
19197 arithmetic and avoided where correct operation is required and the
19198 input operand might be a NaN. A longer sequence of instructions that
19199 manipulate the sign bit of floating-point datum manually is used
19200 instead unless the @option{-ffinite-math-only} option has also been
19203 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
19204 this case these instructions are considered non-arithmetic and therefore
19205 operating correctly in all cases, including in particular where the
19206 input operand is a NaN. These instructions are therefore always used
19207 for the respective operations.
19210 @itemx -mnan=legacy
19212 @opindex mnan=legacy
19213 These options control the encoding of the special not-a-number (NaN)
19214 IEEE 754 floating-point data.
19216 The @option{-mnan=legacy} option selects the legacy encoding. In this
19217 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
19218 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
19219 by the first bit of their trailing significand field being 1.
19221 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
19222 this case qNaNs are denoted by the first bit of their trailing
19223 significand field being 1, whereas sNaNs are denoted by the first bit of
19224 their trailing significand field being 0.
19226 The default is @option{-mnan=legacy} unless GCC has been configured with
19227 @option{--with-nan=2008}.
19233 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
19234 implement atomic memory built-in functions. When neither option is
19235 specified, GCC uses the instructions if the target architecture
19238 @option{-mllsc} is useful if the runtime environment can emulate the
19239 instructions and @option{-mno-llsc} can be useful when compiling for
19240 nonstandard ISAs. You can make either option the default by
19241 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
19242 respectively. @option{--with-llsc} is the default for some
19243 configurations; see the installation documentation for details.
19249 Use (do not use) revision 1 of the MIPS DSP ASE@.
19250 @xref{MIPS DSP Built-in Functions}. This option defines the
19251 preprocessor macro @code{__mips_dsp}. It also defines
19252 @code{__mips_dsp_rev} to 1.
19258 Use (do not use) revision 2 of the MIPS DSP ASE@.
19259 @xref{MIPS DSP Built-in Functions}. This option defines the
19260 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
19261 It also defines @code{__mips_dsp_rev} to 2.
19264 @itemx -mno-smartmips
19265 @opindex msmartmips
19266 @opindex mno-smartmips
19267 Use (do not use) the MIPS SmartMIPS ASE.
19269 @item -mpaired-single
19270 @itemx -mno-paired-single
19271 @opindex mpaired-single
19272 @opindex mno-paired-single
19273 Use (do not use) paired-single floating-point instructions.
19274 @xref{MIPS Paired-Single Support}. This option requires
19275 hardware floating-point support to be enabled.
19281 Use (do not use) MIPS Digital Media Extension instructions.
19282 This option can only be used when generating 64-bit code and requires
19283 hardware floating-point support to be enabled.
19288 @opindex mno-mips3d
19289 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
19290 The option @option{-mips3d} implies @option{-mpaired-single}.
19293 @itemx -mno-micromips
19294 @opindex mmicromips
19295 @opindex mno-mmicromips
19296 Generate (do not generate) microMIPS code.
19298 MicroMIPS code generation can also be controlled on a per-function basis
19299 by means of @code{micromips} and @code{nomicromips} attributes.
19300 @xref{Function Attributes}, for more information.
19306 Use (do not use) MT Multithreading instructions.
19312 Use (do not use) the MIPS MCU ASE instructions.
19318 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
19324 Use (do not use) the MIPS Virtualization Application Specific instructions.
19330 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
19334 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
19335 an explanation of the default and the way that the pointer size is
19340 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
19342 The default size of @code{int}s, @code{long}s and pointers depends on
19343 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
19344 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
19345 32-bit @code{long}s. Pointers are the same size as @code{long}s,
19346 or the same size as integer registers, whichever is smaller.
19352 Assume (do not assume) that all symbols have 32-bit values, regardless
19353 of the selected ABI@. This option is useful in combination with
19354 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
19355 to generate shorter and faster references to symbolic addresses.
19359 Put definitions of externally-visible data in a small data section
19360 if that data is no bigger than @var{num} bytes. GCC can then generate
19361 more efficient accesses to the data; see @option{-mgpopt} for details.
19363 The default @option{-G} option depends on the configuration.
19365 @item -mlocal-sdata
19366 @itemx -mno-local-sdata
19367 @opindex mlocal-sdata
19368 @opindex mno-local-sdata
19369 Extend (do not extend) the @option{-G} behavior to local data too,
19370 such as to static variables in C@. @option{-mlocal-sdata} is the
19371 default for all configurations.
19373 If the linker complains that an application is using too much small data,
19374 you might want to try rebuilding the less performance-critical parts with
19375 @option{-mno-local-sdata}. You might also want to build large
19376 libraries with @option{-mno-local-sdata}, so that the libraries leave
19377 more room for the main program.
19379 @item -mextern-sdata
19380 @itemx -mno-extern-sdata
19381 @opindex mextern-sdata
19382 @opindex mno-extern-sdata
19383 Assume (do not assume) that externally-defined data is in
19384 a small data section if the size of that data is within the @option{-G} limit.
19385 @option{-mextern-sdata} is the default for all configurations.
19387 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
19388 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
19389 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
19390 is placed in a small data section. If @var{Var} is defined by another
19391 module, you must either compile that module with a high-enough
19392 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
19393 definition. If @var{Var} is common, you must link the application
19394 with a high-enough @option{-G} setting.
19396 The easiest way of satisfying these restrictions is to compile
19397 and link every module with the same @option{-G} option. However,
19398 you may wish to build a library that supports several different
19399 small data limits. You can do this by compiling the library with
19400 the highest supported @option{-G} setting and additionally using
19401 @option{-mno-extern-sdata} to stop the library from making assumptions
19402 about externally-defined data.
19408 Use (do not use) GP-relative accesses for symbols that are known to be
19409 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
19410 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
19413 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
19414 might not hold the value of @code{_gp}. For example, if the code is
19415 part of a library that might be used in a boot monitor, programs that
19416 call boot monitor routines pass an unknown value in @code{$gp}.
19417 (In such situations, the boot monitor itself is usually compiled
19418 with @option{-G0}.)
19420 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
19421 @option{-mno-extern-sdata}.
19423 @item -membedded-data
19424 @itemx -mno-embedded-data
19425 @opindex membedded-data
19426 @opindex mno-embedded-data
19427 Allocate variables to the read-only data section first if possible, then
19428 next in the small data section if possible, otherwise in data. This gives
19429 slightly slower code than the default, but reduces the amount of RAM required
19430 when executing, and thus may be preferred for some embedded systems.
19432 @item -muninit-const-in-rodata
19433 @itemx -mno-uninit-const-in-rodata
19434 @opindex muninit-const-in-rodata
19435 @opindex mno-uninit-const-in-rodata
19436 Put uninitialized @code{const} variables in the read-only data section.
19437 This option is only meaningful in conjunction with @option{-membedded-data}.
19439 @item -mcode-readable=@var{setting}
19440 @opindex mcode-readable
19441 Specify whether GCC may generate code that reads from executable sections.
19442 There are three possible settings:
19445 @item -mcode-readable=yes
19446 Instructions may freely access executable sections. This is the
19449 @item -mcode-readable=pcrel
19450 MIPS16 PC-relative load instructions can access executable sections,
19451 but other instructions must not do so. This option is useful on 4KSc
19452 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
19453 It is also useful on processors that can be configured to have a dual
19454 instruction/data SRAM interface and that, like the M4K, automatically
19455 redirect PC-relative loads to the instruction RAM.
19457 @item -mcode-readable=no
19458 Instructions must not access executable sections. This option can be
19459 useful on targets that are configured to have a dual instruction/data
19460 SRAM interface but that (unlike the M4K) do not automatically redirect
19461 PC-relative loads to the instruction RAM.
19464 @item -msplit-addresses
19465 @itemx -mno-split-addresses
19466 @opindex msplit-addresses
19467 @opindex mno-split-addresses
19468 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
19469 relocation operators. This option has been superseded by
19470 @option{-mexplicit-relocs} but is retained for backwards compatibility.
19472 @item -mexplicit-relocs
19473 @itemx -mno-explicit-relocs
19474 @opindex mexplicit-relocs
19475 @opindex mno-explicit-relocs
19476 Use (do not use) assembler relocation operators when dealing with symbolic
19477 addresses. The alternative, selected by @option{-mno-explicit-relocs},
19478 is to use assembler macros instead.
19480 @option{-mexplicit-relocs} is the default if GCC was configured
19481 to use an assembler that supports relocation operators.
19483 @item -mcheck-zero-division
19484 @itemx -mno-check-zero-division
19485 @opindex mcheck-zero-division
19486 @opindex mno-check-zero-division
19487 Trap (do not trap) on integer division by zero.
19489 The default is @option{-mcheck-zero-division}.
19491 @item -mdivide-traps
19492 @itemx -mdivide-breaks
19493 @opindex mdivide-traps
19494 @opindex mdivide-breaks
19495 MIPS systems check for division by zero by generating either a
19496 conditional trap or a break instruction. Using traps results in
19497 smaller code, but is only supported on MIPS II and later. Also, some
19498 versions of the Linux kernel have a bug that prevents trap from
19499 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
19500 allow conditional traps on architectures that support them and
19501 @option{-mdivide-breaks} to force the use of breaks.
19503 The default is usually @option{-mdivide-traps}, but this can be
19504 overridden at configure time using @option{--with-divide=breaks}.
19505 Divide-by-zero checks can be completely disabled using
19506 @option{-mno-check-zero-division}.
19511 @opindex mno-memcpy
19512 Force (do not force) the use of @code{memcpy} for non-trivial block
19513 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
19514 most constant-sized copies.
19517 @itemx -mno-long-calls
19518 @opindex mlong-calls
19519 @opindex mno-long-calls
19520 Disable (do not disable) use of the @code{jal} instruction. Calling
19521 functions using @code{jal} is more efficient but requires the caller
19522 and callee to be in the same 256 megabyte segment.
19524 This option has no effect on abicalls code. The default is
19525 @option{-mno-long-calls}.
19531 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
19532 instructions, as provided by the R4650 ISA@.
19538 Enable (disable) use of the @code{madd} and @code{msub} integer
19539 instructions. The default is @option{-mimadd} on architectures
19540 that support @code{madd} and @code{msub} except for the 74k
19541 architecture where it was found to generate slower code.
19544 @itemx -mno-fused-madd
19545 @opindex mfused-madd
19546 @opindex mno-fused-madd
19547 Enable (disable) use of the floating-point multiply-accumulate
19548 instructions, when they are available. The default is
19549 @option{-mfused-madd}.
19551 On the R8000 CPU when multiply-accumulate instructions are used,
19552 the intermediate product is calculated to infinite precision
19553 and is not subject to the FCSR Flush to Zero bit. This may be
19554 undesirable in some circumstances. On other processors the result
19555 is numerically identical to the equivalent computation using
19556 separate multiply, add, subtract and negate instructions.
19560 Tell the MIPS assembler to not run its preprocessor over user
19561 assembler files (with a @samp{.s} suffix) when assembling them.
19566 @opindex mno-fix-24k
19567 Work around the 24K E48 (lost data on stores during refill) errata.
19568 The workarounds are implemented by the assembler rather than by GCC@.
19571 @itemx -mno-fix-r4000
19572 @opindex mfix-r4000
19573 @opindex mno-fix-r4000
19574 Work around certain R4000 CPU errata:
19577 A double-word or a variable shift may give an incorrect result if executed
19578 immediately after starting an integer division.
19580 A double-word or a variable shift may give an incorrect result if executed
19581 while an integer multiplication is in progress.
19583 An integer division may give an incorrect result if started in a delay slot
19584 of a taken branch or a jump.
19588 @itemx -mno-fix-r4400
19589 @opindex mfix-r4400
19590 @opindex mno-fix-r4400
19591 Work around certain R4400 CPU errata:
19594 A double-word or a variable shift may give an incorrect result if executed
19595 immediately after starting an integer division.
19599 @itemx -mno-fix-r10000
19600 @opindex mfix-r10000
19601 @opindex mno-fix-r10000
19602 Work around certain R10000 errata:
19605 @code{ll}/@code{sc} sequences may not behave atomically on revisions
19606 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
19609 This option can only be used if the target architecture supports
19610 branch-likely instructions. @option{-mfix-r10000} is the default when
19611 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
19615 @itemx -mno-fix-rm7000
19616 @opindex mfix-rm7000
19617 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
19618 workarounds are implemented by the assembler rather than by GCC@.
19621 @itemx -mno-fix-vr4120
19622 @opindex mfix-vr4120
19623 Work around certain VR4120 errata:
19626 @code{dmultu} does not always produce the correct result.
19628 @code{div} and @code{ddiv} do not always produce the correct result if one
19629 of the operands is negative.
19631 The workarounds for the division errata rely on special functions in
19632 @file{libgcc.a}. At present, these functions are only provided by
19633 the @code{mips64vr*-elf} configurations.
19635 Other VR4120 errata require a NOP to be inserted between certain pairs of
19636 instructions. These errata are handled by the assembler, not by GCC itself.
19639 @opindex mfix-vr4130
19640 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
19641 workarounds are implemented by the assembler rather than by GCC,
19642 although GCC avoids using @code{mflo} and @code{mfhi} if the
19643 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
19644 instructions are available instead.
19647 @itemx -mno-fix-sb1
19649 Work around certain SB-1 CPU core errata.
19650 (This flag currently works around the SB-1 revision 2
19651 ``F1'' and ``F2'' floating-point errata.)
19653 @item -mr10k-cache-barrier=@var{setting}
19654 @opindex mr10k-cache-barrier
19655 Specify whether GCC should insert cache barriers to avoid the
19656 side-effects of speculation on R10K processors.
19658 In common with many processors, the R10K tries to predict the outcome
19659 of a conditional branch and speculatively executes instructions from
19660 the ``taken'' branch. It later aborts these instructions if the
19661 predicted outcome is wrong. However, on the R10K, even aborted
19662 instructions can have side effects.
19664 This problem only affects kernel stores and, depending on the system,
19665 kernel loads. As an example, a speculatively-executed store may load
19666 the target memory into cache and mark the cache line as dirty, even if
19667 the store itself is later aborted. If a DMA operation writes to the
19668 same area of memory before the ``dirty'' line is flushed, the cached
19669 data overwrites the DMA-ed data. See the R10K processor manual
19670 for a full description, including other potential problems.
19672 One workaround is to insert cache barrier instructions before every memory
19673 access that might be speculatively executed and that might have side
19674 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
19675 controls GCC's implementation of this workaround. It assumes that
19676 aborted accesses to any byte in the following regions does not have
19681 the memory occupied by the current function's stack frame;
19684 the memory occupied by an incoming stack argument;
19687 the memory occupied by an object with a link-time-constant address.
19690 It is the kernel's responsibility to ensure that speculative
19691 accesses to these regions are indeed safe.
19693 If the input program contains a function declaration such as:
19699 then the implementation of @code{foo} must allow @code{j foo} and
19700 @code{jal foo} to be executed speculatively. GCC honors this
19701 restriction for functions it compiles itself. It expects non-GCC
19702 functions (such as hand-written assembly code) to do the same.
19704 The option has three forms:
19707 @item -mr10k-cache-barrier=load-store
19708 Insert a cache barrier before a load or store that might be
19709 speculatively executed and that might have side effects even
19712 @item -mr10k-cache-barrier=store
19713 Insert a cache barrier before a store that might be speculatively
19714 executed and that might have side effects even if aborted.
19716 @item -mr10k-cache-barrier=none
19717 Disable the insertion of cache barriers. This is the default setting.
19720 @item -mflush-func=@var{func}
19721 @itemx -mno-flush-func
19722 @opindex mflush-func
19723 Specifies the function to call to flush the I and D caches, or to not
19724 call any such function. If called, the function must take the same
19725 arguments as the common @code{_flush_func}, that is, the address of the
19726 memory range for which the cache is being flushed, the size of the
19727 memory range, and the number 3 (to flush both caches). The default
19728 depends on the target GCC was configured for, but commonly is either
19729 @code{_flush_func} or @code{__cpu_flush}.
19731 @item mbranch-cost=@var{num}
19732 @opindex mbranch-cost
19733 Set the cost of branches to roughly @var{num} ``simple'' instructions.
19734 This cost is only a heuristic and is not guaranteed to produce
19735 consistent results across releases. A zero cost redundantly selects
19736 the default, which is based on the @option{-mtune} setting.
19738 @item -mbranch-likely
19739 @itemx -mno-branch-likely
19740 @opindex mbranch-likely
19741 @opindex mno-branch-likely
19742 Enable or disable use of Branch Likely instructions, regardless of the
19743 default for the selected architecture. By default, Branch Likely
19744 instructions may be generated if they are supported by the selected
19745 architecture. An exception is for the MIPS32 and MIPS64 architectures
19746 and processors that implement those architectures; for those, Branch
19747 Likely instructions are not be generated by default because the MIPS32
19748 and MIPS64 architectures specifically deprecate their use.
19750 @item -mcompact-branches=never
19751 @itemx -mcompact-branches=optimal
19752 @itemx -mcompact-branches=always
19753 @opindex mcompact-branches=never
19754 @opindex mcompact-branches=optimal
19755 @opindex mcompact-branches=always
19756 These options control which form of branches will be generated. The
19757 default is @option{-mcompact-branches=optimal}.
19759 The @option{-mcompact-branches=never} option ensures that compact branch
19760 instructions will never be generated.
19762 The @option{-mcompact-branches=always} option ensures that a compact
19763 branch instruction will be generated if available. If a compact branch
19764 instruction is not available, a delay slot form of the branch will be
19767 This option is supported from MIPS Release 6 onwards.
19769 The @option{-mcompact-branches=optimal} option will cause a delay slot
19770 branch to be used if one is available in the current ISA and the delay
19771 slot is successfully filled. If the delay slot is not filled, a compact
19772 branch will be chosen if one is available.
19774 @item -mfp-exceptions
19775 @itemx -mno-fp-exceptions
19776 @opindex mfp-exceptions
19777 Specifies whether FP exceptions are enabled. This affects how
19778 FP instructions are scheduled for some processors.
19779 The default is that FP exceptions are
19782 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
19783 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
19786 @item -mvr4130-align
19787 @itemx -mno-vr4130-align
19788 @opindex mvr4130-align
19789 The VR4130 pipeline is two-way superscalar, but can only issue two
19790 instructions together if the first one is 8-byte aligned. When this
19791 option is enabled, GCC aligns pairs of instructions that it
19792 thinks should execute in parallel.
19794 This option only has an effect when optimizing for the VR4130.
19795 It normally makes code faster, but at the expense of making it bigger.
19796 It is enabled by default at optimization level @option{-O3}.
19801 Enable (disable) generation of @code{synci} instructions on
19802 architectures that support it. The @code{synci} instructions (if
19803 enabled) are generated when @code{__builtin___clear_cache} is
19806 This option defaults to @option{-mno-synci}, but the default can be
19807 overridden by configuring GCC with @option{--with-synci}.
19809 When compiling code for single processor systems, it is generally safe
19810 to use @code{synci}. However, on many multi-core (SMP) systems, it
19811 does not invalidate the instruction caches on all cores and may lead
19812 to undefined behavior.
19814 @item -mrelax-pic-calls
19815 @itemx -mno-relax-pic-calls
19816 @opindex mrelax-pic-calls
19817 Try to turn PIC calls that are normally dispatched via register
19818 @code{$25} into direct calls. This is only possible if the linker can
19819 resolve the destination at link time and if the destination is within
19820 range for a direct call.
19822 @option{-mrelax-pic-calls} is the default if GCC was configured to use
19823 an assembler and a linker that support the @code{.reloc} assembly
19824 directive and @option{-mexplicit-relocs} is in effect. With
19825 @option{-mno-explicit-relocs}, this optimization can be performed by the
19826 assembler and the linker alone without help from the compiler.
19828 @item -mmcount-ra-address
19829 @itemx -mno-mcount-ra-address
19830 @opindex mmcount-ra-address
19831 @opindex mno-mcount-ra-address
19832 Emit (do not emit) code that allows @code{_mcount} to modify the
19833 calling function's return address. When enabled, this option extends
19834 the usual @code{_mcount} interface with a new @var{ra-address}
19835 parameter, which has type @code{intptr_t *} and is passed in register
19836 @code{$12}. @code{_mcount} can then modify the return address by
19837 doing both of the following:
19840 Returning the new address in register @code{$31}.
19842 Storing the new address in @code{*@var{ra-address}},
19843 if @var{ra-address} is nonnull.
19846 The default is @option{-mno-mcount-ra-address}.
19848 @item -mframe-header-opt
19849 @itemx -mno-frame-header-opt
19850 @opindex mframe-header-opt
19851 Enable (disable) frame header optimization in the o32 ABI. When using the
19852 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
19853 function to write out register arguments. When enabled, this optimization
19854 will suppress the allocation of the frame header if it can be determined that
19857 This optimization is off by default at all optimization levels.
19860 @itemx -mno-lxc1-sxc1
19861 @opindex mlxc1-sxc1
19862 When applicable, enable (disable) the generation of @code{lwxc1},
19863 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
19868 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
19869 @code{madd.d} and related instructions. Enabled by default.
19874 @subsection MMIX Options
19875 @cindex MMIX Options
19877 These options are defined for the MMIX:
19881 @itemx -mno-libfuncs
19883 @opindex mno-libfuncs
19884 Specify that intrinsic library functions are being compiled, passing all
19885 values in registers, no matter the size.
19888 @itemx -mno-epsilon
19890 @opindex mno-epsilon
19891 Generate floating-point comparison instructions that compare with respect
19892 to the @code{rE} epsilon register.
19894 @item -mabi=mmixware
19896 @opindex mabi=mmixware
19898 Generate code that passes function parameters and return values that (in
19899 the called function) are seen as registers @code{$0} and up, as opposed to
19900 the GNU ABI which uses global registers @code{$231} and up.
19902 @item -mzero-extend
19903 @itemx -mno-zero-extend
19904 @opindex mzero-extend
19905 @opindex mno-zero-extend
19906 When reading data from memory in sizes shorter than 64 bits, use (do not
19907 use) zero-extending load instructions by default, rather than
19908 sign-extending ones.
19911 @itemx -mno-knuthdiv
19913 @opindex mno-knuthdiv
19914 Make the result of a division yielding a remainder have the same sign as
19915 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
19916 remainder follows the sign of the dividend. Both methods are
19917 arithmetically valid, the latter being almost exclusively used.
19919 @item -mtoplevel-symbols
19920 @itemx -mno-toplevel-symbols
19921 @opindex mtoplevel-symbols
19922 @opindex mno-toplevel-symbols
19923 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
19924 code can be used with the @code{PREFIX} assembly directive.
19928 Generate an executable in the ELF format, rather than the default
19929 @samp{mmo} format used by the @command{mmix} simulator.
19931 @item -mbranch-predict
19932 @itemx -mno-branch-predict
19933 @opindex mbranch-predict
19934 @opindex mno-branch-predict
19935 Use (do not use) the probable-branch instructions, when static branch
19936 prediction indicates a probable branch.
19938 @item -mbase-addresses
19939 @itemx -mno-base-addresses
19940 @opindex mbase-addresses
19941 @opindex mno-base-addresses
19942 Generate (do not generate) code that uses @emph{base addresses}. Using a
19943 base address automatically generates a request (handled by the assembler
19944 and the linker) for a constant to be set up in a global register. The
19945 register is used for one or more base address requests within the range 0
19946 to 255 from the value held in the register. The generally leads to short
19947 and fast code, but the number of different data items that can be
19948 addressed is limited. This means that a program that uses lots of static
19949 data may require @option{-mno-base-addresses}.
19951 @item -msingle-exit
19952 @itemx -mno-single-exit
19953 @opindex msingle-exit
19954 @opindex mno-single-exit
19955 Force (do not force) generated code to have a single exit point in each
19959 @node MN10300 Options
19960 @subsection MN10300 Options
19961 @cindex MN10300 options
19963 These @option{-m} options are defined for Matsushita MN10300 architectures:
19968 Generate code to avoid bugs in the multiply instructions for the MN10300
19969 processors. This is the default.
19971 @item -mno-mult-bug
19972 @opindex mno-mult-bug
19973 Do not generate code to avoid bugs in the multiply instructions for the
19974 MN10300 processors.
19978 Generate code using features specific to the AM33 processor.
19982 Do not generate code using features specific to the AM33 processor. This
19987 Generate code using features specific to the AM33/2.0 processor.
19991 Generate code using features specific to the AM34 processor.
19993 @item -mtune=@var{cpu-type}
19995 Use the timing characteristics of the indicated CPU type when
19996 scheduling instructions. This does not change the targeted processor
19997 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
19998 @samp{am33-2} or @samp{am34}.
20000 @item -mreturn-pointer-on-d0
20001 @opindex mreturn-pointer-on-d0
20002 When generating a function that returns a pointer, return the pointer
20003 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
20004 only in @code{a0}, and attempts to call such functions without a prototype
20005 result in errors. Note that this option is on by default; use
20006 @option{-mno-return-pointer-on-d0} to disable it.
20010 Do not link in the C run-time initialization object file.
20014 Indicate to the linker that it should perform a relaxation optimization pass
20015 to shorten branches, calls and absolute memory addresses. This option only
20016 has an effect when used on the command line for the final link step.
20018 This option makes symbolic debugging impossible.
20022 Allow the compiler to generate @emph{Long Instruction Word}
20023 instructions if the target is the @samp{AM33} or later. This is the
20024 default. This option defines the preprocessor macro @code{__LIW__}.
20028 Do not allow the compiler to generate @emph{Long Instruction Word}
20029 instructions. This option defines the preprocessor macro
20034 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
20035 instructions if the target is the @samp{AM33} or later. This is the
20036 default. This option defines the preprocessor macro @code{__SETLB__}.
20040 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
20041 instructions. This option defines the preprocessor macro
20042 @code{__NO_SETLB__}.
20046 @node Moxie Options
20047 @subsection Moxie Options
20048 @cindex Moxie Options
20054 Generate big-endian code. This is the default for @samp{moxie-*-*}
20059 Generate little-endian code.
20063 Generate mul.x and umul.x instructions. This is the default for
20064 @samp{moxiebox-*-*} configurations.
20068 Do not link in the C run-time initialization object file.
20072 @node MSP430 Options
20073 @subsection MSP430 Options
20074 @cindex MSP430 Options
20076 These options are defined for the MSP430:
20082 Force assembly output to always use hex constants. Normally such
20083 constants are signed decimals, but this option is available for
20084 testsuite and/or aesthetic purposes.
20088 Select the MCU to target. This is used to create a C preprocessor
20089 symbol based upon the MCU name, converted to upper case and pre- and
20090 post-fixed with @samp{__}. This in turn is used by the
20091 @file{msp430.h} header file to select an MCU-specific supplementary
20094 The option also sets the ISA to use. If the MCU name is one that is
20095 known to only support the 430 ISA then that is selected, otherwise the
20096 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
20097 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
20098 name selects the 430X ISA.
20100 In addition an MCU-specific linker script is added to the linker
20101 command line. The script's name is the name of the MCU with
20102 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
20103 command line defines the C preprocessor symbol @code{__XXX__} and
20104 cause the linker to search for a script called @file{xxx.ld}.
20106 This option is also passed on to the assembler.
20109 @itemx -mno-warn-mcu
20111 @opindex mno-warn-mcu
20112 This option enables or disables warnings about conflicts between the
20113 MCU name specified by the @option{-mmcu} option and the ISA set by the
20114 @option{-mcpu} option and/or the hardware multiply support set by the
20115 @option{-mhwmult} option. It also toggles warnings about unrecognized
20116 MCU names. This option is on by default.
20120 Specifies the ISA to use. Accepted values are @samp{msp430},
20121 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
20122 @option{-mmcu=} option should be used to select the ISA.
