1 @c Copyright (C) 1988-2018 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{} 1988-2018 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), dbx(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.
124 Some options take one or more arguments typically separated either
125 by a space or by the equals sign (@samp{=}) from the option name.
126 Unless documented otherwise, an argument can be either numeric or
127 a string. Numeric arguments must typically be small unsigned decimal
128 or hexadecimal integers. Hexadecimal arguments must begin with
129 the @samp{0x} prefix. Arguments to options that specify a size
130 threshold of some sort may be arbitrarily large decimal or hexadecimal
131 integers followed by a byte size suffix designating a multiple of bytes
132 such as @code{kB} and @code{KiB} for kilobyte and kibibyte, respectively,
133 @code{MB} and @code{MiB} for megabyte and mebibyte, @code{GB} and
134 @code{GiB} for gigabyte and gigibyte, and so on. Such arguments are
135 designated by @var{byte-size} in the following text. Refer to the NIST,
136 IEC, and other relevant national and international standards for the full
137 listing and explanation of the binary and decimal byte size prefixes.
141 @xref{Option Index}, for an index to GCC's options.
144 * Option Summary:: Brief list of all options, without explanations.
145 * Overall Options:: Controlling the kind of output:
146 an executable, object files, assembler files,
147 or preprocessed source.
148 * Invoking G++:: Compiling C++ programs.
149 * C Dialect Options:: Controlling the variant of C language compiled.
150 * C++ Dialect Options:: Variations on C++.
151 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
153 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
155 * Warning Options:: How picky should the compiler be?
156 * Debugging Options:: Producing debuggable code.
157 * Optimize Options:: How much optimization?
158 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
159 * Preprocessor Options:: Controlling header files and macro definitions.
160 Also, getting dependency information for Make.
161 * Assembler Options:: Passing options to the assembler.
162 * Link Options:: Specifying libraries and so on.
163 * Directory Options:: Where to find header files and libraries.
164 Where to find the compiler executable files.
165 * Code Gen Options:: Specifying conventions for function calls, data layout
167 * Developer Options:: Printing GCC configuration info, statistics, and
169 * Submodel Options:: Target-specific options, such as compiling for a
170 specific processor variant.
171 * Spec Files:: How to pass switches to sub-processes.
172 * Environment Variables:: Env vars that affect GCC.
173 * Precompiled Headers:: Compiling a header once, and using it many times.
179 @section Option Summary
181 Here is a summary of all the options, grouped by type. Explanations are
182 in the following sections.
185 @item Overall Options
186 @xref{Overall Options,,Options Controlling the Kind of Output}.
187 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
188 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
189 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
190 @@@var{file} -ffile-prefix-map=@var{old}=@var{new} @gol
191 -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
192 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
194 @item C Language Options
195 @xref{C Dialect Options,,Options Controlling C Dialect}.
196 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
197 -fpermitted-flt-eval-methods=@var{standard} @gol
198 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
199 -fno-asm -fno-builtin -fno-builtin-@var{function} -fgimple@gol
200 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
201 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
202 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
203 -fsigned-bitfields -fsigned-char @gol
204 -funsigned-bitfields -funsigned-char}
206 @item C++ Language Options
207 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
208 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
209 -faligned-new=@var{n} -fargs-in-order=@var{n} -fcheck-new @gol
210 -fconstexpr-depth=@var{n} -fconstexpr-loop-limit=@var{n} @gol
211 -fno-elide-constructors @gol
212 -fno-enforce-eh-specs @gol
213 -fno-gnu-keywords @gol
214 -fno-implicit-templates @gol
215 -fno-implicit-inline-templates @gol
216 -fno-implement-inlines -fms-extensions @gol
217 -fnew-inheriting-ctors @gol
218 -fnew-ttp-matching @gol
219 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
220 -fno-optional-diags -fpermissive @gol
221 -fno-pretty-templates @gol
222 -frepo -fno-rtti -fsized-deallocation @gol
223 -ftemplate-backtrace-limit=@var{n} @gol
224 -ftemplate-depth=@var{n} @gol
225 -fno-threadsafe-statics -fuse-cxa-atexit @gol
226 -fno-weak -nostdinc++ @gol
227 -fvisibility-inlines-hidden @gol
228 -fvisibility-ms-compat @gol
229 -fext-numeric-literals @gol
230 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
231 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wliteral-suffix @gol
232 -Wmultiple-inheritance -Wno-init-list-lifetime @gol
233 -Wnamespaces -Wnarrowing @gol
234 -Wpessimizing-move -Wredundant-move @gol
235 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
236 -Wnon-virtual-dtor -Wreorder -Wregister @gol
237 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
238 -Wno-non-template-friend -Wold-style-cast @gol
239 -Woverloaded-virtual -Wno-pmf-conversions @gol
240 -Wno-class-conversion -Wno-terminate @gol
241 -Wsign-promo -Wvirtual-inheritance}
243 @item Objective-C and Objective-C++ Language Options
244 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
245 Objective-C and Objective-C++ Dialects}.
246 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
247 -fgnu-runtime -fnext-runtime @gol
248 -fno-nil-receivers @gol
249 -fobjc-abi-version=@var{n} @gol
250 -fobjc-call-cxx-cdtors @gol
251 -fobjc-direct-dispatch @gol
252 -fobjc-exceptions @gol
255 -fobjc-std=objc1 @gol
256 -fno-local-ivars @gol
257 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
258 -freplace-objc-classes @gol
261 -Wassign-intercept @gol
262 -Wno-protocol -Wselector @gol
263 -Wstrict-selector-match @gol
264 -Wundeclared-selector}
266 @item Diagnostic Message Formatting Options
267 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
268 @gccoptlist{-fmessage-length=@var{n} @gol
269 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
270 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
271 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
272 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol
273 -fdiagnostics-minimum-margin-width=@var{width} @gol
274 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
275 -fdiagnostics-show-template-tree -fno-elide-type @gol
278 @item Warning Options
279 @xref{Warning Options,,Options to Request or Suppress Warnings}.
280 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
281 -pedantic-errors @gol
282 -w -Wextra -Wall -Waddress -Waggregate-return -Waligned-new @gol
283 -Walloc-zero -Walloc-size-larger-than=@var{byte-size}
284 -Walloca -Walloca-larger-than=@var{byte-size} @gol
285 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
286 -Wno-attributes -Wbool-compare -Wbool-operation @gol
287 -Wno-builtin-declaration-mismatch @gol
288 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
289 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol
290 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
291 -Wchar-subscripts -Wcatch-value -Wcatch-value=@var{n} @gol
292 -Wclobbered -Wcomment -Wconditionally-supported @gol
293 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
294 -Wdelete-incomplete @gol
295 -Wno-attribute-warning @gol
296 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
297 -Wdisabled-optimization @gol
298 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
299 -Wno-div-by-zero -Wdouble-promotion @gol
300 -Wduplicated-branches -Wduplicated-cond @gol
301 -Wempty-body -Wenum-compare -Wno-endif-labels -Wexpansion-to-defined @gol
302 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
303 -Wfloat-equal -Wformat -Wformat=2 @gol
304 -Wno-format-contains-nul -Wno-format-extra-args @gol
305 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
306 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
307 -Wformat-y2k -Wframe-address @gol
308 -Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol
309 -Wjump-misses-init @gol
310 -Wif-not-aligned @gol
311 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
312 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
313 -Wimplicit-function-declaration -Wimplicit-int @gol
314 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
315 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
316 -Winvalid-pch -Wlarger-than=@var{byte-size} @gol
317 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
318 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
319 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
320 -Wmissing-field-initializers -Wmissing-include-dirs -Wmissing-profile @gol
321 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
322 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
323 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
324 -Woverride-init-side-effects -Woverlength-strings @gol
325 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
326 -Wparentheses -Wno-pedantic-ms-format @gol
327 -Wplacement-new -Wplacement-new=@var{n} @gol
328 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
329 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol
330 -Wrestrict -Wno-return-local-addr @gol
331 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
332 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
333 -Wshift-overflow -Wshift-overflow=@var{n} @gol
334 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
335 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
336 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
337 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
338 -Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol
339 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
340 -Wstringop-overflow=@var{n} -Wstringop-truncation @gol
341 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
342 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
343 -Wmissing-format-attribute -Wsubobject-linkage @gol
344 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
345 -Wswitch-unreachable -Wsync-nand @gol
346 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
347 -Wtype-limits -Wundef @gol
348 -Wuninitialized -Wunknown-pragmas @gol
349 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
350 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
351 -Wunused-parameter -Wno-unused-result @gol
352 -Wunused-value -Wunused-variable @gol
353 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
354 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
355 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
356 -Wvla -Wvla-larger-than=@var{byte-size} -Wvolatile-register-var @gol
358 -Wzero-as-null-pointer-constant -Whsa}
360 @item C and Objective-C-only Warning Options
361 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
362 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
363 -Wold-style-declaration -Wold-style-definition @gol
364 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
365 -Wdeclaration-after-statement -Wpointer-sign}
367 @item Debugging Options
368 @xref{Debugging Options,,Options for Debugging Your Program}.
369 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
370 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
371 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
372 -gas-loc-support -gno-as-loc-support @gol
373 -gas-locview-support -gno-as-locview-support @gol
374 -gcolumn-info -gno-column-info @gol
375 -gstatement-frontiers -gno-statement-frontiers @gol
376 -gvariable-location-views -gno-variable-location-views @gol
377 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
378 -ginline-points -gno-inline-points @gol
379 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
380 -gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol
381 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
382 -fno-eliminate-unused-debug-types @gol
383 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
384 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
385 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
386 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
387 -fvar-tracking -fvar-tracking-assignments}
389 @item Optimization Options
390 @xref{Optimize Options,,Options that Control Optimization}.
391 @gccoptlist{-faggressive-loop-optimizations @gol
392 -falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
393 -falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
394 -falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
395 -falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
396 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
397 -fauto-inc-dec -fbranch-probabilities @gol
398 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
399 -fbtr-bb-exclusive -fcaller-saves @gol
400 -fcombine-stack-adjustments -fconserve-stack @gol
401 -fcompare-elim -fcprop-registers -fcrossjumping @gol
402 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
403 -fcx-limited-range @gol
404 -fdata-sections -fdce -fdelayed-branch @gol
405 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
406 -fdevirtualize-at-ltrans -fdse @gol
407 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
408 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
409 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
410 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
411 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
412 -fif-conversion2 -findirect-inlining @gol
413 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
414 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
415 -fipa-bit-cp -fipa-vrp @gol
416 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-reference-addressable @gol
417 -fipa-icf -fira-algorithm=@var{algorithm} @gol
418 -fira-region=@var{region} -fira-hoist-pressure @gol
419 -fira-loop-pressure -fno-ira-share-save-slots @gol
420 -fno-ira-share-spill-slots @gol
421 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
422 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
423 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
424 -floop-block -floop-interchange -floop-strip-mine @gol
425 -floop-unroll-and-jam -floop-nest-optimize @gol
426 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
427 -flto-partition=@var{alg} -fmerge-all-constants @gol
428 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
429 -fmove-loop-invariants -fno-branch-count-reg @gol
430 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
431 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
432 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
433 -fno-sched-spec -fno-signed-zeros @gol
434 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
435 -fomit-frame-pointer -foptimize-sibling-calls @gol
436 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
437 -fprefetch-loop-arrays @gol
438 -fprofile-correction @gol
439 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
440 -fprofile-reorder-functions @gol
441 -freciprocal-math -free -frename-registers -freorder-blocks @gol
442 -freorder-blocks-algorithm=@var{algorithm} @gol
443 -freorder-blocks-and-partition -freorder-functions @gol
444 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
445 -frounding-math -fsave-optimization-record @gol
446 -fsched2-use-superblocks -fsched-pressure @gol
447 -fsched-spec-load -fsched-spec-load-dangerous @gol
448 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
449 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
450 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
451 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
452 -fschedule-fusion @gol
453 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
454 -fselective-scheduling -fselective-scheduling2 @gol
455 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
456 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
457 -fsignaling-nans @gol
458 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
460 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
461 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
462 -fthread-jumps -ftracer -ftree-bit-ccp @gol
463 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
464 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
465 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
466 -ftree-loop-if-convert -ftree-loop-im @gol
467 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
468 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
469 -ftree-loop-vectorize @gol
470 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
471 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol
472 -ftree-switch-conversion -ftree-tail-merge @gol
473 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
474 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
475 -funsafe-math-optimizations -funswitch-loops @gol
476 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
477 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
478 --param @var{name}=@var{value}
479 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
481 @item Program Instrumentation Options
482 @xref{Instrumentation Options,,Program Instrumentation Options}.
483 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
484 -fprofile-abs-path @gol
485 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
486 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
487 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
488 -fsanitize-undefined-trap-on-error -fbounds-check @gol
489 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
490 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
491 -fstack-protector-explicit -fstack-check @gol
492 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
493 -fno-stack-limit -fsplit-stack @gol
494 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
495 -fvtv-counts -fvtv-debug @gol
496 -finstrument-functions @gol
497 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
498 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
500 @item Preprocessor Options
501 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
502 @gccoptlist{-A@var{question}=@var{answer} @gol
503 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
504 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
505 -dD -dI -dM -dN -dU @gol
506 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
507 -fexec-charset=@var{charset} -fextended-identifiers @gol
508 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
509 -fno-canonical-system-headers @gol -fpch-deps -fpch-preprocess @gol
510 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
511 -fwide-exec-charset=@var{charset} -fworking-directory @gol
512 -H -imacros @var{file} -include @var{file} @gol
513 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
514 -no-integrated-cpp -P -pthread -remap @gol
515 -traditional -traditional-cpp -trigraphs @gol
516 -U@var{macro} -undef @gol
517 -Wp,@var{option} -Xpreprocessor @var{option}}
519 @item Assembler Options
520 @xref{Assembler Options,,Passing Options to the Assembler}.
521 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
524 @xref{Link Options,,Options for Linking}.
525 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
526 -nostartfiles -nodefaultlibs -nolibc -nostdlib @gol
527 -e @var{entry} --entry=@var{entry} @gol
528 -pie -pthread -r -rdynamic @gol
529 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
530 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
531 -shared -shared-libgcc -symbolic @gol
532 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
533 -u @var{symbol} -z @var{keyword}}
535 @item Directory Options
536 @xref{Directory Options,,Options for Directory Search}.
537 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
538 -idirafter @var{dir} @gol
539 -imacros @var{file} -imultilib @var{dir} @gol
540 -iplugindir=@var{dir} -iprefix @var{file} @gol
541 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
542 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
543 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
544 -nostdinc -nostdinc++ --sysroot=@var{dir}}
546 @item Code Generation Options
547 @xref{Code Gen Options,,Options for Code Generation Conventions}.
548 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
549 -ffixed-@var{reg} -fexceptions @gol
550 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
551 -fasynchronous-unwind-tables @gol
553 -finhibit-size-directive -fno-common -fno-ident @gol
554 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
555 -fno-jump-tables @gol
556 -frecord-gcc-switches @gol
557 -freg-struct-return -fshort-enums -fshort-wchar @gol
558 -fverbose-asm -fpack-struct[=@var{n}] @gol
559 -fleading-underscore -ftls-model=@var{model} @gol
560 -fstack-reuse=@var{reuse_level} @gol
561 -ftrampolines -ftrapv -fwrapv @gol
562 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
563 -fstrict-volatile-bitfields -fsync-libcalls}
565 @item Developer Options
566 @xref{Developer Options,,GCC Developer Options}.
567 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
568 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
569 -fdbg-cnt=@var{counter-value-list} @gol
570 -fdisable-ipa-@var{pass_name} @gol
571 -fdisable-rtl-@var{pass_name} @gol
572 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
573 -fdisable-tree-@var{pass_name} @gol
574 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
575 -fdump-debug -fdump-earlydebug @gol
576 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
577 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
578 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
579 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
581 -fdump-lang-@var{switch} @gol
582 -fdump-lang-@var{switch}-@var{options} @gol
583 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
585 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
586 -fdump-statistics @gol
588 -fdump-tree-@var{switch} @gol
589 -fdump-tree-@var{switch}-@var{options} @gol
590 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
591 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
592 -fenable-@var{kind}-@var{pass} @gol
593 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
594 -fira-verbose=@var{n} @gol
595 -flto-report -flto-report-wpa -fmem-report-wpa @gol
596 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
597 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
598 -fprofile-report @gol
599 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
600 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
601 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
602 -fvar-tracking-assignments-toggle -gtoggle @gol
603 -print-file-name=@var{library} -print-libgcc-file-name @gol
604 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
605 -print-prog-name=@var{program} -print-search-dirs -Q @gol
606 -print-sysroot -print-sysroot-headers-suffix @gol
607 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
609 @item Machine-Dependent Options
610 @xref{Submodel Options,,Machine-Dependent Options}.
611 @c This list is ordered alphanumerically by subsection name.
612 @c Try and put the significant identifier (CPU or system) first,
613 @c so users have a clue at guessing where the ones they want will be.