20126 Link to the simulator runtime libraries and linker script. Overrides
20127 any scripts that would be selected by the @option{-mmcu=} option.
20131 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
20135 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
20139 This option is passed to the assembler and linker, and allows the
20140 linker to perform certain optimizations that cannot be done until
20145 Describes the type of hardware multiply supported by the target.
20146 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
20147 for the original 16-bit-only multiply supported by early MCUs.
20148 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
20149 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
20150 A value of @samp{auto} can also be given. This tells GCC to deduce
20151 the hardware multiply support based upon the MCU name provided by the
20152 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
20153 the MCU name is not recognized then no hardware multiply support is
20154 assumed. @code{auto} is the default setting.
20156 Hardware multiplies are normally performed by calling a library
20157 routine. This saves space in the generated code. When compiling at
20158 @option{-O3} or higher however the hardware multiplier is invoked
20159 inline. This makes for bigger, but faster code.
20161 The hardware multiply routines disable interrupts whilst running and
20162 restore the previous interrupt state when they finish. This makes
20163 them safe to use inside interrupt handlers as well as in normal code.
20167 Enable the use of a minimum runtime environment - no static
20168 initializers or constructors. This is intended for memory-constrained
20169 devices. The compiler includes special symbols in some objects
20170 that tell the linker and runtime which code fragments are required.
20172 @item -mcode-region=
20173 @itemx -mdata-region=
20174 @opindex mcode-region
20175 @opindex mdata-region
20176 These options tell the compiler where to place functions and data that
20177 do not have one of the @code{lower}, @code{upper}, @code{either} or
20178 @code{section} attributes. Possible values are @code{lower},
20179 @code{upper}, @code{either} or @code{any}. The first three behave
20180 like the corresponding attribute. The fourth possible value -
20181 @code{any} - is the default. It leaves placement entirely up to the
20182 linker script and how it assigns the standard sections
20183 (@code{.text}, @code{.data}, etc) to the memory regions.
20185 @item -msilicon-errata=
20186 @opindex msilicon-errata
20187 This option passes on a request to assembler to enable the fixes for
20188 the named silicon errata.
20190 @item -msilicon-errata-warn=
20191 @opindex msilicon-errata-warn
20192 This option passes on a request to the assembler to enable warning
20193 messages when a silicon errata might need to be applied.
20197 @node NDS32 Options
20198 @subsection NDS32 Options
20199 @cindex NDS32 Options
20201 These options are defined for NDS32 implementations:
20206 @opindex mbig-endian
20207 Generate code in big-endian mode.
20209 @item -mlittle-endian
20210 @opindex mlittle-endian
20211 Generate code in little-endian mode.
20213 @item -mreduced-regs
20214 @opindex mreduced-regs
20215 Use reduced-set registers for register allocation.
20218 @opindex mfull-regs
20219 Use full-set registers for register allocation.
20223 Generate conditional move instructions.
20227 Do not generate conditional move instructions.
20231 Generate performance extension instructions.
20233 @item -mno-perf-ext
20234 @opindex mno-perf-ext
20235 Do not generate performance extension instructions.
20239 Generate v3 push25/pop25 instructions.
20242 @opindex mno-v3push
20243 Do not generate v3 push25/pop25 instructions.
20247 Generate 16-bit instructions.
20250 @opindex mno-16-bit
20251 Do not generate 16-bit instructions.
20253 @item -misr-vector-size=@var{num}
20254 @opindex misr-vector-size
20255 Specify the size of each interrupt vector, which must be 4 or 16.
20257 @item -mcache-block-size=@var{num}
20258 @opindex mcache-block-size
20259 Specify the size of each cache block,
20260 which must be a power of 2 between 4 and 512.
20262 @item -march=@var{arch}
20264 Specify the name of the target architecture.
20266 @item -mcmodel=@var{code-model}
20268 Set the code model to one of
20271 All the data and read-only data segments must be within 512KB addressing space.
20272 The text segment must be within 16MB addressing space.
20273 @item @samp{medium}
20274 The data segment must be within 512KB while the read-only data segment can be
20275 within 4GB addressing space. The text segment should be still within 16MB
20278 All the text and data segments can be within 4GB addressing space.
20282 @opindex mctor-dtor
20283 Enable constructor/destructor feature.
20287 Guide linker to relax instructions.
20291 @node Nios II Options
20292 @subsection Nios II Options
20293 @cindex Nios II options
20294 @cindex Altera Nios II options
20296 These are the options defined for the Altera Nios II processor.
20302 @cindex smaller data references
20303 Put global and static objects less than or equal to @var{num} bytes
20304 into the small data or BSS sections instead of the normal data or BSS
20305 sections. The default value of @var{num} is 8.
20307 @item -mgpopt=@var{option}
20312 Generate (do not generate) GP-relative accesses. The following
20313 @var{option} names are recognized:
20318 Do not generate GP-relative accesses.
20321 Generate GP-relative accesses for small data objects that are not
20322 external, weak, or uninitialized common symbols.
20323 Also use GP-relative addressing for objects that
20324 have been explicitly placed in a small data section via a @code{section}
20328 As for @samp{local}, but also generate GP-relative accesses for
20329 small data objects that are external, weak, or common. If you use this option,
20330 you must ensure that all parts of your program (including libraries) are
20331 compiled with the same @option{-G} setting.
20334 Generate GP-relative accesses for all data objects in the program. If you
20335 use this option, the entire data and BSS segments
20336 of your program must fit in 64K of memory and you must use an appropriate
20337 linker script to allocate them within the addressable range of the
20341 Generate GP-relative addresses for function pointers as well as data
20342 pointers. If you use this option, the entire text, data, and BSS segments
20343 of your program must fit in 64K of memory and you must use an appropriate
20344 linker script to allocate them within the addressable range of the
20349 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
20350 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
20352 The default is @option{-mgpopt} except when @option{-fpic} or
20353 @option{-fPIC} is specified to generate position-independent code.
20354 Note that the Nios II ABI does not permit GP-relative accesses from
20357 You may need to specify @option{-mno-gpopt} explicitly when building
20358 programs that include large amounts of small data, including large
20359 GOT data sections. In this case, the 16-bit offset for GP-relative
20360 addressing may not be large enough to allow access to the entire
20361 small data section.
20367 Generate little-endian (default) or big-endian (experimental) code,
20370 @item -march=@var{arch}
20372 This specifies the name of the target Nios II architecture. GCC uses this
20373 name to determine what kind of instructions it can emit when generating
20374 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
20376 The preprocessor macro @code{__nios2_arch__} is available to programs,
20377 with value 1 or 2, indicating the targeted ISA level.
20379 @item -mbypass-cache
20380 @itemx -mno-bypass-cache
20381 @opindex mno-bypass-cache
20382 @opindex mbypass-cache
20383 Force all load and store instructions to always bypass cache by
20384 using I/O variants of the instructions. The default is not to
20387 @item -mno-cache-volatile
20388 @itemx -mcache-volatile
20389 @opindex mcache-volatile
20390 @opindex mno-cache-volatile
20391 Volatile memory access bypass the cache using the I/O variants of
20392 the load and store instructions. The default is not to bypass the cache.
20394 @item -mno-fast-sw-div
20395 @itemx -mfast-sw-div
20396 @opindex mno-fast-sw-div
20397 @opindex mfast-sw-div
20398 Do not use table-based fast divide for small numbers. The default
20399 is to use the fast divide at @option{-O3} and above.
20403 @itemx -mno-hw-mulx
20407 @opindex mno-hw-mul
20409 @opindex mno-hw-mulx
20411 @opindex mno-hw-div
20413 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
20414 instructions by the compiler. The default is to emit @code{mul}
20415 and not emit @code{div} and @code{mulx}.
20421 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
20422 CDX (code density) instructions. Enabling these instructions also
20423 requires @option{-march=r2}. Since these instructions are optional
20424 extensions to the R2 architecture, the default is not to emit them.
20426 @item -mcustom-@var{insn}=@var{N}
20427 @itemx -mno-custom-@var{insn}
20428 @opindex mcustom-@var{insn}
20429 @opindex mno-custom-@var{insn}
20430 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
20431 custom instruction with encoding @var{N} when generating code that uses
20432 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
20433 instruction 253 for single-precision floating-point add operations instead
20434 of the default behavior of using a library call.
20436 The following values of @var{insn} are supported. Except as otherwise
20437 noted, floating-point operations are expected to be implemented with
20438 normal IEEE 754 semantics and correspond directly to the C operators or the
20439 equivalent GCC built-in functions (@pxref{Other Builtins}).
20441 Single-precision floating point:
20444 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
20445 Binary arithmetic operations.
20451 Unary absolute value.
20453 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
20454 Comparison operations.
20456 @item @samp{fmins}, @samp{fmaxs}
20457 Floating-point minimum and maximum. These instructions are only
20458 generated if @option{-ffinite-math-only} is specified.
20460 @item @samp{fsqrts}
20461 Unary square root operation.
20463 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
20464 Floating-point trigonometric and exponential functions. These instructions
20465 are only generated if @option{-funsafe-math-optimizations} is also specified.
20469 Double-precision floating point:
20472 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
20473 Binary arithmetic operations.
20479 Unary absolute value.
20481 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
20482 Comparison operations.
20484 @item @samp{fmind}, @samp{fmaxd}
20485 Double-precision minimum and maximum. These instructions are only
20486 generated if @option{-ffinite-math-only} is specified.
20488 @item @samp{fsqrtd}
20489 Unary square root operation.
20491 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
20492 Double-precision trigonometric and exponential functions. These instructions
20493 are only generated if @option{-funsafe-math-optimizations} is also specified.
20499 @item @samp{fextsd}
20500 Conversion from single precision to double precision.
20502 @item @samp{ftruncds}
20503 Conversion from double precision to single precision.
20505 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
20506 Conversion from floating point to signed or unsigned integer types, with
20507 truncation towards zero.
20510 Conversion from single-precision floating point to signed integer,
20511 rounding to the nearest integer and ties away from zero.
20512 This corresponds to the @code{__builtin_lroundf} function when
20513 @option{-fno-math-errno} is used.
20515 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
20516 Conversion from signed or unsigned integer types to floating-point types.
20520 In addition, all of the following transfer instructions for internal
20521 registers X and Y must be provided to use any of the double-precision
20522 floating-point instructions. Custom instructions taking two
20523 double-precision source operands expect the first operand in the
20524 64-bit register X. The other operand (or only operand of a unary
20525 operation) is given to the custom arithmetic instruction with the
20526 least significant half in source register @var{src1} and the most
20527 significant half in @var{src2}. A custom instruction that returns a
20528 double-precision result returns the most significant 32 bits in the
20529 destination register and the other half in 32-bit register Y.
20530 GCC automatically generates the necessary code sequences to write
20531 register X and/or read register Y when double-precision floating-point
20532 instructions are used.
20537 Write @var{src1} into the least significant half of X and @var{src2} into
20538 the most significant half of X.
20541 Write @var{src1} into Y.
20543 @item @samp{frdxhi}, @samp{frdxlo}
20544 Read the most or least (respectively) significant half of X and store it in
20548 Read the value of Y and store it into @var{dest}.
20551 Note that you can gain more local control over generation of Nios II custom
20552 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
20553 and @code{target("no-custom-@var{insn}")} function attributes
20554 (@pxref{Function Attributes})
20555 or pragmas (@pxref{Function Specific Option Pragmas}).
20557 @item -mcustom-fpu-cfg=@var{name}
20558 @opindex mcustom-fpu-cfg
20560 This option enables a predefined, named set of custom instruction encodings
20561 (see @option{-mcustom-@var{insn}} above).
20562 Currently, the following sets are defined:
20564 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
20565 @gccoptlist{-mcustom-fmuls=252 @gol
20566 -mcustom-fadds=253 @gol
20567 -mcustom-fsubs=254 @gol
20568 -fsingle-precision-constant}
20570 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
20571 @gccoptlist{-mcustom-fmuls=252 @gol
20572 -mcustom-fadds=253 @gol
20573 -mcustom-fsubs=254 @gol
20574 -mcustom-fdivs=255 @gol
20575 -fsingle-precision-constant}
20577 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
20578 @gccoptlist{-mcustom-floatus=243 @gol
20579 -mcustom-fixsi=244 @gol
20580 -mcustom-floatis=245 @gol
20581 -mcustom-fcmpgts=246 @gol
20582 -mcustom-fcmples=249 @gol
20583 -mcustom-fcmpeqs=250 @gol
20584 -mcustom-fcmpnes=251 @gol
20585 -mcustom-fmuls=252 @gol
20586 -mcustom-fadds=253 @gol
20587 -mcustom-fsubs=254 @gol
20588 -mcustom-fdivs=255 @gol
20589 -fsingle-precision-constant}
20591 Custom instruction assignments given by individual
20592 @option{-mcustom-@var{insn}=} options override those given by
20593 @option{-mcustom-fpu-cfg=}, regardless of the
20594 order of the options on the command line.
20596 Note that you can gain more local control over selection of a FPU
20597 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
20598 function attribute (@pxref{Function Attributes})
20599 or pragma (@pxref{Function Specific Option Pragmas}).
20603 These additional @samp{-m} options are available for the Altera Nios II
20604 ELF (bare-metal) target:
20610 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
20611 startup and termination code, and is typically used in conjunction with
20612 @option{-msys-crt0=} to specify the location of the alternate startup code
20613 provided by the HAL BSP.
20617 Link with a limited version of the C library, @option{-lsmallc}, rather than
20620 @item -msys-crt0=@var{startfile}
20622 @var{startfile} is the file name of the startfile (crt0) to use
20623 when linking. This option is only useful in conjunction with @option{-mhal}.
20625 @item -msys-lib=@var{systemlib}
20627 @var{systemlib} is the library name of the library that provides
20628 low-level system calls required by the C library,
20629 e.g. @code{read} and @code{write}.
20630 This option is typically used to link with a library provided by a HAL BSP.
20634 @node Nvidia PTX Options
20635 @subsection Nvidia PTX Options
20636 @cindex Nvidia PTX options
20637 @cindex nvptx options
20639 These options are defined for Nvidia PTX:
20647 Generate code for 32-bit or 64-bit ABI.
20650 @opindex mmainkernel
20651 Link in code for a __main kernel. This is for stand-alone instead of
20652 offloading execution.
20656 Apply partitioned execution optimizations. This is the default when any
20657 level of optimization is selected.
20660 @opindex msoft-stack
20661 Generate code that does not use @code{.local} memory
20662 directly for stack storage. Instead, a per-warp stack pointer is
20663 maintained explicitly. This enables variable-length stack allocation (with
20664 variable-length arrays or @code{alloca}), and when global memory is used for
20665 underlying storage, makes it possible to access automatic variables from other
20666 threads, or with atomic instructions. This code generation variant is used
20667 for OpenMP offloading, but the option is exposed on its own for the purpose
20668 of testing the compiler; to generate code suitable for linking into programs
20669 using OpenMP offloading, use option @option{-mgomp}.
20671 @item -muniform-simt
20672 @opindex muniform-simt
20673 Switch to code generation variant that allows to execute all threads in each
20674 warp, while maintaining memory state and side effects as if only one thread
20675 in each warp was active outside of OpenMP SIMD regions. All atomic operations
20676 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
20677 current lane index equals the master lane index), and the register being
20678 assigned is copied via a shuffle instruction from the master lane. Outside of
20679 SIMD regions lane 0 is the master; inside, each thread sees itself as the
20680 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
20681 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
20682 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
20683 with current lane index to compute the master lane index.
20687 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
20688 @option{-muniform-simt} options, and selects corresponding multilib variant.
20692 @node PDP-11 Options
20693 @subsection PDP-11 Options
20694 @cindex PDP-11 Options
20696 These options are defined for the PDP-11:
20701 Use hardware FPP floating point. This is the default. (FIS floating
20702 point on the PDP-11/40 is not supported.)
20705 @opindex msoft-float
20706 Do not use hardware floating point.
20710 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
20714 Return floating-point results in memory. This is the default.
20718 Generate code for a PDP-11/40.
20722 Generate code for a PDP-11/45. This is the default.
20726 Generate code for a PDP-11/10.
20728 @item -mbcopy-builtin
20729 @opindex mbcopy-builtin
20730 Use inline @code{movmemhi} patterns for copying memory. This is the
20735 Do not use inline @code{movmemhi} patterns for copying memory.
20741 Use 16-bit @code{int}. This is the default.
20747 Use 32-bit @code{int}.
20750 @itemx -mno-float32
20752 @opindex mno-float32
20753 Use 64-bit @code{float}. This is the default.
20756 @itemx -mno-float64
20758 @opindex mno-float64
20759 Use 32-bit @code{float}.
20763 Use @code{abshi2} pattern. This is the default.
20767 Do not use @code{abshi2} pattern.
20769 @item -mbranch-expensive
20770 @opindex mbranch-expensive
20771 Pretend that branches are expensive. This is for experimenting with
20772 code generation only.
20774 @item -mbranch-cheap
20775 @opindex mbranch-cheap
20776 Do not pretend that branches are expensive. This is the default.
20780 Use Unix assembler syntax. This is the default when configured for
20781 @samp{pdp11-*-bsd}.
20785 Use DEC assembler syntax. This is the default when configured for any
20786 PDP-11 target other than @samp{pdp11-*-bsd}.
20789 @node picoChip Options
20790 @subsection picoChip Options
20791 @cindex picoChip options
20793 These @samp{-m} options are defined for picoChip implementations:
20797 @item -mae=@var{ae_type}
20799 Set the instruction set, register set, and instruction scheduling
20800 parameters for array element type @var{ae_type}. Supported values
20801 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
20803 @option{-mae=ANY} selects a completely generic AE type. Code
20804 generated with this option runs on any of the other AE types. The
20805 code is not as efficient as it would be if compiled for a specific
20806 AE type, and some types of operation (e.g., multiplication) do not
20807 work properly on all types of AE.
20809 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
20810 for compiled code, and is the default.
20812 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
20813 option may suffer from poor performance of byte (char) manipulation,
20814 since the DSP AE does not provide hardware support for byte load/stores.
20816 @item -msymbol-as-address
20817 Enable the compiler to directly use a symbol name as an address in a
20818 load/store instruction, without first loading it into a
20819 register. Typically, the use of this option generates larger
20820 programs, which run faster than when the option isn't used. However, the
20821 results vary from program to program, so it is left as a user option,
20822 rather than being permanently enabled.
20824 @item -mno-inefficient-warnings
20825 Disables warnings about the generation of inefficient code. These
20826 warnings can be generated, for example, when compiling code that
20827 performs byte-level memory operations on the MAC AE type. The MAC AE has
20828 no hardware support for byte-level memory operations, so all byte
20829 load/stores must be synthesized from word load/store operations. This is
20830 inefficient and a warning is generated to indicate
20831 that you should rewrite the code to avoid byte operations, or to target
20832 an AE type that has the necessary hardware support. This option disables
20837 @node PowerPC Options
20838 @subsection PowerPC Options
20839 @cindex PowerPC options
20841 These are listed under @xref{RS/6000 and PowerPC Options}.
20843 @node RISC-V Options
20844 @subsection RISC-V Options
20845 @cindex RISC-V Options
20847 These command-line options are defined for RISC-V targets:
20850 @item -mbranch-cost=@var{n}
20851 @opindex mbranch-cost
20852 Set the cost of branches to roughly @var{n} instructions.
20857 Don't optimize block moves.
20862 When generating PIC code, allow the use of PLTs. Ignored for non-PIC.
20864 @item -mabi=@var{ABI-string}
20866 Specify integer and floating-point calling convention. This defaults to the
20867 natural calling convention: e.g.@ LP64 for RV64I, ILP32 for RV32I, LP64D for
20873 Use hardware floating-point divide and square root instructions. This requires
20874 the F or D extensions for floating-point registers.
20879 Use hardware instructions for integer division. This requires the M extension.
20881 @item -march=@var{ISA-string}
20883 Generate code for given RISC-V ISA (e.g.@ @samp{rv64im}). ISA strings must be
20884 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, and @samp{rv32imaf}.
20886 @item -mtune=@var{processor-string}
20888 Optimize the output for the given processor, specified by microarchitecture
20891 @item -msmall-data-limit=@var{n}
20892 @opindex msmall-data-limit
20893 Put global and static data smaller than @var{n} bytes into a special section
20896 @item -msave-restore
20897 @itemx -mno-save-restore
20898 @opindex msave-restore
20899 Use smaller but slower prologue and epilogue code.
20901 @item -mcmodel=@var{code-model}
20903 Specify the code model.
20908 @subsection RL78 Options
20909 @cindex RL78 Options
20915 Links in additional target libraries to support operation within a
20924 Specifies the type of hardware multiplication and division support to
20925 be used. The simplest is @code{none}, which uses software for both
20926 multiplication and division. This is the default. The @code{g13}
20927 value is for the hardware multiply/divide peripheral found on the
20928 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
20929 the multiplication and division instructions supported by the RL78/G14
20930 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
20931 the value @code{mg10} is an alias for @code{none}.
20933 In addition a C preprocessor macro is defined, based upon the setting
20934 of this option. Possible values are: @code{__RL78_MUL_NONE__},
20935 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
20942 Specifies the RL78 core to target. The default is the G14 core, also
20943 known as an S3 core or just RL78. The G13 or S2 core does not have
20944 multiply or divide instructions, instead it uses a hardware peripheral
20945 for these operations. The G10 or S1 core does not have register
20946 banks, so it uses a different calling convention.
20948 If this option is set it also selects the type of hardware multiply
20949 support to use, unless this is overridden by an explicit
20950 @option{-mmul=none} option on the command line. Thus specifying
20951 @option{-mcpu=g13} enables the use of the G13 hardware multiply
20952 peripheral and specifying @option{-mcpu=g10} disables the use of
20953 hardware multiplications altogether.
20955 Note, although the RL78/G14 core is the default target, specifying
20956 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
20957 change the behavior of the toolchain since it also enables G14
20958 hardware multiply support. If these options are not specified on the
20959 command line then software multiplication routines will be used even
20960 though the code targets the RL78 core. This is for backwards
20961 compatibility with older toolchains which did not have hardware
20962 multiply and divide support.
20964 In addition a C preprocessor macro is defined, based upon the setting
20965 of this option. Possible values are: @code{__RL78_G10__},
20966 @code{__RL78_G13__} or @code{__RL78_G14__}.
20976 These are aliases for the corresponding @option{-mcpu=} option. They
20977 are provided for backwards compatibility.
20981 Allow the compiler to use all of the available registers. By default
20982 registers @code{r24..r31} are reserved for use in interrupt handlers.
20983 With this option enabled these registers can be used in ordinary
20986 @item -m64bit-doubles
20987 @itemx -m32bit-doubles
20988 @opindex m64bit-doubles
20989 @opindex m32bit-doubles
20990 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20991 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20992 @option{-m32bit-doubles}.
20994 @item -msave-mduc-in-interrupts
20995 @item -mno-save-mduc-in-interrupts
20996 @opindex msave-mduc-in-interrupts
20997 @opindex mno-save-mduc-in-interrupts
20998 Specifies that interrupt handler functions should preserve the
20999 MDUC registers. This is only necessary if normal code might use
21000 the MDUC registers, for example because it performs multiplication
21001 and division operations. The default is to ignore the MDUC registers
21002 as this makes the interrupt handlers faster. The target option -mg13
21003 needs to be passed for this to work as this feature is only available
21004 on the G13 target (S2 core). The MDUC registers will only be saved
21005 if the interrupt handler performs a multiplication or division
21006 operation or it calls another function.
21010 @node RS/6000 and PowerPC Options
21011 @subsection IBM RS/6000 and PowerPC Options
21012 @cindex RS/6000 and PowerPC Options
21013 @cindex IBM RS/6000 and PowerPC Options
21015 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
21017 @item -mpowerpc-gpopt
21018 @itemx -mno-powerpc-gpopt
21019 @itemx -mpowerpc-gfxopt
21020 @itemx -mno-powerpc-gfxopt
21023 @itemx -mno-powerpc64
21027 @itemx -mno-popcntb
21029 @itemx -mno-popcntd
21038 @itemx -mno-hard-dfp
21039 @opindex mpowerpc-gpopt
21040 @opindex mno-powerpc-gpopt
21041 @opindex mpowerpc-gfxopt
21042 @opindex mno-powerpc-gfxopt
21043 @opindex mpowerpc64
21044 @opindex mno-powerpc64
21048 @opindex mno-popcntb
21050 @opindex mno-popcntd
21056 @opindex mno-mfpgpr
21058 @opindex mno-hard-dfp
21059 You use these options to specify which instructions are available on the
21060 processor you are using. The default value of these options is
21061 determined when configuring GCC@. Specifying the
21062 @option{-mcpu=@var{cpu_type}} overrides the specification of these
21063 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
21064 rather than the options listed above.
21066 Specifying @option{-mpowerpc-gpopt} allows
21067 GCC to use the optional PowerPC architecture instructions in the
21068 General Purpose group, including floating-point square root. Specifying
21069 @option{-mpowerpc-gfxopt} allows GCC to
21070 use the optional PowerPC architecture instructions in the Graphics
21071 group, including floating-point select.
21073 The @option{-mmfcrf} option allows GCC to generate the move from
21074 condition register field instruction implemented on the POWER4
21075 processor and other processors that support the PowerPC V2.01
21077 The @option{-mpopcntb} option allows GCC to generate the popcount and
21078 double-precision FP reciprocal estimate instruction implemented on the
21079 POWER5 processor and other processors that support the PowerPC V2.02
21081 The @option{-mpopcntd} option allows GCC to generate the popcount
21082 instruction implemented on the POWER7 processor and other processors
21083 that support the PowerPC V2.06 architecture.
21084 The @option{-mfprnd} option allows GCC to generate the FP round to
21085 integer instructions implemented on the POWER5+ processor and other
21086 processors that support the PowerPC V2.03 architecture.
21087 The @option{-mcmpb} option allows GCC to generate the compare bytes
21088 instruction implemented on the POWER6 processor and other processors
21089 that support the PowerPC V2.05 architecture.
21090 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
21091 general-purpose register instructions implemented on the POWER6X
21092 processor and other processors that support the extended PowerPC V2.05
21094 The @option{-mhard-dfp} option allows GCC to generate the decimal
21095 floating-point instructions implemented on some POWER processors.
21097 The @option{-mpowerpc64} option allows GCC to generate the additional
21098 64-bit instructions that are found in the full PowerPC64 architecture
21099 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
21100 @option{-mno-powerpc64}.
21102 @item -mcpu=@var{cpu_type}
21104 Set architecture type, register usage, and
21105 instruction scheduling parameters for machine type @var{cpu_type}.
21106 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
21107 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
21108 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
21109 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
21110 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
21111 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
21112 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
21113 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
21114 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
21115 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
21116 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
21119 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
21120 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
21121 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
21122 architecture machine types, with an appropriate, generic processor
21123 model assumed for scheduling purposes.
21125 The other options specify a specific processor. Code generated under
21126 those options runs best on that processor, and may not run at all on
21129 The @option{-mcpu} options automatically enable or disable the
21132 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
21133 -mpopcntb -mpopcntd -mpowerpc64 @gol
21134 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
21135 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
21136 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
21137 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
21139 The particular options set for any particular CPU varies between
21140 compiler versions, depending on what setting seems to produce optimal
21141 code for that CPU; it doesn't necessarily reflect the actual hardware's
21142 capabilities. If you wish to set an individual option to a particular
21143 value, you may specify it after the @option{-mcpu} option, like
21144 @option{-mcpu=970 -mno-altivec}.