615 @emph{AArch64 Options}
616 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
617 -mgeneral-regs-only @gol
618 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
619 -mstrict-align -mno-strict-align @gol
620 -momit-leaf-frame-pointer @gol
621 -mtls-dialect=desc -mtls-dialect=traditional @gol
622 -mtls-size=@var{size} @gol
623 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
624 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
625 -mpc-relative-literal-loads @gol
626 -msign-return-address=@var{scope} @gol
627 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
628 -moverride=@var{string} -mverbose-cost-dump -mtrack-speculation}
630 @emph{Adapteva Epiphany Options}
631 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
632 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
633 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
634 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
635 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
636 -msplit-vecmove-early -m1reg-@var{reg}}
639 @gccoptlist{-mbarrel-shifter -mjli-always @gol
640 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
641 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
642 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
643 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
644 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
645 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
646 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
647 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
648 -mvolatile-cache -mtp-regno=@var{regno} @gol
649 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
650 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
651 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
652 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
653 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
654 -mtune=@var{cpu} -mmultcost=@var{num} @gol
655 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
656 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index}
659 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
660 -mabi=@var{name} @gol
661 -mapcs-stack-check -mno-apcs-stack-check @gol
662 -mapcs-reentrant -mno-apcs-reentrant @gol
663 -msched-prolog -mno-sched-prolog @gol
664 -mlittle-endian -mbig-endian @gol
666 -mfloat-abi=@var{name} @gol
667 -mfp16-format=@var{name}
668 -mthumb-interwork -mno-thumb-interwork @gol
669 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
670 -mtune=@var{name} -mprint-tune-info @gol
671 -mstructure-size-boundary=@var{n} @gol
672 -mabort-on-noreturn @gol
673 -mlong-calls -mno-long-calls @gol
674 -msingle-pic-base -mno-single-pic-base @gol
675 -mpic-register=@var{reg} @gol
676 -mnop-fun-dllimport @gol
677 -mpoke-function-name @gol
678 -mthumb -marm -mflip-thumb @gol
679 -mtpcs-frame -mtpcs-leaf-frame @gol
680 -mcaller-super-interworking -mcallee-super-interworking @gol
681 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
682 -mword-relocations @gol
683 -mfix-cortex-m3-ldrd @gol
684 -munaligned-access @gol
685 -mneon-for-64bits @gol
686 -mslow-flash-data @gol
687 -masm-syntax-unified @gol
689 -mverbose-cost-dump @gol
694 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
695 -mbranch-cost=@var{cost} @gol
696 -mcall-prologues -mgas-isr-prologues -mint8 @gol
697 -mn_flash=@var{size} -mno-interrupts @gol
698 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
699 -mfract-convert-truncate @gol
700 -mshort-calls -nodevicelib @gol
701 -Waddr-space-convert -Wmisspelled-isr}
703 @emph{Blackfin Options}
704 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
705 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
706 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
707 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
708 -mno-id-shared-library -mshared-library-id=@var{n} @gol
709 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
710 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
711 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
715 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
716 -msim -msdata=@var{sdata-type}}
719 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
720 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
721 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
722 -mstack-align -mdata-align -mconst-align @gol
723 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
724 -melf -maout -melinux -mlinux -sim -sim2 @gol
725 -mmul-bug-workaround -mno-mul-bug-workaround}
728 @gccoptlist{-mmac @gol
729 -mcr16cplus -mcr16c @gol
730 -msim -mint32 -mbit-ops
731 -mdata-model=@var{model}}
734 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
735 -mbig-endian -EB -mlittle-endian -EL @gol
736 -mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
737 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
738 -mdsp -medsp -mvdsp @gol
739 -mdiv -msmart -mhigh-registers -manchor @gol
740 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
741 -mbranch-cost=@var{n} -mcse-cc -msched-prolog}
743 @emph{Darwin Options}
744 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
745 -arch_only -bind_at_load -bundle -bundle_loader @gol
746 -client_name -compatibility_version -current_version @gol
748 -dependency-file -dylib_file -dylinker_install_name @gol
749 -dynamic -dynamiclib -exported_symbols_list @gol
750 -filelist -flat_namespace -force_cpusubtype_ALL @gol
751 -force_flat_namespace -headerpad_max_install_names @gol
753 -image_base -init -install_name -keep_private_externs @gol
754 -multi_module -multiply_defined -multiply_defined_unused @gol
755 -noall_load -no_dead_strip_inits_and_terms @gol
756 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
757 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
758 -private_bundle -read_only_relocs -sectalign @gol
759 -sectobjectsymbols -whyload -seg1addr @gol
760 -sectcreate -sectobjectsymbols -sectorder @gol
761 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
762 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
763 -segprot -segs_read_only_addr -segs_read_write_addr @gol
764 -single_module -static -sub_library -sub_umbrella @gol
765 -twolevel_namespace -umbrella -undefined @gol
766 -unexported_symbols_list -weak_reference_mismatches @gol
767 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
768 -mkernel -mone-byte-bool}
770 @emph{DEC Alpha Options}
771 @gccoptlist{-mno-fp-regs -msoft-float @gol
772 -mieee -mieee-with-inexact -mieee-conformant @gol
773 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
774 -mtrap-precision=@var{mode} -mbuild-constants @gol
775 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
776 -mbwx -mmax -mfix -mcix @gol
777 -mfloat-vax -mfloat-ieee @gol
778 -mexplicit-relocs -msmall-data -mlarge-data @gol
779 -msmall-text -mlarge-text @gol
780 -mmemory-latency=@var{time}}
783 @gccoptlist{-msmall-model -mno-lsim}
786 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
789 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
790 -mhard-float -msoft-float @gol
791 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
792 -mdouble -mno-double @gol
793 -mmedia -mno-media -mmuladd -mno-muladd @gol
794 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
795 -mlinked-fp -mlong-calls -malign-labels @gol
796 -mlibrary-pic -macc-4 -macc-8 @gol
797 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
798 -moptimize-membar -mno-optimize-membar @gol
799 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
800 -mvliw-branch -mno-vliw-branch @gol
801 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
802 -mno-nested-cond-exec -mtomcat-stats @gol
806 @emph{GNU/Linux Options}
807 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
808 -tno-android-cc -tno-android-ld}
810 @emph{H8/300 Options}
811 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
814 @gccoptlist{-march=@var{architecture-type} @gol
815 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
816 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
817 -mfixed-range=@var{register-range} @gol
818 -mjump-in-delay -mlinker-opt -mlong-calls @gol
819 -mlong-load-store -mno-disable-fpregs @gol
820 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
821 -mno-jump-in-delay -mno-long-load-store @gol
822 -mno-portable-runtime -mno-soft-float @gol
823 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
824 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
825 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
826 -munix=@var{unix-std} -nolibdld -static -threads}
829 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
830 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
831 -mconstant-gp -mauto-pic -mfused-madd @gol
832 -minline-float-divide-min-latency @gol
833 -minline-float-divide-max-throughput @gol
834 -mno-inline-float-divide @gol
835 -minline-int-divide-min-latency @gol
836 -minline-int-divide-max-throughput @gol
837 -mno-inline-int-divide @gol
838 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
839 -mno-inline-sqrt @gol
840 -mdwarf2-asm -mearly-stop-bits @gol
841 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
842 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
843 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
844 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
845 -msched-spec-ldc -msched-spec-control-ldc @gol
846 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
847 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
848 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
849 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
852 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
853 -msign-extend-enabled -muser-enabled}
855 @emph{M32R/D Options}
856 @gccoptlist{-m32r2 -m32rx -m32r @gol
858 -malign-loops -mno-align-loops @gol
859 -missue-rate=@var{number} @gol
860 -mbranch-cost=@var{number} @gol
861 -mmodel=@var{code-size-model-type} @gol
862 -msdata=@var{sdata-type} @gol
863 -mno-flush-func -mflush-func=@var{name} @gol
864 -mno-flush-trap -mflush-trap=@var{number} @gol
868 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
870 @emph{M680x0 Options}
871 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
872 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
873 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
874 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
875 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
876 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
877 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
878 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
879 -mxgot -mno-xgot -mlong-jump-table-offsets}
882 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
883 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
884 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
885 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
886 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
889 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
890 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
891 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
892 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
895 @emph{MicroBlaze Options}
896 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
897 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
898 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
899 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
900 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}
901 -mpic-data-is-text-relative}
904 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
905 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
906 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
907 -mips16 -mno-mips16 -mflip-mips16 @gol
908 -minterlink-compressed -mno-interlink-compressed @gol
909 -minterlink-mips16 -mno-interlink-mips16 @gol
910 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
911 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
912 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
913 -mno-float -msingle-float -mdouble-float @gol
914 -modd-spreg -mno-odd-spreg @gol
915 -mabs=@var{mode} -mnan=@var{encoding} @gol
916 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
919 -mvirt -mno-virt @gol
922 -mginv -mno-ginv @gol
923 -mmicromips -mno-micromips @gol
925 -mloongson-mmi -mno-loongson-mmi @gol
926 -mloongson-ext -mno-loongson-ext @gol
927 -mloongson-ext2 -mno-loongson-ext2 @gol
928 -mfpu=@var{fpu-type} @gol
929 -msmartmips -mno-smartmips @gol
930 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
931 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
932 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
933 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
934 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
935 -membedded-data -mno-embedded-data @gol
936 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
937 -mcode-readable=@var{setting} @gol
938 -msplit-addresses -mno-split-addresses @gol
939 -mexplicit-relocs -mno-explicit-relocs @gol
940 -mcheck-zero-division -mno-check-zero-division @gol
941 -mdivide-traps -mdivide-breaks @gol
942 -mload-store-pairs -mno-load-store-pairs @gol
943 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
944 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
945 -mfix-24k -mno-fix-24k @gol
946 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
947 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
948 -mfix-vr4120 -mno-fix-vr4120 @gol
949 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
950 -mflush-func=@var{func} -mno-flush-func @gol
951 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
952 -mcompact-branches=@var{policy} @gol
953 -mfp-exceptions -mno-fp-exceptions @gol
954 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
955 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
956 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
957 -mframe-header-opt -mno-frame-header-opt}
960 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
961 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
962 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
963 -mno-base-addresses -msingle-exit -mno-single-exit}
965 @emph{MN10300 Options}
966 @gccoptlist{-mmult-bug -mno-mult-bug @gol
967 -mno-am33 -mam33 -mam33-2 -mam34 @gol
968 -mtune=@var{cpu-type} @gol
969 -mreturn-pointer-on-d0 @gol
970 -mno-crt0 -mrelax -mliw -msetlb}
973 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
975 @emph{MSP430 Options}
976 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
978 -mcode-region= -mdata-region= @gol
979 -msilicon-errata= -msilicon-errata-warn= @gol
983 @gccoptlist{-mbig-endian -mlittle-endian @gol
984 -mreduced-regs -mfull-regs @gol
985 -mcmov -mno-cmov @gol
986 -mext-perf -mno-ext-perf @gol
987 -mext-perf2 -mno-ext-perf2 @gol
988 -mext-string -mno-ext-string @gol
989 -mv3push -mno-v3push @gol
990 -m16bit -mno-16bit @gol
991 -misr-vector-size=@var{num} @gol
992 -mcache-block-size=@var{num} @gol
993 -march=@var{arch} @gol
994 -mcmodel=@var{code-model} @gol
997 @emph{Nios II Options}
998 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
999 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
1001 -mno-bypass-cache -mbypass-cache @gol
1002 -mno-cache-volatile -mcache-volatile @gol
1003 -mno-fast-sw-div -mfast-sw-div @gol
1004 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
1005 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
1006 -mcustom-fpu-cfg=@var{name} @gol
1007 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
1008 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
1010 @emph{Nvidia PTX Options}
1011 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
1013 @emph{OpenRISC Options}
1014 @gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol
1015 -msoft-mul -msoft-div @gol
1016 -mcmov -mror -msext -msfimm -mshftimm}
1018 @emph{PDP-11 Options}
1019 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
1020 -mint32 -mno-int16 -mint16 -mno-int32 @gol
1021 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra}
1023 @emph{picoChip Options}
1024 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
1025 -msymbol-as-address -mno-inefficient-warnings}
1027 @emph{PowerPC Options}
1028 See RS/6000 and PowerPC Options.
1030 @emph{PowerPC SPE Options}
1031 @gccoptlist{-mcpu=@var{cpu-type} @gol
1032 -mtune=@var{cpu-type} @gol
1033 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb @gol
1034 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1035 -m32 -mxl-compat -mno-xl-compat @gol
1036 -malign-power -malign-natural @gol
1037 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1038 -msingle-float -mdouble-float @gol
1039 -mupdate -mno-update @gol
1040 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1041 -mstrict-align -mno-strict-align -mrelocatable @gol
1042 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1043 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1044 -msingle-pic-base @gol
1045 -mprioritize-restricted-insns=@var{priority} @gol
1046 -msched-costly-dep=@var{dependence_type} @gol
1047 -minsert-sched-nops=@var{scheme} @gol
1048 -mcall-sysv -mcall-netbsd @gol
1049 -maix-struct-return -msvr4-struct-return @gol
1050 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1051 -mblock-move-inline-limit=@var{num} @gol
1052 -misel -mno-isel @gol
1053 -misel=yes -misel=no @gol
1055 -mspe=yes -mspe=no @gol
1056 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
1057 -mprototype -mno-prototype @gol
1058 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1059 -msdata=@var{opt} -mvxworks -G @var{num} @gol
1060 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1061 -mno-recip-precision @gol
1062 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1063 -msave-toc-indirect -mno-save-toc-indirect @gol
1064 -mcompat-align-parm -mno-compat-align-parm @gol
1065 -mfloat128 -mno-float128 @gol
1066 -mgnu-attribute -mno-gnu-attribute @gol
1067 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1068 -mstack-protector-guard-offset=@var{offset}}
1070 @emph{RISC-V Options}
1071 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1073 -mabi=@var{ABI-string} @gol
1074 -mfdiv -mno-fdiv @gol
1076 -march=@var{ISA-string} @gol
1077 -mtune=@var{processor-string} @gol
1078 -mpreferred-stack-boundary=@var{num} @gol
1079 -msmall-data-limit=@var{N-bytes} @gol
1080 -msave-restore -mno-save-restore @gol
1081 -mstrict-align -mno-strict-align @gol
1082 -mcmodel=medlow -mcmodel=medany @gol
1083 -mexplicit-relocs -mno-explicit-relocs @gol
1084 -mrelax -mno-relax @gol}
1087 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1088 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1089 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1091 @emph{RS/6000 and PowerPC Options}
1092 @gccoptlist{-mcpu=@var{cpu-type} @gol
1093 -mtune=@var{cpu-type} @gol
1094 -mcmodel=@var{code-model} @gol
1096 -maltivec -mno-altivec @gol
1097 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1098 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1099 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1100 -mfprnd -mno-fprnd @gol
1101 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
1102 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1103 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1104 -malign-power -malign-natural @gol
1105 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1106 -mupdate -mno-update @gol
1107 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1108 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1109 -mstrict-align -mno-strict-align -mrelocatable @gol
1110 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1111 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1112 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
1113 -mprioritize-restricted-insns=@var{priority} @gol
1114 -msched-costly-dep=@var{dependence_type} @gol
1115 -minsert-sched-nops=@var{scheme} @gol
1116 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1117 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1118 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1119 -mtraceback=@var{traceback_type} @gol
1120 -maix-struct-return -msvr4-struct-return @gol
1121 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1122 -mblock-move-inline-limit=@var{num} @gol
1123 -mblock-compare-inline-limit=@var{num} @gol
1124 -mblock-compare-inline-loop-limit=@var{num} @gol
1125 -mstring-compare-inline-limit=@var{num} @gol
1126 -misel -mno-isel @gol
1127 -mvrsave -mno-vrsave @gol
1128 -mmulhw -mno-mulhw @gol
1129 -mdlmzb -mno-dlmzb @gol
1130 -mprototype -mno-prototype @gol
1131 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1132 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1133 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1134 -mno-recip-precision @gol
1135 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1136 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1137 -msave-toc-indirect -mno-save-toc-indirect @gol
1138 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1139 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1140 -mquad-memory -mno-quad-memory @gol
1141 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1142 -mcompat-align-parm -mno-compat-align-parm @gol
1143 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1144 -mgnu-attribute -mno-gnu-attribute @gol
1145 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1146 -mstack-protector-guard-offset=@var{offset}}
1149 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1151 -mbig-endian-data -mlittle-endian-data @gol
1154 -mas100-syntax -mno-as100-syntax@gol
1156 -mmax-constant-size=@gol
1159 -mallow-string-insns -mno-allow-string-insns@gol
1161 -mno-warn-multiple-fast-interrupts@gol
1162 -msave-acc-in-interrupts}
1164 @emph{S/390 and zSeries Options}
1165 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1166 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1167 -mlong-double-64 -mlong-double-128 @gol
1168 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1169 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1170 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1171 -mhtm -mvx -mzvector @gol
1172 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1173 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1174 -mhotpatch=@var{halfwords},@var{halfwords}}
1176 @emph{Score Options}
1177 @gccoptlist{-meb -mel @gol
1181 -mscore5 -mscore5u -mscore7 -mscore7d}
1184 @gccoptlist{-m1 -m2 -m2e @gol
1185 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1187 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1188 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1189 -mb -ml -mdalign -mrelax @gol
1190 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1191 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1192 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1193 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1194 -maccumulate-outgoing-args @gol
1195 -matomic-model=@var{atomic-model} @gol
1196 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1197 -mcbranch-force-delay-slot @gol
1198 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1199 -mpretend-cmove -mtas}
1201 @emph{Solaris 2 Options}
1202 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1205 @emph{SPARC Options}
1206 @gccoptlist{-mcpu=@var{cpu-type} @gol
1207 -mtune=@var{cpu-type} @gol
1208 -mcmodel=@var{code-model} @gol
1209 -mmemory-model=@var{mem-model} @gol
1210 -m32 -m64 -mapp-regs -mno-app-regs @gol
1211 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1212 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1213 -mhard-quad-float -msoft-quad-float @gol
1214 -mstack-bias -mno-stack-bias @gol
1215 -mstd-struct-return -mno-std-struct-return @gol
1216 -munaligned-doubles -mno-unaligned-doubles @gol
1217 -muser-mode -mno-user-mode @gol
1218 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1219 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1220 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1221 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1222 -mpopc -mno-popc -msubxc -mno-subxc @gol
1223 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1227 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1228 -msafe-dma -munsafe-dma @gol
1230 -msmall-mem -mlarge-mem -mstdmain @gol
1231 -mfixed-range=@var{register-range} @gol
1233 -maddress-space-conversion -mno-address-space-conversion @gol
1234 -mcache-size=@var{cache-size} @gol
1235 -matomic-updates -mno-atomic-updates}
1237 @emph{System V Options}
1238 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1240 @emph{TILE-Gx Options}
1241 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1242 -mcmodel=@var{code-model}}
1244 @emph{TILEPro Options}
1245 @gccoptlist{-mcpu=@var{cpu} -m32}
1248 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1249 -mprolog-function -mno-prolog-function -mspace @gol
1250 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1251 -mapp-regs -mno-app-regs @gol
1252 -mdisable-callt -mno-disable-callt @gol
1253 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1254 -mv850e -mv850 -mv850e3v5 @gol
1265 @gccoptlist{-mg -mgnu -munix}
1267 @emph{Visium Options}
1268 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1269 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1272 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1273 -mpointer-size=@var{size}}
1275 @emph{VxWorks Options}
1276 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1277 -Xbind-lazy -Xbind-now}
1280 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1281 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1282 -mfpmath=@var{unit} @gol
1283 -masm=@var{dialect} -mno-fancy-math-387 @gol
1284 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1285 -mno-wide-multiply -mrtd -malign-double @gol
1286 -mpreferred-stack-boundary=@var{num} @gol
1287 -mincoming-stack-boundary=@var{num} @gol
1288 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1289 -mrecip -mrecip=@var{opt} @gol
1290 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1291 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1292 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1293 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1294 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd -mptwrite @gol
1295 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1296 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1297 -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp @gol
1298 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol
1299 -mshstk -mforce-indirect-call -mavx512vbmi2 @gol
1300 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq
1302 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1303 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1304 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1305 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1306 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1307 -mregparm=@var{num} -msseregparm @gol
1308 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1309 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1310 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1311 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1312 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1313 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1314 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1315 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1316 -mstack-protector-guard-reg=@var{reg} @gol
1317 -mstack-protector-guard-offset=@var{offset} @gol
1318 -mstack-protector-guard-symbol=@var{symbol} @gol
1319 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1320 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1321 -mindirect-branch-register}
1323 @emph{x86 Windows Options}
1324 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1325 -mnop-fun-dllimport -mthread @gol
1326 -municode -mwin32 -mwindows -fno-set-stack-executable}
1328 @emph{Xstormy16 Options}
1331 @emph{Xtensa Options}
1332 @gccoptlist{-mconst16 -mno-const16 @gol
1333 -mfused-madd -mno-fused-madd @gol
1335 -mserialize-volatile -mno-serialize-volatile @gol
1336 -mtext-section-literals -mno-text-section-literals @gol
1337 -mauto-litpools -mno-auto-litpools @gol
1338 -mtarget-align -mno-target-align @gol
1339 -mlongcalls -mno-longcalls}
1341 @emph{zSeries Options}
1342 See S/390 and zSeries Options.
1346 @node Overall Options
1347 @section Options Controlling the Kind of Output
1349 Compilation can involve up to four stages: preprocessing, compilation
1350 proper, assembly and linking, always in that order. GCC is capable of
1351 preprocessing and compiling several files either into several
1352 assembler input files, or into one assembler input file; then each
1353 assembler input file produces an object file, and linking combines all
1354 the object files (those newly compiled, and those specified as input)
1355 into an executable file.
1357 @cindex file name suffix
1358 For any given input file, the file name suffix determines what kind of
1359 compilation is done:
1363 C source code that must be preprocessed.
1366 C source code that should not be preprocessed.
1369 C++ source code that should not be preprocessed.
1372 Objective-C source code. Note that you must link with the @file{libobjc}
1373 library to make an Objective-C program work.
1376 Objective-C source code that should not be preprocessed.
1380 Objective-C++ source code. Note that you must link with the @file{libobjc}
1381 library to make an Objective-C++ program work. Note that @samp{.M} refers
1382 to a literal capital M@.
1384 @item @var{file}.mii
1385 Objective-C++ source code that should not be preprocessed.
1388 C, C++, Objective-C or Objective-C++ header file to be turned into a
1389 precompiled header (default), or C, C++ header file to be turned into an
1390 Ada spec (via the @option{-fdump-ada-spec} switch).
1393 @itemx @var{file}.cp
1394 @itemx @var{file}.cxx
1395 @itemx @var{file}.cpp
1396 @itemx @var{file}.CPP
1397 @itemx @var{file}.c++
1399 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1400 the last two letters must both be literally @samp{x}. Likewise,
1401 @samp{.C} refers to a literal capital C@.
1405 Objective-C++ source code that must be preprocessed.
1407 @item @var{file}.mii
1408 Objective-C++ source code that should not be preprocessed.
1412 @itemx @var{file}.hp
1413 @itemx @var{file}.hxx
1414 @itemx @var{file}.hpp
1415 @itemx @var{file}.HPP
1416 @itemx @var{file}.h++
1417 @itemx @var{file}.tcc
1418 C++ header file to be turned into a precompiled header or Ada spec.
1421 @itemx @var{file}.for
1422 @itemx @var{file}.ftn
1423 Fixed form Fortran source code that should not be preprocessed.
1426 @itemx @var{file}.FOR
1427 @itemx @var{file}.fpp
1428 @itemx @var{file}.FPP
1429 @itemx @var{file}.FTN
1430 Fixed form Fortran source code that must be preprocessed (with the traditional
1433 @item @var{file}.f90
1434 @itemx @var{file}.f95
1435 @itemx @var{file}.f03
1436 @itemx @var{file}.f08
1437 Free form Fortran source code that should not be preprocessed.
1439 @item @var{file}.F90
1440 @itemx @var{file}.F95
1441 @itemx @var{file}.F03
1442 @itemx @var{file}.F08
1443 Free form Fortran source code that must be preprocessed (with the
1444 traditional preprocessor).
1449 @item @var{file}.brig
1450 BRIG files (binary representation of HSAIL).
1459 D documentation code (Ddoc).
1461 @item @var{file}.ads
1462 Ada source code file that contains a library unit declaration (a
1463 declaration of a package, subprogram, or generic, or a generic
1464 instantiation), or a library unit renaming declaration (a package,
1465 generic, or subprogram renaming declaration). Such files are also
1468 @item @var{file}.adb
1469 Ada source code file containing a library unit body (a subprogram or
1470 package body). Such files are also called @dfn{bodies}.
1472 @c GCC also knows about some suffixes for languages not yet included:
1480 @itemx @var{file}.sx
1481 Assembler code that must be preprocessed.
1484 An object file to be fed straight into linking.
1485 Any file name with no recognized suffix is treated this way.
1489 You can specify the input language explicitly with the @option{-x} option:
1492 @item -x @var{language}
1493 Specify explicitly the @var{language} for the following input files
1494 (rather than letting the compiler choose a default based on the file
1495 name suffix). This option applies to all following input files until
1496 the next @option{-x} option. Possible values for @var{language} are:
1498 c c-header cpp-output
1499 c++ c++-header c++-cpp-output
1500 objective-c objective-c-header objective-c-cpp-output
1501 objective-c++ objective-c++-header objective-c++-cpp-output
1502 assembler assembler-with-cpp
1505 f77 f77-cpp-input f95 f95-cpp-input
1511 Turn off any specification of a language, so that subsequent files are
1512 handled according to their file name suffixes (as they are if @option{-x}
1513 has not been used at all).
1516 If you only want some of the stages of compilation, you can use
1517 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1518 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1519 @command{gcc} is to stop. Note that some combinations (for example,
1520 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1525 Compile or assemble the source files, but do not link. The linking
1526 stage simply is not done. The ultimate output is in the form of an
1527 object file for each source file.
1529 By default, the object file name for a source file is made by replacing
1530 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1532 Unrecognized input files, not requiring compilation or assembly, are
1537 Stop after the stage of compilation proper; do not assemble. The output
1538 is in the form of an assembler code file for each non-assembler input
1541 By default, the assembler file name for a source file is made by
1542 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1544 Input files that don't require compilation are ignored.
1548 Stop after the preprocessing stage; do not run the compiler proper. The
1549 output is in the form of preprocessed source code, which is sent to the
1552 Input files that don't require preprocessing are ignored.
1554 @cindex output file option
1557 Place output in file @var{file}. This applies to whatever
1558 sort of output is being produced, whether it be an executable file,
1559 an object file, an assembler file or preprocessed C code.
1561 If @option{-o} is not specified, the default is to put an executable
1562 file in @file{a.out}, the object file for
1563 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1564 assembler file in @file{@var{source}.s}, a precompiled header file in
1565 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1570 Print (on standard error output) the commands executed to run the stages
1571 of compilation. Also print the version number of the compiler driver
1572 program and of the preprocessor and the compiler proper.
1576 Like @option{-v} except the commands are not executed and arguments
1577 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1578 This is useful for shell scripts to capture the driver-generated command lines.
1582 Print (on the standard output) a description of the command-line options
1583 understood by @command{gcc}. If the @option{-v} option is also specified
1584 then @option{--help} is also passed on to the various processes
1585 invoked by @command{gcc}, so that they can display the command-line options
1586 they accept. If the @option{-Wextra} option has also been specified
1587 (prior to the @option{--help} option), then command-line options that
1588 have no documentation associated with them are also displayed.
1591 @opindex target-help
1592 Print (on the standard output) a description of target-specific command-line
1593 options for each tool. For some targets extra target-specific
1594 information may also be printed.
1596 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1597 Print (on the standard output) a description of the command-line
1598 options understood by the compiler that fit into all specified classes
1599 and qualifiers. These are the supported classes:
1602 @item @samp{optimizers}
1603 Display all of the optimization options supported by the
1606 @item @samp{warnings}
1607 Display all of the options controlling warning messages
1608 produced by the compiler.
1611 Display target-specific options. Unlike the
1612 @option{--target-help} option however, target-specific options of the
1613 linker and assembler are not displayed. This is because those
1614 tools do not currently support the extended @option{--help=} syntax.
1617 Display the values recognized by the @option{--param}
1620 @item @var{language}
1621 Display the options supported for @var{language}, where
1622 @var{language} is the name of one of the languages supported in this
1626 Display the options that are common to all languages.
1629 These are the supported qualifiers:
1632 @item @samp{undocumented}
1633 Display only those options that are undocumented.
1636 Display options taking an argument that appears after an equal
1637 sign in the same continuous piece of text, such as:
1638 @samp{--help=target}.
1640 @item @samp{separate}
1641 Display options taking an argument that appears as a separate word
1642 following the original option, such as: @samp{-o output-file}.
1645 Thus for example to display all the undocumented target-specific
1646 switches supported by the compiler, use:
1649 --help=target,undocumented
1652 The sense of a qualifier can be inverted by prefixing it with the
1653 @samp{^} character, so for example to display all binary warning
1654 options (i.e., ones that are either on or off and that do not take an
1655 argument) that have a description, use:
1658 --help=warnings,^joined,^undocumented
1661 The argument to @option{--help=} should not consist solely of inverted
1664 Combining several classes is possible, although this usually
1665 restricts the output so much that there is nothing to display. One
1666 case where it does work, however, is when one of the classes is
1667 @var{target}. For example, to display all the target-specific
1668 optimization options, use:
1671 --help=target,optimizers
1674 The @option{--help=} option can be repeated on the command line. Each
1675 successive use displays its requested class of options, skipping
1676 those that have already been displayed.
1678 If the @option{-Q} option appears on the command line before the
1679 @option{--help=} option, then the descriptive text displayed by
1680 @option{--help=} is changed. Instead of describing the displayed
1681 options, an indication is given as to whether the option is enabled,
1682 disabled or set to a specific value (assuming that the compiler
1683 knows this at the point where the @option{--help=} option is used).
1685 Here is a truncated example from the ARM port of @command{gcc}:
1688 % gcc -Q -mabi=2 --help=target -c
1689 The following options are target specific:
1691 -mabort-on-noreturn [disabled]
1695 The output is sensitive to the effects of previous command-line
1696 options, so for example it is possible to find out which optimizations
1697 are enabled at @option{-O2} by using:
1700 -Q -O2 --help=optimizers
1703 Alternatively you can discover which binary optimizations are enabled
1704 by @option{-O3} by using:
1707 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1708 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1709 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1714 Display the version number and copyrights of the invoked GCC@.
1716 @item -pass-exit-codes
1717 @opindex pass-exit-codes
1718 Normally the @command{gcc} program exits with the code of 1 if any
1719 phase of the compiler returns a non-success return code. If you specify
1720 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1721 the numerically highest error produced by any phase returning an error
1722 indication. The C, C++, and Fortran front ends return 4 if an internal
1723 compiler error is encountered.
1727 Use pipes rather than temporary files for communication between the
1728 various stages of compilation. This fails to work on some systems where
1729 the assembler is unable to read from a pipe; but the GNU assembler has
1732 @item -specs=@var{file}
1734 Process @var{file} after the compiler reads in the standard @file{specs}
1735 file, in order to override the defaults which the @command{gcc} driver
1736 program uses when determining what switches to pass to @command{cc1},
1737 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1738 @option{-specs=@var{file}} can be specified on the command line, and they
1739 are processed in order, from left to right. @xref{Spec Files}, for
1740 information about the format of the @var{file}.
1744 Invoke all subcommands under a wrapper program. The name of the
1745 wrapper program and its parameters are passed as a comma separated
1749 gcc -c t.c -wrapper gdb,--args
1753 This invokes all subprograms of @command{gcc} under
1754 @samp{gdb --args}, thus the invocation of @command{cc1} is
1755 @samp{gdb --args cc1 @dots{}}.
1757 @item -ffile-prefix-map=@var{old}=@var{new}
1758 @opindex ffile-prefix-map
1759 When compiling files residing in directory @file{@var{old}}, record
1760 any references to them in the result of the compilation as if the
1761 files resided in directory @file{@var{new}} instead. Specifying this
1762 option is equivalent to specifying all the individual
1763 @option{-f*-prefix-map} options. This can be used to make reproducible
1764 builds that are location independent. See also
1765 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1767 @item -fplugin=@var{name}.so
1769 Load the plugin code in file @var{name}.so, assumed to be a
1770 shared object to be dlopen'd by the compiler. The base name of
1771 the shared object file is used to identify the plugin for the
1772 purposes of argument parsing (See
1773 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1774 Each plugin should define the callback functions specified in the
1777 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1778 @opindex fplugin-arg
1779 Define an argument called @var{key} with a value of @var{value}
1780 for the plugin called @var{name}.
1782 @item -fdump-ada-spec@r{[}-slim@r{]}
1783 @opindex fdump-ada-spec
1784 For C and C++ source and include files, generate corresponding Ada specs.
1785 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1786 GNAT User's Guide}, which provides detailed documentation on this feature.
1788 @item -fada-spec-parent=@var{unit}
1789 @opindex fada-spec-parent
1790 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1791 Ada specs as child units of parent @var{unit}.
1793 @item -fdump-go-spec=@var{file}
1794 @opindex fdump-go-spec
1795 For input files in any language, generate corresponding Go
1796 declarations in @var{file}. This generates Go @code{const},
1797 @code{type}, @code{var}, and @code{func} declarations which may be a
1798 useful way to start writing a Go interface to code written in some
1801 @include @value{srcdir}/../libiberty/at-file.texi
1805 @section Compiling C++ Programs
1807 @cindex suffixes for C++ source
1808 @cindex C++ source file suffixes
1809 C++ source files conventionally use one of the suffixes @samp{.C},
1810 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1811 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1812 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1813 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1814 files with these names and compiles them as C++ programs even if you
1815 call the compiler the same way as for compiling C programs (usually
1816 with the name @command{gcc}).
1820 However, the use of @command{gcc} does not add the C++ library.
1821 @command{g++} is a program that calls GCC and automatically specifies linking
1822 against the C++ library. It treats @samp{.c},
1823 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1824 files unless @option{-x} is used. This program is also useful when
1825 precompiling a C header file with a @samp{.h} extension for use in C++
1826 compilations. On many systems, @command{g++} is also installed with
1827 the name @command{c++}.
1829 @cindex invoking @command{g++}
1830 When you compile C++ programs, you may specify many of the same
1831 command-line options that you use for compiling programs in any
1832 language; or command-line options meaningful for C and related
1833 languages; or options that are meaningful only for C++ programs.
1834 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1835 explanations of options for languages related to C@.
1836 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1837 explanations of options that are meaningful only for C++ programs.
1839 @node C Dialect Options
1840 @section Options Controlling C Dialect
1841 @cindex dialect options
1842 @cindex language dialect options
1843 @cindex options, dialect
1845 The following options control the dialect of C (or languages derived
1846 from C, such as C++, Objective-C and Objective-C++) that the compiler
1850 @cindex ANSI support
1854 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1855 equivalent to @option{-std=c++98}.
1857 This turns off certain features of GCC that are incompatible with ISO
1858 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1859 such as the @code{asm} and @code{typeof} keywords, and
1860 predefined macros such as @code{unix} and @code{vax} that identify the
1861 type of system you are using. It also enables the undesirable and
1862 rarely used ISO trigraph feature. For the C compiler,
1863 it disables recognition of C++ style @samp{//} comments as well as
1864 the @code{inline} keyword.
1866 The alternate keywords @code{__asm__}, @code{__extension__},
1867 @code{__inline__} and @code{__typeof__} continue to work despite
1868 @option{-ansi}. You would not want to use them in an ISO C program, of
1869 course, but it is useful to put them in header files that might be included
1870 in compilations done with @option{-ansi}. Alternate predefined macros
1871 such as @code{__unix__} and @code{__vax__} are also available, with or
1872 without @option{-ansi}.
1874 The @option{-ansi} option does not cause non-ISO programs to be
1875 rejected gratuitously. For that, @option{-Wpedantic} is required in
1876 addition to @option{-ansi}. @xref{Warning Options}.
1878 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1879 option is used. Some header files may notice this macro and refrain
1880 from declaring certain functions or defining certain macros that the
1881 ISO standard doesn't call for; this is to avoid interfering with any
1882 programs that might use these names for other things.
1884 Functions that are normally built in but do not have semantics
1885 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1886 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1887 built-in functions provided by GCC}, for details of the functions
1892 Determine the language standard. @xref{Standards,,Language Standards
1893 Supported by GCC}, for details of these standard versions. This option
1894 is currently only supported when compiling C or C++.
1896 The compiler can accept several base standards, such as @samp{c90} or
1897 @samp{c++98}, and GNU dialects of those standards, such as
1898 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1899 compiler accepts all programs following that standard plus those
1900 using GNU extensions that do not contradict it. For example,
1901 @option{-std=c90} turns off certain features of GCC that are
1902 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1903 keywords, but not other GNU extensions that do not have a meaning in
1904 ISO C90, such as omitting the middle term of a @code{?:}
1905 expression. On the other hand, when a GNU dialect of a standard is
1906 specified, all features supported by the compiler are enabled, even when
1907 those features change the meaning of the base standard. As a result, some
1908 strict-conforming programs may be rejected. The particular standard
1909 is used by @option{-Wpedantic} to identify which features are GNU
1910 extensions given that version of the standard. For example
1911 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1912 comments, while @option{-std=gnu99 -Wpedantic} does not.
1914 A value for this option must be provided; possible values are
1920 Support all ISO C90 programs (certain GNU extensions that conflict
1921 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1923 @item iso9899:199409
1924 ISO C90 as modified in amendment 1.
1930 ISO C99. This standard is substantially completely supported, modulo
1931 bugs and floating-point issues
1932 (mainly but not entirely relating to optional C99 features from
1933 Annexes F and G). See
1934 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1935 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1940 ISO C11, the 2011 revision of the ISO C standard. This standard is
1941 substantially completely supported, modulo bugs, floating-point issues
1942 (mainly but not entirely relating to optional C11 features from
1943 Annexes F and G) and the optional Annexes K (Bounds-checking
1944 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1950 ISO C17, the 2017 revision of the ISO C standard
1951 (published in 2018). This standard is
1952 same as C11 except for corrections of defects (all of which are also
1953 applied with @option{-std=c11}) and a new value of
1954 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1957 The next version of the ISO C standard, still under development. The
1958 support for this version is experimental and incomplete.
1962 GNU dialect of ISO C90 (including some C99 features).
1966 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1970 GNU dialect of ISO C11.
1971 The name @samp{gnu1x} is deprecated.
1975 GNU dialect of ISO C17. This is the default for C code.
1978 The next version of the ISO C standard, still under development, plus
1979 GNU extensions. The support for this version is experimental and
1984 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1985 additional defect reports. Same as @option{-ansi} for C++ code.
1989 GNU dialect of @option{-std=c++98}.
1993 The 2011 ISO C++ standard plus amendments.
1994 The name @samp{c++0x} is deprecated.
1998 GNU dialect of @option{-std=c++11}.
1999 The name @samp{gnu++0x} is deprecated.
2003 The 2014 ISO C++ standard plus amendments.
2004 The name @samp{c++1y} is deprecated.
2008 GNU dialect of @option{-std=c++14}.
2009 This is the default for C++ code.
2010 The name @samp{gnu++1y} is deprecated.
2014 The 2017 ISO C++ standard plus amendments.
2015 The name @samp{c++1z} is deprecated.
2019 GNU dialect of @option{-std=c++17}.
2020 The name @samp{gnu++1z} is deprecated.
2023 The next revision of the ISO C++ standard, tentatively planned for
2024 2020. Support is highly experimental, and will almost certainly
2025 change in incompatible ways in future releases.
2028 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
2029 and will almost certainly change in incompatible ways in future
2033 @item -fgnu89-inline
2034 @opindex fgnu89-inline
2035 The option @option{-fgnu89-inline} tells GCC to use the traditional
2036 GNU semantics for @code{inline} functions when in C99 mode.
2037 @xref{Inline,,An Inline Function is As Fast As a Macro}.
2038 Using this option is roughly equivalent to adding the
2039 @code{gnu_inline} function attribute to all inline functions
2040 (@pxref{Function Attributes}).
2042 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
2043 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
2044 specifies the default behavior).
2045 This option is not supported in @option{-std=c90} or
2046 @option{-std=gnu90} mode.
2048 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
2049 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
2050 in effect for @code{inline} functions. @xref{Common Predefined
2051 Macros,,,cpp,The C Preprocessor}.
2053 @item -fpermitted-flt-eval-methods=@var{style}
2054 @opindex fpermitted-flt-eval-methods
2055 @opindex fpermitted-flt-eval-methods=c11
2056 @opindex fpermitted-flt-eval-methods=ts-18661-3
2057 ISO/IEC TS 18661-3 defines new permissible values for
2058 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2059 a semantic type that is an interchange or extended format should be
2060 evaluated to the precision and range of that type. These new values are
2061 a superset of those permitted under C99/C11, which does not specify the
2062 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2063 conforming to C11 may not have been written expecting the possibility of
2066 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2067 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2068 or the extended set of values specified in ISO/IEC TS 18661-3.
2070 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2072 The default when in a standards compliant mode (@option{-std=c11} or similar)
2073 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2074 dialect (@option{-std=gnu11} or similar) is
2075 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2077 @item -aux-info @var{filename}
2079 Output to the given filename prototyped declarations for all functions
2080 declared and/or defined in a translation unit, including those in header
2081 files. This option is silently ignored in any language other than C@.
2083 Besides declarations, the file indicates, in comments, the origin of
2084 each declaration (source file and line), whether the declaration was
2085 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2086 @samp{O} for old, respectively, in the first character after the line
2087 number and the colon), and whether it came from a declaration or a
2088 definition (@samp{C} or @samp{F}, respectively, in the following
2089 character). In the case of function definitions, a K&R-style list of
2090 arguments followed by their declarations is also provided, inside
2091 comments, after the declaration.
2093 @item -fallow-parameterless-variadic-functions
2094 @opindex fallow-parameterless-variadic-functions
2095 Accept variadic functions without named parameters.
2097 Although it is possible to define such a function, this is not very
2098 useful as it is not possible to read the arguments. This is only
2099 supported for C as this construct is allowed by C++.
2103 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2104 keyword, so that code can use these words as identifiers. You can use
2105 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2106 instead. @option{-ansi} implies @option{-fno-asm}.
2108 In C++, this switch only affects the @code{typeof} keyword, since
2109 @code{asm} and @code{inline} are standard keywords. You may want to
2110 use the @option{-fno-gnu-keywords} flag instead, which has the same
2111 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2112 switch only affects the @code{asm} and @code{typeof} keywords, since
2113 @code{inline} is a standard keyword in ISO C99.
2116 @itemx -fno-builtin-@var{function}
2117 @opindex fno-builtin
2118 @cindex built-in functions
2119 Don't recognize built-in functions that do not begin with
2120 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2121 functions provided by GCC}, for details of the functions affected,
2122 including those which are not built-in functions when @option{-ansi} or
2123 @option{-std} options for strict ISO C conformance are used because they
2124 do not have an ISO standard meaning.
2126 GCC normally generates special code to handle certain built-in functions
2127 more efficiently; for instance, calls to @code{alloca} may become single
2128 instructions which adjust the stack directly, and calls to @code{memcpy}
2129 may become inline copy loops. The resulting code is often both smaller
2130 and faster, but since the function calls no longer appear as such, you
2131 cannot set a breakpoint on those calls, nor can you change the behavior
2132 of the functions by linking with a different library. In addition,
2133 when a function is recognized as a built-in function, GCC may use
2134 information about that function to warn about problems with calls to
2135 that function, or to generate more efficient code, even if the
2136 resulting code still contains calls to that function. For example,
2137 warnings are given with @option{-Wformat} for bad calls to
2138 @code{printf} when @code{printf} is built in and @code{strlen} is
2139 known not to modify global memory.
2141 With the @option{-fno-builtin-@var{function}} option
2142 only the built-in function @var{function} is
2143 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2144 function is named that is not built-in in this version of GCC, this
2145 option is ignored. There is no corresponding
2146 @option{-fbuiltin-@var{function}} option; if you wish to enable
2147 built-in functions selectively when using @option{-fno-builtin} or
2148 @option{-ffreestanding}, you may define macros such as:
2151 #define abs(n) __builtin_abs ((n))
2152 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2158 Enable parsing of function definitions marked with @code{__GIMPLE}.
2159 This is an experimental feature that allows unit testing of GIMPLE
2164 @cindex hosted environment
2166 Assert that compilation targets a hosted environment. This implies
2167 @option{-fbuiltin}. A hosted environment is one in which the
2168 entire standard library is available, and in which @code{main} has a return
2169 type of @code{int}. Examples are nearly everything except a kernel.
2170 This is equivalent to @option{-fno-freestanding}.
2172 @item -ffreestanding
2173 @opindex ffreestanding
2174 @cindex hosted environment
2176 Assert that compilation targets a freestanding environment. This
2177 implies @option{-fno-builtin}. A freestanding environment
2178 is one in which the standard library may not exist, and program startup may
2179 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2180 This is equivalent to @option{-fno-hosted}.
2182 @xref{Standards,,Language Standards Supported by GCC}, for details of
2183 freestanding and hosted environments.
2187 @cindex OpenACC accelerator programming
2188 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2189 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2190 compiler generates accelerated code according to the OpenACC Application
2191 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2192 implies @option{-pthread}, and thus is only supported on targets that
2193 have support for @option{-pthread}.
2195 @item -fopenacc-dim=@var{geom}
2196 @opindex fopenacc-dim
2197 @cindex OpenACC accelerator programming
2198 Specify default compute dimensions for parallel offload regions that do
2199 not explicitly specify. The @var{geom} value is a triple of
2200 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2201 can be omitted, to use a target-specific default value.
2205 @cindex OpenMP parallel
2206 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2207 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2208 compiler generates parallel code according to the OpenMP Application
2209 Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option
2210 implies @option{-pthread}, and thus is only supported on targets that
2211 have support for @option{-pthread}. @option{-fopenmp} implies
2212 @option{-fopenmp-simd}.
2215 @opindex fopenmp-simd
2218 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2219 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2224 When the option @option{-fgnu-tm} is specified, the compiler
2225 generates code for the Linux variant of Intel's current Transactional
2226 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2227 an experimental feature whose interface may change in future versions
2228 of GCC, as the official specification changes. Please note that not
2229 all architectures are supported for this feature.
2231 For more information on GCC's support for transactional memory,
2232 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2233 Transactional Memory Library}.
2235 Note that the transactional memory feature is not supported with
2236 non-call exceptions (@option{-fnon-call-exceptions}).
2238 @item -fms-extensions
2239 @opindex fms-extensions
2240 Accept some non-standard constructs used in Microsoft header files.
2242 In C++ code, this allows member names in structures to be similar
2243 to previous types declarations.
2252 Some cases of unnamed fields in structures and unions are only
2253 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2254 fields within structs/unions}, for details.
2256 Note that this option is off for all targets but x86
2257 targets using ms-abi.
2259 @item -fplan9-extensions
2260 @opindex fplan9-extensions
2261 Accept some non-standard constructs used in Plan 9 code.
2263 This enables @option{-fms-extensions}, permits passing pointers to
2264 structures with anonymous fields to functions that expect pointers to
2265 elements of the type of the field, and permits referring to anonymous
2266 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2267 struct/union fields within structs/unions}, for details. This is only
2268 supported for C, not C++.
2270 @item -fcond-mismatch
2271 @opindex fcond-mismatch
2272 Allow conditional expressions with mismatched types in the second and
2273 third arguments. The value of such an expression is void. This option
2274 is not supported for C++.
2276 @item -flax-vector-conversions
2277 @opindex flax-vector-conversions
2278 Allow implicit conversions between vectors with differing numbers of
2279 elements and/or incompatible element types. This option should not be
2282 @item -funsigned-char
2283 @opindex funsigned-char
2284 Let the type @code{char} be unsigned, like @code{unsigned char}.
2286 Each kind of machine has a default for what @code{char} should
2287 be. It is either like @code{unsigned char} by default or like
2288 @code{signed char} by default.
2290 Ideally, a portable program should always use @code{signed char} or
2291 @code{unsigned char} when it depends on the signedness of an object.
2292 But many programs have been written to use plain @code{char} and
2293 expect it to be signed, or expect it to be unsigned, depending on the
2294 machines they were written for. This option, and its inverse, let you
2295 make such a program work with the opposite default.
2297 The type @code{char} is always a distinct type from each of
2298 @code{signed char} or @code{unsigned char}, even though its behavior
2299 is always just like one of those two.
2302 @opindex fsigned-char
2303 Let the type @code{char} be signed, like @code{signed char}.
2305 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2306 the negative form of @option{-funsigned-char}. Likewise, the option
2307 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2309 @item -fsigned-bitfields
2310 @itemx -funsigned-bitfields
2311 @itemx -fno-signed-bitfields
2312 @itemx -fno-unsigned-bitfields
2313 @opindex fsigned-bitfields
2314 @opindex funsigned-bitfields
2315 @opindex fno-signed-bitfields
2316 @opindex fno-unsigned-bitfields
2317 These options control whether a bit-field is signed or unsigned, when the
2318 declaration does not use either @code{signed} or @code{unsigned}. By
2319 default, such a bit-field is signed, because this is consistent: the
2320 basic integer types such as @code{int} are signed types.
2322 @item -fsso-struct=@var{endianness}
2323 @opindex fsso-struct
2324 Set the default scalar storage order of structures and unions to the
2325 specified endianness. The accepted values are @samp{big-endian},
2326 @samp{little-endian} and @samp{native} for the native endianness of
2327 the target (the default). This option is not supported for C++.
2329 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2330 code that is not binary compatible with code generated without it if the
2331 specified endianness is not the native endianness of the target.
2334 @node C++ Dialect Options
2335 @section Options Controlling C++ Dialect
2337 @cindex compiler options, C++
2338 @cindex C++ options, command-line
2339 @cindex options, C++
2340 This section describes the command-line options that are only meaningful
2341 for C++ programs. You can also use most of the GNU compiler options
2342 regardless of what language your program is in. For example, you
2343 might compile a file @file{firstClass.C} like this:
2346 g++ -g -fstrict-enums -O -c firstClass.C
2350 In this example, only @option{-fstrict-enums} is an option meant
2351 only for C++ programs; you can use the other options with any
2352 language supported by GCC@.
2354 Some options for compiling C programs, such as @option{-std}, are also
2355 relevant for C++ programs.
2356 @xref{C Dialect Options,,Options Controlling C Dialect}.
2358 Here is a list of options that are @emph{only} for compiling C++ programs:
2362 @item -fabi-version=@var{n}
2363 @opindex fabi-version
2364 Use version @var{n} of the C++ ABI@. The default is version 0.
2366 Version 0 refers to the version conforming most closely to
2367 the C++ ABI specification. Therefore, the ABI obtained using version 0
2368 will change in different versions of G++ as ABI bugs are fixed.
2370 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2372 Version 2 is the version of the C++ ABI that first appeared in G++
2373 3.4, and was the default through G++ 4.9.
2375 Version 3 corrects an error in mangling a constant address as a
2378 Version 4, which first appeared in G++ 4.5, implements a standard
2379 mangling for vector types.
2381 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2382 attribute const/volatile on function pointer types, decltype of a
2383 plain decl, and use of a function parameter in the declaration of
2386 Version 6, which first appeared in G++ 4.7, corrects the promotion
2387 behavior of C++11 scoped enums and the mangling of template argument
2388 packs, const/static_cast, prefix ++ and --, and a class scope function
2389 used as a template argument.
2391 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2392 builtin type and corrects the mangling of lambdas in default argument
2395 Version 8, which first appeared in G++ 4.9, corrects the substitution
2396 behavior of function types with function-cv-qualifiers.
2398 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2401 Version 10, which first appeared in G++ 6.1, adds mangling of
2402 attributes that affect type identity, such as ia32 calling convention
2403 attributes (e.g.@: @samp{stdcall}).
2405 Version 11, which first appeared in G++ 7, corrects the mangling of
2406 sizeof... expressions and operator names. For multiple entities with
2407 the same name within a function, that are declared in different scopes,
2408 the mangling now changes starting with the twelfth occurrence. It also
2409 implies @option{-fnew-inheriting-ctors}.
2411 Version 12, which first appeared in G++ 8, corrects the calling
2412 conventions for empty classes on the x86_64 target and for classes
2413 with only deleted copy/move constructors. It accidentally changes the
2414 calling convention for classes with a deleted copy constructor and a
2415 trivial move constructor.
2417 Version 13, which first appeared in G++ 8.2, fixes the accidental
2418 change in version 12.
2420 See also @option{-Wabi}.
2422 @item -fabi-compat-version=@var{n}
2423 @opindex fabi-compat-version
2424 On targets that support strong aliases, G++
2425 works around mangling changes by creating an alias with the correct
2426 mangled name when defining a symbol with an incorrect mangled name.
2427 This switch specifies which ABI version to use for the alias.
2429 With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7
2430 compatibility). If another ABI version is explicitly selected, this
2431 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2432 use @option{-fabi-compat-version=2}.
2434 If this option is not provided but @option{-Wabi=@var{n}} is, that
2435 version is used for compatibility aliases. If this option is provided
2436 along with @option{-Wabi} (without the version), the version from this
2437 option is used for the warning.
2439 @item -fno-access-control
2440 @opindex fno-access-control
2441 Turn off all access checking. This switch is mainly useful for working
2442 around bugs in the access control code.
2445 @opindex faligned-new
2446 Enable support for C++17 @code{new} of types that require more
2447 alignment than @code{void* ::operator new(std::size_t)} provides. A
2448 numeric argument such as @code{-faligned-new=32} can be used to
2449 specify how much alignment (in bytes) is provided by that function,
2450 but few users will need to override the default of
2451 @code{alignof(std::max_align_t)}.
2453 This flag is enabled by default for @option{-std=c++17}.
2457 Check that the pointer returned by @code{operator new} is non-null
2458 before attempting to modify the storage allocated. This check is
2459 normally unnecessary because the C++ standard specifies that
2460 @code{operator new} only returns @code{0} if it is declared
2461 @code{throw()}, in which case the compiler always checks the
2462 return value even without this option. In all other cases, when
2463 @code{operator new} has a non-empty exception specification, memory
2464 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2465 @samp{new (nothrow)}.
2469 Enable support for the C++ Extensions for Concepts Technical
2470 Specification, ISO 19217 (2015), which allows code like
2473 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2474 template <Addable T> T add (T a, T b) @{ return a + b; @}
2477 @item -fconstexpr-depth=@var{n}
2478 @opindex fconstexpr-depth
2479 Set the maximum nested evaluation depth for C++11 constexpr functions
2480 to @var{n}. A limit is needed to detect endless recursion during
2481 constant expression evaluation. The minimum specified by the standard
2484 @item -fconstexpr-loop-limit=@var{n}
2485 @opindex fconstexpr-loop-limit
2486 Set the maximum number of iterations for a loop in C++14 constexpr functions
2487 to @var{n}. A limit is needed to detect infinite loops during
2488 constant expression evaluation. The default is 262144 (1<<18).
2490 @item -fdeduce-init-list
2491 @opindex fdeduce-init-list
2492 Enable deduction of a template type parameter as
2493 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2496 template <class T> auto forward(T t) -> decltype (realfn (t))
2503 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2507 This deduction was implemented as a possible extension to the
2508 originally proposed semantics for the C++11 standard, but was not part
2509 of the final standard, so it is disabled by default. This option is
2510 deprecated, and may be removed in a future version of G++.
2512 @item -fno-elide-constructors
2513 @opindex fno-elide-constructors
2514 The C++ standard allows an implementation to omit creating a temporary
2515 that is only used to initialize another object of the same type.
2516 Specifying this option disables that optimization, and forces G++ to
2517 call the copy constructor in all cases. This option also causes G++
2518 to call trivial member functions which otherwise would be expanded inline.
2520 In C++17, the compiler is required to omit these temporaries, but this
2521 option still affects trivial member functions.
2523 @item -fno-enforce-eh-specs
2524 @opindex fno-enforce-eh-specs
2525 Don't generate code to check for violation of exception specifications
2526 at run time. This option violates the C++ standard, but may be useful
2527 for reducing code size in production builds, much like defining
2528 @code{NDEBUG}. This does not give user code permission to throw
2529 exceptions in violation of the exception specifications; the compiler
2530 still optimizes based on the specifications, so throwing an
2531 unexpected exception results in undefined behavior at run time.
2533 @item -fextern-tls-init
2534 @itemx -fno-extern-tls-init
2535 @opindex fextern-tls-init
2536 @opindex fno-extern-tls-init
2537 The C++11 and OpenMP standards allow @code{thread_local} and
2538 @code{threadprivate} variables to have dynamic (runtime)
2539 initialization. To support this, any use of such a variable goes
2540 through a wrapper function that performs any necessary initialization.
2541 When the use and definition of the variable are in the same
2542 translation unit, this overhead can be optimized away, but when the
2543 use is in a different translation unit there is significant overhead
2544 even if the variable doesn't actually need dynamic initialization. If
2545 the programmer can be sure that no use of the variable in a
2546 non-defining TU needs to trigger dynamic initialization (either
2547 because the variable is statically initialized, or a use of the
2548 variable in the defining TU will be executed before any uses in
2549 another TU), they can avoid this overhead with the
2550 @option{-fno-extern-tls-init} option.
2552 On targets that support symbol aliases, the default is
2553 @option{-fextern-tls-init}. On targets that do not support symbol
2554 aliases, the default is @option{-fno-extern-tls-init}.
2556 @item -fno-gnu-keywords
2557 @opindex fno-gnu-keywords
2558 Do not recognize @code{typeof} as a keyword, so that code can use this
2559 word as an identifier. You can use the keyword @code{__typeof__} instead.
2560 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2561 @option{-std=c++98}, @option{-std=c++11}, etc.
2563 @item -fno-implicit-templates
2564 @opindex fno-implicit-templates
2565 Never emit code for non-inline templates that are instantiated
2566 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2567 @xref{Template Instantiation}, for more information.
2569 @item -fno-implicit-inline-templates
2570 @opindex fno-implicit-inline-templates
2571 Don't emit code for implicit instantiations of inline templates, either.
2572 The default is to handle inlines differently so that compiles with and
2573 without optimization need the same set of explicit instantiations.
2575 @item -fno-implement-inlines
2576 @opindex fno-implement-inlines
2577 To save space, do not emit out-of-line copies of inline functions
2578 controlled by @code{#pragma implementation}. This causes linker
2579 errors if these functions are not inlined everywhere they are called.
2581 @item -fms-extensions
2582 @opindex fms-extensions
2583 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2584 int and getting a pointer to member function via non-standard syntax.
2586 @item -fnew-inheriting-ctors
2587 @opindex fnew-inheriting-ctors
2588 Enable the P0136 adjustment to the semantics of C++11 constructor
2589 inheritance. This is part of C++17 but also considered to be a Defect
2590 Report against C++11 and C++14. This flag is enabled by default
2591 unless @option{-fabi-version=10} or lower is specified.
2593 @item -fnew-ttp-matching
2594 @opindex fnew-ttp-matching
2595 Enable the P0522 resolution to Core issue 150, template template
2596 parameters and default arguments: this allows a template with default
2597 template arguments as an argument for a template template parameter
2598 with fewer template parameters. This flag is enabled by default for
2599 @option{-std=c++17}.
2601 @item -fno-nonansi-builtins
2602 @opindex fno-nonansi-builtins
2603 Disable built-in declarations of functions that are not mandated by
2604 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2605 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2608 @opindex fnothrow-opt
2609 Treat a @code{throw()} exception specification as if it were a
2610 @code{noexcept} specification to reduce or eliminate the text size
2611 overhead relative to a function with no exception specification. If
2612 the function has local variables of types with non-trivial
2613 destructors, the exception specification actually makes the
2614 function smaller because the EH cleanups for those variables can be
2615 optimized away. The semantic effect is that an exception thrown out of
2616 a function with such an exception specification results in a call
2617 to @code{terminate} rather than @code{unexpected}.
2619 @item -fno-operator-names
2620 @opindex fno-operator-names
2621 Do not treat the operator name keywords @code{and}, @code{bitand},
2622 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2623 synonyms as keywords.
2625 @item -fno-optional-diags
2626 @opindex fno-optional-diags
2627 Disable diagnostics that the standard says a compiler does not need to
2628 issue. Currently, the only such diagnostic issued by G++ is the one for
2629 a name having multiple meanings within a class.
2632 @opindex fpermissive
2633 Downgrade some diagnostics about nonconformant code from errors to
2634 warnings. Thus, using @option{-fpermissive} allows some
2635 nonconforming code to compile.
2637 @item -fno-pretty-templates
2638 @opindex fno-pretty-templates
2639 When an error message refers to a specialization of a function
2640 template, the compiler normally prints the signature of the
2641 template followed by the template arguments and any typedefs or
2642 typenames in the signature (e.g.@: @code{void f(T) [with T = int]}
2643 rather than @code{void f(int)}) so that it's clear which template is
2644 involved. When an error message refers to a specialization of a class
2645 template, the compiler omits any template arguments that match
2646 the default template arguments for that template. If either of these
2647 behaviors make it harder to understand the error message rather than
2648 easier, you can use @option{-fno-pretty-templates} to disable them.
2652 Enable automatic template instantiation at link time. This option also
2653 implies @option{-fno-implicit-templates}. @xref{Template
2654 Instantiation}, for more information.
2658 Disable generation of information about every class with virtual
2659 functions for use by the C++ run-time type identification features
2660 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2661 of the language, you can save some space by using this flag. Note that
2662 exception handling uses the same information, but G++ generates it as
2663 needed. The @code{dynamic_cast} operator can still be used for casts that
2664 do not require run-time type information, i.e.@: casts to @code{void *} or to
2665 unambiguous base classes.
2667 @item -fsized-deallocation
2668 @opindex fsized-deallocation
2669 Enable the built-in global declarations
2671 void operator delete (void *, std::size_t) noexcept;
2672 void operator delete[] (void *, std::size_t) noexcept;
2674 as introduced in C++14. This is useful for user-defined replacement
2675 deallocation functions that, for example, use the size of the object
2676 to make deallocation faster. Enabled by default under
2677 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2678 warns about places that might want to add a definition.
2680 @item -fstrict-enums
2681 @opindex fstrict-enums
2682 Allow the compiler to optimize using the assumption that a value of
2683 enumerated type can only be one of the values of the enumeration (as
2684 defined in the C++ standard; basically, a value that can be
2685 represented in the minimum number of bits needed to represent all the
2686 enumerators). This assumption may not be valid if the program uses a
2687 cast to convert an arbitrary integer value to the enumerated type.
2689 @item -fstrong-eval-order
2690 @opindex fstrong-eval-order
2691 Evaluate member access, array subscripting, and shift expressions in
2692 left-to-right order, and evaluate assignment in right-to-left order,
2693 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2694 @option{-fstrong-eval-order=some} enables just the ordering of member
2695 access and shift expressions, and is the default without
2696 @option{-std=c++17}.
2698 @item -ftemplate-backtrace-limit=@var{n}
2699 @opindex ftemplate-backtrace-limit
2700 Set the maximum number of template instantiation notes for a single
2701 warning or error to @var{n}. The default value is 10.
2703 @item -ftemplate-depth=@var{n}
2704 @opindex ftemplate-depth
2705 Set the maximum instantiation depth for template classes to @var{n}.
2706 A limit on the template instantiation depth is needed to detect
2707 endless recursions during template class instantiation. ANSI/ISO C++
2708 conforming programs must not rely on a maximum depth greater than 17
2709 (changed to 1024 in C++11). The default value is 900, as the compiler
2710 can run out of stack space before hitting 1024 in some situations.
2712 @item -fno-threadsafe-statics
2713 @opindex fno-threadsafe-statics
2714 Do not emit the extra code to use the routines specified in the C++
2715 ABI for thread-safe initialization of local statics. You can use this
2716 option to reduce code size slightly in code that doesn't need to be
2719 @item -fuse-cxa-atexit
2720 @opindex fuse-cxa-atexit
2721 Register destructors for objects with static storage duration with the
2722 @code{__cxa_atexit} function rather than the @code{atexit} function.
2723 This option is required for fully standards-compliant handling of static
2724 destructors, but only works if your C library supports
2725 @code{__cxa_atexit}.
2727 @item -fno-use-cxa-get-exception-ptr
2728 @opindex fno-use-cxa-get-exception-ptr
2729 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2730 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2731 if the runtime routine is not available.
2733 @item -fvisibility-inlines-hidden
2734 @opindex fvisibility-inlines-hidden
2735 This switch declares that the user does not attempt to compare
2736 pointers to inline functions or methods where the addresses of the two functions
2737 are taken in different shared objects.
2739 The effect of this is that GCC may, effectively, mark inline methods with
2740 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2741 appear in the export table of a DSO and do not require a PLT indirection
2742 when used within the DSO@. Enabling this option can have a dramatic effect
2743 on load and link times of a DSO as it massively reduces the size of the
2744 dynamic export table when the library makes heavy use of templates.
2746 The behavior of this switch is not quite the same as marking the
2747 methods as hidden directly, because it does not affect static variables
2748 local to the function or cause the compiler to deduce that
2749 the function is defined in only one shared object.
2751 You may mark a method as having a visibility explicitly to negate the
2752 effect of the switch for that method. For example, if you do want to
2753 compare pointers to a particular inline method, you might mark it as
2754 having default visibility. Marking the enclosing class with explicit
2755 visibility has no effect.
2757 Explicitly instantiated inline methods are unaffected by this option
2758 as their linkage might otherwise cross a shared library boundary.
2759 @xref{Template Instantiation}.
2761 @item -fvisibility-ms-compat
2762 @opindex fvisibility-ms-compat
2763 This flag attempts to use visibility settings to make GCC's C++
2764 linkage model compatible with that of Microsoft Visual Studio.
2766 The flag makes these changes to GCC's linkage model:
2770 It sets the default visibility to @code{hidden}, like
2771 @option{-fvisibility=hidden}.
2774 Types, but not their members, are not hidden by default.
2777 The One Definition Rule is relaxed for types without explicit
2778 visibility specifications that are defined in more than one
2779 shared object: those declarations are permitted if they are
2780 permitted when this option is not used.
2783 In new code it is better to use @option{-fvisibility=hidden} and
2784 export those classes that are intended to be externally visible.
2785 Unfortunately it is possible for code to rely, perhaps accidentally,
2786 on the Visual Studio behavior.
2788 Among the consequences of these changes are that static data members
2789 of the same type with the same name but defined in different shared
2790 objects are different, so changing one does not change the other;
2791 and that pointers to function members defined in different shared
2792 objects may not compare equal. When this flag is given, it is a
2793 violation of the ODR to define types with the same name differently.
2797 Do not use weak symbol support, even if it is provided by the linker.
2798 By default, G++ uses weak symbols if they are available. This
2799 option exists only for testing, and should not be used by end-users;
2800 it results in inferior code and has no benefits. This option may
2801 be removed in a future release of G++.
2805 Do not search for header files in the standard directories specific to
2806 C++, but do still search the other standard directories. (This option
2807 is used when building the C++ library.)
2810 In addition, these optimization, warning, and code generation options
2811 have meanings only for C++ programs:
2814 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2817 Warn when G++ it generates code that is probably not compatible with
2818 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2819 ABI with each major release, normally @option{-Wabi} will warn only if
2820 there is a check added later in a release series for an ABI issue
2821 discovered since the initial release. @option{-Wabi} will warn about
2822 more things if an older ABI version is selected (with
2823 @option{-fabi-version=@var{n}}).
2825 @option{-Wabi} can also be used with an explicit version number to
2826 warn about compatibility with a particular @option{-fabi-version}
2827 level, e.g.@: @option{-Wabi=2} to warn about changes relative to
2828 @option{-fabi-version=2}.