21146 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
21147 not enabled or disabled by the @option{-mcpu} option at present because
21148 AIX does not have full support for these options. You may still
21149 enable or disable them individually if you're sure it'll work in your
21152 @item -mtune=@var{cpu_type}
21154 Set the instruction scheduling parameters for machine type
21155 @var{cpu_type}, but do not set the architecture type or register usage,
21156 as @option{-mcpu=@var{cpu_type}} does. The same
21157 values for @var{cpu_type} are used for @option{-mtune} as for
21158 @option{-mcpu}. If both are specified, the code generated uses the
21159 architecture and registers set by @option{-mcpu}, but the
21160 scheduling parameters set by @option{-mtune}.
21162 @item -mcmodel=small
21163 @opindex mcmodel=small
21164 Generate PowerPC64 code for the small model: The TOC is limited to
21167 @item -mcmodel=medium
21168 @opindex mcmodel=medium
21169 Generate PowerPC64 code for the medium model: The TOC and other static
21170 data may be up to a total of 4G in size.
21172 @item -mcmodel=large
21173 @opindex mcmodel=large
21174 Generate PowerPC64 code for the large model: The TOC may be up to 4G
21175 in size. Other data and code is only limited by the 64-bit address
21179 @itemx -mno-altivec
21181 @opindex mno-altivec
21182 Generate code that uses (does not use) AltiVec instructions, and also
21183 enable the use of built-in functions that allow more direct access to
21184 the AltiVec instruction set. You may also need to set
21185 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
21188 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
21189 @option{-maltivec=be}, the element order for AltiVec intrinsics such
21190 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
21191 match array element order corresponding to the endianness of the
21192 target. That is, element zero identifies the leftmost element in a
21193 vector register when targeting a big-endian platform, and identifies
21194 the rightmost element in a vector register when targeting a
21195 little-endian platform.
21198 @opindex maltivec=be
21199 Generate AltiVec instructions using big-endian element order,
21200 regardless of whether the target is big- or little-endian. This is
21201 the default when targeting a big-endian platform.
21203 The element order is used to interpret element numbers in AltiVec
21204 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21205 @code{vec_insert}. By default, these match array element order
21206 corresponding to the endianness for the target.
21209 @opindex maltivec=le
21210 Generate AltiVec instructions using little-endian element order,
21211 regardless of whether the target is big- or little-endian. This is
21212 the default when targeting a little-endian platform. This option is
21213 currently ignored when targeting a big-endian platform.
21215 The element order is used to interpret element numbers in AltiVec
21216 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
21217 @code{vec_insert}. By default, these match array element order
21218 corresponding to the endianness for the target.
21223 @opindex mno-vrsave
21224 Generate VRSAVE instructions when generating AltiVec code.
21226 @item -mgen-cell-microcode
21227 @opindex mgen-cell-microcode
21228 Generate Cell microcode instructions.
21230 @item -mwarn-cell-microcode
21231 @opindex mwarn-cell-microcode
21232 Warn when a Cell microcode instruction is emitted. An example
21233 of a Cell microcode instruction is a variable shift.
21236 @opindex msecure-plt
21237 Generate code that allows @command{ld} and @command{ld.so}
21238 to build executables and shared
21239 libraries with non-executable @code{.plt} and @code{.got} sections.
21241 32-bit SYSV ABI option.
21245 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
21247 requires @code{.plt} and @code{.got}
21248 sections that are both writable and executable.
21249 This is a PowerPC 32-bit SYSV ABI option.
21255 This switch enables or disables the generation of ISEL instructions.
21257 @item -misel=@var{yes/no}
21258 This switch has been deprecated. Use @option{-misel} and
21259 @option{-mno-isel} instead.
21263 Enable Local Register Allocation. By default the port uses LRA.
21264 (i.e. @option{-mno-lra}).
21270 This switch enables or disables the generation of SPE simd
21276 @opindex mno-paired
21277 This switch enables or disables the generation of PAIRED simd
21280 @item -mspe=@var{yes/no}
21281 This option has been deprecated. Use @option{-mspe} and
21282 @option{-mno-spe} instead.
21288 Generate code that uses (does not use) vector/scalar (VSX)
21289 instructions, and also enable the use of built-in functions that allow
21290 more direct access to the VSX instruction set.
21295 @opindex mno-crypto
21296 Enable the use (disable) of the built-in functions that allow direct
21297 access to the cryptographic instructions that were added in version
21298 2.07 of the PowerPC ISA.
21300 @item -mdirect-move
21301 @itemx -mno-direct-move
21302 @opindex mdirect-move
21303 @opindex mno-direct-move
21304 Generate code that uses (does not use) the instructions to move data
21305 between the general purpose registers and the vector/scalar (VSX)
21306 registers that were added in version 2.07 of the PowerPC ISA.
21312 Enable (disable) the use of the built-in functions that allow direct
21313 access to the Hardware Transactional Memory (HTM) instructions that
21314 were added in version 2.07 of the PowerPC ISA.
21316 @item -mpower8-fusion
21317 @itemx -mno-power8-fusion
21318 @opindex mpower8-fusion
21319 @opindex mno-power8-fusion
21320 Generate code that keeps (does not keeps) some integer operations
21321 adjacent so that the instructions can be fused together on power8 and
21324 @item -mpower8-vector
21325 @itemx -mno-power8-vector
21326 @opindex mpower8-vector
21327 @opindex mno-power8-vector
21328 Generate code that uses (does not use) the vector and scalar
21329 instructions that were added in version 2.07 of the PowerPC ISA. Also
21330 enable the use of built-in functions that allow more direct access to
21331 the vector instructions.
21333 @item -mquad-memory
21334 @itemx -mno-quad-memory
21335 @opindex mquad-memory
21336 @opindex mno-quad-memory
21337 Generate code that uses (does not use) the non-atomic quad word memory
21338 instructions. The @option{-mquad-memory} option requires use of
21341 @item -mquad-memory-atomic
21342 @itemx -mno-quad-memory-atomic
21343 @opindex mquad-memory-atomic
21344 @opindex mno-quad-memory-atomic
21345 Generate code that uses (does not use) the atomic quad word memory
21346 instructions. The @option{-mquad-memory-atomic} option requires use of
21349 @item -mupper-regs-di
21350 @itemx -mno-upper-regs-di
21351 @opindex mupper-regs-di
21352 @opindex mno-upper-regs-di
21353 Generate code that uses (does not use) the scalar instructions that
21354 target all 64 registers in the vector/scalar floating point register
21355 set that were added in version 2.06 of the PowerPC ISA when processing
21356 integers. @option{-mupper-regs-di} is turned on by default if you use
21357 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21358 @option{-mcpu=power9}, or @option{-mvsx} options.
21360 @item -mupper-regs-df
21361 @itemx -mno-upper-regs-df
21362 @opindex mupper-regs-df
21363 @opindex mno-upper-regs-df
21364 Generate code that uses (does not use) the scalar double precision
21365 instructions that target all 64 registers in the vector/scalar
21366 floating point register set that were added in version 2.06 of the
21367 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
21368 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
21369 @option{-mcpu=power9}, or @option{-mvsx} options.
21371 @item -mupper-regs-sf
21372 @itemx -mno-upper-regs-sf
21373 @opindex mupper-regs-sf
21374 @opindex mno-upper-regs-sf
21375 Generate code that uses (does not use) the scalar single precision
21376 instructions that target all 64 registers in the vector/scalar
21377 floating point register set that were added in version 2.07 of the
21378 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
21379 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
21380 @option{-mcpu=power9} options.
21383 @itemx -mno-upper-regs
21384 @opindex mupper-regs
21385 @opindex mno-upper-regs
21386 Generate code that uses (does not use) the scalar
21387 instructions that target all 64 registers in the vector/scalar
21388 floating point register set, depending on the model of the machine.
21390 If the @option{-mno-upper-regs} option is used, it turns off both
21391 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
21394 @itemx -mno-float128
21396 @opindex mno-float128
21397 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
21398 and use either software emulation for IEEE 128-bit floating point or
21399 hardware instructions.
21401 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
21402 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
21403 option. The @option{-mfloat128} option only works on PowerPC 64-bit
21406 If you use the ISA 3.0 instruction set (@option{-mcpu=power9}), the
21407 @option{-mfloat128} option will also enable the generation of ISA 3.0
21408 IEEE 128-bit floating point instructions. Otherwise, IEEE 128-bit
21409 floating point will be done with software emulation.
21411 @item -mfloat128-hardware
21412 @itemx -mno-float128-hardware
21413 @opindex mfloat128-hardware
21414 @opindex mno-float128-hardware
21415 Enable/disable using ISA 3.0 hardware instructions to support the
21416 @var{__float128} data type.
21418 If you use @option{-mfloat128-hardware}, it will enable the option
21419 @option{-mfloat128} as well.
21421 If you select ISA 3.0 instructions with @option{-mcpu=power9}, but do
21422 not use either @option{-mfloat128} or @option{-mfloat128-hardware},
21423 the IEEE 128-bit floating point support will not be enabled.
21425 @item -mfloat-gprs=@var{yes/single/double/no}
21426 @itemx -mfloat-gprs
21427 @opindex mfloat-gprs
21428 This switch enables or disables the generation of floating-point
21429 operations on the general-purpose registers for architectures that
21432 The argument @samp{yes} or @samp{single} enables the use of
21433 single-precision floating-point operations.
21435 The argument @samp{double} enables the use of single and
21436 double-precision floating-point operations.
21438 The argument @samp{no} disables floating-point operations on the
21439 general-purpose registers.
21441 This option is currently only available on the MPC854x.
21447 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
21448 targets (including GNU/Linux). The 32-bit environment sets int, long
21449 and pointer to 32 bits and generates code that runs on any PowerPC
21450 variant. The 64-bit environment sets int to 32 bits and long and
21451 pointer to 64 bits, and generates code for PowerPC64, as for
21452 @option{-mpowerpc64}.
21455 @itemx -mno-fp-in-toc
21456 @itemx -mno-sum-in-toc
21457 @itemx -mminimal-toc
21459 @opindex mno-fp-in-toc
21460 @opindex mno-sum-in-toc
21461 @opindex mminimal-toc
21462 Modify generation of the TOC (Table Of Contents), which is created for
21463 every executable file. The @option{-mfull-toc} option is selected by
21464 default. In that case, GCC allocates at least one TOC entry for
21465 each unique non-automatic variable reference in your program. GCC
21466 also places floating-point constants in the TOC@. However, only
21467 16,384 entries are available in the TOC@.
21469 If you receive a linker error message that saying you have overflowed
21470 the available TOC space, you can reduce the amount of TOC space used
21471 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
21472 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
21473 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
21474 generate code to calculate the sum of an address and a constant at
21475 run time instead of putting that sum into the TOC@. You may specify one
21476 or both of these options. Each causes GCC to produce very slightly
21477 slower and larger code at the expense of conserving TOC space.
21479 If you still run out of space in the TOC even when you specify both of
21480 these options, specify @option{-mminimal-toc} instead. This option causes
21481 GCC to make only one TOC entry for every file. When you specify this
21482 option, GCC produces code that is slower and larger but which
21483 uses extremely little TOC space. You may wish to use this option
21484 only on files that contain less frequently-executed code.
21490 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
21491 @code{long} type, and the infrastructure needed to support them.
21492 Specifying @option{-maix64} implies @option{-mpowerpc64},
21493 while @option{-maix32} disables the 64-bit ABI and
21494 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
21497 @itemx -mno-xl-compat
21498 @opindex mxl-compat
21499 @opindex mno-xl-compat
21500 Produce code that conforms more closely to IBM XL compiler semantics
21501 when using AIX-compatible ABI@. Pass floating-point arguments to
21502 prototyped functions beyond the register save area (RSA) on the stack
21503 in addition to argument FPRs. Do not assume that most significant
21504 double in 128-bit long double value is properly rounded when comparing
21505 values and converting to double. Use XL symbol names for long double
21508 The AIX calling convention was extended but not initially documented to
21509 handle an obscure K&R C case of calling a function that takes the
21510 address of its arguments with fewer arguments than declared. IBM XL
21511 compilers access floating-point arguments that do not fit in the
21512 RSA from the stack when a subroutine is compiled without
21513 optimization. Because always storing floating-point arguments on the
21514 stack is inefficient and rarely needed, this option is not enabled by
21515 default and only is necessary when calling subroutines compiled by IBM
21516 XL compilers without optimization.
21520 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
21521 application written to use message passing with special startup code to
21522 enable the application to run. The system must have PE installed in the
21523 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
21524 must be overridden with the @option{-specs=} option to specify the
21525 appropriate directory location. The Parallel Environment does not
21526 support threads, so the @option{-mpe} option and the @option{-pthread}
21527 option are incompatible.
21529 @item -malign-natural
21530 @itemx -malign-power
21531 @opindex malign-natural
21532 @opindex malign-power
21533 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
21534 @option{-malign-natural} overrides the ABI-defined alignment of larger
21535 types, such as floating-point doubles, on their natural size-based boundary.
21536 The option @option{-malign-power} instructs GCC to follow the ABI-specified
21537 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
21539 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
21543 @itemx -mhard-float
21544 @opindex msoft-float
21545 @opindex mhard-float
21546 Generate code that does not use (uses) the floating-point register set.
21547 Software floating-point emulation is provided if you use the
21548 @option{-msoft-float} option, and pass the option to GCC when linking.
21550 @item -msingle-float
21551 @itemx -mdouble-float
21552 @opindex msingle-float
21553 @opindex mdouble-float
21554 Generate code for single- or double-precision floating-point operations.
21555 @option{-mdouble-float} implies @option{-msingle-float}.
21558 @opindex msimple-fpu
21559 Do not generate @code{sqrt} and @code{div} instructions for hardware
21560 floating-point unit.
21562 @item -mfpu=@var{name}
21564 Specify type of floating-point unit. Valid values for @var{name} are
21565 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
21566 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
21567 @samp{sp_full} (equivalent to @option{-msingle-float}),
21568 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
21571 @opindex mxilinx-fpu
21572 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
21575 @itemx -mno-multiple
21577 @opindex mno-multiple
21578 Generate code that uses (does not use) the load multiple word
21579 instructions and the store multiple word instructions. These
21580 instructions are generated by default on POWER systems, and not
21581 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
21582 PowerPC systems, since those instructions do not work when the
21583 processor is in little-endian mode. The exceptions are PPC740 and
21584 PPC750 which permit these instructions in little-endian mode.
21589 @opindex mno-string
21590 Generate code that uses (does not use) the load string instructions
21591 and the store string word instructions to save multiple registers and
21592 do small block moves. These instructions are generated by default on
21593 POWER systems, and not generated on PowerPC systems. Do not use
21594 @option{-mstring} on little-endian PowerPC systems, since those
21595 instructions do not work when the processor is in little-endian mode.
21596 The exceptions are PPC740 and PPC750 which permit these instructions
21597 in little-endian mode.
21602 @opindex mno-update
21603 Generate code that uses (does not use) the load or store instructions
21604 that update the base register to the address of the calculated memory
21605 location. These instructions are generated by default. If you use
21606 @option{-mno-update}, there is a small window between the time that the
21607 stack pointer is updated and the address of the previous frame is
21608 stored, which means code that walks the stack frame across interrupts or
21609 signals may get corrupted data.
21611 @item -mavoid-indexed-addresses
21612 @itemx -mno-avoid-indexed-addresses
21613 @opindex mavoid-indexed-addresses
21614 @opindex mno-avoid-indexed-addresses
21615 Generate code that tries to avoid (not avoid) the use of indexed load
21616 or store instructions. These instructions can incur a performance
21617 penalty on Power6 processors in certain situations, such as when
21618 stepping through large arrays that cross a 16M boundary. This option
21619 is enabled by default when targeting Power6 and disabled otherwise.
21622 @itemx -mno-fused-madd
21623 @opindex mfused-madd
21624 @opindex mno-fused-madd
21625 Generate code that uses (does not use) the floating-point multiply and
21626 accumulate instructions. These instructions are generated by default
21627 if hardware floating point is used. The machine-dependent
21628 @option{-mfused-madd} option is now mapped to the machine-independent
21629 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21630 mapped to @option{-ffp-contract=off}.
21636 Generate code that uses (does not use) the half-word multiply and
21637 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
21638 These instructions are generated by default when targeting those
21645 Generate code that uses (does not use) the string-search @samp{dlmzb}
21646 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
21647 generated by default when targeting those processors.
21649 @item -mno-bit-align
21651 @opindex mno-bit-align
21652 @opindex mbit-align
21653 On System V.4 and embedded PowerPC systems do not (do) force structures
21654 and unions that contain bit-fields to be aligned to the base type of the
21657 For example, by default a structure containing nothing but 8
21658 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
21659 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
21660 the structure is aligned to a 1-byte boundary and is 1 byte in
21663 @item -mno-strict-align
21664 @itemx -mstrict-align
21665 @opindex mno-strict-align
21666 @opindex mstrict-align
21667 On System V.4 and embedded PowerPC systems do not (do) assume that
21668 unaligned memory references are handled by the system.
21670 @item -mrelocatable
21671 @itemx -mno-relocatable
21672 @opindex mrelocatable
21673 @opindex mno-relocatable
21674 Generate code that allows (does not allow) a static executable to be
21675 relocated to a different address at run time. A simple embedded
21676 PowerPC system loader should relocate the entire contents of
21677 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
21678 a table of 32-bit addresses generated by this option. For this to
21679 work, all objects linked together must be compiled with
21680 @option{-mrelocatable} or @option{-mrelocatable-lib}.
21681 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
21683 @item -mrelocatable-lib
21684 @itemx -mno-relocatable-lib
21685 @opindex mrelocatable-lib
21686 @opindex mno-relocatable-lib
21687 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
21688 @code{.fixup} section to allow static executables to be relocated at
21689 run time, but @option{-mrelocatable-lib} does not use the smaller stack
21690 alignment of @option{-mrelocatable}. Objects compiled with
21691 @option{-mrelocatable-lib} may be linked with objects compiled with
21692 any combination of the @option{-mrelocatable} options.
21698 On System V.4 and embedded PowerPC systems do not (do) assume that
21699 register 2 contains a pointer to a global area pointing to the addresses
21700 used in the program.
21703 @itemx -mlittle-endian
21705 @opindex mlittle-endian
21706 On System V.4 and embedded PowerPC systems compile code for the
21707 processor in little-endian mode. The @option{-mlittle-endian} option is
21708 the same as @option{-mlittle}.
21711 @itemx -mbig-endian
21713 @opindex mbig-endian
21714 On System V.4 and embedded PowerPC systems compile code for the
21715 processor in big-endian mode. The @option{-mbig-endian} option is
21716 the same as @option{-mbig}.
21718 @item -mdynamic-no-pic
21719 @opindex mdynamic-no-pic
21720 On Darwin and Mac OS X systems, compile code so that it is not
21721 relocatable, but that its external references are relocatable. The
21722 resulting code is suitable for applications, but not shared
21725 @item -msingle-pic-base
21726 @opindex msingle-pic-base
21727 Treat the register used for PIC addressing as read-only, rather than
21728 loading it in the prologue for each function. The runtime system is
21729 responsible for initializing this register with an appropriate value
21730 before execution begins.
21732 @item -mprioritize-restricted-insns=@var{priority}
21733 @opindex mprioritize-restricted-insns
21734 This option controls the priority that is assigned to
21735 dispatch-slot restricted instructions during the second scheduling
21736 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
21737 or @samp{2} to assign no, highest, or second-highest (respectively)
21738 priority to dispatch-slot restricted
21741 @item -msched-costly-dep=@var{dependence_type}
21742 @opindex msched-costly-dep
21743 This option controls which dependences are considered costly
21744 by the target during instruction scheduling. The argument
21745 @var{dependence_type} takes one of the following values:
21749 No dependence is costly.
21752 All dependences are costly.
21754 @item @samp{true_store_to_load}
21755 A true dependence from store to load is costly.
21757 @item @samp{store_to_load}
21758 Any dependence from store to load is costly.
21761 Any dependence for which the latency is greater than or equal to
21762 @var{number} is costly.
21765 @item -minsert-sched-nops=@var{scheme}
21766 @opindex minsert-sched-nops
21767 This option controls which NOP insertion scheme is used during
21768 the second scheduling pass. The argument @var{scheme} takes one of the
21776 Pad with NOPs any dispatch group that has vacant issue slots,
21777 according to the scheduler's grouping.
21779 @item @samp{regroup_exact}
21780 Insert NOPs to force costly dependent insns into
21781 separate groups. Insert exactly as many NOPs as needed to force an insn
21782 to a new group, according to the estimated processor grouping.
21785 Insert NOPs to force costly dependent insns into
21786 separate groups. Insert @var{number} NOPs to force an insn to a new group.
21790 @opindex mcall-sysv
21791 On System V.4 and embedded PowerPC systems compile code using calling
21792 conventions that adhere to the March 1995 draft of the System V
21793 Application Binary Interface, PowerPC processor supplement. This is the
21794 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
21796 @item -mcall-sysv-eabi
21798 @opindex mcall-sysv-eabi
21799 @opindex mcall-eabi
21800 Specify both @option{-mcall-sysv} and @option{-meabi} options.
21802 @item -mcall-sysv-noeabi
21803 @opindex mcall-sysv-noeabi
21804 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
21806 @item -mcall-aixdesc
21808 On System V.4 and embedded PowerPC systems compile code for the AIX
21812 @opindex mcall-linux
21813 On System V.4 and embedded PowerPC systems compile code for the
21814 Linux-based GNU system.
21816 @item -mcall-freebsd
21817 @opindex mcall-freebsd
21818 On System V.4 and embedded PowerPC systems compile code for the
21819 FreeBSD operating system.
21821 @item -mcall-netbsd
21822 @opindex mcall-netbsd
21823 On System V.4 and embedded PowerPC systems compile code for the
21824 NetBSD operating system.
21826 @item -mcall-openbsd
21827 @opindex mcall-netbsd
21828 On System V.4 and embedded PowerPC systems compile code for the
21829 OpenBSD operating system.
21831 @item -maix-struct-return
21832 @opindex maix-struct-return
21833 Return all structures in memory (as specified by the AIX ABI)@.
21835 @item -msvr4-struct-return
21836 @opindex msvr4-struct-return
21837 Return structures smaller than 8 bytes in registers (as specified by the
21840 @item -mabi=@var{abi-type}
21842 Extend the current ABI with a particular extension, or remove such extension.
21843 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
21844 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
21845 @samp{elfv1}, @samp{elfv2}@.
21849 Extend the current ABI with SPE ABI extensions. This does not change
21850 the default ABI, instead it adds the SPE ABI extensions to the current
21854 @opindex mabi=no-spe
21855 Disable Book-E SPE ABI extensions for the current ABI@.
21857 @item -mabi=ibmlongdouble
21858 @opindex mabi=ibmlongdouble
21859 Change the current ABI to use IBM extended-precision long double.
21860 This is a PowerPC 32-bit SYSV ABI option.
21862 @item -mabi=ieeelongdouble
21863 @opindex mabi=ieeelongdouble
21864 Change the current ABI to use IEEE extended-precision long double.
21865 This is a PowerPC 32-bit Linux ABI option.
21868 @opindex mabi=elfv1
21869 Change the current ABI to use the ELFv1 ABI.
21870 This is the default ABI for big-endian PowerPC 64-bit Linux.
21871 Overriding the default ABI requires special system support and is
21872 likely to fail in spectacular ways.
21875 @opindex mabi=elfv2
21876 Change the current ABI to use the ELFv2 ABI.
21877 This is the default ABI for little-endian PowerPC 64-bit Linux.
21878 Overriding the default ABI requires special system support and is
21879 likely to fail in spectacular ways.
21881 @item -mgnu-attribute
21882 @itemx -mno-gnu-attribute
21883 @opindex mgnu-attribute
21884 @opindex mno-gnu-attribute
21885 Emit .gnu_attribute assembly directives to set tag/value pairs in a
21886 .gnu.attributes section that specify ABI variations in function
21887 parameters or return values.
21890 @itemx -mno-prototype
21891 @opindex mprototype
21892 @opindex mno-prototype
21893 On System V.4 and embedded PowerPC systems assume that all calls to
21894 variable argument functions are properly prototyped. Otherwise, the
21895 compiler must insert an instruction before every non-prototyped call to
21896 set or clear bit 6 of the condition code register (@code{CR}) to
21897 indicate whether floating-point values are passed in the floating-point
21898 registers in case the function takes variable arguments. With
21899 @option{-mprototype}, only calls to prototyped variable argument functions
21900 set or clear the bit.
21904 On embedded PowerPC systems, assume that the startup module is called
21905 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
21906 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
21911 On embedded PowerPC systems, assume that the startup module is called
21912 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
21917 On embedded PowerPC systems, assume that the startup module is called
21918 @file{crt0.o} and the standard C libraries are @file{libads.a} and
21921 @item -myellowknife
21922 @opindex myellowknife
21923 On embedded PowerPC systems, assume that the startup module is called
21924 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
21929 On System V.4 and embedded PowerPC systems, specify that you are
21930 compiling for a VxWorks system.
21934 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
21935 header to indicate that @samp{eabi} extended relocations are used.
21941 On System V.4 and embedded PowerPC systems do (do not) adhere to the
21942 Embedded Applications Binary Interface (EABI), which is a set of
21943 modifications to the System V.4 specifications. Selecting @option{-meabi}
21944 means that the stack is aligned to an 8-byte boundary, a function
21945 @code{__eabi} is called from @code{main} to set up the EABI
21946 environment, and the @option{-msdata} option can use both @code{r2} and
21947 @code{r13} to point to two separate small data areas. Selecting
21948 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
21949 no EABI initialization function is called from @code{main}, and the
21950 @option{-msdata} option only uses @code{r13} to point to a single
21951 small data area. The @option{-meabi} option is on by default if you
21952 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
21955 @opindex msdata=eabi
21956 On System V.4 and embedded PowerPC systems, put small initialized
21957 @code{const} global and static data in the @code{.sdata2} section, which
21958 is pointed to by register @code{r2}. Put small initialized
21959 non-@code{const} global and static data in the @code{.sdata} section,
21960 which is pointed to by register @code{r13}. Put small uninitialized
21961 global and static data in the @code{.sbss} section, which is adjacent to
21962 the @code{.sdata} section. The @option{-msdata=eabi} option is
21963 incompatible with the @option{-mrelocatable} option. The
21964 @option{-msdata=eabi} option also sets the @option{-memb} option.
21967 @opindex msdata=sysv
21968 On System V.4 and embedded PowerPC systems, put small global and static
21969 data in the @code{.sdata} section, which is pointed to by register
21970 @code{r13}. Put small uninitialized global and static data in the
21971 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
21972 The @option{-msdata=sysv} option is incompatible with the
21973 @option{-mrelocatable} option.
21975 @item -msdata=default
21977 @opindex msdata=default
21979 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
21980 compile code the same as @option{-msdata=eabi}, otherwise compile code the
21981 same as @option{-msdata=sysv}.
21984 @opindex msdata=data
21985 On System V.4 and embedded PowerPC systems, put small global
21986 data in the @code{.sdata} section. Put small uninitialized global
21987 data in the @code{.sbss} section. Do not use register @code{r13}
21988 to address small data however. This is the default behavior unless
21989 other @option{-msdata} options are used.
21993 @opindex msdata=none
21995 On embedded PowerPC systems, put all initialized global and static data
21996 in the @code{.data} section, and all uninitialized data in the
21997 @code{.bss} section.