2830 If an explicit version number is provided and
2831 @option{-fabi-compat-version} is not specified, the version number
2832 from this option is used for compatibility aliases. If no explicit
2833 version number is provided with this option, but
2834 @option{-fabi-compat-version} is specified, that version number is
2835 used for ABI warnings.
2837 Although an effort has been made to warn about
2838 all such cases, there are probably some cases that are not warned about,
2839 even though G++ is generating incompatible code. There may also be
2840 cases where warnings are emitted even though the code that is generated
2843 You should rewrite your code to avoid these warnings if you are
2844 concerned about the fact that code generated by G++ may not be binary
2845 compatible with code generated by other compilers.
2847 Known incompatibilities in @option{-fabi-version=2} (which was the
2848 default from GCC 3.4 to 4.9) include:
2853 A template with a non-type template parameter of reference type was
2854 mangled incorrectly:
2857 template <int &> struct S @{@};
2861 This was fixed in @option{-fabi-version=3}.
2864 SIMD vector types declared using @code{__attribute ((vector_size))} were
2865 mangled in a non-standard way that does not allow for overloading of
2866 functions taking vectors of different sizes.
2868 The mangling was changed in @option{-fabi-version=4}.
2871 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2872 qualifiers, and @code{decltype} of a plain declaration was folded away.
2874 These mangling issues were fixed in @option{-fabi-version=5}.
2877 Scoped enumerators passed as arguments to a variadic function are
2878 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2879 On most targets this does not actually affect the parameter passing
2880 ABI, as there is no way to pass an argument smaller than @code{int}.
2882 Also, the ABI changed the mangling of template argument packs,
2883 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2884 a class scope function used as a template argument.
2886 These issues were corrected in @option{-fabi-version=6}.
2889 Lambdas in default argument scope were mangled incorrectly, and the
2890 ABI changed the mangling of @code{nullptr_t}.
2892 These issues were corrected in @option{-fabi-version=7}.
2895 When mangling a function type with function-cv-qualifiers, the
2896 un-qualified function type was incorrectly treated as a substitution
2899 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2902 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2903 unaligned accesses. Note that this did not affect the ABI of a
2904 function with a @code{nullptr_t} parameter, as parameters have a
2907 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2910 Target-specific attributes that affect the identity of a type, such as
2911 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2912 did not affect the mangled name, leading to name collisions when
2913 function pointers were used as template arguments.
2915 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2919 It also warns about psABI-related changes. The known psABI changes at this
2925 For SysV/x86-64, unions with @code{long double} members are
2926 passed in memory as specified in psABI. For example:
2936 @code{union U} is always passed in memory.
2940 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2943 Warn when a type with an ABI tag is used in a context that does not
2944 have that ABI tag. See @ref{C++ Attributes} for more information
2947 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2948 @opindex Wctor-dtor-privacy
2949 @opindex Wno-ctor-dtor-privacy
2950 Warn when a class seems unusable because all the constructors or
2951 destructors in that class are private, and it has neither friends nor
2952 public static member functions. Also warn if there are no non-private
2953 methods, and there's at least one private member function that isn't
2954 a constructor or destructor.
2956 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2957 @opindex Wdelete-non-virtual-dtor
2958 @opindex Wno-delete-non-virtual-dtor
2959 Warn when @code{delete} is used to destroy an instance of a class that
2960 has virtual functions and non-virtual destructor. It is unsafe to delete
2961 an instance of a derived class through a pointer to a base class if the
2962 base class does not have a virtual destructor. This warning is enabled
2965 @item -Wdeprecated-copy @r{(C++ and Objective-C++ only)}
2966 @opindex Wdeprecated-copy
2967 @opindex Wno-deprecated-copy
2968 Warn that the implicit declaration of a copy constructor or copy
2969 assignment operator is deprecated if the class has a user-provided
2970 copy constructor, copy assignment operator, or destructor, in C++11
2971 and up. This warning is enabled by @option{-Wall}.
2973 @item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)}
2974 @opindex Winit-list-lifetime
2975 @opindex Wno-init-list-lifetime
2976 Do not warn about uses of @code{std::initializer_list} that are likely
2977 to result in dangling pointers. Since the underlying array for an
2978 @code{initializer_list} is handled like a normal C++ temporary object,
2979 it is easy to inadvertently keep a pointer to the array past the end
2980 of the array's lifetime. For example:
2984 If a function returns a temporary @code{initializer_list}, or a local
2985 @code{initializer_list} variable, the array's lifetime ends at the end
2986 of the return statement, so the value returned has a dangling pointer.
2989 If a new-expression creates an @code{initializer_list}, the array only
2990 lives until the end of the enclosing full-expression, so the
2991 @code{initializer_list} in the heap has a dangling pointer.
2994 When an @code{initializer_list} variable is assigned from a
2995 brace-enclosed initializer list, the temporary array created for the
2996 right side of the assignment only lives until the end of the
2997 full-expression, so at the next statement the @code{initializer_list}
2998 variable has a dangling pointer.
3001 // li's initial underlying array lives as long as li
3002 std::initializer_list<int> li = @{ 1,2,3 @};
3003 // assignment changes li to point to a temporary array
3005 // now the temporary is gone and li has a dangling pointer
3006 int i = li.begin()[0] // undefined behavior
3010 When a list constructor stores the @code{begin} pointer from the
3011 @code{initializer_list} argument, this doesn't extend the lifetime of
3012 the array, so if a class variable is constructed from a temporary
3013 @code{initializer_list}, the pointer is left dangling by the end of
3014 the variable declaration statement.
3018 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
3019 @opindex Wliteral-suffix
3020 @opindex Wno-literal-suffix
3021 Warn when a string or character literal is followed by a ud-suffix which does
3022 not begin with an underscore. As a conforming extension, GCC treats such
3023 suffixes as separate preprocessing tokens in order to maintain backwards
3024 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
3028 #define __STDC_FORMAT_MACROS
3029 #include <inttypes.h>
3034 printf("My int64: %" PRId64"\n", i64);
3038 In this case, @code{PRId64} is treated as a separate preprocessing token.
3040 Additionally, warn when a user-defined literal operator is declared with
3041 a literal suffix identifier that doesn't begin with an underscore. Literal
3042 suffix identifiers that don't begin with an underscore are reserved for
3043 future standardization.
3045 This warning is enabled by default.
3047 @item -Wlto-type-mismatch
3048 @opindex Wlto-type-mismatch
3049 @opindex Wno-lto-type-mismatch
3051 During the link-time optimization warn about type mismatches in
3052 global declarations from different compilation units.
3053 Requires @option{-flto} to be enabled. Enabled by default.
3055 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
3057 @opindex Wno-narrowing
3058 For C++11 and later standards, narrowing conversions are diagnosed by default,
3059 as required by the standard. A narrowing conversion from a constant produces
3060 an error, and a narrowing conversion from a non-constant produces a warning,
3061 but @option{-Wno-narrowing} suppresses the diagnostic.
3062 Note that this does not affect the meaning of well-formed code;
3063 narrowing conversions are still considered ill-formed in SFINAE contexts.
3065 With @option{-Wnarrowing} in C++98, warn when a narrowing
3066 conversion prohibited by C++11 occurs within
3070 int i = @{ 2.2 @}; // error: narrowing from double to int
3073 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
3075 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
3077 @opindex Wno-noexcept
3078 Warn when a noexcept-expression evaluates to false because of a call
3079 to a function that does not have a non-throwing exception
3080 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
3081 the compiler to never throw an exception.
3083 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
3084 @opindex Wnoexcept-type
3085 @opindex Wno-noexcept-type
3086 Warn if the C++17 feature making @code{noexcept} part of a function
3087 type changes the mangled name of a symbol relative to C++14. Enabled
3088 by @option{-Wabi} and @option{-Wc++17-compat}.
3093 template <class T> void f(T t) @{ t(); @};
3095 void h() @{ f(g); @}
3099 In C++14, @code{f} calls @code{f<void(*)()>}, but in
3100 C++17 it calls @code{f<void(*)()noexcept>}.
3102 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3103 @opindex Wclass-memaccess
3104 @opindex Wno-class-memaccess
3105 Warn when the destination of a call to a raw memory function such as
3106 @code{memset} or @code{memcpy} is an object of class type, and when writing
3107 into such an object might bypass the class non-trivial or deleted constructor
3108 or copy assignment, violate const-correctness or encapsulation, or corrupt
3109 virtual table pointers. Modifying the representation of such objects may
3110 violate invariants maintained by member functions of the class. For example,
3111 the call to @code{memset} below is undefined because it modifies a non-trivial
3112 class object and is, therefore, diagnosed. The safe way to either initialize
3113 or clear the storage of objects of such types is by using the appropriate
3114 constructor or assignment operator, if one is available.
3116 std::string str = "abc";
3117 memset (&str, 0, sizeof str);
3119 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3120 Explicitly casting the pointer to the class object to @code{void *} or
3121 to a type that can be safely accessed by the raw memory function suppresses
3124 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3125 @opindex Wnon-virtual-dtor
3126 @opindex Wno-non-virtual-dtor
3127 Warn when a class has virtual functions and an accessible non-virtual
3128 destructor itself or in an accessible polymorphic base class, in which
3129 case it is possible but unsafe to delete an instance of a derived
3130 class through a pointer to the class itself or base class. This
3131 warning is automatically enabled if @option{-Weffc++} is specified.
3133 @item -Wregister @r{(C++ and Objective-C++ only)}
3135 @opindex Wno-register
3136 Warn on uses of the @code{register} storage class specifier, except
3137 when it is part of the GNU @ref{Explicit Register Variables} extension.
3138 The use of the @code{register} keyword as storage class specifier has
3139 been deprecated in C++11 and removed in C++17.
3140 Enabled by default with @option{-std=c++17}.
3142 @item -Wreorder @r{(C++ and Objective-C++ only)}
3144 @opindex Wno-reorder
3145 @cindex reordering, warning
3146 @cindex warning for reordering of member initializers
3147 Warn when the order of member initializers given in the code does not
3148 match the order in which they must be executed. For instance:
3154 A(): j (0), i (1) @{ @}
3159 The compiler rearranges the member initializers for @code{i}
3160 and @code{j} to match the declaration order of the members, emitting
3161 a warning to that effect. This warning is enabled by @option{-Wall}.
3163 @item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)}
3164 @opindex Wpessimizing-move
3165 @opindex Wno-pessimizing-move
3166 This warning warns when a call to @code{std::move} prevents copy
3167 elision. A typical scenario when copy elision can occur is when returning in
3168 a function with a class return type, when the expression being returned is the
3169 name of a non-volatile automatic object, and is not a function parameter, and
3170 has the same type as the function return type.
3180 return std::move (t);
3184 But in this example, the @code{std::move} call prevents copy elision.
3186 This warning is enabled by @option{-Wall}.
3188 @item -Wno-redundant-move @r{(C++ and Objective-C++ only)}
3189 @opindex Wredundant-move
3190 @opindex Wno-redundant-move
3191 This warning warns about redundant calls to @code{std::move}; that is, when
3192 a move operation would have been performed even without the @code{std::move}
3193 call. This happens because the compiler is forced to treat the object as if
3194 it were an rvalue in certain situations such as returning a local variable,
3195 where copy elision isn't applicable. Consider:
3204 return std::move (t);
3208 Here, the @code{std::move} call is redundant. Because G++ implements Core
3209 Issue 1579, another example is:
3212 struct T @{ // convertible to U
3222 return std::move (t);
3225 In this example, copy elision isn't applicable because the type of the
3226 expression being returned and the function return type differ, yet G++
3227 treats the return value as if it were designated by an rvalue.
3229 This warning is enabled by @option{-Wextra}.
3231 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3232 @opindex fext-numeric-literals
3233 @opindex fno-ext-numeric-literals
3234 Accept imaginary, fixed-point, or machine-defined
3235 literal number suffixes as GNU extensions.
3236 When this option is turned off these suffixes are treated
3237 as C++11 user-defined literal numeric suffixes.
3238 This is on by default for all pre-C++11 dialects and all GNU dialects:
3239 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3240 @option{-std=gnu++14}.
3241 This option is off by default
3242 for ISO C++11 onwards (@option{-std=c++11}, ...).
3245 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3248 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3251 Warn about violations of the following style guidelines from Scott Meyers'
3252 @cite{Effective C++} series of books:
3256 Define a copy constructor and an assignment operator for classes
3257 with dynamically-allocated memory.
3260 Prefer initialization to assignment in constructors.
3263 Have @code{operator=} return a reference to @code{*this}.
3266 Don't try to return a reference when you must return an object.
3269 Distinguish between prefix and postfix forms of increment and
3270 decrement operators.
3273 Never overload @code{&&}, @code{||}, or @code{,}.
3277 This option also enables @option{-Wnon-virtual-dtor}, which is also
3278 one of the effective C++ recommendations. However, the check is
3279 extended to warn about the lack of virtual destructor in accessible
3280 non-polymorphic bases classes too.
3282 When selecting this option, be aware that the standard library
3283 headers do not obey all of these guidelines; use @samp{grep -v}
3284 to filter out those warnings.
3286 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3287 @opindex Wstrict-null-sentinel
3288 @opindex Wno-strict-null-sentinel
3289 Warn about the use of an uncasted @code{NULL} as sentinel. When
3290 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3291 to @code{__null}. Although it is a null pointer constant rather than a
3292 null pointer, it is guaranteed to be of the same size as a pointer.
3293 But this use is not portable across different compilers.
3295 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3296 @opindex Wno-non-template-friend
3297 @opindex Wnon-template-friend
3298 Disable warnings when non-template friend functions are declared
3299 within a template. In very old versions of GCC that predate implementation
3300 of the ISO standard, declarations such as
3301 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3302 could be interpreted as a particular specialization of a template
3303 function; the warning exists to diagnose compatibility problems,
3304 and is enabled by default.
3306 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3307 @opindex Wold-style-cast
3308 @opindex Wno-old-style-cast
3309 Warn if an old-style (C-style) cast to a non-void type is used within
3310 a C++ program. The new-style casts (@code{dynamic_cast},
3311 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3312 less vulnerable to unintended effects and much easier to search for.
3314 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3315 @opindex Woverloaded-virtual
3316 @opindex Wno-overloaded-virtual
3317 @cindex overloaded virtual function, warning
3318 @cindex warning for overloaded virtual function
3319 Warn when a function declaration hides virtual functions from a
3320 base class. For example, in:
3327 struct B: public A @{
3332 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3343 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3344 @opindex Wno-pmf-conversions
3345 @opindex Wpmf-conversions
3346 Disable the diagnostic for converting a bound pointer to member function
3349 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3350 @opindex Wsign-promo
3351 @opindex Wno-sign-promo
3352 Warn when overload resolution chooses a promotion from unsigned or
3353 enumerated type to a signed type, over a conversion to an unsigned type of
3354 the same size. Previous versions of G++ tried to preserve
3355 unsignedness, but the standard mandates the current behavior.
3357 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3359 @opindex Wno-templates
3360 Warn when a primary template declaration is encountered. Some coding
3361 rules disallow templates, and this may be used to enforce that rule.
3362 The warning is inactive inside a system header file, such as the STL, so
3363 one can still use the STL. One may also instantiate or specialize
3366 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3367 @opindex Wmultiple-inheritance
3368 @opindex Wno-multiple-inheritance
3369 Warn when a class is defined with multiple direct base classes. Some
3370 coding rules disallow multiple inheritance, and this may be used to
3371 enforce that rule. The warning is inactive inside a system header file,
3372 such as the STL, so one can still use the STL. One may also define
3373 classes that indirectly use multiple inheritance.
3375 @item -Wvirtual-inheritance
3376 @opindex Wvirtual-inheritance
3377 @opindex Wno-virtual-inheritance
3378 Warn when a class is defined with a virtual direct base class. Some
3379 coding rules disallow multiple inheritance, and this may be used to
3380 enforce that rule. The warning is inactive inside a system header file,
3381 such as the STL, so one can still use the STL. One may also define
3382 classes that indirectly use virtual inheritance.
3385 @opindex Wnamespaces
3386 @opindex Wno-namespaces
3387 Warn when a namespace definition is opened. Some coding rules disallow
3388 namespaces, and this may be used to enforce that rule. The warning is
3389 inactive inside a system header file, such as the STL, so one can still
3390 use the STL. One may also use using directives and qualified names.
3392 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3394 @opindex Wno-terminate
3395 Disable the warning about a throw-expression that will immediately
3396 result in a call to @code{terminate}.
3398 @item -Wno-class-conversion @r{(C++ and Objective-C++ only)}
3399 @opindex Wno-class-conversion
3400 @opindex Wclass-conversion
3401 Disable the warning about the case when a conversion function converts an
3402 object to the same type, to a base class of that type, or to void; such
3403 a conversion function will never be called.
3406 @node Objective-C and Objective-C++ Dialect Options
3407 @section Options Controlling Objective-C and Objective-C++ Dialects
3409 @cindex compiler options, Objective-C and Objective-C++
3410 @cindex Objective-C and Objective-C++ options, command-line
3411 @cindex options, Objective-C and Objective-C++
3412 (NOTE: This manual does not describe the Objective-C and Objective-C++
3413 languages themselves. @xref{Standards,,Language Standards
3414 Supported by GCC}, for references.)
3416 This section describes the command-line options that are only meaningful
3417 for Objective-C and Objective-C++ programs. You can also use most of
3418 the language-independent GNU compiler options.
3419 For example, you might compile a file @file{some_class.m} like this:
3422 gcc -g -fgnu-runtime -O -c some_class.m
3426 In this example, @option{-fgnu-runtime} is an option meant only for
3427 Objective-C and Objective-C++ programs; you can use the other options with
3428 any language supported by GCC@.
3430 Note that since Objective-C is an extension of the C language, Objective-C
3431 compilations may also use options specific to the C front-end (e.g.,
3432 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3433 C++-specific options (e.g., @option{-Wabi}).
3435 Here is a list of options that are @emph{only} for compiling Objective-C
3436 and Objective-C++ programs:
3439 @item -fconstant-string-class=@var{class-name}
3440 @opindex fconstant-string-class
3441 Use @var{class-name} as the name of the class to instantiate for each
3442 literal string specified with the syntax @code{@@"@dots{}"}. The default
3443 class name is @code{NXConstantString} if the GNU runtime is being used, and
3444 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3445 @option{-fconstant-cfstrings} option, if also present, overrides the
3446 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3447 to be laid out as constant CoreFoundation strings.
3450 @opindex fgnu-runtime
3451 Generate object code compatible with the standard GNU Objective-C
3452 runtime. This is the default for most types of systems.
3454 @item -fnext-runtime
3455 @opindex fnext-runtime
3456 Generate output compatible with the NeXT runtime. This is the default
3457 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3458 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3461 @item -fno-nil-receivers
3462 @opindex fno-nil-receivers
3463 Assume that all Objective-C message dispatches (@code{[receiver
3464 message:arg]}) in this translation unit ensure that the receiver is
3465 not @code{nil}. This allows for more efficient entry points in the
3466 runtime to be used. This option is only available in conjunction with
3467 the NeXT runtime and ABI version 0 or 1.
3469 @item -fobjc-abi-version=@var{n}
3470 @opindex fobjc-abi-version
3471 Use version @var{n} of the Objective-C ABI for the selected runtime.
3472 This option is currently supported only for the NeXT runtime. In that
3473 case, Version 0 is the traditional (32-bit) ABI without support for
3474 properties and other Objective-C 2.0 additions. Version 1 is the
3475 traditional (32-bit) ABI with support for properties and other
3476 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3477 nothing is specified, the default is Version 0 on 32-bit target
3478 machines, and Version 2 on 64-bit target machines.
3480 @item -fobjc-call-cxx-cdtors
3481 @opindex fobjc-call-cxx-cdtors
3482 For each Objective-C class, check if any of its instance variables is a
3483 C++ object with a non-trivial default constructor. If so, synthesize a
3484 special @code{- (id) .cxx_construct} instance method which runs
3485 non-trivial default constructors on any such instance variables, in order,
3486 and then return @code{self}. Similarly, check if any instance variable
3487 is a C++ object with a non-trivial destructor, and if so, synthesize a
3488 special @code{- (void) .cxx_destruct} method which runs
3489 all such default destructors, in reverse order.
3491 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3492 methods thusly generated only operate on instance variables
3493 declared in the current Objective-C class, and not those inherited
3494 from superclasses. It is the responsibility of the Objective-C
3495 runtime to invoke all such methods in an object's inheritance
3496 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3497 by the runtime immediately after a new object instance is allocated;
3498 the @code{- (void) .cxx_destruct} methods are invoked immediately
3499 before the runtime deallocates an object instance.
3501 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3502 support for invoking the @code{- (id) .cxx_construct} and
3503 @code{- (void) .cxx_destruct} methods.
3505 @item -fobjc-direct-dispatch
3506 @opindex fobjc-direct-dispatch
3507 Allow fast jumps to the message dispatcher. On Darwin this is
3508 accomplished via the comm page.
3510 @item -fobjc-exceptions
3511 @opindex fobjc-exceptions
3512 Enable syntactic support for structured exception handling in
3513 Objective-C, similar to what is offered by C++. This option
3514 is required to use the Objective-C keywords @code{@@try},
3515 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3516 @code{@@synchronized}. This option is available with both the GNU
3517 runtime and the NeXT runtime (but not available in conjunction with
3518 the NeXT runtime on Mac OS X 10.2 and earlier).
3522 Enable garbage collection (GC) in Objective-C and Objective-C++
3523 programs. This option is only available with the NeXT runtime; the
3524 GNU runtime has a different garbage collection implementation that
3525 does not require special compiler flags.
3527 @item -fobjc-nilcheck
3528 @opindex fobjc-nilcheck
3529 For the NeXT runtime with version 2 of the ABI, check for a nil
3530 receiver in method invocations before doing the actual method call.
3531 This is the default and can be disabled using
3532 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3533 checked for nil in this way no matter what this flag is set to.
3534 Currently this flag does nothing when the GNU runtime, or an older
3535 version of the NeXT runtime ABI, is used.
3537 @item -fobjc-std=objc1
3539 Conform to the language syntax of Objective-C 1.0, the language
3540 recognized by GCC 4.0. This only affects the Objective-C additions to
3541 the C/C++ language; it does not affect conformance to C/C++ standards,
3542 which is controlled by the separate C/C++ dialect option flags. When
3543 this option is used with the Objective-C or Objective-C++ compiler,
3544 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3545 This is useful if you need to make sure that your Objective-C code can
3546 be compiled with older versions of GCC@.
3548 @item -freplace-objc-classes
3549 @opindex freplace-objc-classes
3550 Emit a special marker instructing @command{ld(1)} not to statically link in
3551 the resulting object file, and allow @command{dyld(1)} to load it in at
3552 run time instead. This is used in conjunction with the Fix-and-Continue
3553 debugging mode, where the object file in question may be recompiled and
3554 dynamically reloaded in the course of program execution, without the need
3555 to restart the program itself. Currently, Fix-and-Continue functionality
3556 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3561 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3562 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3563 compile time) with static class references that get initialized at load time,
3564 which improves run-time performance. Specifying the @option{-fzero-link} flag
3565 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3566 to be retained. This is useful in Zero-Link debugging mode, since it allows
3567 for individual class implementations to be modified during program execution.
3568 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3569 regardless of command-line options.
3571 @item -fno-local-ivars
3572 @opindex fno-local-ivars
3573 @opindex flocal-ivars
3574 By default instance variables in Objective-C can be accessed as if
3575 they were local variables from within the methods of the class they're
3576 declared in. This can lead to shadowing between instance variables
3577 and other variables declared either locally inside a class method or
3578 globally with the same name. Specifying the @option{-fno-local-ivars}
3579 flag disables this behavior thus avoiding variable shadowing issues.
3581 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3582 @opindex fivar-visibility
3583 Set the default instance variable visibility to the specified option
3584 so that instance variables declared outside the scope of any access
3585 modifier directives default to the specified visibility.
3589 Dump interface declarations for all classes seen in the source file to a
3590 file named @file{@var{sourcename}.decl}.
3592 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3593 @opindex Wassign-intercept
3594 @opindex Wno-assign-intercept
3595 Warn whenever an Objective-C assignment is being intercepted by the
3598 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3599 @opindex Wno-protocol
3601 If a class is declared to implement a protocol, a warning is issued for
3602 every method in the protocol that is not implemented by the class. The
3603 default behavior is to issue a warning for every method not explicitly
3604 implemented in the class, even if a method implementation is inherited
3605 from the superclass. If you use the @option{-Wno-protocol} option, then
3606 methods inherited from the superclass are considered to be implemented,
3607 and no warning is issued for them.
3609 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3611 @opindex Wno-selector
3612 Warn if multiple methods of different types for the same selector are
3613 found during compilation. The check is performed on the list of methods
3614 in the final stage of compilation. Additionally, a check is performed
3615 for each selector appearing in a @code{@@selector(@dots{})}
3616 expression, and a corresponding method for that selector has been found
3617 during compilation. Because these checks scan the method table only at
3618 the end of compilation, these warnings are not produced if the final
3619 stage of compilation is not reached, for example because an error is
3620 found during compilation, or because the @option{-fsyntax-only} option is
3623 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3624 @opindex Wstrict-selector-match
3625 @opindex Wno-strict-selector-match
3626 Warn if multiple methods with differing argument and/or return types are
3627 found for a given selector when attempting to send a message using this
3628 selector to a receiver of type @code{id} or @code{Class}. When this flag
3629 is off (which is the default behavior), the compiler omits such warnings
3630 if any differences found are confined to types that share the same size
3633 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3634 @opindex Wundeclared-selector
3635 @opindex Wno-undeclared-selector
3636 Warn if a @code{@@selector(@dots{})} expression referring to an
3637 undeclared selector is found. A selector is considered undeclared if no
3638 method with that name has been declared before the
3639 @code{@@selector(@dots{})} expression, either explicitly in an
3640 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3641 an @code{@@implementation} section. This option always performs its
3642 checks as soon as a @code{@@selector(@dots{})} expression is found,
3643 while @option{-Wselector} only performs its checks in the final stage of
3644 compilation. This also enforces the coding style convention
3645 that methods and selectors must be declared before being used.
3647 @item -print-objc-runtime-info
3648 @opindex print-objc-runtime-info
3649 Generate C header describing the largest structure that is passed by
3654 @node Diagnostic Message Formatting Options
3655 @section Options to Control Diagnostic Messages Formatting
3656 @cindex options to control diagnostics formatting
3657 @cindex diagnostic messages
3658 @cindex message formatting
3660 Traditionally, diagnostic messages have been formatted irrespective of
3661 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3662 options described below
3663 to control the formatting algorithm for diagnostic messages,
3664 e.g.@: how many characters per line, how often source location
3665 information should be reported. Note that some language front ends may not
3666 honor these options.
3669 @item -fmessage-length=@var{n}
3670 @opindex fmessage-length
3671 Try to format error messages so that they fit on lines of about
3672 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3673 done; each error message appears on a single line. This is the
3674 default for all front ends.
3676 Note - this option also affects the display of the @samp{#error} and
3677 @samp{#warning} pre-processor directives, and the @samp{deprecated}
3678 function/type/variable attribute. It does not however affect the
3679 @samp{pragma GCC warning} and @samp{pragma GCC error} pragmas.
3681 @item -fdiagnostics-show-location=once
3682 @opindex fdiagnostics-show-location
3683 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3684 reporter to emit source location information @emph{once}; that is, in
3685 case the message is too long to fit on a single physical line and has to
3686 be wrapped, the source location won't be emitted (as prefix) again,
3687 over and over, in subsequent continuation lines. This is the default
3690 @item -fdiagnostics-show-location=every-line
3691 Only meaningful in line-wrapping mode. Instructs the diagnostic
3692 messages reporter to emit the same source location information (as
3693 prefix) for physical lines that result from the process of breaking
3694 a message which is too long to fit on a single line.
3696 @item -fdiagnostics-color[=@var{WHEN}]
3697 @itemx -fno-diagnostics-color
3698 @opindex fdiagnostics-color
3699 @cindex highlight, color
3700 @vindex GCC_COLORS @r{environment variable}
3701 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3702 or @samp{auto}. The default depends on how the compiler has been configured,
3703 it can be any of the above @var{WHEN} options or also @samp{never}
3704 if @env{GCC_COLORS} environment variable isn't present in the environment,
3705 and @samp{auto} otherwise.