21999 @item -mblock-move-inline-limit=@var{num}
22000 @opindex mblock-move-inline-limit
22001 Inline all block moves (such as calls to @code{memcpy} or structure
22002 copies) less than or equal to @var{num} bytes. The minimum value for
22003 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
22004 targets. The default value is target-specific.
22008 @cindex smaller data references (PowerPC)
22009 @cindex .sdata/.sdata2 references (PowerPC)
22010 On embedded PowerPC systems, put global and static items less than or
22011 equal to @var{num} bytes into the small data or BSS sections instead of
22012 the normal data or BSS section. By default, @var{num} is 8. The
22013 @option{-G @var{num}} switch is also passed to the linker.
22014 All modules should be compiled with the same @option{-G @var{num}} value.
22017 @itemx -mno-regnames
22019 @opindex mno-regnames
22020 On System V.4 and embedded PowerPC systems do (do not) emit register
22021 names in the assembly language output using symbolic forms.
22024 @itemx -mno-longcall
22026 @opindex mno-longcall
22027 By default assume that all calls are far away so that a longer and more
22028 expensive calling sequence is required. This is required for calls
22029 farther than 32 megabytes (33,554,432 bytes) from the current location.
22030 A short call is generated if the compiler knows
22031 the call cannot be that far away. This setting can be overridden by
22032 the @code{shortcall} function attribute, or by @code{#pragma
22035 Some linkers are capable of detecting out-of-range calls and generating
22036 glue code on the fly. On these systems, long calls are unnecessary and
22037 generate slower code. As of this writing, the AIX linker can do this,
22038 as can the GNU linker for PowerPC/64. It is planned to add this feature
22039 to the GNU linker for 32-bit PowerPC systems as well.
22041 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
22042 callee, L42}, plus a @dfn{branch island} (glue code). The two target
22043 addresses represent the callee and the branch island. The
22044 Darwin/PPC linker prefers the first address and generates a @code{bl
22045 callee} if the PPC @code{bl} instruction reaches the callee directly;
22046 otherwise, the linker generates @code{bl L42} to call the branch
22047 island. The branch island is appended to the body of the
22048 calling function; it computes the full 32-bit address of the callee
22051 On Mach-O (Darwin) systems, this option directs the compiler emit to
22052 the glue for every direct call, and the Darwin linker decides whether
22053 to use or discard it.
22055 In the future, GCC may ignore all longcall specifications
22056 when the linker is known to generate glue.
22058 @item -mtls-markers
22059 @itemx -mno-tls-markers
22060 @opindex mtls-markers
22061 @opindex mno-tls-markers
22062 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
22063 specifying the function argument. The relocation allows the linker to
22064 reliably associate function call with argument setup instructions for
22065 TLS optimization, which in turn allows GCC to better schedule the
22071 This option enables use of the reciprocal estimate and
22072 reciprocal square root estimate instructions with additional
22073 Newton-Raphson steps to increase precision instead of doing a divide or
22074 square root and divide for floating-point arguments. You should use
22075 the @option{-ffast-math} option when using @option{-mrecip} (or at
22076 least @option{-funsafe-math-optimizations},
22077 @option{-ffinite-math-only}, @option{-freciprocal-math} and
22078 @option{-fno-trapping-math}). Note that while the throughput of the
22079 sequence is generally higher than the throughput of the non-reciprocal
22080 instruction, the precision of the sequence can be decreased by up to 2
22081 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
22084 @item -mrecip=@var{opt}
22085 @opindex mrecip=opt
22086 This option controls which reciprocal estimate instructions
22087 may be used. @var{opt} is a comma-separated list of options, which may
22088 be preceded by a @code{!} to invert the option:
22093 Enable all estimate instructions.
22096 Enable the default instructions, equivalent to @option{-mrecip}.
22099 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22102 Enable the reciprocal approximation instructions for both
22103 single and double precision.
22106 Enable the single-precision reciprocal approximation instructions.
22109 Enable the double-precision reciprocal approximation instructions.
22112 Enable the reciprocal square root approximation instructions for both
22113 single and double precision.
22116 Enable the single-precision reciprocal square root approximation instructions.
22119 Enable the double-precision reciprocal square root approximation instructions.
22123 So, for example, @option{-mrecip=all,!rsqrtd} enables
22124 all of the reciprocal estimate instructions, except for the
22125 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
22126 which handle the double-precision reciprocal square root calculations.
22128 @item -mrecip-precision
22129 @itemx -mno-recip-precision
22130 @opindex mrecip-precision
22131 Assume (do not assume) that the reciprocal estimate instructions
22132 provide higher-precision estimates than is mandated by the PowerPC
22133 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
22134 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
22135 The double-precision square root estimate instructions are not generated by
22136 default on low-precision machines, since they do not provide an
22137 estimate that converges after three steps.
22139 @item -mveclibabi=@var{type}
22140 @opindex mveclibabi
22141 Specifies the ABI type to use for vectorizing intrinsics using an
22142 external library. The only type supported at present is @samp{mass},
22143 which specifies to use IBM's Mathematical Acceleration Subsystem
22144 (MASS) libraries for vectorizing intrinsics using external libraries.
22145 GCC currently emits calls to @code{acosd2}, @code{acosf4},
22146 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
22147 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
22148 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
22149 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
22150 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
22151 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
22152 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
22153 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
22154 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
22155 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
22156 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
22157 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
22158 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
22159 for power7. Both @option{-ftree-vectorize} and
22160 @option{-funsafe-math-optimizations} must also be enabled. The MASS
22161 libraries must be specified at link time.
22166 Generate (do not generate) the @code{friz} instruction when the
22167 @option{-funsafe-math-optimizations} option is used to optimize
22168 rounding of floating-point values to 64-bit integer and back to floating
22169 point. The @code{friz} instruction does not return the same value if
22170 the floating-point number is too large to fit in an integer.
22172 @item -mpointers-to-nested-functions
22173 @itemx -mno-pointers-to-nested-functions
22174 @opindex mpointers-to-nested-functions
22175 Generate (do not generate) code to load up the static chain register
22176 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
22177 systems where a function pointer points to a 3-word descriptor giving
22178 the function address, TOC value to be loaded in register @code{r2}, and
22179 static chain value to be loaded in register @code{r11}. The
22180 @option{-mpointers-to-nested-functions} is on by default. You cannot
22181 call through pointers to nested functions or pointers
22182 to functions compiled in other languages that use the static chain if
22183 you use @option{-mno-pointers-to-nested-functions}.
22185 @item -msave-toc-indirect
22186 @itemx -mno-save-toc-indirect
22187 @opindex msave-toc-indirect
22188 Generate (do not generate) code to save the TOC value in the reserved
22189 stack location in the function prologue if the function calls through
22190 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
22191 saved in the prologue, it is saved just before the call through the
22192 pointer. The @option{-mno-save-toc-indirect} option is the default.
22194 @item -mcompat-align-parm
22195 @itemx -mno-compat-align-parm
22196 @opindex mcompat-align-parm
22197 Generate (do not generate) code to pass structure parameters with a
22198 maximum alignment of 64 bits, for compatibility with older versions
22201 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
22202 structure parameter on a 128-bit boundary when that structure contained
22203 a member requiring 128-bit alignment. This is corrected in more
22204 recent versions of GCC. This option may be used to generate code
22205 that is compatible with functions compiled with older versions of
22208 The @option{-mno-compat-align-parm} option is the default.
22210 @item -mstack-protector-guard=@var{guard}
22211 @itemx -mstack-protector-guard-reg=@var{reg}
22212 @itemx -mstack-protector-guard-offset=@var{offset}
22213 @opindex mstack-protector-guard
22214 @opindex mstack-protector-guard-reg
22215 @opindex mstack-protector-guard-offset
22216 Generate stack protection code using canary at @var{guard}. Supported
22217 locations are @samp{global} for global canary or @samp{tls} for per-thread
22218 canary in the TLS block (the default with GNU libc version 2.4 or later).
22220 With the latter choice the options
22221 @option{-mstack-protector-guard-reg=@var{reg}} and
22222 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
22223 which register to use as base register for reading the canary, and from what
22224 offset from that base register. The default for those is as specified in the
22229 @subsection RX Options
22232 These command-line options are defined for RX targets:
22235 @item -m64bit-doubles
22236 @itemx -m32bit-doubles
22237 @opindex m64bit-doubles
22238 @opindex m32bit-doubles
22239 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
22240 or 32 bits (@option{-m32bit-doubles}) in size. The default is
22241 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
22242 works on 32-bit values, which is why the default is
22243 @option{-m32bit-doubles}.
22249 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
22250 floating-point hardware. The default is enabled for the RX600
22251 series and disabled for the RX200 series.
22253 Floating-point instructions are only generated for 32-bit floating-point
22254 values, however, so the FPU hardware is not used for doubles if the
22255 @option{-m64bit-doubles} option is used.
22257 @emph{Note} If the @option{-fpu} option is enabled then
22258 @option{-funsafe-math-optimizations} is also enabled automatically.
22259 This is because the RX FPU instructions are themselves unsafe.
22261 @item -mcpu=@var{name}
22263 Selects the type of RX CPU to be targeted. Currently three types are
22264 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
22265 the specific @samp{RX610} CPU. The default is @samp{RX600}.
22267 The only difference between @samp{RX600} and @samp{RX610} is that the
22268 @samp{RX610} does not support the @code{MVTIPL} instruction.
22270 The @samp{RX200} series does not have a hardware floating-point unit
22271 and so @option{-nofpu} is enabled by default when this type is
22274 @item -mbig-endian-data
22275 @itemx -mlittle-endian-data
22276 @opindex mbig-endian-data
22277 @opindex mlittle-endian-data
22278 Store data (but not code) in the big-endian format. The default is
22279 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
22282 @item -msmall-data-limit=@var{N}
22283 @opindex msmall-data-limit
22284 Specifies the maximum size in bytes of global and static variables
22285 which can be placed into the small data area. Using the small data
22286 area can lead to smaller and faster code, but the size of area is
22287 limited and it is up to the programmer to ensure that the area does
22288 not overflow. Also when the small data area is used one of the RX's
22289 registers (usually @code{r13}) is reserved for use pointing to this
22290 area, so it is no longer available for use by the compiler. This
22291 could result in slower and/or larger code if variables are pushed onto
22292 the stack instead of being held in this register.
22294 Note, common variables (variables that have not been initialized) and
22295 constants are not placed into the small data area as they are assigned
22296 to other sections in the output executable.
22298 The default value is zero, which disables this feature. Note, this
22299 feature is not enabled by default with higher optimization levels
22300 (@option{-O2} etc) because of the potentially detrimental effects of
22301 reserving a register. It is up to the programmer to experiment and
22302 discover whether this feature is of benefit to their program. See the
22303 description of the @option{-mpid} option for a description of how the
22304 actual register to hold the small data area pointer is chosen.
22310 Use the simulator runtime. The default is to use the libgloss
22311 board-specific runtime.
22313 @item -mas100-syntax
22314 @itemx -mno-as100-syntax
22315 @opindex mas100-syntax
22316 @opindex mno-as100-syntax
22317 When generating assembler output use a syntax that is compatible with
22318 Renesas's AS100 assembler. This syntax can also be handled by the GAS
22319 assembler, but it has some restrictions so it is not generated by default.
22321 @item -mmax-constant-size=@var{N}
22322 @opindex mmax-constant-size
22323 Specifies the maximum size, in bytes, of a constant that can be used as
22324 an operand in a RX instruction. Although the RX instruction set does
22325 allow constants of up to 4 bytes in length to be used in instructions,
22326 a longer value equates to a longer instruction. Thus in some
22327 circumstances it can be beneficial to restrict the size of constants
22328 that are used in instructions. Constants that are too big are instead
22329 placed into a constant pool and referenced via register indirection.
22331 The value @var{N} can be between 0 and 4. A value of 0 (the default)
22332 or 4 means that constants of any size are allowed.
22336 Enable linker relaxation. Linker relaxation is a process whereby the
22337 linker attempts to reduce the size of a program by finding shorter
22338 versions of various instructions. Disabled by default.
22340 @item -mint-register=@var{N}
22341 @opindex mint-register
22342 Specify the number of registers to reserve for fast interrupt handler
22343 functions. The value @var{N} can be between 0 and 4. A value of 1
22344 means that register @code{r13} is reserved for the exclusive use
22345 of fast interrupt handlers. A value of 2 reserves @code{r13} and
22346 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
22347 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
22348 A value of 0, the default, does not reserve any registers.
22350 @item -msave-acc-in-interrupts
22351 @opindex msave-acc-in-interrupts
22352 Specifies that interrupt handler functions should preserve the
22353 accumulator register. This is only necessary if normal code might use
22354 the accumulator register, for example because it performs 64-bit
22355 multiplications. The default is to ignore the accumulator as this
22356 makes the interrupt handlers faster.
22362 Enables the generation of position independent data. When enabled any
22363 access to constant data is done via an offset from a base address
22364 held in a register. This allows the location of constant data to be
22365 determined at run time without requiring the executable to be
22366 relocated, which is a benefit to embedded applications with tight
22367 memory constraints. Data that can be modified is not affected by this
22370 Note, using this feature reserves a register, usually @code{r13}, for
22371 the constant data base address. This can result in slower and/or
22372 larger code, especially in complicated functions.
22374 The actual register chosen to hold the constant data base address
22375 depends upon whether the @option{-msmall-data-limit} and/or the
22376 @option{-mint-register} command-line options are enabled. Starting
22377 with register @code{r13} and proceeding downwards, registers are
22378 allocated first to satisfy the requirements of @option{-mint-register},
22379 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
22380 is possible for the small data area register to be @code{r8} if both
22381 @option{-mint-register=4} and @option{-mpid} are specified on the
22384 By default this feature is not enabled. The default can be restored
22385 via the @option{-mno-pid} command-line option.
22387 @item -mno-warn-multiple-fast-interrupts
22388 @itemx -mwarn-multiple-fast-interrupts
22389 @opindex mno-warn-multiple-fast-interrupts
22390 @opindex mwarn-multiple-fast-interrupts
22391 Prevents GCC from issuing a warning message if it finds more than one
22392 fast interrupt handler when it is compiling a file. The default is to
22393 issue a warning for each extra fast interrupt handler found, as the RX
22394 only supports one such interrupt.
22396 @item -mallow-string-insns
22397 @itemx -mno-allow-string-insns
22398 @opindex mallow-string-insns
22399 @opindex mno-allow-string-insns
22400 Enables or disables the use of the string manipulation instructions
22401 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
22402 @code{SWHILE} and also the @code{RMPA} instruction. These
22403 instructions may prefetch data, which is not safe to do if accessing
22404 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
22405 for more information).
22407 The default is to allow these instructions, but it is not possible for
22408 GCC to reliably detect all circumstances where a string instruction
22409 might be used to access an I/O register, so their use cannot be
22410 disabled automatically. Instead it is reliant upon the programmer to
22411 use the @option{-mno-allow-string-insns} option if their program
22412 accesses I/O space.
22414 When the instructions are enabled GCC defines the C preprocessor
22415 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
22416 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
22422 Use only (or not only) @code{JSR} instructions to access functions.
22423 This option can be used when code size exceeds the range of @code{BSR}
22424 instructions. Note that @option{-mno-jsr} does not mean to not use
22425 @code{JSR} but instead means that any type of branch may be used.
22428 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
22429 has special significance to the RX port when used with the
22430 @code{interrupt} function attribute. This attribute indicates a
22431 function intended to process fast interrupts. GCC ensures
22432 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
22433 and/or @code{r13} and only provided that the normal use of the
22434 corresponding registers have been restricted via the
22435 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
22438 @node S/390 and zSeries Options
22439 @subsection S/390 and zSeries Options
22440 @cindex S/390 and zSeries Options
22442 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
22446 @itemx -msoft-float
22447 @opindex mhard-float
22448 @opindex msoft-float
22449 Use (do not use) the hardware floating-point instructions and registers
22450 for floating-point operations. When @option{-msoft-float} is specified,
22451 functions in @file{libgcc.a} are used to perform floating-point
22452 operations. When @option{-mhard-float} is specified, the compiler
22453 generates IEEE floating-point instructions. This is the default.
22456 @itemx -mno-hard-dfp
22458 @opindex mno-hard-dfp
22459 Use (do not use) the hardware decimal-floating-point instructions for
22460 decimal-floating-point operations. When @option{-mno-hard-dfp} is
22461 specified, functions in @file{libgcc.a} are used to perform
22462 decimal-floating-point operations. When @option{-mhard-dfp} is
22463 specified, the compiler generates decimal-floating-point hardware
22464 instructions. This is the default for @option{-march=z9-ec} or higher.
22466 @item -mlong-double-64
22467 @itemx -mlong-double-128
22468 @opindex mlong-double-64
22469 @opindex mlong-double-128
22470 These switches control the size of @code{long double} type. A size
22471 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22472 type. This is the default.
22475 @itemx -mno-backchain
22476 @opindex mbackchain
22477 @opindex mno-backchain
22478 Store (do not store) the address of the caller's frame as backchain pointer
22479 into the callee's stack frame.
22480 A backchain may be needed to allow debugging using tools that do not understand
22481 DWARF call frame information.
22482 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
22483 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
22484 the backchain is placed into the topmost word of the 96/160 byte register
22487 In general, code compiled with @option{-mbackchain} is call-compatible with
22488 code compiled with @option{-mmo-backchain}; however, use of the backchain
22489 for debugging purposes usually requires that the whole binary is built with
22490 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
22491 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22492 to build a linux kernel use @option{-msoft-float}.
22494 The default is to not maintain the backchain.
22496 @item -mpacked-stack
22497 @itemx -mno-packed-stack
22498 @opindex mpacked-stack
22499 @opindex mno-packed-stack
22500 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
22501 specified, the compiler uses the all fields of the 96/160 byte register save
22502 area only for their default purpose; unused fields still take up stack space.
22503 When @option{-mpacked-stack} is specified, register save slots are densely
22504 packed at the top of the register save area; unused space is reused for other
22505 purposes, allowing for more efficient use of the available stack space.
22506 However, when @option{-mbackchain} is also in effect, the topmost word of
22507 the save area is always used to store the backchain, and the return address
22508 register is always saved two words below the backchain.
22510 As long as the stack frame backchain is not used, code generated with
22511 @option{-mpacked-stack} is call-compatible with code generated with
22512 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
22513 S/390 or zSeries generated code that uses the stack frame backchain at run
22514 time, not just for debugging purposes. Such code is not call-compatible
22515 with code compiled with @option{-mpacked-stack}. Also, note that the
22516 combination of @option{-mbackchain},
22517 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
22518 to build a linux kernel use @option{-msoft-float}.
22520 The default is to not use the packed stack layout.
22523 @itemx -mno-small-exec
22524 @opindex msmall-exec
22525 @opindex mno-small-exec
22526 Generate (or do not generate) code using the @code{bras} instruction
22527 to do subroutine calls.
22528 This only works reliably if the total executable size does not
22529 exceed 64k. The default is to use the @code{basr} instruction instead,
22530 which does not have this limitation.
22536 When @option{-m31} is specified, generate code compliant to the
22537 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
22538 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
22539 particular to generate 64-bit instructions. For the @samp{s390}
22540 targets, the default is @option{-m31}, while the @samp{s390x}
22541 targets default to @option{-m64}.
22547 When @option{-mzarch} is specified, generate code using the
22548 instructions available on z/Architecture.
22549 When @option{-mesa} is specified, generate code using the
22550 instructions available on ESA/390. Note that @option{-mesa} is
22551 not possible with @option{-m64}.
22552 When generating code compliant to the GNU/Linux for S/390 ABI,
22553 the default is @option{-mesa}. When generating code compliant
22554 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
22560 The @option{-mhtm} option enables a set of builtins making use of
22561 instructions available with the transactional execution facility
22562 introduced with the IBM zEnterprise EC12 machine generation
22563 @ref{S/390 System z Built-in Functions}.
22564 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
22570 When @option{-mvx} is specified, generate code using the instructions
22571 available with the vector extension facility introduced with the IBM
22572 z13 machine generation.
22573 This option changes the ABI for some vector type values with regard to
22574 alignment and calling conventions. In case vector type values are
22575 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
22576 command will be added to mark the resulting binary with the ABI used.
22577 @option{-mvx} is enabled by default when using @option{-march=z13}.
22580 @itemx -mno-zvector
22582 @opindex mno-zvector
22583 The @option{-mzvector} option enables vector language extensions and
22584 builtins using instructions available with the vector extension
22585 facility introduced with the IBM z13 machine generation.
22586 This option adds support for @samp{vector} to be used as a keyword to
22587 define vector type variables and arguments. @samp{vector} is only
22588 available when GNU extensions are enabled. It will not be expanded
22589 when requesting strict standard compliance e.g. with @option{-std=c99}.
22590 In addition to the GCC low-level builtins @option{-mzvector} enables
22591 a set of builtins added for compatibility with AltiVec-style
22592 implementations like Power and Cell. In order to make use of these
22593 builtins the header file @file{vecintrin.h} needs to be included.
22594 @option{-mzvector} is disabled by default.
22600 Generate (or do not generate) code using the @code{mvcle} instruction
22601 to perform block moves. When @option{-mno-mvcle} is specified,
22602 use a @code{mvc} loop instead. This is the default unless optimizing for
22609 Print (or do not print) additional debug information when compiling.
22610 The default is to not print debug information.
22612 @item -march=@var{cpu-type}
22614 Generate code that runs on @var{cpu-type}, which is the name of a
22615 system representing a certain processor type. Possible values for
22616 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
22617 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
22618 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
22621 The default is @option{-march=z900}. @samp{g5}/@samp{arch3} and
22622 @samp{g6} are deprecated and will be removed with future releases.
22624 Specifying @samp{native} as cpu type can be used to select the best
22625 architecture option for the host processor.
22626 @option{-march=native} has no effect if GCC does not recognize the
22629 @item -mtune=@var{cpu-type}
22631 Tune to @var{cpu-type} everything applicable about the generated code,
22632 except for the ABI and the set of available instructions.
22633 The list of @var{cpu-type} values is the same as for @option{-march}.
22634 The default is the value used for @option{-march}.
22637 @itemx -mno-tpf-trace
22638 @opindex mtpf-trace
22639 @opindex mno-tpf-trace
22640 Generate code that adds (does not add) in TPF OS specific branches to trace
22641 routines in the operating system. This option is off by default, even
22642 when compiling for the TPF OS@.
22645 @itemx -mno-fused-madd
22646 @opindex mfused-madd
22647 @opindex mno-fused-madd
22648 Generate code that uses (does not use) the floating-point multiply and
22649 accumulate instructions. These instructions are generated by default if
22650 hardware floating point is used.
22652 @item -mwarn-framesize=@var{framesize}
22653 @opindex mwarn-framesize
22654 Emit a warning if the current function exceeds the given frame size. Because
22655 this is a compile-time check it doesn't need to be a real problem when the program
22656 runs. It is intended to identify functions that most probably cause
22657 a stack overflow. It is useful to be used in an environment with limited stack
22658 size e.g.@: the linux kernel.
22660 @item -mwarn-dynamicstack
22661 @opindex mwarn-dynamicstack
22662 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
22663 arrays. This is generally a bad idea with a limited stack size.
22665 @item -mstack-guard=@var{stack-guard}
22666 @itemx -mstack-size=@var{stack-size}
22667 @opindex mstack-guard
22668 @opindex mstack-size
22669 If these options are provided the S/390 back end emits additional instructions in
22670 the function prologue that trigger a trap if the stack size is @var{stack-guard}
22671 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
22672 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
22673 the frame size of the compiled function is chosen.
22674 These options are intended to be used to help debugging stack overflow problems.
22675 The additionally emitted code causes only little overhead and hence can also be
22676 used in production-like systems without greater performance degradation. The given
22677 values have to be exact powers of 2 and @var{stack-size} has to be greater than
22678 @var{stack-guard} without exceeding 64k.
22679 In order to be efficient the extra code makes the assumption that the stack starts
22680 at an address aligned to the value given by @var{stack-size}.
22681 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
22683 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
22685 If the hotpatch option is enabled, a ``hot-patching'' function
22686 prologue is generated for all functions in the compilation unit.
22687 The funtion label is prepended with the given number of two-byte
22688 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
22689 the label, 2 * @var{post-halfwords} bytes are appended, using the
22690 largest NOP like instructions the architecture allows (maximum
22693 If both arguments are zero, hotpatching is disabled.
22695 This option can be overridden for individual functions with the
22696 @code{hotpatch} attribute.
22699 @node Score Options
22700 @subsection Score Options
22701 @cindex Score Options
22703 These options are defined for Score implementations:
22708 Compile code for big-endian mode. This is the default.
22712 Compile code for little-endian mode.
22716 Disable generation of @code{bcnz} instructions.
22720 Enable generation of unaligned load and store instructions.
22724 Enable the use of multiply-accumulate instructions. Disabled by default.
22728 Specify the SCORE5 as the target architecture.
22732 Specify the SCORE5U of the target architecture.
22736 Specify the SCORE7 as the target architecture. This is the default.
22740 Specify the SCORE7D as the target architecture.
22744 @subsection SH Options
22746 These @samp{-m} options are defined for the SH implementations:
22751 Generate code for the SH1.
22755 Generate code for the SH2.
22758 Generate code for the SH2e.
22762 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
22763 that the floating-point unit is not used.
22765 @item -m2a-single-only
22766 @opindex m2a-single-only
22767 Generate code for the SH2a-FPU, in such a way that no double-precision
22768 floating-point operations are used.
22771 @opindex m2a-single
22772 Generate code for the SH2a-FPU assuming the floating-point unit is in
22773 single-precision mode by default.
22777 Generate code for the SH2a-FPU assuming the floating-point unit is in
22778 double-precision mode by default.
22782 Generate code for the SH3.
22786 Generate code for the SH3e.
22790 Generate code for the SH4 without a floating-point unit.
22792 @item -m4-single-only
22793 @opindex m4-single-only
22794 Generate code for the SH4 with a floating-point unit that only
22795 supports single-precision arithmetic.
22799 Generate code for the SH4 assuming the floating-point unit is in
22800 single-precision mode by default.
22804 Generate code for the SH4.
22808 Generate code for SH4-100.
22810 @item -m4-100-nofpu
22811 @opindex m4-100-nofpu
22812 Generate code for SH4-100 in such a way that the
22813 floating-point unit is not used.
22815 @item -m4-100-single
22816 @opindex m4-100-single
22817 Generate code for SH4-100 assuming the floating-point unit is in
22818 single-precision mode by default.
22820 @item -m4-100-single-only
22821 @opindex m4-100-single-only
22822 Generate code for SH4-100 in such a way that no double-precision
22823 floating-point operations are used.
22827 Generate code for SH4-200.
22829 @item -m4-200-nofpu
22830 @opindex m4-200-nofpu
22831 Generate code for SH4-200 without in such a way that the
22832 floating-point unit is not used.
22834 @item -m4-200-single
22835 @opindex m4-200-single
22836 Generate code for SH4-200 assuming the floating-point unit is in
22837 single-precision mode by default.
22839 @item -m4-200-single-only
22840 @opindex m4-200-single-only
22841 Generate code for SH4-200 in such a way that no double-precision
22842 floating-point operations are used.
22846 Generate code for SH4-300.
22848 @item -m4-300-nofpu
22849 @opindex m4-300-nofpu
22850 Generate code for SH4-300 without in such a way that the
22851 floating-point unit is not used.
22853 @item -m4-300-single
22854 @opindex m4-300-single
22855 Generate code for SH4-300 in such a way that no double-precision
22856 floating-point operations are used.
22858 @item -m4-300-single-only
22859 @opindex m4-300-single-only
22860 Generate code for SH4-300 in such a way that no double-precision
22861 floating-point operations are used.
22865 Generate code for SH4-340 (no MMU, no FPU).
22869 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
22874 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
22875 floating-point unit is not used.
22877 @item -m4a-single-only
22878 @opindex m4a-single-only
22879 Generate code for the SH4a, in such a way that no double-precision
22880 floating-point operations are used.
22883 @opindex m4a-single
22884 Generate code for the SH4a assuming the floating-point unit is in
22885 single-precision mode by default.
22889 Generate code for the SH4a.
22893 Same as @option{-m4a-nofpu}, except that it implicitly passes
22894 @option{-dsp} to the assembler. GCC doesn't generate any DSP
22895 instructions at the moment.
22899 Compile code for the processor in big-endian mode.
22903 Compile code for the processor in little-endian mode.
22907 Align doubles at 64-bit boundaries. Note that this changes the calling
22908 conventions, and thus some functions from the standard C library do
22909 not work unless you recompile it first with @option{-mdalign}.
22913 Shorten some address references at link time, when possible; uses the
22914 linker option @option{-relax}.
22918 Use 32-bit offsets in @code{switch} tables. The default is to use
22923 Enable the use of bit manipulation instructions on SH2A.
22927 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
22928 alignment constraints.
22932 Comply with the calling conventions defined by Renesas.
22935 @opindex mno-renesas
22936 Comply with the calling conventions defined for GCC before the Renesas
22937 conventions were available. This option is the default for all
22938 targets of the SH toolchain.
22941 @opindex mnomacsave
22942 Mark the @code{MAC} register as call-clobbered, even if
22943 @option{-mrenesas} is given.
22949 Control the IEEE compliance of floating-point comparisons, which affects the
22950 handling of cases where the result of a comparison is unordered. By default
22951 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
22952 enabled @option{-mno-ieee} is implicitly set, which results in faster
22953 floating-point greater-equal and less-equal comparisons. The implicit settings
22954 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
22956 @item -minline-ic_invalidate
22957 @opindex minline-ic_invalidate
22958 Inline code to invalidate instruction cache entries after setting up
22959 nested function trampolines.
22960 This option has no effect if @option{-musermode} is in effect and the selected
22961 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
22963 If the selected code generation option does not allow the use of the @code{icbi}
22964 instruction, and @option{-musermode} is not in effect, the inlined code
22965 manipulates the instruction cache address array directly with an associative
22966 write. This not only requires privileged mode at run time, but it also
22967 fails if the cache line had been mapped via the TLB and has become unmapped.
22971 Dump instruction size and location in the assembly code.
22974 @opindex mpadstruct
22975 This option is deprecated. It pads structures to multiple of 4 bytes,
22976 which is incompatible with the SH ABI@.
22978 @item -matomic-model=@var{model}
22979 @opindex matomic-model=@var{model}
22980 Sets the model of atomic operations and additional parameters as a comma
22981 separated list. For details on the atomic built-in functions see
22982 @ref{__atomic Builtins}. The following models and parameters are supported:
22987 Disable compiler generated atomic sequences and emit library calls for atomic
22988 operations. This is the default if the target is not @code{sh*-*-linux*}.
22991 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
22992 built-in functions. The generated atomic sequences require additional support
22993 from the interrupt/exception handling code of the system and are only suitable
22994 for SH3* and SH4* single-core systems. This option is enabled by default when
22995 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
22996 this option also partially utilizes the hardware atomic instructions
22997 @code{movli.l} and @code{movco.l} to create more efficient code, unless
22998 @samp{strict} is specified.
23001 Generate software atomic sequences that use a variable in the thread control
23002 block. This is a variation of the gUSA sequences which can also be used on
23003 SH1* and SH2* targets. The generated atomic sequences require additional
23004 support from the interrupt/exception handling code of the system and are only
23005 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
23006 parameter has to be specified as well.
23009 Generate software atomic sequences that temporarily disable interrupts by
23010 setting @code{SR.IMASK = 1111}. This model works only when the program runs
23011 in privileged mode and is only suitable for single-core systems. Additional
23012 support from the interrupt/exception handling code of the system is not
23013 required. This model is enabled by default when the target is
23014 @code{sh*-*-linux*} and SH1* or SH2*.
23017 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
23018 instructions only. This is only available on SH4A and is suitable for
23019 multi-core systems. Since the hardware instructions support only 32 bit atomic
23020 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
23021 Code compiled with this option is also compatible with other software
23022 atomic model interrupt/exception handling systems if executed on an SH4A
23023 system. Additional support from the interrupt/exception handling code of the
23024 system is not required for this model.
23027 This parameter specifies the offset in bytes of the variable in the thread
23028 control block structure that should be used by the generated atomic sequences
23029 when the @samp{soft-tcb} model has been selected. For other models this
23030 parameter is ignored. The specified value must be an integer multiple of four
23031 and in the range 0-1020.
23034 This parameter prevents mixed usage of multiple atomic models, even if they
23035 are compatible, and makes the compiler generate atomic sequences of the
23036 specified model only.
23042 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
23043 Notice that depending on the particular hardware and software configuration
23044 this can degrade overall performance due to the operand cache line flushes
23045 that are implied by the @code{tas.b} instruction. On multi-core SH4A
23046 processors the @code{tas.b} instruction must be used with caution since it
23047 can result in data corruption for certain cache configurations.
23050 @opindex mprefergot
23051 When generating position-independent code, emit function calls using
23052 the Global Offset Table instead of the Procedure Linkage Table.
23055 @itemx -mno-usermode
23057 @opindex mno-usermode
23058 Don't allow (allow) the compiler generating privileged mode code. Specifying
23059 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
23060 inlined code would not work in user mode. @option{-musermode} is the default
23061 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
23062 @option{-musermode} has no effect, since there is no user mode.
23064 @item -multcost=@var{number}
23065 @opindex multcost=@var{number}
23066 Set the cost to assume for a multiply insn.
23068 @item -mdiv=@var{strategy}
23069 @opindex mdiv=@var{strategy}
23070 Set the division strategy to be used for integer division operations.
23071 @var{strategy} can be one of:
23076 Calls a library function that uses the single-step division instruction
23077 @code{div1} to perform the operation. Division by zero calculates an
23078 unspecified result and does not trap. This is the default except for SH4,
23079 SH2A and SHcompact.
23082 Calls a library function that performs the operation in double precision
23083 floating point. Division by zero causes a floating-point exception. This is
23084 the default for SHcompact with FPU. Specifying this for targets that do not
23085 have a double precision FPU defaults to @code{call-div1}.
23088 Calls a library function that uses a lookup table for small divisors and
23089 the @code{div1} instruction with case distinction for larger divisors. Division
23090 by zero calculates an unspecified result and does not trap. This is the default
23091 for SH4. Specifying this for targets that do not have dynamic shift
23092 instructions defaults to @code{call-div1}.
23096 When a division strategy has not been specified the default strategy is
23097 selected based on the current target. For SH2A the default strategy is to
23098 use the @code{divs} and @code{divu} instructions instead of library function
23101 @item -maccumulate-outgoing-args
23102 @opindex maccumulate-outgoing-args
23103 Reserve space once for outgoing arguments in the function prologue rather
23104 than around each call. Generally beneficial for performance and size. Also
23105 needed for unwinding to avoid changing the stack frame around conditional code.
23107 @item -mdivsi3_libfunc=@var{name}
23108 @opindex mdivsi3_libfunc=@var{name}
23109 Set the name of the library function used for 32-bit signed division to
23111 This only affects the name used in the @samp{call} division strategies, and
23112 the compiler still expects the same sets of input/output/clobbered registers as
23113 if this option were not present.
23115 @item -mfixed-range=@var{register-range}
23116 @opindex mfixed-range
23117 Generate code treating the given register range as fixed registers.
23118 A fixed register is one that the register allocator can not use. This is
23119 useful when compiling kernel code. A register range is specified as
23120 two registers separated by a dash. Multiple register ranges can be
23121 specified separated by a comma.
23123 @item -mbranch-cost=@var{num}
23124 @opindex mbranch-cost=@var{num}
23125 Assume @var{num} to be the cost for a branch instruction. Higher numbers
23126 make the compiler try to generate more branch-free code if possible.
23127 If not specified the value is selected depending on the processor type that
23128 is being compiled for.
23131 @itemx -mno-zdcbranch
23132 @opindex mzdcbranch
23133 @opindex mno-zdcbranch
23134 Assume (do not assume) that zero displacement conditional branch instructions
23135 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
23136 compiler prefers zero displacement branch code sequences. This is
23137 enabled by default when generating code for SH4 and SH4A. It can be explicitly
23138 disabled by specifying @option{-mno-zdcbranch}.
23140 @item -mcbranch-force-delay-slot
23141 @opindex mcbranch-force-delay-slot
23142 Force the usage of delay slots for conditional branches, which stuffs the delay
23143 slot with a @code{nop} if a suitable instruction can't be found. By default
23144 this option is disabled. It can be enabled to work around hardware bugs as
23145 found in the original SH7055.
23148 @itemx -mno-fused-madd
23149 @opindex mfused-madd
23150 @opindex mno-fused-madd
23151 Generate code that uses (does not use) the floating-point multiply and
23152 accumulate instructions. These instructions are generated by default
23153 if hardware floating point is used. The machine-dependent
23154 @option{-mfused-madd} option is now mapped to the machine-independent
23155 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23156 mapped to @option{-ffp-contract=off}.
23162 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
23163 and cosine approximations. The option @option{-mfsca} must be used in
23164 combination with @option{-funsafe-math-optimizations}. It is enabled by default
23165 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
23166 approximations even if @option{-funsafe-math-optimizations} is in effect.
23172 Allow or disallow the compiler to emit the @code{fsrra} instruction for
23173 reciprocal square root approximations. The option @option{-mfsrra} must be used
23174 in combination with @option{-funsafe-math-optimizations} and
23175 @option{-ffinite-math-only}. It is enabled by default when generating code for
23176 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
23177 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
23180 @item -mpretend-cmove
23181 @opindex mpretend-cmove
23182 Prefer zero-displacement conditional branches for conditional move instruction
23183 patterns. This can result in faster code on the SH4 processor.
23187 Generate code using the FDPIC ABI.
23191 @node Solaris 2 Options
23192 @subsection Solaris 2 Options
23193 @cindex Solaris 2 options
23195 These @samp{-m} options are supported on Solaris 2:
23198 @item -mclear-hwcap
23199 @opindex mclear-hwcap
23200 @option{-mclear-hwcap} tells the compiler to remove the hardware
23201 capabilities generated by the Solaris assembler. This is only necessary
23202 when object files use ISA extensions not supported by the current
23203 machine, but check at runtime whether or not to use them.
23205 @item -mimpure-text
23206 @opindex mimpure-text
23207 @option{-mimpure-text}, used in addition to @option{-shared}, tells
23208 the compiler to not pass @option{-z text} to the linker when linking a
23209 shared object. Using this option, you can link position-dependent
23210 code into a shared object.
23212 @option{-mimpure-text} suppresses the ``relocations remain against
23213 allocatable but non-writable sections'' linker error message.
23214 However, the necessary relocations trigger copy-on-write, and the
23215 shared object is not actually shared across processes. Instead of
23216 using @option{-mimpure-text}, you should compile all source code with
23217 @option{-fpic} or @option{-fPIC}.
23221 These switches are supported in addition to the above on Solaris 2:
23226 This is a synonym for @option{-pthread}.
23229 @node SPARC Options
23230 @subsection SPARC Options
23231 @cindex SPARC options
23233 These @samp{-m} options are supported on the SPARC:
23236 @item -mno-app-regs
23238 @opindex mno-app-regs
23240 Specify @option{-mapp-regs} to generate output using the global registers
23241 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
23242 global register 1, each global register 2 through 4 is then treated as an
23243 allocable register that is clobbered by function calls. This is the default.
23245 To be fully SVR4 ABI-compliant at the cost of some performance loss,
23246 specify @option{-mno-app-regs}. You should compile libraries and system
23247 software with this option.
23253 With @option{-mflat}, the compiler does not generate save/restore instructions
23254 and uses a ``flat'' or single register window model. This model is compatible
23255 with the regular register window model. The local registers and the input
23256 registers (0--5) are still treated as ``call-saved'' registers and are
23257 saved on the stack as needed.
23259 With @option{-mno-flat} (the default), the compiler generates save/restore
23260 instructions (except for leaf functions). This is the normal operating mode.
23263 @itemx -mhard-float
23265 @opindex mhard-float
23266 Generate output containing floating-point instructions. This is the
23270 @itemx -msoft-float
23272 @opindex msoft-float
23273 Generate output containing library calls for floating point.
23274 @strong{Warning:} the requisite libraries are not available for all SPARC
23275 targets. Normally the facilities of the machine's usual C compiler are
23276 used, but this cannot be done directly in cross-compilation. You must make
23277 your own arrangements to provide suitable library functions for
23278 cross-compilation. The embedded targets @samp{sparc-*-aout} and
23279 @samp{sparclite-*-*} do provide software floating-point support.
23281 @option{-msoft-float} changes the calling convention in the output file;
23282 therefore, it is only useful if you compile @emph{all} of a program with
23283 this option. In particular, you need to compile @file{libgcc.a}, the
23284 library that comes with GCC, with @option{-msoft-float} in order for
23287 @item -mhard-quad-float
23288 @opindex mhard-quad-float
23289 Generate output containing quad-word (long double) floating-point
23292 @item -msoft-quad-float
23293 @opindex msoft-quad-float
23294 Generate output containing library calls for quad-word (long double)
23295 floating-point instructions. The functions called are those specified
23296 in the SPARC ABI@. This is the default.
23298 As of this writing, there are no SPARC implementations that have hardware
23299 support for the quad-word floating-point instructions. They all invoke
23300 a trap handler for one of these instructions, and then the trap handler
23301 emulates the effect of the instruction. Because of the trap handler overhead,
23302 this is much slower than calling the ABI library routines. Thus the
23303 @option{-msoft-quad-float} option is the default.
23305 @item -mno-unaligned-doubles
23306 @itemx -munaligned-doubles
23307 @opindex mno-unaligned-doubles
23308 @opindex munaligned-doubles
23309 Assume that doubles have 8-byte alignment. This is the default.
23311 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
23312 alignment only if they are contained in another type, or if they have an
23313 absolute address. Otherwise, it assumes they have 4-byte alignment.
23314 Specifying this option avoids some rare compatibility problems with code
23315 generated by other compilers. It is not the default because it results
23316 in a performance loss, especially for floating-point code.
23319 @itemx -mno-user-mode
23320 @opindex muser-mode
23321 @opindex mno-user-mode
23322 Do not generate code that can only run in supervisor mode. This is relevant
23323 only for the @code{casa} instruction emitted for the LEON3 processor. This
23326 @item -mfaster-structs
23327 @itemx -mno-faster-structs
23328 @opindex mfaster-structs
23329 @opindex mno-faster-structs
23330 With @option{-mfaster-structs}, the compiler assumes that structures
23331 should have 8-byte alignment. This enables the use of pairs of
23332 @code{ldd} and @code{std} instructions for copies in structure
23333 assignment, in place of twice as many @code{ld} and @code{st} pairs.
23334 However, the use of this changed alignment directly violates the SPARC
23335 ABI@. Thus, it's intended only for use on targets where the developer
23336 acknowledges that their resulting code is not directly in line with
23337 the rules of the ABI@.
23339 @item -mstd-struct-return
23340 @itemx -mno-std-struct-return
23341 @opindex mstd-struct-return
23342 @opindex mno-std-struct-return
23343 With @option{-mstd-struct-return}, the compiler generates checking code
23344 in functions returning structures or unions to detect size mismatches
23345 between the two sides of function calls, as per the 32-bit ABI@.
23347 The default is @option{-mno-std-struct-return}. This option has no effect
23354 Enable Local Register Allocation. This is the default for SPARC since GCC 7
23355 so @option{-mno-lra} needs to be passed to get old Reload.
23357 @item -mcpu=@var{cpu_type}
23359 Set the instruction set, register set, and instruction scheduling parameters
23360 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23361 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
23362 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
23363 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
23364 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
23365 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
23367 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
23368 which selects the best architecture option for the host processor.
23369 @option{-mcpu=native} has no effect if GCC does not recognize
23372 Default instruction scheduling parameters are used for values that select
23373 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
23374 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
23376 Here is a list of each supported architecture and their supported
23384 supersparc, hypersparc, leon, leon3
23387 f930, f934, sparclite86x
23393 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
23396 By default (unless configured otherwise), GCC generates code for the V7
23397 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
23398 additionally optimizes it for the Cypress CY7C602 chip, as used in the
23399 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
23400 SPARCStation 1, 2, IPX etc.
23402 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
23403 architecture. The only difference from V7 code is that the compiler emits
23404 the integer multiply and integer divide instructions which exist in SPARC-V8
23405 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
23406 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
23409 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
23410 the SPARC architecture. This adds the integer multiply, integer divide step
23411 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
23412 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
23413 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
23414 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
23415 MB86934 chip, which is the more recent SPARClite with FPU@.
23417 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
23418 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
23419 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
23420 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
23421 optimizes it for the TEMIC SPARClet chip.
23423 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
23424 architecture. This adds 64-bit integer and floating-point move instructions,
23425 3 additional floating-point condition code registers and conditional move
23426 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
23427 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
23428 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
23429 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
23430 @option{-mcpu=niagara}, the compiler additionally optimizes it for
23431 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
23432 additionally optimizes it for Sun UltraSPARC T2 chips. With
23433 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
23434 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
23435 additionally optimizes it for Sun UltraSPARC T4 chips. With
23436 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
23437 Oracle SPARC M7 chips.
23439 @item -mtune=@var{cpu_type}
23441 Set the instruction scheduling parameters for machine type
23442 @var{cpu_type}, but do not set the instruction set or register set that the
23443 option @option{-mcpu=@var{cpu_type}} does.
23445 The same values for @option{-mcpu=@var{cpu_type}} can be used for
23446 @option{-mtune=@var{cpu_type}}, but the only useful values are those
23447 that select a particular CPU implementation. Those are
23448 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
23449 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
23450 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
23451 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
23452 @samp{niagara4} and @samp{niagara7}. With native Solaris and
23453 GNU/Linux toolchains, @samp{native} can also be used.
23458 @opindex mno-v8plus
23459 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
23460 difference from the V8 ABI is that the global and out registers are
23461 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
23462 mode for all SPARC-V9 processors.
23468 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
23469 Visual Instruction Set extensions. The default is @option{-mno-vis}.
23475 With @option{-mvis2}, GCC generates code that takes advantage of
23476 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
23477 default is @option{-mvis2} when targeting a cpu that supports such
23478 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
23479 also sets @option{-mvis}.
23485 With @option{-mvis3}, GCC generates code that takes advantage of
23486 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
23487 default is @option{-mvis3} when targeting a cpu that supports such
23488 instructions, such as niagara-3 and later. Setting @option{-mvis3}
23489 also sets @option{-mvis2} and @option{-mvis}.
23495 With @option{-mvis4}, GCC generates code that takes advantage of
23496 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
23497 default is @option{-mvis4} when targeting a cpu that supports such
23498 instructions, such as niagara-7 and later. Setting @option{-mvis4}
23499 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
23504 @opindex mno-cbcond
23505 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
23506 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
23507 when targeting a CPU that supports such instructions, such as Niagara-4 and
23514 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
23515 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
23516 when targeting a CPU that supports such instructions, such as Niagara-3 and
23523 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
23524 Population Count instruction. The default is @option{-mpopc}
23525 when targeting a CPU that supports such an instruction, such as Niagara-2 and
23532 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
23533 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
23534 when targeting a CPU that supports such an instruction, such as Niagara-7 and
23538 @opindex mfix-at697f
23539 Enable the documented workaround for the single erratum of the Atmel AT697F
23540 processor (which corresponds to erratum #13 of the AT697E processor).
23543 @opindex mfix-ut699
23544 Enable the documented workarounds for the floating-point errata and the data
23545 cache nullify errata of the UT699 processor.
23548 These @samp{-m} options are supported in addition to the above
23549 on SPARC-V9 processors in 64-bit environments:
23556 Generate code for a 32-bit or 64-bit environment.
23557 The 32-bit environment sets int, long and pointer to 32 bits.
23558 The 64-bit environment sets int to 32 bits and long and pointer
23561 @item -mcmodel=@var{which}
23563 Set the code model to one of
23567 The Medium/Low code model: 64-bit addresses, programs
23568 must be linked in the low 32 bits of memory. Programs can be statically
23569 or dynamically linked.
23572 The Medium/Middle code model: 64-bit addresses, programs
23573 must be linked in the low 44 bits of memory, the text and data segments must
23574 be less than 2GB in size and the data segment must be located within 2GB of
23578 The Medium/Anywhere code model: 64-bit addresses, programs
23579 may be linked anywhere in memory, the text and data segments must be less
23580 than 2GB in size and the data segment must be located within 2GB of the
23584 The Medium/Anywhere code model for embedded systems:
23585 64-bit addresses, the text and data segments must be less than 2GB in
23586 size, both starting anywhere in memory (determined at link time). The
23587 global register %g4 points to the base of the data segment. Programs
23588 are statically linked and PIC is not supported.
23591 @item -mmemory-model=@var{mem-model}
23592 @opindex mmemory-model
23593 Set the memory model in force on the processor to one of
23597 The default memory model for the processor and operating system.
23600 Relaxed Memory Order
23603 Partial Store Order
23609 Sequential Consistency
23612 These memory models are formally defined in Appendix D of the SPARC-V9
23613 architecture manual, as set in the processor's @code{PSTATE.MM} field.
23616 @itemx -mno-stack-bias
23617 @opindex mstack-bias
23618 @opindex mno-stack-bias
23619 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
23620 frame pointer if present, are offset by @minus{}2047 which must be added back
23621 when making stack frame references. This is the default in 64-bit mode.
23622 Otherwise, assume no such offset is present.
23626 @subsection SPU Options
23627 @cindex SPU options
23629 These @samp{-m} options are supported on the SPU:
23633 @itemx -merror-reloc
23634 @opindex mwarn-reloc
23635 @opindex merror-reloc
23637 The loader for SPU does not handle dynamic relocations. By default, GCC
23638 gives an error when it generates code that requires a dynamic
23639 relocation. @option{-mno-error-reloc} disables the error,
23640 @option{-mwarn-reloc} generates a warning instead.
23643 @itemx -munsafe-dma
23645 @opindex munsafe-dma
23647 Instructions that initiate or test completion of DMA must not be
23648 reordered with respect to loads and stores of the memory that is being
23650 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
23651 memory accesses, but that can lead to inefficient code in places where the
23652 memory is known to not change. Rather than mark the memory as volatile,
23653 you can use @option{-msafe-dma} to tell the compiler to treat
23654 the DMA instructions as potentially affecting all memory.
23656 @item -mbranch-hints
23657 @opindex mbranch-hints
23659 By default, GCC generates a branch hint instruction to avoid
23660 pipeline stalls for always-taken or probably-taken branches. A hint
23661 is not generated closer than 8 instructions away from its branch.
23662 There is little reason to disable them, except for debugging purposes,
23663 or to make an object a little bit smaller.
23667 @opindex msmall-mem
23668 @opindex mlarge-mem
23670 By default, GCC generates code assuming that addresses are never larger
23671 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
23672 a full 32-bit address.
23677 By default, GCC links against startup code that assumes the SPU-style
23678 main function interface (which has an unconventional parameter list).
23679 With @option{-mstdmain}, GCC links your program against startup
23680 code that assumes a C99-style interface to @code{main}, including a
23681 local copy of @code{argv} strings.
23683 @item -mfixed-range=@var{register-range}
23684 @opindex mfixed-range
23685 Generate code treating the given register range as fixed registers.
23686 A fixed register is one that the register allocator cannot use. This is
23687 useful when compiling kernel code. A register range is specified as
23688 two registers separated by a dash. Multiple register ranges can be
23689 specified separated by a comma.
23695 Compile code assuming that pointers to the PPU address space accessed
23696 via the @code{__ea} named address space qualifier are either 32 or 64
23697 bits wide. The default is 32 bits. As this is an ABI-changing option,
23698 all object code in an executable must be compiled with the same setting.
23700 @item -maddress-space-conversion
23701 @itemx -mno-address-space-conversion
23702 @opindex maddress-space-conversion
23703 @opindex mno-address-space-conversion
23704 Allow/disallow treating the @code{__ea} address space as superset
23705 of the generic address space. This enables explicit type casts
23706 between @code{__ea} and generic pointer as well as implicit
23707 conversions of generic pointers to @code{__ea} pointers. The
23708 default is to allow address space pointer conversions.
23710 @item -mcache-size=@var{cache-size}
23711 @opindex mcache-size
23712 This option controls the version of libgcc that the compiler links to an
23713 executable and selects a software-managed cache for accessing variables
23714 in the @code{__ea} address space with a particular cache size. Possible
23715 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
23716 and @samp{128}. The default cache size is 64KB.
23718 @item -matomic-updates
23719 @itemx -mno-atomic-updates
23720 @opindex matomic-updates
23721 @opindex mno-atomic-updates
23722 This option controls the version of libgcc that the compiler links to an
23723 executable and selects whether atomic updates to the software-managed
23724 cache of PPU-side variables are used. If you use atomic updates, changes
23725 to a PPU variable from SPU code using the @code{__ea} named address space
23726 qualifier do not interfere with changes to other PPU variables residing
23727 in the same cache line from PPU code. If you do not use atomic updates,
23728 such interference may occur; however, writing back cache lines is
23729 more efficient. The default behavior is to use atomic updates.
23732 @itemx -mdual-nops=@var{n}
23733 @opindex mdual-nops
23734 By default, GCC inserts NOPs to increase dual issue when it expects
23735 it to increase performance. @var{n} can be a value from 0 to 10. A
23736 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
23737 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
23739 @item -mhint-max-nops=@var{n}
23740 @opindex mhint-max-nops
23741 Maximum number of NOPs to insert for a branch hint. A branch hint must
23742 be at least 8 instructions away from the branch it is affecting. GCC
23743 inserts up to @var{n} NOPs to enforce this, otherwise it does not
23744 generate the branch hint.
23746 @item -mhint-max-distance=@var{n}
23747 @opindex mhint-max-distance
23748 The encoding of the branch hint instruction limits the hint to be within
23749 256 instructions of the branch it is affecting. By default, GCC makes
23750 sure it is within 125.
23753 @opindex msafe-hints
23754 Work around a hardware bug that causes the SPU to stall indefinitely.
23755 By default, GCC inserts the @code{hbrp} instruction to make sure
23756 this stall won't happen.
23760 @node System V Options
23761 @subsection Options for System V
23763 These additional options are available on System V Release 4 for
23764 compatibility with other compilers on those systems:
23769 Create a shared object.
23770 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
23774 Identify the versions of each tool used by the compiler, in a
23775 @code{.ident} assembler directive in the output.
23779 Refrain from adding @code{.ident} directives to the output file (this is
23782 @item -YP,@var{dirs}
23784 Search the directories @var{dirs}, and no others, for libraries
23785 specified with @option{-l}.
23787 @item -Ym,@var{dir}
23789 Look in the directory @var{dir} to find the M4 preprocessor.
23790 The assembler uses this option.
23791 @c This is supposed to go with a -Yd for predefined M4 macro files, but
23792 @c the generic assembler that comes with Solaris takes just -Ym.