3706 @samp{auto} means to use color only when the standard error is a terminal.
3707 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3708 aliases for @option{-fdiagnostics-color=always} and
3709 @option{-fdiagnostics-color=never}, respectively.
3711 The colors are defined by the environment variable @env{GCC_COLORS}.
3712 Its value is a colon-separated list of capabilities and Select Graphic
3713 Rendition (SGR) substrings. SGR commands are interpreted by the
3714 terminal or terminal emulator. (See the section in the documentation
3715 of your text terminal for permitted values and their meanings as
3716 character attributes.) These substring values are integers in decimal
3717 representation and can be concatenated with semicolons.
3718 Common values to concatenate include
3720 @samp{4} for underline,
3722 @samp{7} for inverse,
3723 @samp{39} for default foreground color,
3724 @samp{30} to @samp{37} for foreground colors,
3725 @samp{90} to @samp{97} for 16-color mode foreground colors,
3726 @samp{38;5;0} to @samp{38;5;255}
3727 for 88-color and 256-color modes foreground colors,
3728 @samp{49} for default background color,
3729 @samp{40} to @samp{47} for background colors,
3730 @samp{100} to @samp{107} for 16-color mode background colors,
3731 and @samp{48;5;0} to @samp{48;5;255}
3732 for 88-color and 256-color modes background colors.
3734 The default @env{GCC_COLORS} is
3736 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3737 quote=01:fixit-insert=32:fixit-delete=31:\
3738 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3742 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3743 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3744 @samp{01} is bold, and @samp{31} is red.
3745 Setting @env{GCC_COLORS} to the empty string disables colors.
3746 Supported capabilities are as follows.
3750 @vindex error GCC_COLORS @r{capability}
3751 SGR substring for error: markers.
3754 @vindex warning GCC_COLORS @r{capability}
3755 SGR substring for warning: markers.
3758 @vindex note GCC_COLORS @r{capability}
3759 SGR substring for note: markers.
3762 @vindex range1 GCC_COLORS @r{capability}
3763 SGR substring for first additional range.
3766 @vindex range2 GCC_COLORS @r{capability}
3767 SGR substring for second additional range.
3770 @vindex locus GCC_COLORS @r{capability}
3771 SGR substring for location information, @samp{file:line} or
3772 @samp{file:line:column} etc.
3775 @vindex quote GCC_COLORS @r{capability}
3776 SGR substring for information printed within quotes.
3779 @vindex fixit-insert GCC_COLORS @r{capability}
3780 SGR substring for fix-it hints suggesting text to
3781 be inserted or replaced.
3784 @vindex fixit-delete GCC_COLORS @r{capability}
3785 SGR substring for fix-it hints suggesting text to
3788 @item diff-filename=
3789 @vindex diff-filename GCC_COLORS @r{capability}
3790 SGR substring for filename headers within generated patches.
3793 @vindex diff-hunk GCC_COLORS @r{capability}
3794 SGR substring for the starts of hunks within generated patches.
3797 @vindex diff-delete GCC_COLORS @r{capability}
3798 SGR substring for deleted lines within generated patches.
3801 @vindex diff-insert GCC_COLORS @r{capability}
3802 SGR substring for inserted lines within generated patches.
3805 @vindex type-diff GCC_COLORS @r{capability}
3806 SGR substring for highlighting mismatching types within template
3807 arguments in the C++ frontend.
3810 @item -fno-diagnostics-show-option
3811 @opindex fno-diagnostics-show-option
3812 @opindex fdiagnostics-show-option
3813 By default, each diagnostic emitted includes text indicating the
3814 command-line option that directly controls the diagnostic (if such an
3815 option is known to the diagnostic machinery). Specifying the
3816 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3818 @item -fno-diagnostics-show-caret
3819 @opindex fno-diagnostics-show-caret
3820 @opindex fdiagnostics-show-caret
3821 By default, each diagnostic emitted includes the original source line
3822 and a caret @samp{^} indicating the column. This option suppresses this
3823 information. The source line is truncated to @var{n} characters, if
3824 the @option{-fmessage-length=n} option is given. When the output is done
3825 to the terminal, the width is limited to the width given by the
3826 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3828 @item -fno-diagnostics-show-labels
3829 @opindex fno-diagnostics-show-labels
3830 @opindex fdiagnostics-show-labels
3831 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3832 diagnostics can label ranges of source code with pertinent information, such
3833 as the types of expressions:
3836 printf ("foo %s bar", long_i + long_j);
3842 This option suppresses the printing of these labels (in the example above,
3843 the vertical bars and the ``char *'' and ``long int'' text).
3845 @item -fno-diagnostics-show-line-numbers
3846 @opindex fno-diagnostics-show-line-numbers
3847 @opindex fdiagnostics-show-line-numbers
3848 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3849 a left margin is printed, showing line numbers. This option suppresses this
3852 @item -fdiagnostics-minimum-margin-width=@var{width}
3853 @opindex fdiagnostics-minimum-margin-width
3854 This option controls the minimum width of the left margin printed by
3855 @option{-fdiagnostics-show-line-numbers}. It defaults to 6.
3857 @item -fdiagnostics-parseable-fixits
3858 @opindex fdiagnostics-parseable-fixits
3859 Emit fix-it hints in a machine-parseable format, suitable for consumption
3860 by IDEs. For each fix-it, a line will be printed after the relevant
3861 diagnostic, starting with the string ``fix-it:''. For example:
3864 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3867 The location is expressed as a half-open range, expressed as a count of
3868 bytes, starting at byte 1 for the initial column. In the above example,
3869 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3873 00000000011111111112222222222
3874 12345678901234567890123456789
3875 gtk_widget_showall (dlg);
3880 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3881 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3882 (e.g. vertical tab as ``\013'').
3884 An empty replacement string indicates that the given range is to be removed.
3885 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3886 be inserted at the given position.
3888 @item -fdiagnostics-generate-patch
3889 @opindex fdiagnostics-generate-patch
3890 Print fix-it hints to stderr in unified diff format, after any diagnostics
3891 are printed. For example:
3898 void show_cb(GtkDialog *dlg)
3900 - gtk_widget_showall(dlg);
3901 + gtk_widget_show_all(dlg);
3906 The diff may or may not be colorized, following the same rules
3907 as for diagnostics (see @option{-fdiagnostics-color}).
3909 @item -fdiagnostics-show-template-tree
3910 @opindex fdiagnostics-show-template-tree
3912 In the C++ frontend, when printing diagnostics showing mismatching
3913 template types, such as:
3916 could not convert 'std::map<int, std::vector<double> >()'
3917 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3920 the @option{-fdiagnostics-show-template-tree} flag enables printing a
3921 tree-like structure showing the common and differing parts of the types,
3931 The parts that differ are highlighted with color (``double'' and
3932 ``float'' in this case).
3934 @item -fno-elide-type
3935 @opindex fno-elide-type
3936 @opindex felide-type
3937 By default when the C++ frontend prints diagnostics showing mismatching
3938 template types, common parts of the types are printed as ``[...]'' to
3939 simplify the error message. For example:
3942 could not convert 'std::map<int, std::vector<double> >()'
3943 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
3946 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
3947 This flag also affects the output of the
3948 @option{-fdiagnostics-show-template-tree} flag.
3950 @item -fno-show-column
3951 @opindex fno-show-column
3952 Do not print column numbers in diagnostics. This may be necessary if
3953 diagnostics are being scanned by a program that does not understand the
3954 column numbers, such as @command{dejagnu}.
3958 @node Warning Options
3959 @section Options to Request or Suppress Warnings
3960 @cindex options to control warnings
3961 @cindex warning messages
3962 @cindex messages, warning
3963 @cindex suppressing warnings
3965 Warnings are diagnostic messages that report constructions that
3966 are not inherently erroneous but that are risky or suggest there
3967 may have been an error.
3969 The following language-independent options do not enable specific
3970 warnings but control the kinds of diagnostics produced by GCC@.
3973 @cindex syntax checking
3975 @opindex fsyntax-only
3976 Check the code for syntax errors, but don't do anything beyond that.
3978 @item -fmax-errors=@var{n}
3979 @opindex fmax-errors
3980 Limits the maximum number of error messages to @var{n}, at which point
3981 GCC bails out rather than attempting to continue processing the source
3982 code. If @var{n} is 0 (the default), there is no limit on the number
3983 of error messages produced. If @option{-Wfatal-errors} is also
3984 specified, then @option{-Wfatal-errors} takes precedence over this
3989 Inhibit all warning messages.
3994 Make all warnings into errors.
3999 Make the specified warning into an error. The specifier for a warning
4000 is appended; for example @option{-Werror=switch} turns the warnings
4001 controlled by @option{-Wswitch} into errors. This switch takes a
4002 negative form, to be used to negate @option{-Werror} for specific
4003 warnings; for example @option{-Wno-error=switch} makes
4004 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
4007 The warning message for each controllable warning includes the
4008 option that controls the warning. That option can then be used with
4009 @option{-Werror=} and @option{-Wno-error=} as described above.
4010 (Printing of the option in the warning message can be disabled using the
4011 @option{-fno-diagnostics-show-option} flag.)
4013 Note that specifying @option{-Werror=}@var{foo} automatically implies
4014 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
4017 @item -Wfatal-errors
4018 @opindex Wfatal-errors
4019 @opindex Wno-fatal-errors
4020 This option causes the compiler to abort compilation on the first error
4021 occurred rather than trying to keep going and printing further error
4026 You can request many specific warnings with options beginning with
4027 @samp{-W}, for example @option{-Wimplicit} to request warnings on
4028 implicit declarations. Each of these specific warning options also
4029 has a negative form beginning @samp{-Wno-} to turn off warnings; for
4030 example, @option{-Wno-implicit}. This manual lists only one of the
4031 two forms, whichever is not the default. For further
4032 language-specific options also refer to @ref{C++ Dialect Options} and
4033 @ref{Objective-C and Objective-C++ Dialect Options}.
4035 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
4036 options, such as @option{-Wunused}, which may turn on further options,
4037 such as @option{-Wunused-value}. The combined effect of positive and
4038 negative forms is that more specific options have priority over less
4039 specific ones, independently of their position in the command-line. For
4040 options of the same specificity, the last one takes effect. Options
4041 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
4042 as if they appeared at the end of the command-line.
4044 When an unrecognized warning option is requested (e.g.,
4045 @option{-Wunknown-warning}), GCC emits a diagnostic stating
4046 that the option is not recognized. However, if the @option{-Wno-} form
4047 is used, the behavior is slightly different: no diagnostic is
4048 produced for @option{-Wno-unknown-warning} unless other diagnostics
4049 are being produced. This allows the use of new @option{-Wno-} options
4050 with old compilers, but if something goes wrong, the compiler
4051 warns that an unrecognized option is present.
4058 @opindex Wno-pedantic
4059 Issue all the warnings demanded by strict ISO C and ISO C++;
4060 reject all programs that use forbidden extensions, and some other
4061 programs that do not follow ISO C and ISO C++. For ISO C, follows the
4062 version of the ISO C standard specified by any @option{-std} option used.
4064 Valid ISO C and ISO C++ programs should compile properly with or without
4065 this option (though a rare few require @option{-ansi} or a
4066 @option{-std} option specifying the required version of ISO C)@. However,
4067 without this option, certain GNU extensions and traditional C and C++
4068 features are supported as well. With this option, they are rejected.
4070 @option{-Wpedantic} does not cause warning messages for use of the
4071 alternate keywords whose names begin and end with @samp{__}. Pedantic
4072 warnings are also disabled in the expression that follows
4073 @code{__extension__}. However, only system header files should use
4074 these escape routes; application programs should avoid them.
4075 @xref{Alternate Keywords}.
4077 Some users try to use @option{-Wpedantic} to check programs for strict ISO
4078 C conformance. They soon find that it does not do quite what they want:
4079 it finds some non-ISO practices, but not all---only those for which
4080 ISO C @emph{requires} a diagnostic, and some others for which
4081 diagnostics have been added.
4083 A feature to report any failure to conform to ISO C might be useful in
4084 some instances, but would require considerable additional work and would
4085 be quite different from @option{-Wpedantic}. We don't have plans to
4086 support such a feature in the near future.
4088 Where the standard specified with @option{-std} represents a GNU
4089 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
4090 corresponding @dfn{base standard}, the version of ISO C on which the GNU
4091 extended dialect is based. Warnings from @option{-Wpedantic} are given
4092 where they are required by the base standard. (It does not make sense
4093 for such warnings to be given only for features not in the specified GNU
4094 C dialect, since by definition the GNU dialects of C include all
4095 features the compiler supports with the given option, and there would be
4096 nothing to warn about.)
4098 @item -pedantic-errors
4099 @opindex pedantic-errors
4100 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
4101 requires a diagnostic, in some cases where there is undefined behavior
4102 at compile-time and in some other cases that do not prevent compilation
4103 of programs that are valid according to the standard. This is not
4104 equivalent to @option{-Werror=pedantic}, since there are errors enabled
4105 by this option and not enabled by the latter and vice versa.
4110 This enables all the warnings about constructions that some users
4111 consider questionable, and that are easy to avoid (or modify to
4112 prevent the warning), even in conjunction with macros. This also
4113 enables some language-specific warnings described in @ref{C++ Dialect
4114 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
4116 @option{-Wall} turns on the following warning flags:
4118 @gccoptlist{-Waddress @gol
4119 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
4121 -Wbool-operation @gol
4122 -Wc++11-compat -Wc++14-compat @gol
4123 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
4124 -Wchar-subscripts @gol
4126 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
4127 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
4129 -Wint-in-bool-context @gol
4130 -Wimplicit @r{(C and Objective-C only)} @gol
4131 -Wimplicit-int @r{(C and Objective-C only)} @gol
4132 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
4133 -Winit-self @r{(only for C++)} @gol
4134 -Wlogical-not-parentheses @gol
4135 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
4136 -Wmaybe-uninitialized @gol
4137 -Wmemset-elt-size @gol
4138 -Wmemset-transposed-args @gol
4139 -Wmisleading-indentation @r{(only for C/C++)} @gol
4140 -Wmissing-attributes @gol
4141 -Wmissing-braces @r{(only for C/ObjC)} @gol
4142 -Wmultistatement-macros @gol
4143 -Wnarrowing @r{(only for C++)} @gol
4145 -Wnonnull-compare @gol
4148 -Wpessimizing-move @r{(only for C++)} @gol
4153 -Wsequence-point @gol
4154 -Wsign-compare @r{(only in C++)} @gol
4155 -Wsizeof-pointer-div @gol
4156 -Wsizeof-pointer-memaccess @gol
4157 -Wstrict-aliasing @gol
4158 -Wstrict-overflow=1 @gol
4160 -Wtautological-compare @gol
4162 -Wuninitialized @gol
4163 -Wunknown-pragmas @gol
4164 -Wunused-function @gol
4167 -Wunused-variable @gol
4168 -Wvolatile-register-var @gol
4171 Note that some warning flags are not implied by @option{-Wall}. Some of
4172 them warn about constructions that users generally do not consider
4173 questionable, but which occasionally you might wish to check for;
4174 others warn about constructions that are necessary or hard to avoid in
4175 some cases, and there is no simple way to modify the code to suppress
4176 the warning. Some of them are enabled by @option{-Wextra} but many of
4177 them must be enabled individually.
4183 This enables some extra warning flags that are not enabled by
4184 @option{-Wall}. (This option used to be called @option{-W}. The older
4185 name is still supported, but the newer name is more descriptive.)
4187 @gccoptlist{-Wclobbered @gol
4188 -Wcast-function-type @gol
4190 -Wignored-qualifiers @gol
4191 -Wimplicit-fallthrough=3 @gol
4192 -Wmissing-field-initializers @gol
4193 -Wmissing-parameter-type @r{(C only)} @gol
4194 -Wold-style-declaration @r{(C only)} @gol
4195 -Woverride-init @gol
4196 -Wsign-compare @r{(C only)} @gol
4197 -Wredundant-move @r{(only for C++)} @gol
4199 -Wuninitialized @gol
4200 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
4201 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4202 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4205 The option @option{-Wextra} also prints warning messages for the
4211 A pointer is compared against integer zero with @code{<}, @code{<=},
4212 @code{>}, or @code{>=}.
4215 (C++ only) An enumerator and a non-enumerator both appear in a
4216 conditional expression.
4219 (C++ only) Ambiguous virtual bases.
4222 (C++ only) Subscripting an array that has been declared @code{register}.
4225 (C++ only) Taking the address of a variable that has been declared
4229 (C++ only) A base class is not initialized in the copy constructor
4234 @item -Wchar-subscripts
4235 @opindex Wchar-subscripts
4236 @opindex Wno-char-subscripts
4237 Warn if an array subscript has type @code{char}. This is a common cause
4238 of error, as programmers often forget that this type is signed on some
4240 This warning is enabled by @option{-Wall}.
4245 Warn about an invalid memory access that is found by Pointer Bounds Checker
4246 (@option{-fcheck-pointer-bounds}).
4248 @item -Wno-coverage-mismatch
4249 @opindex Wno-coverage-mismatch
4250 @opindex Wcoverage-mismatch
4251 Warn if feedback profiles do not match when using the
4252 @option{-fprofile-use} option.
4253 If a source file is changed between compiling with @option{-fprofile-generate}
4254 and with @option{-fprofile-use}, the files with the profile feedback can fail
4255 to match the source file and GCC cannot use the profile feedback
4256 information. By default, this warning is enabled and is treated as an
4257 error. @option{-Wno-coverage-mismatch} can be used to disable the
4258 warning or @option{-Wno-error=coverage-mismatch} can be used to
4259 disable the error. Disabling the error for this warning can result in
4260 poorly optimized code and is useful only in the
4261 case of very minor changes such as bug fixes to an existing code-base.
4262 Completely disabling the warning is not recommended.
4265 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4267 Suppress warning messages emitted by @code{#warning} directives.
4269 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4270 @opindex Wdouble-promotion
4271 @opindex Wno-double-promotion
4272 Give a warning when a value of type @code{float} is implicitly
4273 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4274 floating-point unit implement @code{float} in hardware, but emulate
4275 @code{double} in software. On such a machine, doing computations
4276 using @code{double} values is much more expensive because of the
4277 overhead required for software emulation.
4279 It is easy to accidentally do computations with @code{double} because
4280 floating-point literals are implicitly of type @code{double}. For
4284 float area(float radius)
4286 return 3.14159 * radius * radius;
4290 the compiler performs the entire computation with @code{double}
4291 because the floating-point literal is a @code{double}.
4293 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4294 @opindex Wduplicate-decl-specifier
4295 @opindex Wno-duplicate-decl-specifier
4296 Warn if a declaration has duplicate @code{const}, @code{volatile},
4297 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4301 @itemx -Wformat=@var{n}
4304 @opindex ffreestanding
4305 @opindex fno-builtin
4307 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4308 the arguments supplied have types appropriate to the format string
4309 specified, and that the conversions specified in the format string make
4310 sense. This includes standard functions, and others specified by format
4311 attributes (@pxref{Function Attributes}), in the @code{printf},
4312 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4313 not in the C standard) families (or other target-specific families).
4314 Which functions are checked without format attributes having been
4315 specified depends on the standard version selected, and such checks of
4316 functions without the attribute specified are disabled by
4317 @option{-ffreestanding} or @option{-fno-builtin}.
4319 The formats are checked against the format features supported by GNU
4320 libc version 2.2. These include all ISO C90 and C99 features, as well
4321 as features from the Single Unix Specification and some BSD and GNU
4322 extensions. Other library implementations may not support all these
4323 features; GCC does not support warning about features that go beyond a
4324 particular library's limitations. However, if @option{-Wpedantic} is used
4325 with @option{-Wformat}, warnings are given about format features not
4326 in the selected standard version (but not for @code{strfmon} formats,
4327 since those are not in any version of the C standard). @xref{C Dialect
4328 Options,,Options Controlling C Dialect}.
4335 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4336 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4337 @option{-Wformat} also checks for null format arguments for several
4338 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4339 aspects of this level of format checking can be disabled by the
4340 options: @option{-Wno-format-contains-nul},
4341 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4342 @option{-Wformat} is enabled by @option{-Wall}.
4344 @item -Wno-format-contains-nul
4345 @opindex Wno-format-contains-nul
4346 @opindex Wformat-contains-nul
4347 If @option{-Wformat} is specified, do not warn about format strings that
4350 @item -Wno-format-extra-args
4351 @opindex Wno-format-extra-args
4352 @opindex Wformat-extra-args
4353 If @option{-Wformat} is specified, do not warn about excess arguments to a
4354 @code{printf} or @code{scanf} format function. The C standard specifies
4355 that such arguments are ignored.
4357 Where the unused arguments lie between used arguments that are
4358 specified with @samp{$} operand number specifications, normally
4359 warnings are still given, since the implementation could not know what
4360 type to pass to @code{va_arg} to skip the unused arguments. However,
4361 in the case of @code{scanf} formats, this option suppresses the
4362 warning if the unused arguments are all pointers, since the Single
4363 Unix Specification says that such unused arguments are allowed.
4365 @item -Wformat-overflow
4366 @itemx -Wformat-overflow=@var{level}
4367 @opindex Wformat-overflow
4368 @opindex Wno-format-overflow
4369 Warn about calls to formatted input/output functions such as @code{sprintf}
4370 and @code{vsprintf} that might overflow the destination buffer. When the
4371 exact number of bytes written by a format directive cannot be determined
4372 at compile-time it is estimated based on heuristics that depend on the
4373 @var{level} argument and on optimization. While enabling optimization
4374 will in most cases improve the accuracy of the warning, it may also
4375 result in false positives.
4378 @item -Wformat-overflow
4379 @itemx -Wformat-overflow=1
4380 @opindex Wformat-overflow
4381 @opindex Wno-format-overflow
4382 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4383 employs a conservative approach that warns only about calls that most
4384 likely overflow the buffer. At this level, numeric arguments to format
4385 directives with unknown values are assumed to have the value of one, and
4386 strings of unknown length to be empty. Numeric arguments that are known
4387 to be bounded to a subrange of their type, or string arguments whose output
4388 is bounded either by their directive's precision or by a finite set of
4389 string literals, are assumed to take on the value within the range that
4390 results in the most bytes on output. For example, the call to @code{sprintf}
4391 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4392 the terminating NUL character (@code{'\0'}) appended by the function
4393 to the destination buffer will be written past its end. Increasing
4394 the size of the buffer by a single byte is sufficient to avoid the
4395 warning, though it may not be sufficient to avoid the overflow.
4398 void f (int a, int b)
4401 sprintf (buf, "a = %i, b = %i\n", a, b);
4405 @item -Wformat-overflow=2
4406 Level @var{2} warns also about calls that might overflow the destination
4407 buffer given an argument of sufficient length or magnitude. At level
4408 @var{2}, unknown numeric arguments are assumed to have the minimum
4409 representable value for signed types with a precision greater than 1, and
4410 the maximum representable value otherwise. Unknown string arguments whose
4411 length cannot be assumed to be bounded either by the directive's precision,
4412 or by a finite set of string literals they may evaluate to, or the character
4413 array they may point to, are assumed to be 1 character long.
4415 At level @var{2}, the call in the example above is again diagnosed, but
4416 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4417 @code{%i} directive will write some of its digits beyond the end of
4418 the destination buffer. To make the call safe regardless of the values
4419 of the two variables, the size of the destination buffer must be increased
4420 to at least 34 bytes. GCC includes the minimum size of the buffer in
4421 an informational note following the warning.
4423 An alternative to increasing the size of the destination buffer is to
4424 constrain the range of formatted values. The maximum length of string
4425 arguments can be bounded by specifying the precision in the format
4426 directive. When numeric arguments of format directives can be assumed
4427 to be bounded by less than the precision of their type, choosing
4428 an appropriate length modifier to the format specifier will reduce
4429 the required buffer size. For example, if @var{a} and @var{b} in the
4430 example above can be assumed to be within the precision of
4431 the @code{short int} type then using either the @code{%hi} format
4432 directive or casting the argument to @code{short} reduces the maximum
4433 required size of the buffer to 24 bytes.
4436 void f (int a, int b)
4439 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4444 @item -Wno-format-zero-length
4445 @opindex Wno-format-zero-length
4446 @opindex Wformat-zero-length
4447 If @option{-Wformat} is specified, do not warn about zero-length formats.
4448 The C standard specifies that zero-length formats are allowed.
4453 Enable @option{-Wformat} plus additional format checks. Currently
4454 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4457 @item -Wformat-nonliteral
4458 @opindex Wformat-nonliteral
4459 @opindex Wno-format-nonliteral
4460 If @option{-Wformat} is specified, also warn if the format string is not a
4461 string literal and so cannot be checked, unless the format function
4462 takes its format arguments as a @code{va_list}.
4464 @item -Wformat-security
4465 @opindex Wformat-security
4466 @opindex Wno-format-security
4467 If @option{-Wformat} is specified, also warn about uses of format
4468 functions that represent possible security problems. At present, this
4469 warns about calls to @code{printf} and @code{scanf} functions where the
4470 format string is not a string literal and there are no format arguments,
4471 as in @code{printf (foo);}. This may be a security hole if the format
4472 string came from untrusted input and contains @samp{%n}. (This is
4473 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4474 in future warnings may be added to @option{-Wformat-security} that are not
4475 included in @option{-Wformat-nonliteral}.)
4477 @item -Wformat-signedness
4478 @opindex Wformat-signedness
4479 @opindex Wno-format-signedness
4480 If @option{-Wformat} is specified, also warn if the format string
4481 requires an unsigned argument and the argument is signed and vice versa.
4483 @item -Wformat-truncation
4484 @itemx -Wformat-truncation=@var{level}
4485 @opindex Wformat-truncation
4486 @opindex Wno-format-truncation
4487 Warn about calls to formatted input/output functions such as @code{snprintf}
4488 and @code{vsnprintf} that might result in output truncation. When the exact
4489 number of bytes written by a format directive cannot be determined at
4490 compile-time it is estimated based on heuristics that depend on
4491 the @var{level} argument and on optimization. While enabling optimization
4492 will in most cases improve the accuracy of the warning, it may also result
4493 in false positives. Except as noted otherwise, the option uses the same
4494 logic @option{-Wformat-overflow}.
4497 @item -Wformat-truncation
4498 @itemx -Wformat-truncation=1
4499 @opindex Wformat-truncation
4500 @opindex Wno-format-truncation
4501 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4502 employs a conservative approach that warns only about calls to bounded
4503 functions whose return value is unused and that will most likely result
4504 in output truncation.
4506 @item -Wformat-truncation=2
4507 Level @var{2} warns also about calls to bounded functions whose return
4508 value is used and that might result in truncation given an argument of
4509 sufficient length or magnitude.
4513 @opindex Wformat-y2k
4514 @opindex Wno-format-y2k
4515 If @option{-Wformat} is specified, also warn about @code{strftime}
4516 formats that may yield only a two-digit year.
4521 @opindex Wno-nonnull
4522 Warn about passing a null pointer for arguments marked as
4523 requiring a non-null value by the @code{nonnull} function attribute.
4525 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4526 can be disabled with the @option{-Wno-nonnull} option.
4528 @item -Wnonnull-compare
4529 @opindex Wnonnull-compare
4530 @opindex Wno-nonnull-compare
4531 Warn when comparing an argument marked with the @code{nonnull}
4532 function attribute against null inside the function.
4534 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4535 can be disabled with the @option{-Wno-nonnull-compare} option.
4537 @item -Wnull-dereference
4538 @opindex Wnull-dereference
4539 @opindex Wno-null-dereference
4540 Warn if the compiler detects paths that trigger erroneous or
4541 undefined behavior due to dereferencing a null pointer. This option
4542 is only active when @option{-fdelete-null-pointer-checks} is active,
4543 which is enabled by optimizations in most targets. The precision of
4544 the warnings depends on the optimization options used.
4546 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4548 @opindex Wno-init-self
4549 Warn about uninitialized variables that are initialized with themselves.
4550 Note this option can only be used with the @option{-Wuninitialized} option.
4552 For example, GCC warns about @code{i} being uninitialized in the
4553 following snippet only when @option{-Winit-self} has been specified:
4564 This warning is enabled by @option{-Wall} in C++.
4566 @item -Wimplicit-int @r{(C and Objective-C only)}
4567 @opindex Wimplicit-int
4568 @opindex Wno-implicit-int
4569 Warn when a declaration does not specify a type.
4570 This warning is enabled by @option{-Wall}.
4572 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4573 @opindex Wimplicit-function-declaration
4574 @opindex Wno-implicit-function-declaration
4575 Give a warning whenever a function is used before being declared. In
4576 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4577 enabled by default and it is made into an error by
4578 @option{-pedantic-errors}. This warning is also enabled by
4581 @item -Wimplicit @r{(C and Objective-C only)}
4583 @opindex Wno-implicit
4584 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4585 This warning is enabled by @option{-Wall}.
4587 @item -Wimplicit-fallthrough
4588 @opindex Wimplicit-fallthrough
4589 @opindex Wno-implicit-fallthrough
4590 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4591 and @option{-Wno-implicit-fallthrough} is the same as
4592 @option{-Wimplicit-fallthrough=0}.
4594 @item -Wimplicit-fallthrough=@var{n}
4595 @opindex Wimplicit-fallthrough=
4596 Warn when a switch case falls through. For example:
4614 This warning does not warn when the last statement of a case cannot
4615 fall through, e.g. when there is a return statement or a call to function
4616 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4617 also takes into account control flow statements, such as ifs, and only
4618 warns when appropriate. E.g.@:
4628 @} else if (i < 1) @{
4638 Since there are occasions where a switch case fall through is desirable,
4639 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4640 to be used along with a null statement to suppress this warning that
4641 would normally occur:
4649 __attribute__ ((fallthrough));
4656 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4657 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4658 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4659 Instead of these attributes, it is also possible to add a fallthrough comment
4660 to silence the warning. The whole body of the C or C++ style comment should
4661 match the given regular expressions listed below. The option argument @var{n}
4662 specifies what kind of comments are accepted:
4666 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4668 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4669 expression, any comment is used as fallthrough comment.
4671 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4672 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4674 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4675 following regular expressions:
4679 @item @code{-fallthrough}
4681 @item @code{@@fallthrough@@}
4683 @item @code{lint -fallthrough[ \t]*}
4685 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4687 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4689 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4693 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4694 following regular expressions:
4698 @item @code{-fallthrough}
4700 @item @code{@@fallthrough@@}
4702 @item @code{lint -fallthrough[ \t]*}
4704 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
4708 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
4709 fallthrough comments, only attributes disable the warning.
4713 The comment needs to be followed after optional whitespace and other comments
4714 by @code{case} or @code{default} keywords or by a user label that precedes some
4715 @code{case} or @code{default} label.
4730 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
4732 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
4733 @opindex Wif-not-aligned
4734 @opindex Wno-if-not-aligned
4735 Control if warning triggered by the @code{warn_if_not_aligned} attribute
4736 should be issued. This is enabled by default.
4737 Use @option{-Wno-if-not-aligned} to disable it.
4739 @item -Wignored-qualifiers @r{(C and C++ only)}
4740 @opindex Wignored-qualifiers
4741 @opindex Wno-ignored-qualifiers
4742 Warn if the return type of a function has a type qualifier
4743 such as @code{const}. For ISO C such a type qualifier has no effect,
4744 since the value returned by a function is not an lvalue.
4745 For C++, the warning is only emitted for scalar types or @code{void}.
4746 ISO C prohibits qualified @code{void} return types on function
4747 definitions, so such return types always receive a warning
4748 even without this option.
4750 This warning is also enabled by @option{-Wextra}.
4752 @item -Wignored-attributes @r{(C and C++ only)}
4753 @opindex Wignored-attributes
4754 @opindex Wno-ignored-attributes
4755 Warn when an attribute is ignored. This is different from the
4756 @option{-Wattributes} option in that it warns whenever the compiler decides
4757 to drop an attribute, not that the attribute is either unknown, used in a
4758 wrong place, etc. This warning is enabled by default.
4763 Warn if the type of @code{main} is suspicious. @code{main} should be
4764 a function with external linkage, returning int, taking either zero
4765 arguments, two, or three arguments of appropriate types. This warning
4766 is enabled by default in C++ and is enabled by either @option{-Wall}
4767 or @option{-Wpedantic}.
4769 @item -Wmisleading-indentation @r{(C and C++ only)}
4770 @opindex Wmisleading-indentation
4771 @opindex Wno-misleading-indentation
4772 Warn when the indentation of the code does not reflect the block structure.
4773 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
4774 @code{for} clauses with a guarded statement that does not use braces,
4775 followed by an unguarded statement with the same indentation.
4777 In the following example, the call to ``bar'' is misleadingly indented as
4778 if it were guarded by the ``if'' conditional.
4781 if (some_condition ())
4783 bar (); /* Gotcha: this is not guarded by the "if". */
4786 In the case of mixed tabs and spaces, the warning uses the
4787 @option{-ftabstop=} option to determine if the statements line up
4790 The warning is not issued for code involving multiline preprocessor logic
4791 such as the following example.
4796 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
4802 The warning is not issued after a @code{#line} directive, since this
4803 typically indicates autogenerated code, and no assumptions can be made
4804 about the layout of the file that the directive references.
4806 This warning is enabled by @option{-Wall} in C and C++.
4808 @item -Wmissing-attributes
4809 @opindex Wmissing-attributes
4810 @opindex Wno-missing-attributes
4811 Warn when a declaration of a function is missing one or more attributes
4812 that a related function is declared with and whose absence may adversely
4813 affect the correctness or efficiency of generated code. For example, in
4814 C++, the warning is issued when an explicit specialization of a primary
4815 template declared with attribute @code{alloc_align}, @code{alloc_size},
4816 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
4817 or @code{nonnull} is declared without it. Attributes @code{deprecated},
4818 @code{error}, and @code{warning} suppress the warning.
4819 (@pxref{Function Attributes}).
4821 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
4823 For example, since the declaration of the primary function template
4824 below makes use of both attribute @code{malloc} and @code{alloc_size}
4825 the declaration of the explicit specialization of the template is
4826 diagnosed because it is missing one of the attributes.
4830 T* __attribute__ ((malloc, alloc_size (1)))
4834 void* __attribute__ ((malloc)) // missing alloc_size
4835 allocate<void> (size_t);
4838 @item -Wmissing-braces
4839 @opindex Wmissing-braces
4840 @opindex Wno-missing-braces
4841 Warn if an aggregate or union initializer is not fully bracketed. In
4842 the following example, the initializer for @code{a} is not fully
4843 bracketed, but that for @code{b} is fully bracketed. This warning is
4844 enabled by @option{-Wall} in C.
4847 int a[2][2] = @{ 0, 1, 2, 3 @};
4848 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4851 This warning is enabled by @option{-Wall}.
4853 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4854 @opindex Wmissing-include-dirs
4855 @opindex Wno-missing-include-dirs
4856 Warn if a user-supplied include directory does not exist.
4858 @item -Wmissing-profile
4859 @opindex Wmissing-profile
4860 @opindex Wno-missing-profile
4861 Warn if feedback profiles are missing when using the
4862 @option{-fprofile-use} option.
4863 This option diagnoses those cases where a new function or a new file is added
4864 to the user code between compiling with @option{-fprofile-generate} and with
4865 @option{-fprofile-use}, without regenerating the profiles. In these cases, the
4866 profile feedback data files do not contain any profile feedback information for
4867 the newly added function or file respectively. Also, in the case when profile
4868 count data (.gcda) files are removed, GCC cannot use any profile feedback
4869 information. In all these cases, warnings are issued to inform the user that a
4870 profile generation step is due. @option{-Wno-missing-profile} can be used to
4871 disable the warning. Ignoring the warning can result in poorly optimized code.
4872 Completely disabling the warning is not recommended and should be done only
4873 when non-existent profile data is justified.
4875 @item -Wmultistatement-macros
4876 @opindex Wmultistatement-macros
4877 @opindex Wno-multistatement-macros
4878 Warn about unsafe multiple statement macros that appear to be guarded
4879 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
4880 @code{while}, in which only the first statement is actually guarded after
4881 the macro is expanded.
4886 #define DOIT x++; y++
4891 will increment @code{y} unconditionally, not just when @code{c} holds.
4892 The can usually be fixed by wrapping the macro in a do-while loop:
4894 #define DOIT do @{ x++; y++; @} while (0)
4899 This warning is enabled by @option{-Wall} in C and C++.
4902 @opindex Wparentheses
4903 @opindex Wno-parentheses
4904 Warn if parentheses are omitted in certain contexts, such
4905 as when there is an assignment in a context where a truth value
4906 is expected, or when operators are nested whose precedence people
4907 often get confused about.
4909 Also warn if a comparison like @code{x<=y<=z} appears; this is
4910 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4911 interpretation from that of ordinary mathematical notation.
4913 Also warn for dangerous uses of the GNU extension to
4914 @code{?:} with omitted middle operand. When the condition
4915 in the @code{?}: operator is a boolean expression, the omitted value is
4916 always 1. Often programmers expect it to be a value computed
4917 inside the conditional expression instead.
4919 For C++ this also warns for some cases of unnecessary parentheses in
4920 declarations, which can indicate an attempt at a function call instead
4924 // Declares a local variable called mymutex.
4925 std::unique_lock<std::mutex> (mymutex);
4926 // User meant std::unique_lock<std::mutex> lock (mymutex);
4930 This warning is enabled by @option{-Wall}.
4932 @item -Wsequence-point
4933 @opindex Wsequence-point
4934 @opindex Wno-sequence-point
4935 Warn about code that may have undefined semantics because of violations
4936 of sequence point rules in the C and C++ standards.
4938 The C and C++ standards define the order in which expressions in a C/C++
4939 program are evaluated in terms of @dfn{sequence points}, which represent
4940 a partial ordering between the execution of parts of the program: those
4941 executed before the sequence point, and those executed after it. These
4942 occur after the evaluation of a full expression (one which is not part
4943 of a larger expression), after the evaluation of the first operand of a
4944 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4945 function is called (but after the evaluation of its arguments and the
4946 expression denoting the called function), and in certain other places.
4947 Other than as expressed by the sequence point rules, the order of
4948 evaluation of subexpressions of an expression is not specified. All
4949 these rules describe only a partial order rather than a total order,
4950 since, for example, if two functions are called within one expression
4951 with no sequence point between them, the order in which the functions
4952 are called is not specified. However, the standards committee have
4953 ruled that function calls do not overlap.
4955 It is not specified when between sequence points modifications to the
4956 values of objects take effect. Programs whose behavior depends on this
4957 have undefined behavior; the C and C++ standards specify that ``Between
4958 the previous and next sequence point an object shall have its stored
4959 value modified at most once by the evaluation of an expression.
4960 Furthermore, the prior value shall be read only to determine the value
4961 to be stored.''. If a program breaks these rules, the results on any
4962 particular implementation are entirely unpredictable.
4964 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4965 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4966 diagnosed by this option, and it may give an occasional false positive
4967 result, but in general it has been found fairly effective at detecting
4968 this sort of problem in programs.
4970 The C++17 standard will define the order of evaluation of operands in
4971 more cases: in particular it requires that the right-hand side of an
4972 assignment be evaluated before the left-hand side, so the above
4973 examples are no longer undefined. But this warning will still warn
4974 about them, to help people avoid writing code that is undefined in C
4975 and earlier revisions of C++.
4977 The standard is worded confusingly, therefore there is some debate
4978 over the precise meaning of the sequence point rules in subtle cases.
4979 Links to discussions of the problem, including proposed formal
4980 definitions, may be found on the GCC readings page, at
4981 @uref{http://gcc.gnu.org/@/readings.html}.
4983 This warning is enabled by @option{-Wall} for C and C++.
4985 @item -Wno-return-local-addr
4986 @opindex Wno-return-local-addr
4987 @opindex Wreturn-local-addr
4988 Do not warn about returning a pointer (or in C++, a reference) to a
4989 variable that goes out of scope after the function returns.
4992 @opindex Wreturn-type
4993 @opindex Wno-return-type
4994 Warn whenever a function is defined with a return type that defaults
4995 to @code{int}. Also warn about any @code{return} statement with no
4996 return value in a function whose return type is not @code{void}
4997 (falling off the end of the function body is considered returning
5000 For C only, warn about a @code{return} statement with an expression in a
5001 function whose return type is @code{void}, unless the expression type is
5002 also @code{void}. As a GNU extension, the latter case is accepted
5003 without a warning unless @option{-Wpedantic} is used.
5005 For C++, a function without return type always produces a diagnostic
5006 message, even when @option{-Wno-return-type} is specified. The only
5007 exceptions are @code{main} and functions defined in system headers.
5009 This warning is enabled by default for C++ and is enabled by @option{-Wall}.
5011 @item -Wshift-count-negative
5012 @opindex Wshift-count-negative
5013 @opindex Wno-shift-count-negative
5014 Warn if shift count is negative. This warning is enabled by default.
5016 @item -Wshift-count-overflow
5017 @opindex Wshift-count-overflow
5018 @opindex Wno-shift-count-overflow
5019 Warn if shift count >= width of type. This warning is enabled by default.
5021 @item -Wshift-negative-value
5022 @opindex Wshift-negative-value
5023 @opindex Wno-shift-negative-value
5024 Warn if left shifting a negative value. This warning is enabled by
5025 @option{-Wextra} in C99 and C++11 modes (and newer).
5027 @item -Wshift-overflow
5028 @itemx -Wshift-overflow=@var{n}
5029 @opindex Wshift-overflow
5030 @opindex Wno-shift-overflow
5031 Warn about left shift overflows. This warning is enabled by
5032 default in C99 and C++11 modes (and newer).
5035 @item -Wshift-overflow=1
5036 This is the warning level of @option{-Wshift-overflow} and is enabled
5037 by default in C99 and C++11 modes (and newer). This warning level does
5038 not warn about left-shifting 1 into the sign bit. (However, in C, such
5039 an overflow is still rejected in contexts where an integer constant expression
5042 @item -Wshift-overflow=2
5043 This warning level also warns about left-shifting 1 into the sign bit,
5044 unless C++14 mode is active.
5050 Warn whenever a @code{switch} statement has an index of enumerated type
5051 and lacks a @code{case} for one or more of the named codes of that
5052 enumeration. (The presence of a @code{default} label prevents this
5053 warning.) @code{case} labels outside the enumeration range also
5054 provoke warnings when this option is used (even if there is a
5055 @code{default} label).
5056 This warning is enabled by @option{-Wall}.
5058 @item -Wswitch-default
5059 @opindex Wswitch-default
5060 @opindex Wno-switch-default
5061 Warn whenever a @code{switch} statement does not have a @code{default}
5065 @opindex Wswitch-enum
5066 @opindex Wno-switch-enum
5067 Warn whenever a @code{switch} statement has an index of enumerated type
5068 and lacks a @code{case} for one or more of the named codes of that
5069 enumeration. @code{case} labels outside the enumeration range also
5070 provoke warnings when this option is used. The only difference
5071 between @option{-Wswitch} and this option is that this option gives a
5072 warning about an omitted enumeration code even if there is a
5073 @code{default} label.
5076 @opindex Wswitch-bool
5077 @opindex Wno-switch-bool
5078 Warn whenever a @code{switch} statement has an index of boolean type
5079 and the case values are outside the range of a boolean type.
5080 It is possible to suppress this warning by casting the controlling
5081 expression to a type other than @code{bool}. For example:
5084 switch ((int) (a == 4))
5090 This warning is enabled by default for C and C++ programs.
5092 @item -Wswitch-unreachable
5093 @opindex Wswitch-unreachable
5094 @opindex Wno-switch-unreachable
5095 Warn whenever a @code{switch} statement contains statements between the
5096 controlling expression and the first case label, which will never be
5097 executed. For example:
5109 @option{-Wswitch-unreachable} does not warn if the statement between the
5110 controlling expression and the first case label is just a declaration:
5123 This warning is enabled by default for C and C++ programs.
5125 @item -Wsync-nand @r{(C and C++ only)}
5127 @opindex Wno-sync-nand
5128 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
5129 built-in functions are used. These functions changed semantics in GCC 4.4.
5131 @item -Wunused-but-set-parameter
5132 @opindex Wunused-but-set-parameter
5133 @opindex Wno-unused-but-set-parameter
5134 Warn whenever a function parameter is assigned to, but otherwise unused
5135 (aside from its declaration).
5137 To suppress this warning use the @code{unused} attribute
5138 (@pxref{Variable Attributes}).
5140 This warning is also enabled by @option{-Wunused} together with
5143 @item -Wunused-but-set-variable
5144 @opindex Wunused-but-set-variable
5145 @opindex Wno-unused-but-set-variable
5146 Warn whenever a local variable is assigned to, but otherwise unused
5147 (aside from its declaration).
5148 This warning is enabled by @option{-Wall}.
5150 To suppress this warning use the @code{unused} attribute
5151 (@pxref{Variable Attributes}).
5153 This warning is also enabled by @option{-Wunused}, which is enabled
5156 @item -Wunused-function
5157 @opindex Wunused-function
5158 @opindex Wno-unused-function
5159 Warn whenever a static function is declared but not defined or a
5160 non-inline static function is unused.
5161 This warning is enabled by @option{-Wall}.
5163 @item -Wunused-label
5164 @opindex Wunused-label
5165 @opindex Wno-unused-label
5166 Warn whenever a label is declared but not used.
5167 This warning is enabled by @option{-Wall}.
5169 To suppress this warning use the @code{unused} attribute
5170 (@pxref{Variable Attributes}).
5172 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
5173 @opindex Wunused-local-typedefs
5174 @opindex Wno-unused-local-typedefs
5175 Warn when a typedef locally defined in a function is not used.
5176 This warning is enabled by @option{-Wall}.
5178 @item -Wunused-parameter
5179 @opindex Wunused-parameter
5180 @opindex Wno-unused-parameter
5181 Warn whenever a function parameter is unused aside from its declaration.
5183 To suppress this warning use the @code{unused} attribute
5184 (@pxref{Variable Attributes}).
5186 @item -Wno-unused-result
5187 @opindex Wunused-result
5188 @opindex Wno-unused-result
5189 Do not warn if a caller of a function marked with attribute
5190 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
5191 its return value. The default is @option{-Wunused-result}.
5193 @item -Wunused-variable
5194 @opindex Wunused-variable
5195 @opindex Wno-unused-variable
5196 Warn whenever a local or static variable is unused aside from its
5197 declaration. This option implies @option{-Wunused-const-variable=1} for C,
5198 but not for C++. This warning is enabled by @option{-Wall}.
5200 To suppress this warning use the @code{unused} attribute
5201 (@pxref{Variable Attributes}).
5203 @item -Wunused-const-variable
5204 @itemx -Wunused-const-variable=@var{n}
5205 @opindex Wunused-const-variable
5206 @opindex Wno-unused-const-variable
5207 Warn whenever a constant static variable is unused aside from its declaration.
5208 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
5209 for C, but not for C++. In C this declares variable storage, but in C++ this
5210 is not an error since const variables take the place of @code{#define}s.
5212 To suppress this warning use the @code{unused} attribute
5213 (@pxref{Variable Attributes}).
5216 @item -Wunused-const-variable=1
5217 This is the warning level that is enabled by @option{-Wunused-variable} for
5218 C. It warns only about unused static const variables defined in the main
5219 compilation unit, but not about static const variables declared in any
5222 @item -Wunused-const-variable=2
5223 This warning level also warns for unused constant static variables in
5224 headers (excluding system headers). This is the warning level of
5225 @option{-Wunused-const-variable} and must be explicitly requested since
5226 in C++ this isn't an error and in C it might be harder to clean up all
5230 @item -Wunused-value
5231 @opindex Wunused-value
5232 @opindex Wno-unused-value
5233 Warn whenever a statement computes a result that is explicitly not
5234 used. To suppress this warning cast the unused expression to
5235 @code{void}. This includes an expression-statement or the left-hand
5236 side of a comma expression that contains no side effects. For example,
5237 an expression such as @code{x[i,j]} causes a warning, while
5238 @code{x[(void)i,j]} does not.
5240 This warning is enabled by @option{-Wall}.
5245 All the above @option{-Wunused} options combined.
5247 In order to get a warning about an unused function parameter, you must
5248 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5249 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5251 @item -Wuninitialized
5252 @opindex Wuninitialized
5253 @opindex Wno-uninitialized
5254 Warn if an automatic variable is used without first being initialized
5255 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5256 warn if a non-static reference or non-static @code{const} member
5257 appears in a class without constructors.
5259 If you want to warn about code that uses the uninitialized value of the
5260 variable in its own initializer, use the @option{-Winit-self} option.
5262 These warnings occur for individual uninitialized or clobbered
5263 elements of structure, union or array variables as well as for
5264 variables that are uninitialized or clobbered as a whole. They do
5265 not occur for variables or elements declared @code{volatile}. Because
5266 these warnings depend on optimization, the exact variables or elements
5267 for which there are warnings depends on the precise optimization
5268 options and version of GCC used.
5270 Note that there may be no warning about a variable that is used only
5271 to compute a value that itself is never used, because such
5272 computations may be deleted by data flow analysis before the warnings
5275 @item -Winvalid-memory-model
5276 @opindex Winvalid-memory-model
5277 @opindex Wno-invalid-memory-model
5278 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5279 and the C11 atomic generic functions with a memory consistency argument
5280 that is either invalid for the operation or outside the range of values
5281 of the @code{memory_order} enumeration. For example, since the
5282 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5283 defined for the relaxed, release, and sequentially consistent memory
5284 orders the following code is diagnosed:
5289 __atomic_store_n (i, 0, memory_order_consume);
5293 @option{-Winvalid-memory-model} is enabled by default.
5295 @item -Wmaybe-uninitialized
5296 @opindex Wmaybe-uninitialized
5297 @opindex Wno-maybe-uninitialized
5298 For an automatic (i.e.@: local) variable, if there exists a path from the
5299 function entry to a use of the variable that is initialized, but there exist
5300 some other paths for which the variable is not initialized, the compiler
5301 emits a warning if it cannot prove the uninitialized paths are not
5302 executed at run time.
5304 These warnings are only possible in optimizing compilation, because otherwise
5305 GCC does not keep track of the state of variables.
5307 These warnings are made optional because GCC may not be able to determine when
5308 the code is correct in spite of appearing to have an error. Here is one
5309 example of how this can happen:
5329 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5330 always initialized, but GCC doesn't know this. To suppress the
5331 warning, you need to provide a default case with assert(0) or
5334 @cindex @code{longjmp} warnings
5335 This option also warns when a non-volatile automatic variable might be
5336 changed by a call to @code{longjmp}.
5337 The compiler sees only the calls to @code{setjmp}. It cannot know
5338 where @code{longjmp} will be called; in fact, a signal handler could
5339 call it at any point in the code. As a result, you may get a warning
5340 even when there is in fact no problem because @code{longjmp} cannot
5341 in fact be called at the place that would cause a problem.
5343 Some spurious warnings can be avoided if you declare all the functions
5344 you use that never return as @code{noreturn}. @xref{Function
5347 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5349 @item -Wunknown-pragmas
5350 @opindex Wunknown-pragmas
5351 @opindex Wno-unknown-pragmas
5352 @cindex warning for unknown pragmas
5353 @cindex unknown pragmas, warning
5354 @cindex pragmas, warning of unknown
5355 Warn when a @code{#pragma} directive is encountered that is not understood by
5356 GCC@. If this command-line option is used, warnings are even issued
5357 for unknown pragmas in system header files. This is not the case if
5358 the warnings are only enabled by the @option{-Wall} command-line option.
5361 @opindex Wno-pragmas
5363 Do not warn about misuses of pragmas, such as incorrect parameters,
5364 invalid syntax, or conflicts between pragmas. See also
5365 @option{-Wunknown-pragmas}.
5367 @item -Wno-prio-ctor-dtor
5368 @opindex Wno-prio-ctor-dtor
5369 @opindex Wprio-ctor-dtor
5370 Do not warn if a priority from 0 to 100 is used for constructor or destructor.
5371 The use of constructor and destructor attributes allow you to assign a
5372 priority to the constructor/destructor to control its order of execution
5373 before @code{main} is called or after it returns. The priority values must be
5374 greater than 100 as the compiler reserves priority values between 0--100 for
5377 @item -Wstrict-aliasing
5378 @opindex Wstrict-aliasing
5379 @opindex Wno-strict-aliasing
5380 This option is only active when @option{-fstrict-aliasing} is active.
5381 It warns about code that might break the strict aliasing rules that the
5382 compiler is using for optimization. The warning does not catch all
5383 cases, but does attempt to catch the more common pitfalls. It is
5384 included in @option{-Wall}.
5385 It is equivalent to @option{-Wstrict-aliasing=3}
5387 @item -Wstrict-aliasing=n
5388 @opindex Wstrict-aliasing=n
5389 This option is only active when @option{-fstrict-aliasing} is active.
5390 It warns about code that might break the strict aliasing rules that the
5391 compiler is using for optimization.
5392 Higher levels correspond to higher accuracy (fewer false positives).
5393 Higher levels also correspond to more effort, similar to the way @option{-O}
5395 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5397 Level 1: Most aggressive, quick, least accurate.
5398 Possibly useful when higher levels
5399 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5400 false negatives. However, it has many false positives.
5401 Warns for all pointer conversions between possibly incompatible types,
5402 even if never dereferenced. Runs in the front end only.
5404 Level 2: Aggressive, quick, not too precise.
5405 May still have many false positives (not as many as level 1 though),
5406 and few false negatives (but possibly more than level 1).
5407 Unlike level 1, it only warns when an address is taken. Warns about
5408 incomplete types. Runs in the front end only.
5410 Level 3 (default for @option{-Wstrict-aliasing}):
5411 Should have very few false positives and few false
5412 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5413 Takes care of the common pun+dereference pattern in the front end:
5414 @code{*(int*)&some_float}.
5415 If optimization is enabled, it also runs in the back end, where it deals
5416 with multiple statement cases using flow-sensitive points-to information.
5417 Only warns when the converted pointer is dereferenced.
5418 Does not warn about incomplete types.
5420 @item -Wstrict-overflow
5421 @itemx -Wstrict-overflow=@var{n}
5422 @opindex Wstrict-overflow
5423 @opindex Wno-strict-overflow
5424 This option is only active when signed overflow is undefined.
5425 It warns about cases where the compiler optimizes based on the
5426 assumption that signed overflow does not occur. Note that it does not
5427 warn about all cases where the code might overflow: it only warns
5428 about cases where the compiler implements some optimization. Thus
5429 this warning depends on the optimization level.
5431 An optimization that assumes that signed overflow does not occur is
5432 perfectly safe if the values of the variables involved are such that
5433 overflow never does, in fact, occur. Therefore this warning can
5434 easily give a false positive: a warning about code that is not
5435 actually a problem. To help focus on important issues, several
5436 warning levels are defined. No warnings are issued for the use of
5437 undefined signed overflow when estimating how many iterations a loop
5438 requires, in particular when determining whether a loop will be
5442 @item -Wstrict-overflow=1
5443 Warn about cases that are both questionable and easy to avoid. For
5444 example the compiler simplifies
5445 @code{x + 1 > x} to @code{1}. This level of
5446 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5447 are not, and must be explicitly requested.
5449 @item -Wstrict-overflow=2
5450 Also warn about other cases where a comparison is simplified to a
5451 constant. For example: @code{abs (x) >= 0}. This can only be
5452 simplified when signed integer overflow is undefined, because
5453 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5454 zero. @option{-Wstrict-overflow} (with no level) is the same as
5455 @option{-Wstrict-overflow=2}.
5457 @item -Wstrict-overflow=3
5458 Also warn about other cases where a comparison is simplified. For
5459 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5461 @item -Wstrict-overflow=4
5462 Also warn about other simplifications not covered by the above cases.
5463 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5465 @item -Wstrict-overflow=5
5466 Also warn about cases where the compiler reduces the magnitude of a
5467 constant involved in a comparison. For example: @code{x + 2 > y} is
5468 simplified to @code{x + 1 >= y}. This is reported only at the
5469 highest warning level because this simplification applies to many
5470 comparisons, so this warning level gives a very large number of
5474 @item -Wstringop-overflow
5475 @itemx -Wstringop-overflow=@var{type}
5476 @opindex Wstringop-overflow
5477 @opindex Wno-stringop-overflow
5478 Warn for calls to string manipulation functions such as @code{memcpy} and
5479 @code{strcpy} that are determined to overflow the destination buffer. The
5480 optional argument is one greater than the type of Object Size Checking to
5481 perform to determine the size of the destination. @xref{Object Size Checking}.