23795 @node TILE-Gx Options
23796 @subsection TILE-Gx Options
23797 @cindex TILE-Gx options
23799 These @samp{-m} options are supported on the TILE-Gx:
23802 @item -mcmodel=small
23803 @opindex mcmodel=small
23804 Generate code for the small model. The distance for direct calls is
23805 limited to 500M in either direction. PC-relative addresses are 32
23806 bits. Absolute addresses support the full address range.
23808 @item -mcmodel=large
23809 @opindex mcmodel=large
23810 Generate code for the large model. There is no limitation on call
23811 distance, pc-relative addresses, or absolute addresses.
23813 @item -mcpu=@var{name}
23815 Selects the type of CPU to be targeted. Currently the only supported
23816 type is @samp{tilegx}.
23822 Generate code for a 32-bit or 64-bit environment. The 32-bit
23823 environment sets int, long, and pointer to 32 bits. The 64-bit
23824 environment sets int to 32 bits and long and pointer to 64 bits.
23827 @itemx -mlittle-endian
23828 @opindex mbig-endian
23829 @opindex mlittle-endian
23830 Generate code in big/little endian mode, respectively.
23833 @node TILEPro Options
23834 @subsection TILEPro Options
23835 @cindex TILEPro options
23837 These @samp{-m} options are supported on the TILEPro:
23840 @item -mcpu=@var{name}
23842 Selects the type of CPU to be targeted. Currently the only supported
23843 type is @samp{tilepro}.
23847 Generate code for a 32-bit environment, which sets int, long, and
23848 pointer to 32 bits. This is the only supported behavior so the flag
23849 is essentially ignored.
23853 @subsection V850 Options
23854 @cindex V850 Options
23856 These @samp{-m} options are defined for V850 implementations:
23860 @itemx -mno-long-calls
23861 @opindex mlong-calls
23862 @opindex mno-long-calls
23863 Treat all calls as being far away (near). If calls are assumed to be
23864 far away, the compiler always loads the function's address into a
23865 register, and calls indirect through the pointer.
23871 Do not optimize (do optimize) basic blocks that use the same index
23872 pointer 4 or more times to copy pointer into the @code{ep} register, and
23873 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
23874 option is on by default if you optimize.
23876 @item -mno-prolog-function
23877 @itemx -mprolog-function
23878 @opindex mno-prolog-function
23879 @opindex mprolog-function
23880 Do not use (do use) external functions to save and restore registers
23881 at the prologue and epilogue of a function. The external functions
23882 are slower, but use less code space if more than one function saves
23883 the same number of registers. The @option{-mprolog-function} option
23884 is on by default if you optimize.
23888 Try to make the code as small as possible. At present, this just turns
23889 on the @option{-mep} and @option{-mprolog-function} options.
23891 @item -mtda=@var{n}
23893 Put static or global variables whose size is @var{n} bytes or less into
23894 the tiny data area that register @code{ep} points to. The tiny data
23895 area can hold up to 256 bytes in total (128 bytes for byte references).
23897 @item -msda=@var{n}
23899 Put static or global variables whose size is @var{n} bytes or less into
23900 the small data area that register @code{gp} points to. The small data
23901 area can hold up to 64 kilobytes.
23903 @item -mzda=@var{n}
23905 Put static or global variables whose size is @var{n} bytes or less into
23906 the first 32 kilobytes of memory.
23910 Specify that the target processor is the V850.
23914 Specify that the target processor is the V850E3V5. The preprocessor
23915 constant @code{__v850e3v5__} is defined if this option is used.
23919 Specify that the target processor is the V850E3V5. This is an alias for
23920 the @option{-mv850e3v5} option.
23924 Specify that the target processor is the V850E2V3. The preprocessor
23925 constant @code{__v850e2v3__} is defined if this option is used.
23929 Specify that the target processor is the V850E2. The preprocessor
23930 constant @code{__v850e2__} is defined if this option is used.
23934 Specify that the target processor is the V850E1. The preprocessor
23935 constants @code{__v850e1__} and @code{__v850e__} are defined if
23936 this option is used.
23940 Specify that the target processor is the V850ES. This is an alias for
23941 the @option{-mv850e1} option.
23945 Specify that the target processor is the V850E@. The preprocessor
23946 constant @code{__v850e__} is defined if this option is used.
23948 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
23949 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
23950 are defined then a default target processor is chosen and the
23951 relevant @samp{__v850*__} preprocessor constant is defined.
23953 The preprocessor constants @code{__v850} and @code{__v851__} are always
23954 defined, regardless of which processor variant is the target.
23956 @item -mdisable-callt
23957 @itemx -mno-disable-callt
23958 @opindex mdisable-callt
23959 @opindex mno-disable-callt
23960 This option suppresses generation of the @code{CALLT} instruction for the
23961 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
23964 This option is enabled by default when the RH850 ABI is
23965 in use (see @option{-mrh850-abi}), and disabled by default when the
23966 GCC ABI is in use. If @code{CALLT} instructions are being generated
23967 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
23973 Pass on (or do not pass on) the @option{-mrelax} command-line option
23977 @itemx -mno-long-jumps
23978 @opindex mlong-jumps
23979 @opindex mno-long-jumps
23980 Disable (or re-enable) the generation of PC-relative jump instructions.
23983 @itemx -mhard-float
23984 @opindex msoft-float
23985 @opindex mhard-float
23986 Disable (or re-enable) the generation of hardware floating point
23987 instructions. This option is only significant when the target
23988 architecture is @samp{V850E2V3} or higher. If hardware floating point
23989 instructions are being generated then the C preprocessor symbol
23990 @code{__FPU_OK__} is defined, otherwise the symbol
23991 @code{__NO_FPU__} is defined.
23995 Enables the use of the e3v5 LOOP instruction. The use of this
23996 instruction is not enabled by default when the e3v5 architecture is
23997 selected because its use is still experimental.
24001 @opindex mrh850-abi
24003 Enables support for the RH850 version of the V850 ABI. This is the
24004 default. With this version of the ABI the following rules apply:
24008 Integer sized structures and unions are returned via a memory pointer
24009 rather than a register.
24012 Large structures and unions (more than 8 bytes in size) are passed by
24016 Functions are aligned to 16-bit boundaries.
24019 The @option{-m8byte-align} command-line option is supported.
24022 The @option{-mdisable-callt} command-line option is enabled by
24023 default. The @option{-mno-disable-callt} command-line option is not
24027 When this version of the ABI is enabled the C preprocessor symbol
24028 @code{__V850_RH850_ABI__} is defined.
24032 Enables support for the old GCC version of the V850 ABI. With this
24033 version of the ABI the following rules apply:
24037 Integer sized structures and unions are returned in register @code{r10}.
24040 Large structures and unions (more than 8 bytes in size) are passed by
24044 Functions are aligned to 32-bit boundaries, unless optimizing for
24048 The @option{-m8byte-align} command-line option is not supported.
24051 The @option{-mdisable-callt} command-line option is supported but not
24052 enabled by default.
24055 When this version of the ABI is enabled the C preprocessor symbol
24056 @code{__V850_GCC_ABI__} is defined.
24058 @item -m8byte-align
24059 @itemx -mno-8byte-align
24060 @opindex m8byte-align
24061 @opindex mno-8byte-align
24062 Enables support for @code{double} and @code{long long} types to be
24063 aligned on 8-byte boundaries. The default is to restrict the
24064 alignment of all objects to at most 4-bytes. When
24065 @option{-m8byte-align} is in effect the C preprocessor symbol
24066 @code{__V850_8BYTE_ALIGN__} is defined.
24069 @opindex mbig-switch
24070 Generate code suitable for big switch tables. Use this option only if
24071 the assembler/linker complain about out of range branches within a switch
24076 This option causes r2 and r5 to be used in the code generated by
24077 the compiler. This setting is the default.
24079 @item -mno-app-regs
24080 @opindex mno-app-regs
24081 This option causes r2 and r5 to be treated as fixed registers.
24086 @subsection VAX Options
24087 @cindex VAX options
24089 These @samp{-m} options are defined for the VAX:
24094 Do not output certain jump instructions (@code{aobleq} and so on)
24095 that the Unix assembler for the VAX cannot handle across long
24100 Do output those jump instructions, on the assumption that the
24101 GNU assembler is being used.
24105 Output code for G-format floating-point numbers instead of D-format.
24108 @node Visium Options
24109 @subsection Visium Options
24110 @cindex Visium options
24116 A program which performs file I/O and is destined to run on an MCM target
24117 should be linked with this option. It causes the libraries libc.a and
24118 libdebug.a to be linked. The program should be run on the target under
24119 the control of the GDB remote debugging stub.
24123 A program which performs file I/O and is destined to run on the simulator
24124 should be linked with option. This causes libraries libc.a and libsim.a to
24128 @itemx -mhard-float
24130 @opindex mhard-float
24131 Generate code containing floating-point instructions. This is the
24135 @itemx -msoft-float
24137 @opindex msoft-float
24138 Generate code containing library calls for floating-point.
24140 @option{-msoft-float} changes the calling convention in the output file;
24141 therefore, it is only useful if you compile @emph{all} of a program with
24142 this option. In particular, you need to compile @file{libgcc.a}, the
24143 library that comes with GCC, with @option{-msoft-float} in order for
24146 @item -mcpu=@var{cpu_type}
24148 Set the instruction set, register set, and instruction scheduling parameters
24149 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
24150 @samp{mcm}, @samp{gr5} and @samp{gr6}.
24152 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
24154 By default (unless configured otherwise), GCC generates code for the GR5
24155 variant of the Visium architecture.
24157 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
24158 architecture. The only difference from GR5 code is that the compiler will
24159 generate block move instructions.
24161 @item -mtune=@var{cpu_type}
24163 Set the instruction scheduling parameters for machine type @var{cpu_type},
24164 but do not set the instruction set or register set that the option
24165 @option{-mcpu=@var{cpu_type}} would.
24169 Generate code for the supervisor mode, where there are no restrictions on
24170 the access to general registers. This is the default.
24173 @opindex muser-mode
24174 Generate code for the user mode, where the access to some general registers
24175 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
24176 mode; on the GR6, only registers r29 to r31 are affected.
24180 @subsection VMS Options
24182 These @samp{-m} options are defined for the VMS implementations:
24185 @item -mvms-return-codes
24186 @opindex mvms-return-codes
24187 Return VMS condition codes from @code{main}. The default is to return POSIX-style
24188 condition (e.g.@ error) codes.
24190 @item -mdebug-main=@var{prefix}
24191 @opindex mdebug-main=@var{prefix}
24192 Flag the first routine whose name starts with @var{prefix} as the main
24193 routine for the debugger.
24197 Default to 64-bit memory allocation routines.
24199 @item -mpointer-size=@var{size}
24200 @opindex mpointer-size=@var{size}
24201 Set the default size of pointers. Possible options for @var{size} are
24202 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
24203 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
24204 The later option disables @code{pragma pointer_size}.
24207 @node VxWorks Options
24208 @subsection VxWorks Options
24209 @cindex VxWorks Options
24211 The options in this section are defined for all VxWorks targets.
24212 Options specific to the target hardware are listed with the other
24213 options for that target.
24218 GCC can generate code for both VxWorks kernels and real time processes
24219 (RTPs). This option switches from the former to the latter. It also
24220 defines the preprocessor macro @code{__RTP__}.
24223 @opindex non-static
24224 Link an RTP executable against shared libraries rather than static
24225 libraries. The options @option{-static} and @option{-shared} can
24226 also be used for RTPs (@pxref{Link Options}); @option{-static}
24233 These options are passed down to the linker. They are defined for
24234 compatibility with Diab.
24237 @opindex Xbind-lazy
24238 Enable lazy binding of function calls. This option is equivalent to
24239 @option{-Wl,-z,now} and is defined for compatibility with Diab.
24243 Disable lazy binding of function calls. This option is the default and
24244 is defined for compatibility with Diab.
24248 @subsection x86 Options
24249 @cindex x86 Options
24251 These @samp{-m} options are defined for the x86 family of computers.
24255 @item -march=@var{cpu-type}
24257 Generate instructions for the machine type @var{cpu-type}. In contrast to
24258 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
24259 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
24260 to generate code that may not run at all on processors other than the one
24261 indicated. Specifying @option{-march=@var{cpu-type}} implies
24262 @option{-mtune=@var{cpu-type}}.
24264 The choices for @var{cpu-type} are:
24268 This selects the CPU to generate code for at compilation time by determining
24269 the processor type of the compiling machine. Using @option{-march=native}
24270 enables all instruction subsets supported by the local machine (hence
24271 the result might not run on different machines). Using @option{-mtune=native}
24272 produces code optimized for the local machine under the constraints
24273 of the selected instruction set.
24276 Original Intel i386 CPU@.
24279 Intel i486 CPU@. (No scheduling is implemented for this chip.)
24283 Intel Pentium CPU with no MMX support.
24286 Intel Lakemont MCU, based on Intel Pentium CPU.
24289 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
24292 Intel Pentium Pro CPU@.
24295 When used with @option{-march}, the Pentium Pro
24296 instruction set is used, so the code runs on all i686 family chips.
24297 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
24300 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
24305 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
24309 Intel Pentium M; low-power version of Intel Pentium III CPU
24310 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
24314 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
24317 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
24321 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
24322 SSE2 and SSE3 instruction set support.
24325 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
24326 instruction set support.
24329 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24330 SSE4.1, SSE4.2 and POPCNT instruction set support.
24333 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24334 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
24337 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24338 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
24341 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
24342 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
24343 instruction set support.
24346 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24347 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24348 BMI, BMI2 and F16C instruction set support.
24351 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24352 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24353 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
24356 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24357 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24358 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
24359 XSAVES instruction set support.
24362 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
24363 instruction set support.
24366 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
24367 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
24370 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24371 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24372 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
24373 AVX512CD instruction set support.
24375 @item skylake-avx512
24376 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
24377 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
24378 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
24379 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
24382 AMD K6 CPU with MMX instruction set support.
24386 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
24389 @itemx athlon-tbird
24390 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
24396 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
24397 instruction set support.
24403 Processors based on the AMD K8 core with x86-64 instruction set support,
24404 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
24405 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
24406 instruction set extensions.)
24409 @itemx opteron-sse3
24410 @itemx athlon64-sse3
24411 Improved versions of AMD K8 cores with SSE3 instruction set support.
24415 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
24416 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
24417 instruction set extensions.)
24420 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
24421 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
24422 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
24424 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24425 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
24426 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
24429 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24430 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
24431 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
24432 64-bit instruction set extensions.
24434 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
24435 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
24436 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
24437 SSE4.2, ABM and 64-bit instruction set extensions.
24440 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
24441 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
24442 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
24443 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
24444 instruction set extensions.
24447 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
24448 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
24449 instruction set extensions.)
24452 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
24453 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
24454 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
24457 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
24461 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
24462 instruction set support.
24465 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
24466 (No scheduling is implemented for this chip.)
24469 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
24470 (No scheduling is implemented for this chip.)
24473 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24474 (No scheduling is implemented for this chip.)
24477 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
24478 (No scheduling is implemented for this chip.)
24481 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
24482 (No scheduling is implemented for this chip.)
24485 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
24486 (No scheduling is implemented for this chip.)
24489 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
24490 (No scheduling is implemented for this chip.)
24493 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
24494 AVX and AVX2 instruction set support.
24495 (No scheduling is implemented for this chip.)
24498 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24499 instruction set support.
24500 (No scheduling is implemented for this chip.)
24503 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24504 instruction set support.
24505 (No scheduling is implemented for this chip.)
24508 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
24509 instruction set support.
24510 (No scheduling is implemented for this chip.)
24513 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24514 instruction set support.
24515 (No scheduling is implemented for this chip.)
24518 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24519 instruction set support.
24520 (No scheduling is implemented for this chip.)
24523 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
24524 instruction set support.
24525 (No scheduling is implemented for this chip.)
24528 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
24531 @item -mtune=@var{cpu-type}
24533 Tune to @var{cpu-type} everything applicable about the generated code, except
24534 for the ABI and the set of available instructions.
24535 While picking a specific @var{cpu-type} schedules things appropriately
24536 for that particular chip, the compiler does not generate any code that
24537 cannot run on the default machine type unless you use a
24538 @option{-march=@var{cpu-type}} option.
24539 For example, if GCC is configured for i686-pc-linux-gnu
24540 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
24541 but still runs on i686 machines.
24543 The choices for @var{cpu-type} are the same as for @option{-march}.
24544 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
24548 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
24549 If you know the CPU on which your code will run, then you should use
24550 the corresponding @option{-mtune} or @option{-march} option instead of
24551 @option{-mtune=generic}. But, if you do not know exactly what CPU users
24552 of your application will have, then you should use this option.
24554 As new processors are deployed in the marketplace, the behavior of this
24555 option will change. Therefore, if you upgrade to a newer version of
24556 GCC, code generation controlled by this option will change to reflect
24558 that are most common at the time that version of GCC is released.
24560 There is no @option{-march=generic} option because @option{-march}
24561 indicates the instruction set the compiler can use, and there is no
24562 generic instruction set applicable to all processors. In contrast,
24563 @option{-mtune} indicates the processor (or, in this case, collection of
24564 processors) for which the code is optimized.
24567 Produce code optimized for the most current Intel processors, which are
24568 Haswell and Silvermont for this version of GCC. If you know the CPU
24569 on which your code will run, then you should use the corresponding
24570 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
24571 But, if you want your application performs better on both Haswell and
24572 Silvermont, then you should use this option.
24574 As new Intel processors are deployed in the marketplace, the behavior of
24575 this option will change. Therefore, if you upgrade to a newer version of
24576 GCC, code generation controlled by this option will change to reflect
24577 the most current Intel processors at the time that version of GCC is
24580 There is no @option{-march=intel} option because @option{-march} indicates
24581 the instruction set the compiler can use, and there is no common
24582 instruction set applicable to all processors. In contrast,
24583 @option{-mtune} indicates the processor (or, in this case, collection of
24584 processors) for which the code is optimized.
24587 @item -mcpu=@var{cpu-type}
24589 A deprecated synonym for @option{-mtune}.
24591 @item -mfpmath=@var{unit}
24593 Generate floating-point arithmetic for selected unit @var{unit}. The choices
24594 for @var{unit} are:
24598 Use the standard 387 floating-point coprocessor present on the majority of chips and
24599 emulated otherwise. Code compiled with this option runs almost everywhere.
24600 The temporary results are computed in 80-bit precision instead of the precision
24601 specified by the type, resulting in slightly different results compared to most
24602 of other chips. See @option{-ffloat-store} for more detailed description.
24604 This is the default choice for x86-32 targets.
24607 Use scalar floating-point instructions present in the SSE instruction set.
24608 This instruction set is supported by Pentium III and newer chips,
24609 and in the AMD line
24610 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
24611 instruction set supports only single-precision arithmetic, thus the double and
24612 extended-precision arithmetic are still done using 387. A later version, present
24613 only in Pentium 4 and AMD x86-64 chips, supports double-precision
24616 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
24617 or @option{-msse2} switches to enable SSE extensions and make this option
24618 effective. For the x86-64 compiler, these extensions are enabled by default.
24620 The resulting code should be considerably faster in the majority of cases and avoid
24621 the numerical instability problems of 387 code, but may break some existing
24622 code that expects temporaries to be 80 bits.
24624 This is the default choice for the x86-64 compiler.
24629 Attempt to utilize both instruction sets at once. This effectively doubles the
24630 amount of available registers, and on chips with separate execution units for
24631 387 and SSE the execution resources too. Use this option with care, as it is
24632 still experimental, because the GCC register allocator does not model separate
24633 functional units well, resulting in unstable performance.
24636 @item -masm=@var{dialect}
24637 @opindex masm=@var{dialect}
24638 Output assembly instructions using selected @var{dialect}. Also affects
24639 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
24640 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
24641 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
24642 not support @samp{intel}.
24645 @itemx -mno-ieee-fp
24647 @opindex mno-ieee-fp
24648 Control whether or not the compiler uses IEEE floating-point
24649 comparisons. These correctly handle the case where the result of a
24650 comparison is unordered.
24655 @opindex mhard-float
24656 Generate output containing 80387 instructions for floating point.
24661 @opindex msoft-float
24662 Generate output containing library calls for floating point.
24664 @strong{Warning:} the requisite libraries are not part of GCC@.
24665 Normally the facilities of the machine's usual C compiler are used, but
24666 this can't be done directly in cross-compilation. You must make your
24667 own arrangements to provide suitable library functions for
24670 On machines where a function returns floating-point results in the 80387
24671 register stack, some floating-point opcodes may be emitted even if
24672 @option{-msoft-float} is used.
24674 @item -mno-fp-ret-in-387
24675 @opindex mno-fp-ret-in-387
24676 Do not use the FPU registers for return values of functions.
24678 The usual calling convention has functions return values of types
24679 @code{float} and @code{double} in an FPU register, even if there
24680 is no FPU@. The idea is that the operating system should emulate
24683 The option @option{-mno-fp-ret-in-387} causes such values to be returned
24684 in ordinary CPU registers instead.
24686 @item -mno-fancy-math-387
24687 @opindex mno-fancy-math-387
24688 Some 387 emulators do not support the @code{sin}, @code{cos} and
24689 @code{sqrt} instructions for the 387. Specify this option to avoid
24690 generating those instructions. This option is the default on
24691 OpenBSD and NetBSD@. This option is overridden when @option{-march}
24692 indicates that the target CPU always has an FPU and so the
24693 instruction does not need emulation. These
24694 instructions are not generated unless you also use the
24695 @option{-funsafe-math-optimizations} switch.
24697 @item -malign-double
24698 @itemx -mno-align-double
24699 @opindex malign-double
24700 @opindex mno-align-double
24701 Control whether GCC aligns @code{double}, @code{long double}, and
24702 @code{long long} variables on a two-word boundary or a one-word
24703 boundary. Aligning @code{double} variables on a two-word boundary
24704 produces code that runs somewhat faster on a Pentium at the
24705 expense of more memory.
24707 On x86-64, @option{-malign-double} is enabled by default.
24709 @strong{Warning:} if you use the @option{-malign-double} switch,
24710 structures containing the above types are aligned differently than
24711 the published application binary interface specifications for the x86-32
24712 and are not binary compatible with structures in code compiled
24713 without that switch.
24715 @item -m96bit-long-double
24716 @itemx -m128bit-long-double
24717 @opindex m96bit-long-double
24718 @opindex m128bit-long-double
24719 These switches control the size of @code{long double} type. The x86-32
24720 application binary interface specifies the size to be 96 bits,
24721 so @option{-m96bit-long-double} is the default in 32-bit mode.
24723 Modern architectures (Pentium and newer) prefer @code{long double}
24724 to be aligned to an 8- or 16-byte boundary. In arrays or structures
24725 conforming to the ABI, this is not possible. So specifying
24726 @option{-m128bit-long-double} aligns @code{long double}
24727 to a 16-byte boundary by padding the @code{long double} with an additional
24730 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
24731 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
24733 Notice that neither of these options enable any extra precision over the x87
24734 standard of 80 bits for a @code{long double}.
24736 @strong{Warning:} if you override the default value for your target ABI, this
24737 changes the size of
24738 structures and arrays containing @code{long double} variables,
24739 as well as modifying the function calling convention for functions taking
24740 @code{long double}. Hence they are not binary-compatible
24741 with code compiled without that switch.
24743 @item -mlong-double-64
24744 @itemx -mlong-double-80
24745 @itemx -mlong-double-128
24746 @opindex mlong-double-64
24747 @opindex mlong-double-80
24748 @opindex mlong-double-128
24749 These switches control the size of @code{long double} type. A size
24750 of 64 bits makes the @code{long double} type equivalent to the @code{double}
24751 type. This is the default for 32-bit Bionic C library. A size
24752 of 128 bits makes the @code{long double} type equivalent to the
24753 @code{__float128} type. This is the default for 64-bit Bionic C library.
24755 @strong{Warning:} if you override the default value for your target ABI, this
24756 changes the size of
24757 structures and arrays containing @code{long double} variables,
24758 as well as modifying the function calling convention for functions taking
24759 @code{long double}. Hence they are not binary-compatible
24760 with code compiled without that switch.
24762 @item -malign-data=@var{type}
24763 @opindex malign-data
24764 Control how GCC aligns variables. Supported values for @var{type} are
24765 @samp{compat} uses increased alignment value compatible uses GCC 4.8
24766 and earlier, @samp{abi} uses alignment value as specified by the
24767 psABI, and @samp{cacheline} uses increased alignment value to match
24768 the cache line size. @samp{compat} is the default.
24770 @item -mlarge-data-threshold=@var{threshold}
24771 @opindex mlarge-data-threshold
24772 When @option{-mcmodel=medium} is specified, data objects larger than
24773 @var{threshold} are placed in the large data section. This value must be the
24774 same across all objects linked into the binary, and defaults to 65535.
24778 Use a different function-calling convention, in which functions that
24779 take a fixed number of arguments return with the @code{ret @var{num}}
24780 instruction, which pops their arguments while returning. This saves one
24781 instruction in the caller since there is no need to pop the arguments
24784 You can specify that an individual function is called with this calling
24785 sequence with the function attribute @code{stdcall}. You can also
24786 override the @option{-mrtd} option by using the function attribute
24787 @code{cdecl}. @xref{Function Attributes}.
24789 @strong{Warning:} this calling convention is incompatible with the one
24790 normally used on Unix, so you cannot use it if you need to call
24791 libraries compiled with the Unix compiler.
24793 Also, you must provide function prototypes for all functions that
24794 take variable numbers of arguments (including @code{printf});
24795 otherwise incorrect code is generated for calls to those
24798 In addition, seriously incorrect code results if you call a
24799 function with too many arguments. (Normally, extra arguments are
24800 harmlessly ignored.)
24802 @item -mregparm=@var{num}
24804 Control how many registers are used to pass integer arguments. By
24805 default, no registers are used to pass arguments, and at most 3
24806 registers can be used. You can control this behavior for a specific
24807 function by using the function attribute @code{regparm}.
24808 @xref{Function Attributes}.
24810 @strong{Warning:} if you use this switch, and
24811 @var{num} is nonzero, then you must build all modules with the same
24812 value, including any libraries. This includes the system libraries and
24816 @opindex msseregparm
24817 Use SSE register passing conventions for float and double arguments
24818 and return values. You can control this behavior for a specific
24819 function by using the function attribute @code{sseregparm}.
24820 @xref{Function Attributes}.
24822 @strong{Warning:} if you use this switch then you must build all
24823 modules with the same value, including any libraries. This includes
24824 the system libraries and startup modules.
24826 @item -mvect8-ret-in-mem
24827 @opindex mvect8-ret-in-mem
24828 Return 8-byte vectors in memory instead of MMX registers. This is the
24829 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
24830 Studio compilers until version 12. Later compiler versions (starting
24831 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
24832 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
24833 you need to remain compatible with existing code produced by those
24834 previous compiler versions or older versions of GCC@.
24843 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
24844 is specified, the significands of results of floating-point operations are
24845 rounded to 24 bits (single precision); @option{-mpc64} rounds the
24846 significands of results of floating-point operations to 53 bits (double
24847 precision) and @option{-mpc80} rounds the significands of results of
24848 floating-point operations to 64 bits (extended double precision), which is
24849 the default. When this option is used, floating-point operations in higher
24850 precisions are not available to the programmer without setting the FPU
24851 control word explicitly.
24853 Setting the rounding of floating-point operations to less than the default
24854 80 bits can speed some programs by 2% or more. Note that some mathematical
24855 libraries assume that extended-precision (80-bit) floating-point operations
24856 are enabled by default; routines in such libraries could suffer significant
24857 loss of accuracy, typically through so-called ``catastrophic cancellation'',
24858 when this option is used to set the precision to less than extended precision.
24860 @item -mstackrealign
24861 @opindex mstackrealign
24862 Realign the stack at entry. On the x86, the @option{-mstackrealign}
24863 option generates an alternate prologue and epilogue that realigns the
24864 run-time stack if necessary. This supports mixing legacy codes that keep
24865 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
24866 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
24867 applicable to individual functions.
24869 @item -mpreferred-stack-boundary=@var{num}
24870 @opindex mpreferred-stack-boundary
24871 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24872 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24873 the default is 4 (16 bytes or 128 bits).
24875 @strong{Warning:} When generating code for the x86-64 architecture with
24876 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
24877 used to keep the stack boundary aligned to 8 byte boundary. Since
24878 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
24879 intended to be used in controlled environment where stack space is
24880 important limitation. This option leads to wrong code when functions
24881 compiled with 16 byte stack alignment (such as functions from a standard
24882 library) are called with misaligned stack. In this case, SSE
24883 instructions may lead to misaligned memory access traps. In addition,
24884 variable arguments are handled incorrectly for 16 byte aligned
24885 objects (including x87 long double and __int128), leading to wrong
24886 results. You must build all modules with
24887 @option{-mpreferred-stack-boundary=3}, including any libraries. This
24888 includes the system libraries and startup modules.
24890 @item -mincoming-stack-boundary=@var{num}
24891 @opindex mincoming-stack-boundary
24892 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
24893 boundary. If @option{-mincoming-stack-boundary} is not specified,
24894 the one specified by @option{-mpreferred-stack-boundary} is used.
24896 On Pentium and Pentium Pro, @code{double} and @code{long double} values
24897 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
24898 suffer significant run time performance penalties. On Pentium III, the
24899 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
24900 properly if it is not 16-byte aligned.
24902 To ensure proper alignment of this values on the stack, the stack boundary
24903 must be as aligned as that required by any value stored on the stack.
24904 Further, every function must be generated such that it keeps the stack
24905 aligned. Thus calling a function compiled with a higher preferred
24906 stack boundary from a function compiled with a lower preferred stack
24907 boundary most likely misaligns the stack. It is recommended that
24908 libraries that use callbacks always use the default setting.
24910 This extra alignment does consume extra stack space, and generally
24911 increases code size. Code that is sensitive to stack space usage, such
24912 as embedded systems and operating system kernels, may want to reduce the
24913 preferred alignment to @option{-mpreferred-stack-boundary=2}.
24970 @itemx -mavx512ifma
24971 @opindex mavx512ifma
24973 @itemx -mavx512vbmi
24974 @opindex mavx512vbmi
24986 @opindex mclfushopt
25003 @itemx -mprefetchwt1
25004 @opindex mprefetchwt1
25063 These switches enable the use of instructions in the MMX, SSE,
25064 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
25065 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
25066 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
25067 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, 3DNow!@: or enhanced 3DNow!@:
25068 extended instruction sets. Each has a corresponding @option{-mno-} option
25069 to disable use of these instructions.
25071 These extensions are also available as built-in functions: see
25072 @ref{x86 Built-in Functions}, for details of the functions enabled and
25073 disabled by these switches.
25075 To generate SSE/SSE2 instructions automatically from floating-point
25076 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
25078 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
25079 generates new AVX instructions or AVX equivalence for all SSEx instructions
25082 These options enable GCC to use these extended instructions in
25083 generated code, even without @option{-mfpmath=sse}. Applications that
25084 perform run-time CPU detection must compile separate files for each
25085 supported architecture, using the appropriate flags. In particular,
25086 the file containing the CPU detection code should be compiled without
25089 @item -mdump-tune-features
25090 @opindex mdump-tune-features
25091 This option instructs GCC to dump the names of the x86 performance
25092 tuning features and default settings. The names can be used in
25093 @option{-mtune-ctrl=@var{feature-list}}.
25095 @item -mtune-ctrl=@var{feature-list}
25096 @opindex mtune-ctrl=@var{feature-list}
25097 This option is used to do fine grain control of x86 code generation features.
25098 @var{feature-list} is a comma separated list of @var{feature} names. See also
25099 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
25100 on if it is not preceded with @samp{^}, otherwise, it is turned off.
25101 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
25102 developers. Using it may lead to code paths not covered by testing and can
25103 potentially result in compiler ICEs or runtime errors.
25106 @opindex mno-default
25107 This option instructs GCC to turn off all tunable features. See also
25108 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
25112 This option instructs GCC to emit a @code{cld} instruction in the prologue
25113 of functions that use string instructions. String instructions depend on
25114 the DF flag to select between autoincrement or autodecrement mode. While the
25115 ABI specifies the DF flag to be cleared on function entry, some operating
25116 systems violate this specification by not clearing the DF flag in their
25117 exception dispatchers. The exception handler can be invoked with the DF flag
25118 set, which leads to wrong direction mode when string instructions are used.
25119 This option can be enabled by default on 32-bit x86 targets by configuring
25120 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
25121 instructions can be suppressed with the @option{-mno-cld} compiler option
25125 @opindex mvzeroupper
25126 This option instructs GCC to emit a @code{vzeroupper} instruction
25127 before a transfer of control flow out of the function to minimize
25128 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
25131 @item -mprefer-avx128
25132 @opindex mprefer-avx128
25133 This option instructs GCC to use 128-bit AVX instructions instead of
25134 256-bit AVX instructions in the auto-vectorizer.
25138 This option enables GCC to generate @code{CMPXCHG16B} instructions.
25139 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
25140 (or oword) data types.
25141 This is useful for high-resolution counters that can be updated
25142 by multiple processors (or cores). This instruction is generated as part of
25143 atomic built-in functions: see @ref{__sync Builtins} or
25144 @ref{__atomic Builtins} for details.
25148 This option enables generation of @code{SAHF} instructions in 64-bit code.
25149 Early Intel Pentium 4 CPUs with Intel 64 support,
25150 prior to the introduction of Pentium 4 G1 step in December 2005,
25151 lacked the @code{LAHF} and @code{SAHF} instructions
25152 which are supported by AMD64.
25153 These are load and store instructions, respectively, for certain status flags.
25154 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
25155 @code{drem}, and @code{remainder} built-in functions;
25156 see @ref{Other Builtins} for details.
25160 This option enables use of the @code{movbe} instruction to implement
25161 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
25165 This option enables built-in functions @code{__builtin_ia32_crc32qi},
25166 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
25167 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
25171 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
25172 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
25173 with an additional Newton-Raphson step
25174 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
25175 (and their vectorized
25176 variants) for single-precision floating-point arguments. These instructions
25177 are generated only when @option{-funsafe-math-optimizations} is enabled
25178 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
25179 Note that while the throughput of the sequence is higher than the throughput
25180 of the non-reciprocal instruction, the precision of the sequence can be
25181 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
25183 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
25184 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
25185 combination), and doesn't need @option{-mrecip}.
25187 Also note that GCC emits the above sequence with additional Newton-Raphson step
25188 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
25189 already with @option{-ffast-math} (or the above option combination), and
25190 doesn't need @option{-mrecip}.
25192 @item -mrecip=@var{opt}
25193 @opindex mrecip=opt
25194 This option controls which reciprocal estimate instructions
25195 may be used. @var{opt} is a comma-separated list of options, which may
25196 be preceded by a @samp{!} to invert the option:
25200 Enable all estimate instructions.
25203 Enable the default instructions, equivalent to @option{-mrecip}.
25206 Disable all estimate instructions, equivalent to @option{-mno-recip}.
25209 Enable the approximation for scalar division.
25212 Enable the approximation for vectorized division.
25215 Enable the approximation for scalar square root.
25218 Enable the approximation for vectorized square root.
25221 So, for example, @option{-mrecip=all,!sqrt} enables
25222 all of the reciprocal approximations, except for square root.
25224 @item -mveclibabi=@var{type}
25225 @opindex mveclibabi
25226 Specifies the ABI type to use for vectorizing intrinsics using an
25227 external library. Supported values for @var{type} are @samp{svml}
25228 for the Intel short
25229 vector math library and @samp{acml} for the AMD math core library.
25230 To use this option, both @option{-ftree-vectorize} and
25231 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
25232 ABI-compatible library must be specified at link time.
25234 GCC currently emits calls to @code{vmldExp2},
25235 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
25236 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
25237 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
25238 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
25239 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
25240 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
25241 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
25242 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
25243 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
25244 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
25245 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
25246 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
25247 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
25248 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
25249 when @option{-mveclibabi=acml} is used.
25251 @item -mabi=@var{name}
25253 Generate code for the specified calling convention. Permissible values
25254 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
25255 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
25256 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
25257 You can control this behavior for specific functions by
25258 using the function attributes @code{ms_abi} and @code{sysv_abi}.
25259 @xref{Function Attributes}.
25261 @item -mtls-dialect=@var{type}
25262 @opindex mtls-dialect
25263 Generate code to access thread-local storage using the @samp{gnu} or
25264 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
25265 @samp{gnu2} is more efficient, but it may add compile- and run-time
25266 requirements that cannot be satisfied on all systems.
25269 @itemx -mno-push-args
25270 @opindex mpush-args
25271 @opindex mno-push-args
25272 Use PUSH operations to store outgoing parameters. This method is shorter
25273 and usually equally fast as method using SUB/MOV operations and is enabled
25274 by default. In some cases disabling it may improve performance because of
25275 improved scheduling and reduced dependencies.
25277 @item -maccumulate-outgoing-args
25278 @opindex maccumulate-outgoing-args
25279 If enabled, the maximum amount of space required for outgoing arguments is
25280 computed in the function prologue. This is faster on most modern CPUs
25281 because of reduced dependencies, improved scheduling and reduced stack usage
25282 when the preferred stack boundary is not equal to 2. The drawback is a notable
25283 increase in code size. This switch implies @option{-mno-push-args}.
25287 Support thread-safe exception handling on MinGW. Programs that rely
25288 on thread-safe exception handling must compile and link all code with the
25289 @option{-mthreads} option. When compiling, @option{-mthreads} defines
25290 @option{-D_MT}; when linking, it links in a special thread helper library
25291 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
25293 @item -mms-bitfields
25294 @itemx -mno-ms-bitfields
25295 @opindex mms-bitfields
25296 @opindex mno-ms-bitfields
25298 Enable/disable bit-field layout compatible with the native Microsoft
25301 If @code{packed} is used on a structure, or if bit-fields are used,
25302 it may be that the Microsoft ABI lays out the structure differently
25303 than the way GCC normally does. Particularly when moving packed
25304 data between functions compiled with GCC and the native Microsoft compiler
25305 (either via function call or as data in a file), it may be necessary to access
25308 This option is enabled by default for Microsoft Windows
25309 targets. This behavior can also be controlled locally by use of variable
25310 or type attributes. For more information, see @ref{x86 Variable Attributes}
25311 and @ref{x86 Type Attributes}.
25313 The Microsoft structure layout algorithm is fairly simple with the exception
25314 of the bit-field packing.
25315 The padding and alignment of members of structures and whether a bit-field
25316 can straddle a storage-unit boundary are determine by these rules:
25319 @item Structure members are stored sequentially in the order in which they are
25320 declared: the first member has the lowest memory address and the last member
25323 @item Every data object has an alignment requirement. The alignment requirement
25324 for all data except structures, unions, and arrays is either the size of the
25325 object or the current packing size (specified with either the
25326 @code{aligned} attribute or the @code{pack} pragma),
25327 whichever is less. For structures, unions, and arrays,
25328 the alignment requirement is the largest alignment requirement of its members.
25329 Every object is allocated an offset so that:
25332 offset % alignment_requirement == 0
25335 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
25336 unit if the integral types are the same size and if the next bit-field fits
25337 into the current allocation unit without crossing the boundary imposed by the
25338 common alignment requirements of the bit-fields.
25341 MSVC interprets zero-length bit-fields in the following ways:
25344 @item If a zero-length bit-field is inserted between two bit-fields that
25345 are normally coalesced, the bit-fields are not coalesced.
25352 unsigned long bf_1 : 12;
25354 unsigned long bf_2 : 12;
25359 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
25360 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
25362 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
25363 alignment of the zero-length bit-field is greater than the member that follows it,
25364 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
25385 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
25386 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
25387 bit-field does not affect the alignment of @code{bar} or, as a result, the size
25390 Taking this into account, it is important to note the following:
25393 @item If a zero-length bit-field follows a normal bit-field, the type of the
25394 zero-length bit-field may affect the alignment of the structure as whole. For
25395 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
25396 normal bit-field, and is of type short.
25398 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
25399 still affect the alignment of the structure:
25410 Here, @code{t4} takes up 4 bytes.
25413 @item Zero-length bit-fields following non-bit-field members are ignored:
25425 Here, @code{t5} takes up 2 bytes.
25429 @item -mno-align-stringops
25430 @opindex mno-align-stringops
25431 Do not align the destination of inlined string operations. This switch reduces
25432 code size and improves performance in case the destination is already aligned,
25433 but GCC doesn't know about it.
25435 @item -minline-all-stringops
25436 @opindex minline-all-stringops
25437 By default GCC inlines string operations only when the destination is
25438 known to be aligned to least a 4-byte boundary.
25439 This enables more inlining and increases code
25440 size, but may improve performance of code that depends on fast
25441 @code{memcpy}, @code{strlen},
25442 and @code{memset} for short lengths.
25444 @item -minline-stringops-dynamically
25445 @opindex minline-stringops-dynamically
25446 For string operations of unknown size, use run-time checks with
25447 inline code for small blocks and a library call for large blocks.
25449 @item -mstringop-strategy=@var{alg}
25450 @opindex mstringop-strategy=@var{alg}
25451 Override the internal decision heuristic for the particular algorithm to use
25452 for inlining string operations. The allowed values for @var{alg} are:
25458 Expand using i386 @code{rep} prefix of the specified size.
25462 @itemx unrolled_loop
25463 Expand into an inline loop.
25466 Always use a library call.
25469 @item -mmemcpy-strategy=@var{strategy}
25470 @opindex mmemcpy-strategy=@var{strategy}
25471 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
25472 should be inlined and what inline algorithm to use when the expected size
25473 of the copy operation is known. @var{strategy}
25474 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
25475 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
25476 the max byte size with which inline algorithm @var{alg} is allowed. For the last
25477 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
25478 in the list must be specified in increasing order. The minimal byte size for
25479 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
25482 @item -mmemset-strategy=@var{strategy}
25483 @opindex mmemset-strategy=@var{strategy}
25484 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
25485 @code{__builtin_memset} expansion.
25487 @item -momit-leaf-frame-pointer
25488 @opindex momit-leaf-frame-pointer
25489 Don't keep the frame pointer in a register for leaf functions. This
25490 avoids the instructions to save, set up, and restore frame pointers and
25491 makes an extra register available in leaf functions. The option
25492 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
25493 which might make debugging harder.
25495 @item -mtls-direct-seg-refs
25496 @itemx -mno-tls-direct-seg-refs
25497 @opindex mtls-direct-seg-refs
25498 Controls whether TLS variables may be accessed with offsets from the
25499 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
25500 or whether the thread base pointer must be added. Whether or not this
25501 is valid depends on the operating system, and whether it maps the
25502 segment to cover the entire TLS area.
25504 For systems that use the GNU C Library, the default is on.
25507 @itemx -mno-sse2avx
25509 Specify that the assembler should encode SSE instructions with VEX
25510 prefix. The option @option{-mavx} turns this on by default.
25515 If profiling is active (@option{-pg}), put the profiling
25516 counter call before the prologue.
25517 Note: On x86 architectures the attribute @code{ms_hook_prologue}
25518 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
25520 @item -mrecord-mcount
25521 @itemx -mno-record-mcount
25522 @opindex mrecord-mcount
25523 If profiling is active (@option{-pg}), generate a __mcount_loc section
25524 that contains pointers to each profiling call. This is useful for
25525 automatically patching and out calls.
25528 @itemx -mno-nop-mcount
25529 @opindex mnop-mcount
25530 If profiling is active (@option{-pg}), generate the calls to
25531 the profiling functions as NOPs. This is useful when they
25532 should be patched in later dynamically. This is likely only
25533 useful together with @option{-mrecord-mcount}.
25535 @item -mskip-rax-setup
25536 @itemx -mno-skip-rax-setup
25537 @opindex mskip-rax-setup
25538 When generating code for the x86-64 architecture with SSE extensions
25539 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
25540 register when there are no variable arguments passed in vector registers.
25542 @strong{Warning:} Since RAX register is used to avoid unnecessarily
25543 saving vector registers on stack when passing variable arguments, the
25544 impacts of this option are callees may waste some stack space,
25545 misbehave or jump to a random location. GCC 4.4 or newer don't have
25546 those issues, regardless the RAX register value.
25549 @itemx -mno-8bit-idiv
25550 @opindex m8bit-idiv
25551 On some processors, like Intel Atom, 8-bit unsigned integer divide is
25552 much faster than 32-bit/64-bit integer divide. This option generates a
25553 run-time check. If both dividend and divisor are within range of 0
25554 to 255, 8-bit unsigned integer divide is used instead of
25555 32-bit/64-bit integer divide.
25557 @item -mavx256-split-unaligned-load
25558 @itemx -mavx256-split-unaligned-store
25559 @opindex mavx256-split-unaligned-load
25560 @opindex mavx256-split-unaligned-store
25561 Split 32-byte AVX unaligned load and store.
25563 @item -mstack-protector-guard=@var{guard}
25564 @opindex mstack-protector-guard=@var{guard}
25565 Generate stack protection code using canary at @var{guard}. Supported
25566 locations are @samp{global} for global canary or @samp{tls} for per-thread
25567 canary in the TLS block (the default). This option has effect only when
25568 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
25570 @item -mmitigate-rop
25571 @opindex mmitigate-rop
25572 Try to avoid generating code sequences that contain unintended return
25573 opcodes, to mitigate against certain forms of attack. At the moment,
25574 this option is limited in what it can do and should not be relied
25575 on to provide serious protection.
25577 @item -mgeneral-regs-only
25578 @opindex mgeneral-regs-only
25579 Generate code that uses only the general-purpose registers. This
25580 prevents the compiler from using floating-point, vector, mask and bound
25585 These @samp{-m} switches are supported in addition to the above
25586 on x86-64 processors in 64-bit environments.
25599 Generate code for a 16-bit, 32-bit or 64-bit environment.
25600 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
25602 generates code that runs on any i386 system.
25604 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
25605 types to 64 bits, and generates code for the x86-64 architecture.
25606 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
25607 and @option{-mdynamic-no-pic} options.
25609 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
25611 generates code for the x86-64 architecture.
25613 The @option{-m16} option is the same as @option{-m32}, except for that
25614 it outputs the @code{.code16gcc} assembly directive at the beginning of
25615 the assembly output so that the binary can run in 16-bit mode.
25617 The @option{-miamcu} option generates code which conforms to Intel MCU
25618 psABI. It requires the @option{-m32} option to be turned on.
25620 @item -mno-red-zone
25621 @opindex mno-red-zone
25622 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
25623 by the x86-64 ABI; it is a 128-byte area beyond the location of the
25624 stack pointer that is not modified by signal or interrupt handlers
25625 and therefore can be used for temporary data without adjusting the stack
25626 pointer. The flag @option{-mno-red-zone} disables this red zone.
25628 @item -mcmodel=small
25629 @opindex mcmodel=small
25630 Generate code for the small code model: the program and its symbols must
25631 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
25632 Programs can be statically or dynamically linked. This is the default
25635 @item -mcmodel=kernel
25636 @opindex mcmodel=kernel
25637 Generate code for the kernel code model. The kernel runs in the
25638 negative 2 GB of the address space.
25639 This model has to be used for Linux kernel code.
25641 @item -mcmodel=medium
25642 @opindex mcmodel=medium
25643 Generate code for the medium model: the program is linked in the lower 2
25644 GB of the address space. Small symbols are also placed there. Symbols
25645 with sizes larger than @option{-mlarge-data-threshold} are put into
25646 large data or BSS sections and can be located above 2GB. Programs can
25647 be statically or dynamically linked.
25649 @item -mcmodel=large
25650 @opindex mcmodel=large
25651 Generate code for the large model. This model makes no assumptions
25652 about addresses and sizes of sections.
25654 @item -maddress-mode=long
25655 @opindex maddress-mode=long
25656 Generate code for long address mode. This is only supported for 64-bit
25657 and x32 environments. It is the default address mode for 64-bit
25660 @item -maddress-mode=short
25661 @opindex maddress-mode=short
25662 Generate code for short address mode. This is only supported for 32-bit
25663 and x32 environments. It is the default address mode for 32-bit and
25667 @node x86 Windows Options
25668 @subsection x86 Windows Options
25669 @cindex x86 Windows Options
25670 @cindex Windows Options for x86
25672 These additional options are available for Microsoft Windows targets:
25678 specifies that a console application is to be generated, by
25679 instructing the linker to set the PE header subsystem type
25680 required for console applications.
25681 This option is available for Cygwin and MinGW targets and is
25682 enabled by default on those targets.
25686 This option is available for Cygwin and MinGW targets. It
25687 specifies that a DLL---a dynamic link library---is to be
25688 generated, enabling the selection of the required runtime
25689 startup object and entry point.
25691 @item -mnop-fun-dllimport
25692 @opindex mnop-fun-dllimport
25693 This option is available for Cygwin and MinGW targets. It
25694 specifies that the @code{dllimport} attribute should be ignored.
25698 This option is available for MinGW targets. It specifies
25699 that MinGW-specific thread support is to be used.
25703 This option is available for MinGW-w64 targets. It causes
25704 the @code{UNICODE} preprocessor macro to be predefined, and
25705 chooses Unicode-capable runtime startup code.
25709 This option is available for Cygwin and MinGW targets. It
25710 specifies that the typical Microsoft Windows predefined macros are to
25711 be set in the pre-processor, but does not influence the choice
25712 of runtime library/startup code.
25716 This option is available for Cygwin and MinGW targets. It
25717 specifies that a GUI application is to be generated by
25718 instructing the linker to set the PE header subsystem type
25721 @item -fno-set-stack-executable
25722 @opindex fno-set-stack-executable
25723 This option is available for MinGW targets. It specifies that
25724 the executable flag for the stack used by nested functions isn't
25725 set. This is necessary for binaries running in kernel mode of
25726 Microsoft Windows, as there the User32 API, which is used to set executable
25727 privileges, isn't available.
25729 @item -fwritable-relocated-rdata
25730 @opindex fno-writable-relocated-rdata
25731 This option is available for MinGW and Cygwin targets. It specifies
25732 that relocated-data in read-only section is put into the @code{.data}
25733 section. This is a necessary for older runtimes not supporting
25734 modification of @code{.rdata} sections for pseudo-relocation.
25736 @item -mpe-aligned-commons
25737 @opindex mpe-aligned-commons
25738 This option is available for Cygwin and MinGW targets. It
25739 specifies that the GNU extension to the PE file format that
25740 permits the correct alignment of COMMON variables should be
25741 used when generating code. It is enabled by default if
25742 GCC detects that the target assembler found during configuration
25743 supports the feature.
25746 See also under @ref{x86 Options} for standard options.
25748 @node Xstormy16 Options
25749 @subsection Xstormy16 Options
25750 @cindex Xstormy16 Options
25752 These options are defined for Xstormy16:
25757 Choose startup files and linker script suitable for the simulator.
25760 @node Xtensa Options
25761 @subsection Xtensa Options
25762 @cindex Xtensa Options
25764 These options are supported for Xtensa targets:
25768 @itemx -mno-const16
25770 @opindex mno-const16
25771 Enable or disable use of @code{CONST16} instructions for loading
25772 constant values. The @code{CONST16} instruction is currently not a
25773 standard option from Tensilica. When enabled, @code{CONST16}
25774 instructions are always used in place of the standard @code{L32R}
25775 instructions. The use of @code{CONST16} is enabled by default only if
25776 the @code{L32R} instruction is not available.
25779 @itemx -mno-fused-madd
25780 @opindex mfused-madd
25781 @opindex mno-fused-madd
25782 Enable or disable use of fused multiply/add and multiply/subtract
25783 instructions in the floating-point option. This has no effect if the
25784 floating-point option is not also enabled. Disabling fused multiply/add
25785 and multiply/subtract instructions forces the compiler to use separate
25786 instructions for the multiply and add/subtract operations. This may be
25787 desirable in some cases where strict IEEE 754-compliant results are
25788 required: the fused multiply add/subtract instructions do not round the
25789 intermediate result, thereby producing results with @emph{more} bits of
25790 precision than specified by the IEEE standard. Disabling fused multiply
25791 add/subtract instructions also ensures that the program output is not
25792 sensitive to the compiler's ability to combine multiply and add/subtract
25795 @item -mserialize-volatile
25796 @itemx -mno-serialize-volatile
25797 @opindex mserialize-volatile
25798 @opindex mno-serialize-volatile
25799 When this option is enabled, GCC inserts @code{MEMW} instructions before
25800 @code{volatile} memory references to guarantee sequential consistency.
25801 The default is @option{-mserialize-volatile}. Use
25802 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
25804 @item -mforce-no-pic
25805 @opindex mforce-no-pic
25806 For targets, like GNU/Linux, where all user-mode Xtensa code must be
25807 position-independent code (PIC), this option disables PIC for compiling
25810 @item -mtext-section-literals
25811 @itemx -mno-text-section-literals
25812 @opindex mtext-section-literals
25813 @opindex mno-text-section-literals
25814 These options control the treatment of literal pools. The default is
25815 @option{-mno-text-section-literals}, which places literals in a separate
25816 section in the output file. This allows the literal pool to be placed
25817 in a data RAM/ROM, and it also allows the linker to combine literal
25818 pools from separate object files to remove redundant literals and
25819 improve code size. With @option{-mtext-section-literals}, the literals
25820 are interspersed in the text section in order to keep them as close as
25821 possible to their references. This may be necessary for large assembly
25822 files. Literals for each function are placed right before that function.
25824 @item -mauto-litpools
25825 @itemx -mno-auto-litpools
25826 @opindex mauto-litpools
25827 @opindex mno-auto-litpools
25828 These options control the treatment of literal pools. The default is
25829 @option{-mno-auto-litpools}, which places literals in a separate
25830 section in the output file unless @option{-mtext-section-literals} is
25831 used. With @option{-mauto-litpools} the literals are interspersed in
25832 the text section by the assembler. Compiler does not produce explicit
25833 @code{.literal} directives and loads literals into registers with
25834 @code{MOVI} instructions instead of @code{L32R} to let the assembler
25835 do relaxation and place literals as necessary. This option allows
25836 assembler to create several literal pools per function and assemble
25837 very big functions, which may not be possible with
25838 @option{-mtext-section-literals}.
25840 @item -mtarget-align
25841 @itemx -mno-target-align
25842 @opindex mtarget-align
25843 @opindex mno-target-align
25844 When this option is enabled, GCC instructs the assembler to
25845 automatically align instructions to reduce branch penalties at the
25846 expense of some code density. The assembler attempts to widen density
25847 instructions to align branch targets and the instructions following call
25848 instructions. If there are not enough preceding safe density
25849 instructions to align a target, no widening is performed. The
25850 default is @option{-mtarget-align}. These options do not affect the
25851 treatment of auto-aligned instructions like @code{LOOP}, which the
25852 assembler always aligns, either by widening density instructions or
25853 by inserting NOP instructions.