5482 The argument is meaningful only for functions that operate on character arrays
5483 but not for raw memory functions like @code{memcpy} which always make use
5484 of Object Size type-0. The option also warns for calls that specify a size
5485 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5486 The option produces the best results with optimization enabled but can detect
5487 a small subset of simple buffer overflows even without optimization in
5488 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5489 correspond to the standard functions. In any case, the option warns about
5490 just a subset of buffer overflows detected by the corresponding overflow
5491 checking built-ins. For example, the option will issue a warning for
5492 the @code{strcpy} call below because it copies at least 5 characters
5493 (the string @code{"blue"} including the terminating NUL) into the buffer
5497 enum Color @{ blue, purple, yellow @};
5498 const char* f (enum Color clr)
5500 static char buf [4];
5504 case blue: str = "blue"; break;
5505 case purple: str = "purple"; break;
5506 case yellow: str = "yellow"; break;
5509 return strcpy (buf, str); // warning here
5513 Option @option{-Wstringop-overflow=2} is enabled by default.
5516 @item -Wstringop-overflow
5517 @itemx -Wstringop-overflow=1
5518 @opindex Wstringop-overflow
5519 @opindex Wno-stringop-overflow
5520 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5521 to determine the sizes of destination objects. This is the default setting
5522 of the option. At this setting the option will not warn for writes past
5523 the end of subobjects of larger objects accessed by pointers unless the
5524 size of the largest surrounding object is known. When the destination may
5525 be one of several objects it is assumed to be the largest one of them. On
5526 Linux systems, when optimization is enabled at this setting the option warns
5527 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5530 @item -Wstringop-overflow=2
5531 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5532 to determine the sizes of destination objects. At this setting the option
5533 will warn about overflows when writing to members of the largest complete
5534 objects whose exact size is known. It will, however, not warn for excessive
5535 writes to the same members of unknown objects referenced by pointers since
5536 they may point to arrays containing unknown numbers of elements.
5538 @item -Wstringop-overflow=3
5539 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5540 to determine the sizes of destination objects. At this setting the option
5541 warns about overflowing the smallest object or data member. This is the
5542 most restrictive setting of the option that may result in warnings for safe
5545 @item -Wstringop-overflow=4
5546 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5547 to determine the sizes of destination objects. At this setting the option
5548 will warn about overflowing any data members, and when the destination is
5549 one of several objects it uses the size of the largest of them to decide
5550 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5551 setting of the option may result in warnings for benign code.
5554 @item -Wstringop-truncation
5555 @opindex Wstringop-truncation
5556 @opindex Wno-stringop-truncation
5557 Warn for calls to bounded string manipulation functions such as @code{strncat},
5558 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5559 or leave the destination unchanged.
5561 In the following example, the call to @code{strncat} specifies a bound that
5562 is less than the length of the source string. As a result, the copy of
5563 the source will be truncated and so the call is diagnosed. To avoid the
5564 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5567 void append (char *buf, size_t bufsize)
5569 strncat (buf, ".txt", 3);
5573 As another example, the following call to @code{strncpy} results in copying
5574 to @code{d} just the characters preceding the terminating NUL, without
5575 appending the NUL to the end. Assuming the result of @code{strncpy} is
5576 necessarily a NUL-terminated string is a common mistake, and so the call
5577 is diagnosed. To avoid the warning when the result is not expected to be
5578 NUL-terminated, call @code{memcpy} instead.
5581 void copy (char *d, const char *s)
5583 strncpy (d, s, strlen (s));
5587 In the following example, the call to @code{strncpy} specifies the size
5588 of the destination buffer as the bound. If the length of the source
5589 string is equal to or greater than this size the result of the copy will
5590 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5591 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5592 element of the buffer to @code{NUL}.
5595 void copy (const char *s)
5598 strncpy (buf, s, sizeof buf);
5603 In situations where a character array is intended to store a sequence
5604 of bytes with no terminating @code{NUL} such an array may be annotated
5605 with attribute @code{nonstring} to avoid this warning. Such arrays,
5606 however, are not suitable arguments to functions that expect
5607 @code{NUL}-terminated strings. To help detect accidental misuses of
5608 such arrays GCC issues warnings unless it can prove that the use is
5609 safe. @xref{Common Variable Attributes}.
5611 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5612 @opindex Wsuggest-attribute=
5613 @opindex Wno-suggest-attribute=
5614 Warn for cases where adding an attribute may be beneficial. The
5615 attributes currently supported are listed below.
5618 @item -Wsuggest-attribute=pure
5619 @itemx -Wsuggest-attribute=const
5620 @itemx -Wsuggest-attribute=noreturn
5621 @itemx -Wsuggest-attribute=malloc
5622 @opindex Wsuggest-attribute=pure
5623 @opindex Wno-suggest-attribute=pure
5624 @opindex Wsuggest-attribute=const
5625 @opindex Wno-suggest-attribute=const
5626 @opindex Wsuggest-attribute=noreturn
5627 @opindex Wno-suggest-attribute=noreturn
5628 @opindex Wsuggest-attribute=malloc
5629 @opindex Wno-suggest-attribute=malloc
5631 Warn about functions that might be candidates for attributes
5632 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5633 only warns for functions visible in other compilation units or (in the case of
5634 @code{pure} and @code{const}) if it cannot prove that the function returns
5635 normally. A function returns normally if it doesn't contain an infinite loop or
5636 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5637 requires option @option{-fipa-pure-const}, which is enabled by default at
5638 @option{-O} and higher. Higher optimization levels improve the accuracy
5641 @item -Wsuggest-attribute=format
5642 @itemx -Wmissing-format-attribute
5643 @opindex Wsuggest-attribute=format
5644 @opindex Wmissing-format-attribute
5645 @opindex Wno-suggest-attribute=format
5646 @opindex Wno-missing-format-attribute
5650 Warn about function pointers that might be candidates for @code{format}
5651 attributes. Note these are only possible candidates, not absolute ones.
5652 GCC guesses that function pointers with @code{format} attributes that
5653 are used in assignment, initialization, parameter passing or return
5654 statements should have a corresponding @code{format} attribute in the
5655 resulting type. I.e.@: the left-hand side of the assignment or
5656 initialization, the type of the parameter variable, or the return type
5657 of the containing function respectively should also have a @code{format}
5658 attribute to avoid the warning.
5660 GCC also warns about function definitions that might be
5661 candidates for @code{format} attributes. Again, these are only
5662 possible candidates. GCC guesses that @code{format} attributes
5663 might be appropriate for any function that calls a function like
5664 @code{vprintf} or @code{vscanf}, but this might not always be the
5665 case, and some functions for which @code{format} attributes are
5666 appropriate may not be detected.
5668 @item -Wsuggest-attribute=cold
5669 @opindex Wsuggest-attribute=cold
5670 @opindex Wno-suggest-attribute=cold
5672 Warn about functions that might be candidates for @code{cold} attribute. This
5673 is based on static detection and generally will only warn about functions which
5674 always leads to a call to another @code{cold} function such as wrappers of
5675 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
5678 @item -Wsuggest-final-types
5679 @opindex Wno-suggest-final-types
5680 @opindex Wsuggest-final-types
5681 Warn about types with virtual methods where code quality would be improved
5682 if the type were declared with the C++11 @code{final} specifier,
5684 declared in an anonymous namespace. This allows GCC to more aggressively
5685 devirtualize the polymorphic calls. This warning is more effective with link
5686 time optimization, where the information about the class hierarchy graph is
5689 @item -Wsuggest-final-methods
5690 @opindex Wno-suggest-final-methods
5691 @opindex Wsuggest-final-methods
5692 Warn about virtual methods where code quality would be improved if the method
5693 were declared with the C++11 @code{final} specifier,
5694 or, if possible, its type were
5695 declared in an anonymous namespace or with the @code{final} specifier.
5697 more effective with link-time optimization, where the information about the
5698 class hierarchy graph is more complete. It is recommended to first consider
5699 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
5702 @item -Wsuggest-override
5703 Warn about overriding virtual functions that are not marked with the override
5707 @opindex Wno-alloc-zero
5708 @opindex Walloc-zero
5709 Warn about calls to allocation functions decorated with attribute
5710 @code{alloc_size} that specify zero bytes, including those to the built-in
5711 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
5712 @code{malloc}, and @code{realloc}. Because the behavior of these functions
5713 when called with a zero size differs among implementations (and in the case
5714 of @code{realloc} has been deprecated) relying on it may result in subtle
5715 portability bugs and should be avoided.
5717 @item -Walloc-size-larger-than=@var{byte-size}
5718 @opindex Walloc-size-larger-than=
5719 @opindex Wno-alloc-size-larger-than
5720 Warn about calls to functions decorated with attribute @code{alloc_size}
5721 that attempt to allocate objects larger than the specified number of bytes,
5722 or where the result of the size computation in an integer type with infinite
5723 precision would exceed the value of @samp{PTRDIFF_MAX} on the target.
5724 @option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
5725 Warnings controlled by the option can be disabled either by specifying
5726 @var{byte-size} of @samp{SIZE_MAX} or more or by
5727 @option{-Wno-alloc-size-larger-than}.
5728 @xref{Function Attributes}.
5730 @item -Wno-alloc-size-larger-than
5731 @opindex Wno-alloc-size-larger-than
5732 Disable @option{-Walloc-size-larger-than=} warnings. The option is
5733 equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or
5739 This option warns on all uses of @code{alloca} in the source.
5741 @item -Walloca-larger-than=@var{byte-size}
5742 @opindex Walloca-larger-than=
5743 @opindex Wno-alloca-larger-than
5744 This option warns on calls to @code{alloca} with an integer argument whose
5745 value is either zero, or that is not bounded by a controlling predicate
5746 that limits its value to at most @var{byte-size}. It also warns for calls
5747 to @code{alloca} where the bound value is unknown. Arguments of non-integer
5748 types are considered unbounded even if they appear to be constrained to
5751 For example, a bounded case of @code{alloca} could be:
5754 void func (size_t n)
5765 In the above example, passing @code{-Walloca-larger-than=1000} would not
5766 issue a warning because the call to @code{alloca} is known to be at most
5767 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
5768 the compiler would emit a warning.
5770 Unbounded uses, on the other hand, are uses of @code{alloca} with no
5771 controlling predicate constraining its integer argument. For example:
5776 void *p = alloca (n);
5781 If @code{-Walloca-larger-than=500} were passed, the above would trigger
5782 a warning, but this time because of the lack of bounds checking.
5784 Note, that even seemingly correct code involving signed integers could
5788 void func (signed int n)
5798 In the above example, @var{n} could be negative, causing a larger than
5799 expected argument to be implicitly cast into the @code{alloca} call.
5801 This option also warns when @code{alloca} is used in a loop.
5803 @option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default
5804 but is usually only effective when @option{-ftree-vrp} is active (default
5805 for @option{-O2} and above).
5807 See also @option{-Wvla-larger-than=}@samp{byte-size}.
5809 @item -Wno-alloca-larger-than
5810 @opindex Wno-alloca-larger-than
5811 Disable @option{-Walloca-larger-than=} warnings. The option is
5812 equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger.
5814 @item -Warray-bounds
5815 @itemx -Warray-bounds=@var{n}
5816 @opindex Wno-array-bounds
5817 @opindex Warray-bounds
5818 This option is only active when @option{-ftree-vrp} is active
5819 (default for @option{-O2} and above). It warns about subscripts to arrays
5820 that are always out of bounds. This warning is enabled by @option{-Wall}.
5823 @item -Warray-bounds=1
5824 This is the warning level of @option{-Warray-bounds} and is enabled
5825 by @option{-Wall}; higher levels are not, and must be explicitly requested.
5827 @item -Warray-bounds=2
5828 This warning level also warns about out of bounds access for
5829 arrays at the end of a struct and for arrays accessed through
5830 pointers. This warning level may give a larger number of
5831 false positives and is deactivated by default.
5834 @item -Wattribute-alias
5835 Warn about declarations using the @code{alias} and similar attributes whose
5836 target is incompatible with the type of the alias. @xref{Function Attributes,
5837 ,Declaring Attributes of Functions}.
5839 @item -Wbool-compare
5840 @opindex Wno-bool-compare
5841 @opindex Wbool-compare
5842 Warn about boolean expression compared with an integer value different from
5843 @code{true}/@code{false}. For instance, the following comparison is
5848 if ((n > 1) == 2) @{ @dots{} @}
5850 This warning is enabled by @option{-Wall}.
5852 @item -Wbool-operation
5853 @opindex Wno-bool-operation
5854 @opindex Wbool-operation
5855 Warn about suspicious operations on expressions of a boolean type. For
5856 instance, bitwise negation of a boolean is very likely a bug in the program.
5857 For C, this warning also warns about incrementing or decrementing a boolean,
5858 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
5859 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
5861 This warning is enabled by @option{-Wall}.
5863 @item -Wduplicated-branches
5864 @opindex Wno-duplicated-branches
5865 @opindex Wduplicated-branches
5866 Warn when an if-else has identical branches. This warning detects cases like
5873 It doesn't warn when both branches contain just a null statement. This warning
5874 also warn for conditional operators:
5876 int i = x ? *p : *p;
5879 @item -Wduplicated-cond
5880 @opindex Wno-duplicated-cond
5881 @opindex Wduplicated-cond
5882 Warn about duplicated conditions in an if-else-if chain. For instance,
5883 warn for the following code:
5885 if (p->q != NULL) @{ @dots{} @}
5886 else if (p->q != NULL) @{ @dots{} @}
5889 @item -Wframe-address
5890 @opindex Wno-frame-address
5891 @opindex Wframe-address
5892 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
5893 is called with an argument greater than 0. Such calls may return indeterminate
5894 values or crash the program. The warning is included in @option{-Wall}.
5896 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
5897 @opindex Wno-discarded-qualifiers
5898 @opindex Wdiscarded-qualifiers
5899 Do not warn if type qualifiers on pointers are being discarded.
5900 Typically, the compiler warns if a @code{const char *} variable is
5901 passed to a function that takes a @code{char *} parameter. This option
5902 can be used to suppress such a warning.
5904 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
5905 @opindex Wno-discarded-array-qualifiers
5906 @opindex Wdiscarded-array-qualifiers
5907 Do not warn if type qualifiers on arrays which are pointer targets
5908 are being discarded. Typically, the compiler warns if a
5909 @code{const int (*)[]} variable is passed to a function that
5910 takes a @code{int (*)[]} parameter. This option can be used to
5911 suppress such a warning.
5913 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
5914 @opindex Wno-incompatible-pointer-types
5915 @opindex Wincompatible-pointer-types
5916 Do not warn when there is a conversion between pointers that have incompatible
5917 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
5918 which warns for pointer argument passing or assignment with different
5921 @item -Wno-int-conversion @r{(C and Objective-C only)}
5922 @opindex Wno-int-conversion
5923 @opindex Wint-conversion
5924 Do not warn about incompatible integer to pointer and pointer to integer
5925 conversions. This warning is about implicit conversions; for explicit
5926 conversions the warnings @option{-Wno-int-to-pointer-cast} and
5927 @option{-Wno-pointer-to-int-cast} may be used.
5929 @item -Wno-div-by-zero
5930 @opindex Wno-div-by-zero
5931 @opindex Wdiv-by-zero
5932 Do not warn about compile-time integer division by zero. Floating-point
5933 division by zero is not warned about, as it can be a legitimate way of
5934 obtaining infinities and NaNs.
5936 @item -Wsystem-headers
5937 @opindex Wsystem-headers
5938 @opindex Wno-system-headers
5939 @cindex warnings from system headers
5940 @cindex system headers, warnings from
5941 Print warning messages for constructs found in system header files.
5942 Warnings from system headers are normally suppressed, on the assumption
5943 that they usually do not indicate real problems and would only make the
5944 compiler output harder to read. Using this command-line option tells
5945 GCC to emit warnings from system headers as if they occurred in user
5946 code. However, note that using @option{-Wall} in conjunction with this
5947 option does @emph{not} warn about unknown pragmas in system
5948 headers---for that, @option{-Wunknown-pragmas} must also be used.
5950 @item -Wtautological-compare
5951 @opindex Wtautological-compare
5952 @opindex Wno-tautological-compare
5953 Warn if a self-comparison always evaluates to true or false. This
5954 warning detects various mistakes such as:
5958 if (i > i) @{ @dots{} @}
5961 This warning also warns about bitwise comparisons that always evaluate
5962 to true or false, for instance:
5964 if ((a & 16) == 10) @{ @dots{} @}
5966 will always be false.
5968 This warning is enabled by @option{-Wall}.
5971 @opindex Wtrampolines
5972 @opindex Wno-trampolines
5973 Warn about trampolines generated for pointers to nested functions.
5974 A trampoline is a small piece of data or code that is created at run
5975 time on the stack when the address of a nested function is taken, and is
5976 used to call the nested function indirectly. For some targets, it is
5977 made up of data only and thus requires no special treatment. But, for
5978 most targets, it is made up of code and thus requires the stack to be
5979 made executable in order for the program to work properly.
5982 @opindex Wfloat-equal
5983 @opindex Wno-float-equal
5984 Warn if floating-point values are used in equality comparisons.
5986 The idea behind this is that sometimes it is convenient (for the
5987 programmer) to consider floating-point values as approximations to
5988 infinitely precise real numbers. If you are doing this, then you need
5989 to compute (by analyzing the code, or in some other way) the maximum or
5990 likely maximum error that the computation introduces, and allow for it
5991 when performing comparisons (and when producing output, but that's a
5992 different problem). In particular, instead of testing for equality, you
5993 should check to see whether the two values have ranges that overlap; and
5994 this is done with the relational operators, so equality comparisons are
5997 @item -Wtraditional @r{(C and Objective-C only)}
5998 @opindex Wtraditional
5999 @opindex Wno-traditional
6000 Warn about certain constructs that behave differently in traditional and
6001 ISO C@. Also warn about ISO C constructs that have no traditional C
6002 equivalent, and/or problematic constructs that should be avoided.
6006 Macro parameters that appear within string literals in the macro body.
6007 In traditional C macro replacement takes place within string literals,
6008 but in ISO C it does not.
6011 In traditional C, some preprocessor directives did not exist.
6012 Traditional preprocessors only considered a line to be a directive
6013 if the @samp{#} appeared in column 1 on the line. Therefore
6014 @option{-Wtraditional} warns about directives that traditional C
6015 understands but ignores because the @samp{#} does not appear as the
6016 first character on the line. It also suggests you hide directives like
6017 @code{#pragma} not understood by traditional C by indenting them. Some
6018 traditional implementations do not recognize @code{#elif}, so this option
6019 suggests avoiding it altogether.
6022 A function-like macro that appears without arguments.
6025 The unary plus operator.
6028 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
6029 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
6030 constants.) Note, these suffixes appear in macros defined in the system
6031 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
6032 Use of these macros in user code might normally lead to spurious
6033 warnings, however GCC's integrated preprocessor has enough context to
6034 avoid warning in these cases.
6037 A function declared external in one block and then used after the end of
6041 A @code{switch} statement has an operand of type @code{long}.
6044 A non-@code{static} function declaration follows a @code{static} one.
6045 This construct is not accepted by some traditional C compilers.
6048 The ISO type of an integer constant has a different width or
6049 signedness from its traditional type. This warning is only issued if
6050 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
6051 typically represent bit patterns, are not warned about.
6054 Usage of ISO string concatenation is detected.
6057 Initialization of automatic aggregates.
6060 Identifier conflicts with labels. Traditional C lacks a separate
6061 namespace for labels.
6064 Initialization of unions. If the initializer is zero, the warning is
6065 omitted. This is done under the assumption that the zero initializer in
6066 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
6067 initializer warnings and relies on default initialization to zero in the
6071 Conversions by prototypes between fixed/floating-point values and vice
6072 versa. The absence of these prototypes when compiling with traditional
6073 C causes serious problems. This is a subset of the possible
6074 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
6077 Use of ISO C style function definitions. This warning intentionally is
6078 @emph{not} issued for prototype declarations or variadic functions
6079 because these ISO C features appear in your code when using
6080 libiberty's traditional C compatibility macros, @code{PARAMS} and
6081 @code{VPARAMS}. This warning is also bypassed for nested functions
6082 because that feature is already a GCC extension and thus not relevant to
6083 traditional C compatibility.
6086 @item -Wtraditional-conversion @r{(C and Objective-C only)}
6087 @opindex Wtraditional-conversion
6088 @opindex Wno-traditional-conversion
6089 Warn if a prototype causes a type conversion that is different from what
6090 would happen to the same argument in the absence of a prototype. This
6091 includes conversions of fixed point to floating and vice versa, and
6092 conversions changing the width or signedness of a fixed-point argument
6093 except when the same as the default promotion.
6095 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
6096 @opindex Wdeclaration-after-statement
6097 @opindex Wno-declaration-after-statement
6098 Warn when a declaration is found after a statement in a block. This
6099 construct, known from C++, was introduced with ISO C99 and is by default
6100 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
6105 Warn whenever a local variable or type declaration shadows another
6106 variable, parameter, type, class member (in C++), or instance variable
6107 (in Objective-C) or whenever a built-in function is shadowed. Note
6108 that in C++, the compiler warns if a local variable shadows an
6109 explicit typedef, but not if it shadows a struct/class/enum.
6110 Same as @option{-Wshadow=global}.
6112 @item -Wno-shadow-ivar @r{(Objective-C only)}
6113 @opindex Wno-shadow-ivar
6114 @opindex Wshadow-ivar
6115 Do not warn whenever a local variable shadows an instance variable in an
6118 @item -Wshadow=global
6119 @opindex Wshadow=local
6120 The default for @option{-Wshadow}. Warns for any (global) shadowing.
6122 @item -Wshadow=local
6123 @opindex Wshadow=local
6124 Warn when a local variable shadows another local variable or parameter.
6125 This warning is enabled by @option{-Wshadow=global}.
6127 @item -Wshadow=compatible-local
6128 @opindex Wshadow=compatible-local
6129 Warn when a local variable shadows another local variable or parameter
6130 whose type is compatible with that of the shadowing variable. In C++,
6131 type compatibility here means the type of the shadowing variable can be
6132 converted to that of the shadowed variable. The creation of this flag
6133 (in addition to @option{-Wshadow=local}) is based on the idea that when
6134 a local variable shadows another one of incompatible type, it is most
6135 likely intentional, not a bug or typo, as shown in the following example:
6139 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
6141 for (int i = 0; i < N; ++i)
6150 Since the two variable @code{i} in the example above have incompatible types,
6151 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
6152 Because their types are incompatible, if a programmer accidentally uses one
6153 in place of the other, type checking will catch that and emit an error or
6154 warning. So not warning (about shadowing) in this case will not lead to
6155 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
6156 possibly reduce the number of warnings triggered by intentional shadowing.
6158 This warning is enabled by @option{-Wshadow=local}.
6160 @item -Wlarger-than=@var{byte-size}
6161 @opindex Wlarger-than=
6162 @opindex Wlarger-than-@var{byte-size}
6163 Warn whenever an object is defined whose size exceeds @var{byte-size}.
6164 @option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6165 Warnings controlled by the option can be disabled either by specifying
6166 @var{byte-size} of @samp{SIZE_MAX} or more or by
6167 @option{-Wno-larger-than}.
6169 @item -Wno-larger-than
6170 @opindex Wno-larger-than
6171 Disable @option{-Wlarger-than=} warnings. The option is equivalent
6172 to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger.
6174 @item -Wframe-larger-than=@var{byte-size}
6175 @opindex Wframe-larger-than=
6176 @opindex Wno-frame-larger-than
6177 Warn if the size of a function frame exceeds @var{byte-size}.
6178 The computation done to determine the stack frame size is approximate
6179 and not conservative.
6180 The actual requirements may be somewhat greater than @var{byte-size}
6181 even if you do not get a warning. In addition, any space allocated
6182 via @code{alloca}, variable-length arrays, or related constructs
6183 is not included by the compiler when determining
6184 whether or not to issue a warning.
6185 @option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6186 Warnings controlled by the option can be disabled either by specifying
6187 @var{byte-size} of @samp{SIZE_MAX} or more or by
6188 @option{-Wno-frame-larger-than}.
6190 @item -Wno-frame-larger-than
6191 @opindex Wno-frame-larger-than
6192 Disable @option{-Wframe-larger-than=} warnings. The option is equivalent
6193 to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger.
6195 @item -Wno-free-nonheap-object
6196 @opindex Wno-free-nonheap-object
6197 @opindex Wfree-nonheap-object
6198 Do not warn when attempting to free an object that was not allocated
6201 @item -Wstack-usage=@var{byte-size}
6202 @opindex Wstack-usage
6203 @opindex Wno-stack-usage
6204 Warn if the stack usage of a function might exceed @var{byte-size}.
6205 The computation done to determine the stack usage is conservative.
6206 Any space allocated via @code{alloca}, variable-length arrays, or related
6207 constructs is included by the compiler when determining whether or not to
6210 The message is in keeping with the output of @option{-fstack-usage}.
6214 If the stack usage is fully static but exceeds the specified amount, it's:
6217 warning: stack usage is 1120 bytes
6220 If the stack usage is (partly) dynamic but bounded, it's:
6223 warning: stack usage might be 1648 bytes
6226 If the stack usage is (partly) dynamic and not bounded, it's:
6229 warning: stack usage might be unbounded
6233 @option{-Wstack-usage=}@samp{PTRDIFF_MAX} is enabled by default.
6234 Warnings controlled by the option can be disabled either by specifying
6235 @var{byte-size} of @samp{SIZE_MAX} or more or by
6236 @option{-Wno-stack-usage}.
6238 @item -Wno-stack-usage
6239 @opindex Wno-stack-usage
6240 Disable @option{-Wstack-usage=} warnings. The option is equivalent
6241 to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger.
6243 @item -Wunsafe-loop-optimizations
6244 @opindex Wunsafe-loop-optimizations
6245 @opindex Wno-unsafe-loop-optimizations
6246 Warn if the loop cannot be optimized because the compiler cannot
6247 assume anything on the bounds of the loop indices. With
6248 @option{-funsafe-loop-optimizations} warn if the compiler makes
6251 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
6252 @opindex Wno-pedantic-ms-format
6253 @opindex Wpedantic-ms-format
6254 When used in combination with @option{-Wformat}
6255 and @option{-pedantic} without GNU extensions, this option
6256 disables the warnings about non-ISO @code{printf} / @code{scanf} format
6257 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
6258 which depend on the MS runtime.
6261 @opindex Waligned-new
6262 @opindex Wno-aligned-new
6263 Warn about a new-expression of a type that requires greater alignment
6264 than the @code{alignof(std::max_align_t)} but uses an allocation
6265 function without an explicit alignment parameter. This option is
6266 enabled by @option{-Wall}.
6268 Normally this only warns about global allocation functions, but
6269 @option{-Waligned-new=all} also warns about class member allocation
6272 @item -Wplacement-new
6273 @itemx -Wplacement-new=@var{n}
6274 @opindex Wplacement-new
6275 @opindex Wno-placement-new
6276 Warn about placement new expressions with undefined behavior, such as
6277 constructing an object in a buffer that is smaller than the type of
6278 the object. For example, the placement new expression below is diagnosed
6279 because it attempts to construct an array of 64 integers in a buffer only
6285 This warning is enabled by default.
6288 @item -Wplacement-new=1
6289 This is the default warning level of @option{-Wplacement-new}. At this
6290 level the warning is not issued for some strictly undefined constructs that
6291 GCC allows as extensions for compatibility with legacy code. For example,
6292 the following @code{new} expression is not diagnosed at this level even
6293 though it has undefined behavior according to the C++ standard because
6294 it writes past the end of the one-element array.
6296 struct S @{ int n, a[1]; @};
6297 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6298 new (s->a)int [32]();
6301 @item -Wplacement-new=2
6302 At this level, in addition to diagnosing all the same constructs as at level
6303 1, a diagnostic is also issued for placement new expressions that construct
6304 an object in the last member of structure whose type is an array of a single
6305 element and whose size is less than the size of the object being constructed.
6306 While the previous example would be diagnosed, the following construct makes
6307 use of the flexible member array extension to avoid the warning at level 2.