25856 @itemx -mno-longcalls
25857 @opindex mlongcalls
25858 @opindex mno-longcalls
25859 When this option is enabled, GCC instructs the assembler to translate
25860 direct calls to indirect calls unless it can determine that the target
25861 of a direct call is in the range allowed by the call instruction. This
25862 translation typically occurs for calls to functions in other source
25863 files. Specifically, the assembler translates a direct @code{CALL}
25864 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
25865 The default is @option{-mno-longcalls}. This option should be used in
25866 programs where the call target can potentially be out of range. This
25867 option is implemented in the assembler, not the compiler, so the
25868 assembly code generated by GCC still shows direct call
25869 instructions---look at the disassembled object code to see the actual
25870 instructions. Note that the assembler uses an indirect call for
25871 every cross-file call, not just those that really are out of range.
25874 @node zSeries Options
25875 @subsection zSeries Options
25876 @cindex zSeries options
25878 These are listed under @xref{S/390 and zSeries Options}.
25884 @section Specifying Subprocesses and the Switches to Pass to Them
25887 @command{gcc} is a driver program. It performs its job by invoking a
25888 sequence of other programs to do the work of compiling, assembling and
25889 linking. GCC interprets its command-line parameters and uses these to
25890 deduce which programs it should invoke, and which command-line options
25891 it ought to place on their command lines. This behavior is controlled
25892 by @dfn{spec strings}. In most cases there is one spec string for each
25893 program that GCC can invoke, but a few programs have multiple spec
25894 strings to control their behavior. The spec strings built into GCC can
25895 be overridden by using the @option{-specs=} command-line switch to specify
25898 @dfn{Spec files} are plain-text files that are used to construct spec
25899 strings. They consist of a sequence of directives separated by blank
25900 lines. The type of directive is determined by the first non-whitespace
25901 character on the line, which can be one of the following:
25904 @item %@var{command}
25905 Issues a @var{command} to the spec file processor. The commands that can
25909 @item %include <@var{file}>
25910 @cindex @code{%include}
25911 Search for @var{file} and insert its text at the current point in the
25914 @item %include_noerr <@var{file}>
25915 @cindex @code{%include_noerr}
25916 Just like @samp{%include}, but do not generate an error message if the include
25917 file cannot be found.
25919 @item %rename @var{old_name} @var{new_name}
25920 @cindex @code{%rename}
25921 Rename the spec string @var{old_name} to @var{new_name}.
25925 @item *[@var{spec_name}]:
25926 This tells the compiler to create, override or delete the named spec
25927 string. All lines after this directive up to the next directive or
25928 blank line are considered to be the text for the spec string. If this
25929 results in an empty string then the spec is deleted. (Or, if the
25930 spec did not exist, then nothing happens.) Otherwise, if the spec
25931 does not currently exist a new spec is created. If the spec does
25932 exist then its contents are overridden by the text of this
25933 directive, unless the first character of that text is the @samp{+}
25934 character, in which case the text is appended to the spec.
25936 @item [@var{suffix}]:
25937 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
25938 and up to the next directive or blank line are considered to make up the
25939 spec string for the indicated suffix. When the compiler encounters an
25940 input file with the named suffix, it processes the spec string in
25941 order to work out how to compile that file. For example:
25945 z-compile -input %i
25948 This says that any input file whose name ends in @samp{.ZZ} should be
25949 passed to the program @samp{z-compile}, which should be invoked with the
25950 command-line switch @option{-input} and with the result of performing the
25951 @samp{%i} substitution. (See below.)
25953 As an alternative to providing a spec string, the text following a
25954 suffix directive can be one of the following:
25957 @item @@@var{language}
25958 This says that the suffix is an alias for a known @var{language}. This is
25959 similar to using the @option{-x} command-line switch to GCC to specify a
25960 language explicitly. For example:
25967 Says that .ZZ files are, in fact, C++ source files.
25970 This causes an error messages saying:
25973 @var{name} compiler not installed on this system.
25977 GCC already has an extensive list of suffixes built into it.
25978 This directive adds an entry to the end of the list of suffixes, but
25979 since the list is searched from the end backwards, it is effectively
25980 possible to override earlier entries using this technique.
25984 GCC has the following spec strings built into it. Spec files can
25985 override these strings or create their own. Note that individual
25986 targets can also add their own spec strings to this list.
25989 asm Options to pass to the assembler
25990 asm_final Options to pass to the assembler post-processor
25991 cpp Options to pass to the C preprocessor
25992 cc1 Options to pass to the C compiler
25993 cc1plus Options to pass to the C++ compiler
25994 endfile Object files to include at the end of the link
25995 link Options to pass to the linker
25996 lib Libraries to include on the command line to the linker
25997 libgcc Decides which GCC support library to pass to the linker
25998 linker Sets the name of the linker
25999 predefines Defines to be passed to the C preprocessor
26000 signed_char Defines to pass to CPP to say whether @code{char} is signed
26002 startfile Object files to include at the start of the link
26005 Here is a small example of a spec file:
26008 %rename lib old_lib
26011 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
26014 This example renames the spec called @samp{lib} to @samp{old_lib} and
26015 then overrides the previous definition of @samp{lib} with a new one.
26016 The new definition adds in some extra command-line options before
26017 including the text of the old definition.
26019 @dfn{Spec strings} are a list of command-line options to be passed to their
26020 corresponding program. In addition, the spec strings can contain
26021 @samp{%}-prefixed sequences to substitute variable text or to
26022 conditionally insert text into the command line. Using these constructs
26023 it is possible to generate quite complex command lines.
26025 Here is a table of all defined @samp{%}-sequences for spec
26026 strings. Note that spaces are not generated automatically around the
26027 results of expanding these sequences. Therefore you can concatenate them
26028 together or combine them with constant text in a single argument.
26032 Substitute one @samp{%} into the program name or argument.
26035 Substitute the name of the input file being processed.
26038 Substitute the basename of the input file being processed.
26039 This is the substring up to (and not including) the last period
26040 and not including the directory.
26043 This is the same as @samp{%b}, but include the file suffix (text after
26047 Marks the argument containing or following the @samp{%d} as a
26048 temporary file name, so that that file is deleted if GCC exits
26049 successfully. Unlike @samp{%g}, this contributes no text to the
26052 @item %g@var{suffix}
26053 Substitute a file name that has suffix @var{suffix} and is chosen
26054 once per compilation, and mark the argument in the same way as
26055 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
26056 name is now chosen in a way that is hard to predict even when previously
26057 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
26058 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
26059 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
26060 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
26061 was simply substituted with a file name chosen once per compilation,
26062 without regard to any appended suffix (which was therefore treated
26063 just like ordinary text), making such attacks more likely to succeed.
26065 @item %u@var{suffix}
26066 Like @samp{%g}, but generates a new temporary file name
26067 each time it appears instead of once per compilation.
26069 @item %U@var{suffix}
26070 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
26071 new one if there is no such last file name. In the absence of any
26072 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
26073 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
26074 involves the generation of two distinct file names, one
26075 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
26076 simply substituted with a file name chosen for the previous @samp{%u},
26077 without regard to any appended suffix.
26079 @item %j@var{suffix}
26080 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
26081 writable, and if @option{-save-temps} is not used;
26082 otherwise, substitute the name
26083 of a temporary file, just like @samp{%u}. This temporary file is not
26084 meant for communication between processes, but rather as a junk
26085 disposal mechanism.
26087 @item %|@var{suffix}
26088 @itemx %m@var{suffix}
26089 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
26090 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
26091 all. These are the two most common ways to instruct a program that it
26092 should read from standard input or write to standard output. If you
26093 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
26094 construct: see for example @file{f/lang-specs.h}.
26096 @item %.@var{SUFFIX}
26097 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
26098 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
26099 terminated by the next space or %.
26102 Marks the argument containing or following the @samp{%w} as the
26103 designated output file of this compilation. This puts the argument
26104 into the sequence of arguments that @samp{%o} substitutes.
26107 Substitutes the names of all the output files, with spaces
26108 automatically placed around them. You should write spaces
26109 around the @samp{%o} as well or the results are undefined.
26110 @samp{%o} is for use in the specs for running the linker.
26111 Input files whose names have no recognized suffix are not compiled
26112 at all, but they are included among the output files, so they are
26116 Substitutes the suffix for object files. Note that this is
26117 handled specially when it immediately follows @samp{%g, %u, or %U},
26118 because of the need for those to form complete file names. The
26119 handling is such that @samp{%O} is treated exactly as if it had already
26120 been substituted, except that @samp{%g, %u, and %U} do not currently
26121 support additional @var{suffix} characters following @samp{%O} as they do
26122 following, for example, @samp{.o}.
26125 Substitutes the standard macro predefinitions for the
26126 current target machine. Use this when running @command{cpp}.
26129 Like @samp{%p}, but puts @samp{__} before and after the name of each
26130 predefined macro, except for macros that start with @samp{__} or with
26131 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
26135 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
26136 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
26137 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
26138 and @option{-imultilib} as necessary.
26141 Current argument is the name of a library or startup file of some sort.
26142 Search for that file in a standard list of directories and substitute
26143 the full name found. The current working directory is included in the
26144 list of directories scanned.
26147 Current argument is the name of a linker script. Search for that file
26148 in the current list of directories to scan for libraries. If the file
26149 is located insert a @option{--script} option into the command line
26150 followed by the full path name found. If the file is not found then
26151 generate an error message. Note: the current working directory is not
26155 Print @var{str} as an error message. @var{str} is terminated by a newline.
26156 Use this when inconsistent options are detected.
26158 @item %(@var{name})
26159 Substitute the contents of spec string @var{name} at this point.
26161 @item %x@{@var{option}@}
26162 Accumulate an option for @samp{%X}.
26165 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
26169 Output the accumulated assembler options specified by @option{-Wa}.
26172 Output the accumulated preprocessor options specified by @option{-Wp}.
26175 Process the @code{asm} spec. This is used to compute the
26176 switches to be passed to the assembler.
26179 Process the @code{asm_final} spec. This is a spec string for
26180 passing switches to an assembler post-processor, if such a program is
26184 Process the @code{link} spec. This is the spec for computing the
26185 command line passed to the linker. Typically it makes use of the
26186 @samp{%L %G %S %D and %E} sequences.
26189 Dump out a @option{-L} option for each directory that GCC believes might
26190 contain startup files. If the target supports multilibs then the
26191 current multilib directory is prepended to each of these paths.
26194 Process the @code{lib} spec. This is a spec string for deciding which
26195 libraries are included on the command line to the linker.
26198 Process the @code{libgcc} spec. This is a spec string for deciding
26199 which GCC support library is included on the command line to the linker.
26202 Process the @code{startfile} spec. This is a spec for deciding which
26203 object files are the first ones passed to the linker. Typically
26204 this might be a file named @file{crt0.o}.
26207 Process the @code{endfile} spec. This is a spec string that specifies
26208 the last object files that are passed to the linker.
26211 Process the @code{cpp} spec. This is used to construct the arguments
26212 to be passed to the C preprocessor.
26215 Process the @code{cc1} spec. This is used to construct the options to be
26216 passed to the actual C compiler (@command{cc1}).
26219 Process the @code{cc1plus} spec. This is used to construct the options to be
26220 passed to the actual C++ compiler (@command{cc1plus}).
26223 Substitute the variable part of a matched option. See below.
26224 Note that each comma in the substituted string is replaced by
26228 Remove all occurrences of @code{-S} from the command line. Note---this
26229 command is position dependent. @samp{%} commands in the spec string
26230 before this one see @code{-S}, @samp{%} commands in the spec string
26231 after this one do not.
26233 @item %:@var{function}(@var{args})
26234 Call the named function @var{function}, passing it @var{args}.
26235 @var{args} is first processed as a nested spec string, then split
26236 into an argument vector in the usual fashion. The function returns
26237 a string which is processed as if it had appeared literally as part
26238 of the current spec.
26240 The following built-in spec functions are provided:
26243 @item @code{getenv}
26244 The @code{getenv} spec function takes two arguments: an environment
26245 variable name and a string. If the environment variable is not
26246 defined, a fatal error is issued. Otherwise, the return value is the
26247 value of the environment variable concatenated with the string. For
26248 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
26251 %:getenv(TOPDIR /include)
26254 expands to @file{/path/to/top/include}.
26256 @item @code{if-exists}
26257 The @code{if-exists} spec function takes one argument, an absolute
26258 pathname to a file. If the file exists, @code{if-exists} returns the
26259 pathname. Here is a small example of its usage:
26263 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
26266 @item @code{if-exists-else}
26267 The @code{if-exists-else} spec function is similar to the @code{if-exists}
26268 spec function, except that it takes two arguments. The first argument is
26269 an absolute pathname to a file. If the file exists, @code{if-exists-else}
26270 returns the pathname. If it does not exist, it returns the second argument.
26271 This way, @code{if-exists-else} can be used to select one file or another,
26272 based on the existence of the first. Here is a small example of its usage:
26276 crt0%O%s %:if-exists(crti%O%s) \
26277 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
26280 @item @code{replace-outfile}
26281 The @code{replace-outfile} spec function takes two arguments. It looks for the
26282 first argument in the outfiles array and replaces it with the second argument. Here
26283 is a small example of its usage:
26286 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
26289 @item @code{remove-outfile}
26290 The @code{remove-outfile} spec function takes one argument. It looks for the
26291 first argument in the outfiles array and removes it. Here is a small example
26295 %:remove-outfile(-lm)
26298 @item @code{pass-through-libs}
26299 The @code{pass-through-libs} spec function takes any number of arguments. It
26300 finds any @option{-l} options and any non-options ending in @file{.a} (which it
26301 assumes are the names of linker input library archive files) and returns a
26302 result containing all the found arguments each prepended by
26303 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
26304 intended to be passed to the LTO linker plugin.
26307 %:pass-through-libs(%G %L %G)
26310 @item @code{print-asm-header}
26311 The @code{print-asm-header} function takes no arguments and simply
26312 prints a banner like:
26318 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
26321 It is used to separate compiler options from assembler options
26322 in the @option{--target-help} output.
26325 @item %@{@code{S}@}
26326 Substitutes the @code{-S} switch, if that switch is given to GCC@.
26327 If that switch is not specified, this substitutes nothing. Note that
26328 the leading dash is omitted when specifying this option, and it is
26329 automatically inserted if the substitution is performed. Thus the spec
26330 string @samp{%@{foo@}} matches the command-line option @option{-foo}
26331 and outputs the command-line option @option{-foo}.
26333 @item %W@{@code{S}@}
26334 Like %@{@code{S}@} but mark last argument supplied within as a file to be
26335 deleted on failure.
26337 @item %@{@code{S}*@}
26338 Substitutes all the switches specified to GCC whose names start
26339 with @code{-S}, but which also take an argument. This is used for
26340 switches like @option{-o}, @option{-D}, @option{-I}, etc.
26341 GCC considers @option{-o foo} as being
26342 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
26343 text, including the space. Thus two arguments are generated.
26345 @item %@{@code{S}*&@code{T}*@}
26346 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
26347 (the order of @code{S} and @code{T} in the spec is not significant).
26348 There can be any number of ampersand-separated variables; for each the
26349 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
26351 @item %@{@code{S}:@code{X}@}
26352 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
26354 @item %@{!@code{S}:@code{X}@}
26355 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
26357 @item %@{@code{S}*:@code{X}@}
26358 Substitutes @code{X} if one or more switches whose names start with
26359 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
26360 once, no matter how many such switches appeared. However, if @code{%*}
26361 appears somewhere in @code{X}, then @code{X} is substituted once
26362 for each matching switch, with the @code{%*} replaced by the part of
26363 that switch matching the @code{*}.
26365 If @code{%*} appears as the last part of a spec sequence then a space
26366 is added after the end of the last substitution. If there is more
26367 text in the sequence, however, then a space is not generated. This
26368 allows the @code{%*} substitution to be used as part of a larger
26369 string. For example, a spec string like this:
26372 %@{mcu=*:--script=%*/memory.ld@}
26376 when matching an option like @option{-mcu=newchip} produces:
26379 --script=newchip/memory.ld
26382 @item %@{.@code{S}:@code{X}@}
26383 Substitutes @code{X}, if processing a file with suffix @code{S}.
26385 @item %@{!.@code{S}:@code{X}@}
26386 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
26388 @item %@{,@code{S}:@code{X}@}
26389 Substitutes @code{X}, if processing a file for language @code{S}.
26391 @item %@{!,@code{S}:@code{X}@}
26392 Substitutes @code{X}, if not processing a file for language @code{S}.
26394 @item %@{@code{S}|@code{P}:@code{X}@}
26395 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
26396 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
26397 @code{*} sequences as well, although they have a stronger binding than
26398 the @samp{|}. If @code{%*} appears in @code{X}, all of the
26399 alternatives must be starred, and only the first matching alternative
26402 For example, a spec string like this:
26405 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
26409 outputs the following command-line options from the following input
26410 command-line options:
26415 -d fred.c -foo -baz -boggle
26416 -d jim.d -bar -baz -boggle
26419 @item %@{S:X; T:Y; :D@}
26421 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
26422 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
26423 be as many clauses as you need. This may be combined with @code{.},
26424 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
26429 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
26430 construct may contain other nested @samp{%} constructs or spaces, or
26431 even newlines. They are processed as usual, as described above.
26432 Trailing white space in @code{X} is ignored. White space may also
26433 appear anywhere on the left side of the colon in these constructs,
26434 except between @code{.} or @code{*} and the corresponding word.
26436 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
26437 handled specifically in these constructs. If another value of
26438 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
26439 @option{-W} switch is found later in the command line, the earlier
26440 switch value is ignored, except with @{@code{S}*@} where @code{S} is
26441 just one letter, which passes all matching options.
26443 The character @samp{|} at the beginning of the predicate text is used to
26444 indicate that a command should be piped to the following command, but
26445 only if @option{-pipe} is specified.
26447 It is built into GCC which switches take arguments and which do not.
26448 (You might think it would be useful to generalize this to allow each
26449 compiler's spec to say which switches take arguments. But this cannot
26450 be done in a consistent fashion. GCC cannot even decide which input
26451 files have been specified without knowing which switches take arguments,
26452 and it must know which input files to compile in order to tell which
26455 GCC also knows implicitly that arguments starting in @option{-l} are to be
26456 treated as compiler output files, and passed to the linker in their
26457 proper position among the other output files.
26459 @node Environment Variables
26460 @section Environment Variables Affecting GCC
26461 @cindex environment variables
26463 @c man begin ENVIRONMENT
26464 This section describes several environment variables that affect how GCC
26465 operates. Some of them work by specifying directories or prefixes to use
26466 when searching for various kinds of files. Some are used to specify other
26467 aspects of the compilation environment.
26469 Note that you can also specify places to search using options such as
26470 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
26471 take precedence over places specified using environment variables, which
26472 in turn take precedence over those specified by the configuration of GCC@.
26473 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
26474 GNU Compiler Collection (GCC) Internals}.
26479 @c @itemx LC_COLLATE
26481 @c @itemx LC_MONETARY
26482 @c @itemx LC_NUMERIC
26487 @c @findex LC_COLLATE
26488 @findex LC_MESSAGES
26489 @c @findex LC_MONETARY
26490 @c @findex LC_NUMERIC
26494 These environment variables control the way that GCC uses
26495 localization information which allows GCC to work with different
26496 national conventions. GCC inspects the locale categories
26497 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
26498 so. These locale categories can be set to any value supported by your
26499 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
26500 Kingdom encoded in UTF-8.
26502 The @env{LC_CTYPE} environment variable specifies character
26503 classification. GCC uses it to determine the character boundaries in
26504 a string; this is needed for some multibyte encodings that contain quote
26505 and escape characters that are otherwise interpreted as a string
26508 The @env{LC_MESSAGES} environment variable specifies the language to
26509 use in diagnostic messages.
26511 If the @env{LC_ALL} environment variable is set, it overrides the value
26512 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
26513 and @env{LC_MESSAGES} default to the value of the @env{LANG}
26514 environment variable. If none of these variables are set, GCC
26515 defaults to traditional C English behavior.
26519 If @env{TMPDIR} is set, it specifies the directory to use for temporary
26520 files. GCC uses temporary files to hold the output of one stage of
26521 compilation which is to be used as input to the next stage: for example,
26522 the output of the preprocessor, which is the input to the compiler
26525 @item GCC_COMPARE_DEBUG
26526 @findex GCC_COMPARE_DEBUG
26527 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
26528 @option{-fcompare-debug} to the compiler driver. See the documentation
26529 of this option for more details.
26531 @item GCC_EXEC_PREFIX
26532 @findex GCC_EXEC_PREFIX
26533 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
26534 names of the subprograms executed by the compiler. No slash is added
26535 when this prefix is combined with the name of a subprogram, but you can
26536 specify a prefix that ends with a slash if you wish.
26538 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
26539 an appropriate prefix to use based on the pathname it is invoked with.
26541 If GCC cannot find the subprogram using the specified prefix, it
26542 tries looking in the usual places for the subprogram.
26544 The default value of @env{GCC_EXEC_PREFIX} is
26545 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
26546 the installed compiler. In many cases @var{prefix} is the value
26547 of @code{prefix} when you ran the @file{configure} script.
26549 Other prefixes specified with @option{-B} take precedence over this prefix.
26551 This prefix is also used for finding files such as @file{crt0.o} that are
26554 In addition, the prefix is used in an unusual way in finding the
26555 directories to search for header files. For each of the standard
26556 directories whose name normally begins with @samp{/usr/local/lib/gcc}
26557 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
26558 replacing that beginning with the specified prefix to produce an
26559 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
26560 @file{foo/bar} just before it searches the standard directory
26561 @file{/usr/local/lib/bar}.
26562 If a standard directory begins with the configured
26563 @var{prefix} then the value of @var{prefix} is replaced by
26564 @env{GCC_EXEC_PREFIX} when looking for header files.
26566 @item COMPILER_PATH
26567 @findex COMPILER_PATH
26568 The value of @env{COMPILER_PATH} is a colon-separated list of
26569 directories, much like @env{PATH}. GCC tries the directories thus
26570 specified when searching for subprograms, if it can't find the
26571 subprograms using @env{GCC_EXEC_PREFIX}.
26574 @findex LIBRARY_PATH
26575 The value of @env{LIBRARY_PATH} is a colon-separated list of
26576 directories, much like @env{PATH}. When configured as a native compiler,
26577 GCC tries the directories thus specified when searching for special
26578 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
26579 using GCC also uses these directories when searching for ordinary
26580 libraries for the @option{-l} option (but directories specified with
26581 @option{-L} come first).
26585 @cindex locale definition
26586 This variable is used to pass locale information to the compiler. One way in
26587 which this information is used is to determine the character set to be used
26588 when character literals, string literals and comments are parsed in C and C++.
26589 When the compiler is configured to allow multibyte characters,
26590 the following values for @env{LANG} are recognized:
26594 Recognize JIS characters.
26596 Recognize SJIS characters.
26598 Recognize EUCJP characters.
26601 If @env{LANG} is not defined, or if it has some other value, then the
26602 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
26603 recognize and translate multibyte characters.
26607 Some additional environment variables affect the behavior of the
26610 @include cppenv.texi
26614 @node Precompiled Headers
26615 @section Using Precompiled Headers
26616 @cindex precompiled headers
26617 @cindex speed of compilation
26619 Often large projects have many header files that are included in every
26620 source file. The time the compiler takes to process these header files
26621 over and over again can account for nearly all of the time required to
26622 build the project. To make builds faster, GCC allows you to
26623 @dfn{precompile} a header file.
26625 To create a precompiled header file, simply compile it as you would any
26626 other file, if necessary using the @option{-x} option to make the driver
26627 treat it as a C or C++ header file. You may want to use a
26628 tool like @command{make} to keep the precompiled header up-to-date when
26629 the headers it contains change.
26631 A precompiled header file is searched for when @code{#include} is
26632 seen in the compilation. As it searches for the included file
26633 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
26634 compiler looks for a precompiled header in each directory just before it
26635 looks for the include file in that directory. The name searched for is
26636 the name specified in the @code{#include} with @samp{.gch} appended. If
26637 the precompiled header file can't be used, it is ignored.
26639 For instance, if you have @code{#include "all.h"}, and you have
26640 @file{all.h.gch} in the same directory as @file{all.h}, then the
26641 precompiled header file is used if possible, and the original
26642 header is used otherwise.
26644 Alternatively, you might decide to put the precompiled header file in a
26645 directory and use @option{-I} to ensure that directory is searched
26646 before (or instead of) the directory containing the original header.
26647 Then, if you want to check that the precompiled header file is always
26648 used, you can put a file of the same name as the original header in this
26649 directory containing an @code{#error} command.
26651 This also works with @option{-include}. So yet another way to use
26652 precompiled headers, good for projects not designed with precompiled
26653 header files in mind, is to simply take most of the header files used by
26654 a project, include them from another header file, precompile that header
26655 file, and @option{-include} the precompiled header. If the header files
26656 have guards against multiple inclusion, they are skipped because
26657 they've already been included (in the precompiled header).
26659 If you need to precompile the same header file for different
26660 languages, targets, or compiler options, you can instead make a
26661 @emph{directory} named like @file{all.h.gch}, and put each precompiled
26662 header in the directory, perhaps using @option{-o}. It doesn't matter
26663 what you call the files in the directory; every precompiled header in
26664 the directory is considered. The first precompiled header
26665 encountered in the directory that is valid for this compilation is
26666 used; they're searched in no particular order.
26668 There are many other possibilities, limited only by your imagination,
26669 good sense, and the constraints of your build system.
26671 A precompiled header file can be used only when these conditions apply:
26675 Only one precompiled header can be used in a particular compilation.
26678 A precompiled header can't be used once the first C token is seen. You
26679 can have preprocessor directives before a precompiled header; you cannot
26680 include a precompiled header from inside another header.
26683 The precompiled header file must be produced for the same language as
26684 the current compilation. You can't use a C precompiled header for a C++
26688 The precompiled header file must have been produced by the same compiler
26689 binary as the current compilation is using.
26692 Any macros defined before the precompiled header is included must
26693 either be defined in the same way as when the precompiled header was
26694 generated, or must not affect the precompiled header, which usually
26695 means that they don't appear in the precompiled header at all.
26697 The @option{-D} option is one way to define a macro before a
26698 precompiled header is included; using a @code{#define} can also do it.
26699 There are also some options that define macros implicitly, like
26700 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
26703 @item If debugging information is output when using the precompiled
26704 header, using @option{-g} or similar, the same kind of debugging information
26705 must have been output when building the precompiled header. However,
26706 a precompiled header built using @option{-g} can be used in a compilation
26707 when no debugging information is being output.
26709 @item The same @option{-m} options must generally be used when building
26710 and using the precompiled header. @xref{Submodel Options},
26711 for any cases where this rule is relaxed.
26713 @item Each of the following options must be the same when building and using
26714 the precompiled header:
26716 @gccoptlist{-fexceptions}
26719 Some other command-line options starting with @option{-f},
26720 @option{-p}, or @option{-O} must be defined in the same way as when
26721 the precompiled header was generated. At present, it's not clear
26722 which options are safe to change and which are not; the safest choice
26723 is to use exactly the same options when generating and using the
26724 precompiled header. The following are known to be safe:
26726 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
26727 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
26728 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
26733 For all of these except the last, the compiler automatically
26734 ignores the precompiled header if the conditions aren't met. If you
26735 find an option combination that doesn't work and doesn't cause the
26736 precompiled header to be ignored, please consider filing a bug report,
26739 If you do use differing options when generating and using the
26740 precompiled header, the actual behavior is a mixture of the
26741 behavior for the options. For instance, if you use @option{-g} to
26742 generate the precompiled header but not when using it, you may or may
26743 not get debugging information for routines in the precompiled header.