6309 struct S @{ int n, a[]; @};
6310 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6311 new (s->a)int [32]();
6316 @item -Wpointer-arith
6317 @opindex Wpointer-arith
6318 @opindex Wno-pointer-arith
6319 Warn about anything that depends on the ``size of'' a function type or
6320 of @code{void}. GNU C assigns these types a size of 1, for
6321 convenience in calculations with @code{void *} pointers and pointers
6322 to functions. In C++, warn also when an arithmetic operation involves
6323 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6325 @item -Wpointer-compare
6326 @opindex Wpointer-compare
6327 @opindex Wno-pointer-compare
6328 Warn if a pointer is compared with a zero character constant. This usually
6329 means that the pointer was meant to be dereferenced. For example:
6332 const char *p = foo ();
6337 Note that the code above is invalid in C++11.
6339 This warning is enabled by default.
6342 @opindex Wtype-limits
6343 @opindex Wno-type-limits
6344 Warn if a comparison is always true or always false due to the limited
6345 range of the data type, but do not warn for constant expressions. For
6346 example, warn if an unsigned variable is compared against zero with
6347 @code{<} or @code{>=}. This warning is also enabled by
6350 @item -Wabsolute-value @r{(C and Objective-C only)}
6351 @opindex Wabsolute-value
6352 @opindex Wno-absolute-value
6353 Warn when a wrong absolute value function seems to be used or when it
6354 does not have any effect because its argument is an unsigned type.
6355 This warning be suppressed with an explicit type cast and it is also
6356 enabled by @option{-Wextra}.
6358 @include cppwarnopts.texi
6360 @item -Wbad-function-cast @r{(C and Objective-C only)}
6361 @opindex Wbad-function-cast
6362 @opindex Wno-bad-function-cast
6363 Warn when a function call is cast to a non-matching type.
6364 For example, warn if a call to a function returning an integer type
6365 is cast to a pointer type.
6367 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6368 @opindex Wc90-c99-compat
6369 @opindex Wno-c90-c99-compat
6370 Warn about features not present in ISO C90, but present in ISO C99.
6371 For instance, warn about use of variable length arrays, @code{long long}
6372 type, @code{bool} type, compound literals, designated initializers, and so
6373 on. This option is independent of the standards mode. Warnings are disabled
6374 in the expression that follows @code{__extension__}.
6376 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6377 @opindex Wc99-c11-compat
6378 @opindex Wno-c99-c11-compat
6379 Warn about features not present in ISO C99, but present in ISO C11.
6380 For instance, warn about use of anonymous structures and unions,
6381 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6382 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6383 and so on. This option is independent of the standards mode. Warnings are
6384 disabled in the expression that follows @code{__extension__}.
6386 @item -Wc++-compat @r{(C and Objective-C only)}
6387 @opindex Wc++-compat
6388 @opindex Wno-c++-compat
6389 Warn about ISO C constructs that are outside of the common subset of
6390 ISO C and ISO C++, e.g.@: request for implicit conversion from
6391 @code{void *} to a pointer to non-@code{void} type.
6393 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6394 @opindex Wc++11-compat
6395 @opindex Wno-c++11-compat
6396 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6397 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6398 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6399 enabled by @option{-Wall}.
6401 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6402 @opindex Wc++14-compat
6403 @opindex Wno-c++14-compat
6404 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6405 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6407 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6408 @opindex Wc++17-compat
6409 @opindex Wno-c++17-compat
6410 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6411 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6415 @opindex Wno-cast-qual
6416 Warn whenever a pointer is cast so as to remove a type qualifier from
6417 the target type. For example, warn if a @code{const char *} is cast
6418 to an ordinary @code{char *}.
6420 Also warn when making a cast that introduces a type qualifier in an
6421 unsafe way. For example, casting @code{char **} to @code{const char **}
6422 is unsafe, as in this example:
6425 /* p is char ** value. */
6426 const char **q = (const char **) p;
6427 /* Assignment of readonly string to const char * is OK. */
6429 /* Now char** pointer points to read-only memory. */
6434 @opindex Wcast-align
6435 @opindex Wno-cast-align
6436 Warn whenever a pointer is cast such that the required alignment of the
6437 target is increased. For example, warn if a @code{char *} is cast to
6438 an @code{int *} on machines where integers can only be accessed at
6439 two- or four-byte boundaries.
6441 @item -Wcast-align=strict
6442 @opindex Wcast-align=strict
6443 Warn whenever a pointer is cast such that the required alignment of the
6444 target is increased. For example, warn if a @code{char *} is cast to
6445 an @code{int *} regardless of the target machine.
6447 @item -Wcast-function-type
6448 @opindex Wcast-function-type
6449 @opindex Wno-cast-function-type
6450 Warn when a function pointer is cast to an incompatible function pointer.
6451 In a cast involving function types with a variable argument list only
6452 the types of initial arguments that are provided are considered.
6453 Any parameter of pointer-type matches any other pointer-type. Any benign
6454 differences in integral types are ignored, like @code{int} vs.@: @code{long}
6455 on ILP32 targets. Likewise type qualifiers are ignored. The function
6456 type @code{void (*) (void)} is special and matches everything, which can
6457 be used to suppress this warning.
6458 In a cast involving pointer to member types this warning warns whenever
6459 the type cast is changing the pointer to member type.
6460 This warning is enabled by @option{-Wextra}.
6462 @item -Wwrite-strings
6463 @opindex Wwrite-strings
6464 @opindex Wno-write-strings
6465 When compiling C, give string constants the type @code{const
6466 char[@var{length}]} so that copying the address of one into a
6467 non-@code{const} @code{char *} pointer produces a warning. These
6468 warnings help you find at compile time code that can try to write
6469 into a string constant, but only if you have been very careful about
6470 using @code{const} in declarations and prototypes. Otherwise, it is
6471 just a nuisance. This is why we did not make @option{-Wall} request
6474 When compiling C++, warn about the deprecated conversion from string
6475 literals to @code{char *}. This warning is enabled by default for C++
6479 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6480 @opindex Wcatch-value
6481 @opindex Wno-catch-value
6482 Warn about catch handlers that do not catch via reference.
6483 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6484 warn about polymorphic class types that are caught by value.
6485 With @option{-Wcatch-value=2} warn about all class types that are caught
6486 by value. With @option{-Wcatch-value=3} warn about all types that are
6487 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6491 @opindex Wno-clobbered
6492 Warn for variables that might be changed by @code{longjmp} or
6493 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6495 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6496 @opindex Wconditionally-supported
6497 @opindex Wno-conditionally-supported
6498 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6501 @opindex Wconversion
6502 @opindex Wno-conversion
6503 Warn for implicit conversions that may alter a value. This includes
6504 conversions between real and integer, like @code{abs (x)} when
6505 @code{x} is @code{double}; conversions between signed and unsigned,
6506 like @code{unsigned ui = -1}; and conversions to smaller types, like
6507 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6508 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6509 changed by the conversion like in @code{abs (2.0)}. Warnings about
6510 conversions between signed and unsigned integers can be disabled by
6511 using @option{-Wno-sign-conversion}.
6513 For C++, also warn for confusing overload resolution for user-defined
6514 conversions; and conversions that never use a type conversion
6515 operator: conversions to @code{void}, the same type, a base class or a
6516 reference to them. Warnings about conversions between signed and
6517 unsigned integers are disabled by default in C++ unless
6518 @option{-Wsign-conversion} is explicitly enabled.
6520 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6521 @opindex Wconversion-null
6522 @opindex Wno-conversion-null
6523 Do not warn for conversions between @code{NULL} and non-pointer
6524 types. @option{-Wconversion-null} is enabled by default.
6526 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6527 @opindex Wzero-as-null-pointer-constant
6528 @opindex Wno-zero-as-null-pointer-constant
6529 Warn when a literal @samp{0} is used as null pointer constant. This can
6530 be useful to facilitate the conversion to @code{nullptr} in C++11.
6532 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6533 @opindex Wsubobject-linkage
6534 @opindex Wno-subobject-linkage
6535 Warn if a class type has a base or a field whose type uses the anonymous
6536 namespace or depends on a type with no linkage. If a type A depends on
6537 a type B with no or internal linkage, defining it in multiple
6538 translation units would be an ODR violation because the meaning of B
6539 is different in each translation unit. If A only appears in a single
6540 translation unit, the best way to silence the warning is to give it
6541 internal linkage by putting it in an anonymous namespace as well. The
6542 compiler doesn't give this warning for types defined in the main .C
6543 file, as those are unlikely to have multiple definitions.
6544 @option{-Wsubobject-linkage} is enabled by default.
6546 @item -Wdangling-else
6547 @opindex Wdangling-else
6548 @opindex Wno-dangling-else
6549 Warn about constructions where there may be confusion to which
6550 @code{if} statement an @code{else} branch belongs. Here is an example of
6565 In C/C++, every @code{else} branch belongs to the innermost possible
6566 @code{if} statement, which in this example is @code{if (b)}. This is
6567 often not what the programmer expected, as illustrated in the above
6568 example by indentation the programmer chose. When there is the
6569 potential for this confusion, GCC issues a warning when this flag
6570 is specified. To eliminate the warning, add explicit braces around
6571 the innermost @code{if} statement so there is no way the @code{else}
6572 can belong to the enclosing @code{if}. The resulting code
6589 This warning is enabled by @option{-Wparentheses}.
6593 @opindex Wno-date-time
6594 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6595 are encountered as they might prevent bit-wise-identical reproducible
6598 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6599 @opindex Wdelete-incomplete
6600 @opindex Wno-delete-incomplete
6601 Warn when deleting a pointer to incomplete type, which may cause
6602 undefined behavior at runtime. This warning is enabled by default.
6604 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6605 @opindex Wuseless-cast
6606 @opindex Wno-useless-cast
6607 Warn when an expression is casted to its own type.
6610 @opindex Wempty-body
6611 @opindex Wno-empty-body
6612 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6613 while} statement. This warning is also enabled by @option{-Wextra}.
6615 @item -Wenum-compare
6616 @opindex Wenum-compare
6617 @opindex Wno-enum-compare
6618 Warn about a comparison between values of different enumerated types.
6619 In C++ enumerated type mismatches in conditional expressions are also
6620 diagnosed and the warning is enabled by default. In C this warning is
6621 enabled by @option{-Wall}.
6623 @item -Wextra-semi @r{(C++, Objective-C++ only)}
6624 @opindex Wextra-semi
6625 @opindex Wno-extra-semi
6626 Warn about redundant semicolon after in-class function definition.
6628 @item -Wjump-misses-init @r{(C, Objective-C only)}
6629 @opindex Wjump-misses-init
6630 @opindex Wno-jump-misses-init
6631 Warn if a @code{goto} statement or a @code{switch} statement jumps
6632 forward across the initialization of a variable, or jumps backward to a
6633 label after the variable has been initialized. This only warns about
6634 variables that are initialized when they are declared. This warning is
6635 only supported for C and Objective-C; in C++ this sort of branch is an
6638 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
6639 can be disabled with the @option{-Wno-jump-misses-init} option.
6641 @item -Wsign-compare
6642 @opindex Wsign-compare
6643 @opindex Wno-sign-compare
6644 @cindex warning for comparison of signed and unsigned values
6645 @cindex comparison of signed and unsigned values, warning
6646 @cindex signed and unsigned values, comparison warning
6647 Warn when a comparison between signed and unsigned values could produce
6648 an incorrect result when the signed value is converted to unsigned.
6649 In C++, this warning is also enabled by @option{-Wall}. In C, it is
6650 also enabled by @option{-Wextra}.
6652 @item -Wsign-conversion
6653 @opindex Wsign-conversion
6654 @opindex Wno-sign-conversion
6655 Warn for implicit conversions that may change the sign of an integer
6656 value, like assigning a signed integer expression to an unsigned
6657 integer variable. An explicit cast silences the warning. In C, this
6658 option is enabled also by @option{-Wconversion}.
6660 @item -Wfloat-conversion
6661 @opindex Wfloat-conversion
6662 @opindex Wno-float-conversion
6663 Warn for implicit conversions that reduce the precision of a real value.
6664 This includes conversions from real to integer, and from higher precision
6665 real to lower precision real values. This option is also enabled by
6666 @option{-Wconversion}.
6668 @item -Wno-scalar-storage-order
6669 @opindex Wno-scalar-storage-order
6670 @opindex Wscalar-storage-order
6671 Do not warn on suspicious constructs involving reverse scalar storage order.
6673 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
6674 @opindex Wsized-deallocation
6675 @opindex Wno-sized-deallocation
6676 Warn about a definition of an unsized deallocation function
6678 void operator delete (void *) noexcept;
6679 void operator delete[] (void *) noexcept;
6681 without a definition of the corresponding sized deallocation function
6683 void operator delete (void *, std::size_t) noexcept;
6684 void operator delete[] (void *, std::size_t) noexcept;
6686 or vice versa. Enabled by @option{-Wextra} along with
6687 @option{-fsized-deallocation}.
6689 @item -Wsizeof-pointer-div
6690 @opindex Wsizeof-pointer-div
6691 @opindex Wno-sizeof-pointer-div
6692 Warn for suspicious divisions of two sizeof expressions that divide
6693 the pointer size by the element size, which is the usual way to compute
6694 the array size but won't work out correctly with pointers. This warning
6695 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
6696 not an array, but a pointer. This warning is enabled by @option{-Wall}.
6698 @item -Wsizeof-pointer-memaccess
6699 @opindex Wsizeof-pointer-memaccess
6700 @opindex Wno-sizeof-pointer-memaccess
6701 Warn for suspicious length parameters to certain string and memory built-in
6702 functions if the argument uses @code{sizeof}. This warning triggers for
6703 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
6704 an array, but a pointer, and suggests a possible fix, or about
6705 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
6706 also warns about calls to bounded string copy functions like @code{strncat}
6707 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
6708 the source array. For example, in the following function the call to
6709 @code{strncat} specifies the size of the source string as the bound. That
6710 is almost certainly a mistake and so the call is diagnosed.
6712 void make_file (const char *name)
6714 char path[PATH_MAX];
6715 strncpy (path, name, sizeof path - 1);
6716 strncat (path, ".text", sizeof ".text");
6721 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
6723 @item -Wsizeof-array-argument
6724 @opindex Wsizeof-array-argument
6725 @opindex Wno-sizeof-array-argument
6726 Warn when the @code{sizeof} operator is applied to a parameter that is
6727 declared as an array in a function definition. This warning is enabled by
6728 default for C and C++ programs.
6730 @item -Wmemset-elt-size
6731 @opindex Wmemset-elt-size
6732 @opindex Wno-memset-elt-size
6733 Warn for suspicious calls to the @code{memset} built-in function, if the
6734 first argument references an array, and the third argument is a number
6735 equal to the number of elements, but not equal to the size of the array
6736 in memory. This indicates that the user has omitted a multiplication by
6737 the element size. This warning is enabled by @option{-Wall}.
6739 @item -Wmemset-transposed-args
6740 @opindex Wmemset-transposed-args
6741 @opindex Wno-memset-transposed-args
6742 Warn for suspicious calls to the @code{memset} built-in function, if the
6743 second argument is not zero and the third argument is zero. This warns e.g.@:
6744 about @code{memset (buf, sizeof buf, 0)} where most probably
6745 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
6746 is only emitted if the third argument is literal zero. If it is some
6747 expression that is folded to zero, a cast of zero to some type, etc.,
6748 it is far less likely that the user has mistakenly exchanged the arguments
6749 and no warning is emitted. This warning is enabled by @option{-Wall}.
6753 @opindex Wno-address
6754 Warn about suspicious uses of memory addresses. These include using
6755 the address of a function in a conditional expression, such as
6756 @code{void func(void); if (func)}, and comparisons against the memory
6757 address of a string literal, such as @code{if (x == "abc")}. Such
6758 uses typically indicate a programmer error: the address of a function
6759 always evaluates to true, so their use in a conditional usually
6760 indicate that the programmer forgot the parentheses in a function
6761 call; and comparisons against string literals result in unspecified
6762 behavior and are not portable in C, so they usually indicate that the
6763 programmer intended to use @code{strcmp}. This warning is enabled by
6767 @opindex Wlogical-op
6768 @opindex Wno-logical-op
6769 Warn about suspicious uses of logical operators in expressions.
6770 This includes using logical operators in contexts where a
6771 bit-wise operator is likely to be expected. Also warns when
6772 the operands of a logical operator are the same:
6775 if (a < 0 && a < 0) @{ @dots{} @}
6778 @item -Wlogical-not-parentheses
6779 @opindex Wlogical-not-parentheses
6780 @opindex Wno-logical-not-parentheses
6781 Warn about logical not used on the left hand side operand of a comparison.
6782 This option does not warn if the right operand is considered to be a boolean
6783 expression. Its purpose is to detect suspicious code like the following:
6787 if (!a > 1) @{ @dots{} @}
6790 It is possible to suppress the warning by wrapping the LHS into
6793 if ((!a) > 1) @{ @dots{} @}
6796 This warning is enabled by @option{-Wall}.
6798 @item -Waggregate-return
6799 @opindex Waggregate-return
6800 @opindex Wno-aggregate-return
6801 Warn if any functions that return structures or unions are defined or
6802 called. (In languages where you can return an array, this also elicits
6805 @item -Wno-aggressive-loop-optimizations
6806 @opindex Wno-aggressive-loop-optimizations
6807 @opindex Waggressive-loop-optimizations
6808 Warn if in a loop with constant number of iterations the compiler detects
6809 undefined behavior in some statement during one or more of the iterations.
6811 @item -Wno-attributes
6812 @opindex Wno-attributes
6813 @opindex Wattributes
6814 Do not warn if an unexpected @code{__attribute__} is used, such as
6815 unrecognized attributes, function attributes applied to variables,
6816 etc. This does not stop errors for incorrect use of supported
6819 @item -Wno-builtin-declaration-mismatch
6820 @opindex Wno-builtin-declaration-mismatch
6821 @opindex Wbuiltin-declaration-mismatch
6822 Warn if a built-in function is declared with the wrong signature or
6824 This warning is enabled by default.
6826 @item -Wno-builtin-macro-redefined
6827 @opindex Wno-builtin-macro-redefined
6828 @opindex Wbuiltin-macro-redefined
6829 Do not warn if certain built-in macros are redefined. This suppresses
6830 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
6831 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
6833 @item -Wstrict-prototypes @r{(C and Objective-C only)}
6834 @opindex Wstrict-prototypes
6835 @opindex Wno-strict-prototypes
6836 Warn if a function is declared or defined without specifying the
6837 argument types. (An old-style function definition is permitted without
6838 a warning if preceded by a declaration that specifies the argument
6841 @item -Wold-style-declaration @r{(C and Objective-C only)}
6842 @opindex Wold-style-declaration
6843 @opindex Wno-old-style-declaration
6844 Warn for obsolescent usages, according to the C Standard, in a
6845 declaration. For example, warn if storage-class specifiers like
6846 @code{static} are not the first things in a declaration. This warning
6847 is also enabled by @option{-Wextra}.
6849 @item -Wold-style-definition @r{(C and Objective-C only)}
6850 @opindex Wold-style-definition
6851 @opindex Wno-old-style-definition
6852 Warn if an old-style function definition is used. A warning is given
6853 even if there is a previous prototype.
6855 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
6856 @opindex Wmissing-parameter-type
6857 @opindex Wno-missing-parameter-type
6858 A function parameter is declared without a type specifier in K&R-style
6865 This warning is also enabled by @option{-Wextra}.
6867 @item -Wmissing-prototypes @r{(C and Objective-C only)}
6868 @opindex Wmissing-prototypes
6869 @opindex Wno-missing-prototypes
6870 Warn if a global function is defined without a previous prototype
6871 declaration. This warning is issued even if the definition itself
6872 provides a prototype. Use this option to detect global functions
6873 that do not have a matching prototype declaration in a header file.
6874 This option is not valid for C++ because all function declarations
6875 provide prototypes and a non-matching declaration declares an
6876 overload rather than conflict with an earlier declaration.
6877 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
6879 @item -Wmissing-declarations
6880 @opindex Wmissing-declarations
6881 @opindex Wno-missing-declarations
6882 Warn if a global function is defined without a previous declaration.
6883 Do so even if the definition itself provides a prototype.
6884 Use this option to detect global functions that are not declared in
6885 header files. In C, no warnings are issued for functions with previous
6886 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
6887 missing prototypes. In C++, no warnings are issued for function templates,
6888 or for inline functions, or for functions in anonymous namespaces.
6890 @item -Wmissing-field-initializers
6891 @opindex Wmissing-field-initializers
6892 @opindex Wno-missing-field-initializers
6896 Warn if a structure's initializer has some fields missing. For
6897 example, the following code causes such a warning, because
6898 @code{x.h} is implicitly zero:
6901 struct s @{ int f, g, h; @};
6902 struct s x = @{ 3, 4 @};
6905 This option does not warn about designated initializers, so the following
6906 modification does not trigger a warning:
6909 struct s @{ int f, g, h; @};
6910 struct s x = @{ .f = 3, .g = 4 @};
6913 In C this option does not warn about the universal zero initializer
6917 struct s @{ int f, g, h; @};
6918 struct s x = @{ 0 @};
6921 Likewise, in C++ this option does not warn about the empty @{ @}
6922 initializer, for example:
6925 struct s @{ int f, g, h; @};
6929 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
6930 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
6932 @item -Wno-multichar
6933 @opindex Wno-multichar
6935 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
6936 Usually they indicate a typo in the user's code, as they have
6937 implementation-defined values, and should not be used in portable code.
6939 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
6940 @opindex Wnormalized=
6941 @opindex Wnormalized
6942 @opindex Wno-normalized
6945 @cindex character set, input normalization
6946 In ISO C and ISO C++, two identifiers are different if they are
6947 different sequences of characters. However, sometimes when characters
6948 outside the basic ASCII character set are used, you can have two
6949 different character sequences that look the same. To avoid confusion,
6950 the ISO 10646 standard sets out some @dfn{normalization rules} which
6951 when applied ensure that two sequences that look the same are turned into
6952 the same sequence. GCC can warn you if you are using identifiers that
6953 have not been normalized; this option controls that warning.
6955 There are four levels of warning supported by GCC@. The default is
6956 @option{-Wnormalized=nfc}, which warns about any identifier that is
6957 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
6958 recommended form for most uses. It is equivalent to
6959 @option{-Wnormalized}.
6961 Unfortunately, there are some characters allowed in identifiers by
6962 ISO C and ISO C++ that, when turned into NFC, are not allowed in
6963 identifiers. That is, there's no way to use these symbols in portable
6964 ISO C or C++ and have all your identifiers in NFC@.
6965 @option{-Wnormalized=id} suppresses the warning for these characters.
6966 It is hoped that future versions of the standards involved will correct
6967 this, which is why this option is not the default.
6969 You can switch the warning off for all characters by writing
6970 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
6971 only do this if you are using some other normalization scheme (like
6972 ``D''), because otherwise you can easily create bugs that are
6973 literally impossible to see.
6975 Some characters in ISO 10646 have distinct meanings but look identical
6976 in some fonts or display methodologies, especially once formatting has
6977 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
6978 LETTER N'', displays just like a regular @code{n} that has been
6979 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
6980 normalization scheme to convert all these into a standard form as
6981 well, and GCC warns if your code is not in NFKC if you use
6982 @option{-Wnormalized=nfkc}. This warning is comparable to warning
6983 about every identifier that contains the letter O because it might be
6984 confused with the digit 0, and so is not the default, but may be
6985 useful as a local coding convention if the programming environment
6986 cannot be fixed to display these characters distinctly.
6988 @item -Wno-attribute-warning
6989 @opindex Wno-attribute-warning
6990 @opindex Wattribute-warning
6991 Do not warn about usage of functions (@pxref{Function Attributes})
6992 declared with @code{warning} attribute. By default, this warning is
6993 enabled. @option{-Wno-attribute-warning} can be used to disable the
6994 warning or @option{-Wno-error=attribute-warning} can be used to
6995 disable the error when compiled with @option{-Werror} flag.
6997 @item -Wno-deprecated
6998 @opindex Wno-deprecated
6999 @opindex Wdeprecated
7000 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
7002 @item -Wno-deprecated-declarations
7003 @opindex Wno-deprecated-declarations
7004 @opindex Wdeprecated-declarations
7005 Do not warn about uses of functions (@pxref{Function Attributes}),
7006 variables (@pxref{Variable Attributes}), and types (@pxref{Type
7007 Attributes}) marked as deprecated by using the @code{deprecated}
7011 @opindex Wno-overflow
7013 Do not warn about compile-time overflow in constant expressions.
7018 Warn about One Definition Rule violations during link-time optimization.
7019 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
7022 @opindex Wopenmp-simd
7023 @opindex Wno-openmp-simd
7024 Warn if the vectorizer cost model overrides the OpenMP
7025 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
7026 option can be used to relax the cost model.
7028 @item -Woverride-init @r{(C and Objective-C only)}
7029 @opindex Woverride-init
7030 @opindex Wno-override-init
7034 Warn if an initialized field without side effects is overridden when
7035 using designated initializers (@pxref{Designated Inits, , Designated
7038 This warning is included in @option{-Wextra}. To get other
7039 @option{-Wextra} warnings without this one, use @option{-Wextra
7040 -Wno-override-init}.
7042 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
7043 @opindex Woverride-init-side-effects
7044 @opindex Wno-override-init-side-effects
7045 Warn if an initialized field with side effects is overridden when
7046 using designated initializers (@pxref{Designated Inits, , Designated
7047 Initializers}). This warning is enabled by default.
7052 Warn if a structure is given the packed attribute, but the packed
7053 attribute has no effect on the layout or size of the structure.
7054 Such structures may be mis-aligned for little benefit. For
7055 instance, in this code, the variable @code{f.x} in @code{struct bar}
7056 is misaligned even though @code{struct bar} does not itself
7057 have the packed attribute:
7064 @} __attribute__((packed));
7072 @item -Wpacked-bitfield-compat
7073 @opindex Wpacked-bitfield-compat
7074 @opindex Wno-packed-bitfield-compat
7075 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
7076 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
7077 the change can lead to differences in the structure layout. GCC
7078 informs you when the offset of such a field has changed in GCC 4.4.
7079 For example there is no longer a 4-bit padding between field @code{a}
7080 and @code{b} in this structure:
7087 @} __attribute__ ((packed));
7090 This warning is enabled by default. Use
7091 @option{-Wno-packed-bitfield-compat} to disable this warning.
7093 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
7094 @opindex Wpacked-not-aligned
7095 @opindex Wno-packed-not-aligned
7096 Warn if a structure field with explicitly specified alignment in a
7097 packed struct or union is misaligned. For example, a warning will
7098 be issued on @code{struct S}, like, @code{warning: alignment 1 of
7099 'struct S' is less than 8}, in this code:
7103 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
7104 struct __attribute__ ((packed)) S @{
7110 This warning is enabled by @option{-Wall}.
7115 Warn if padding is included in a structure, either to align an element
7116 of the structure or to align the whole structure. Sometimes when this
7117 happens it is possible to rearrange the fields of the structure to
7118 reduce the padding and so make the structure smaller.
7120 @item -Wredundant-decls
7121 @opindex Wredundant-decls
7122 @opindex Wno-redundant-decls
7123 Warn if anything is declared more than once in the same scope, even in
7124 cases where multiple declaration is valid and changes nothing.
7128 @opindex Wno-restrict
7129 Warn when an object referenced by a @code{restrict}-qualified parameter
7130 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
7131 argument, or when copies between such objects overlap. For example,
7132 the call to the @code{strcpy} function below attempts to truncate the string
7133 by replacing its initial characters with the last four. However, because
7134 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
7135 the call is diagnosed.
7140 char a[] = "abcd1234";
7145 The @option{-Wrestrict} option detects some instances of simple overlap
7146 even without optimization but works best at @option{-O2} and above. It
7147 is included in @option{-Wall}.
7149 @item -Wnested-externs @r{(C and Objective-C only)}
7150 @opindex Wnested-externs
7151 @opindex Wno-nested-externs
7152 Warn if an @code{extern} declaration is encountered within a function.
7154 @item -Wno-inherited-variadic-ctor
7155 @opindex Winherited-variadic-ctor
7156 @opindex Wno-inherited-variadic-ctor
7157 Suppress warnings about use of C++11 inheriting constructors when the
7158 base class inherited from has a C variadic constructor; the warning is
7159 on by default because the ellipsis is not inherited.
7164 Warn if a function that is declared as inline cannot be inlined.
7165 Even with this option, the compiler does not warn about failures to
7166 inline functions declared in system headers.
7168 The compiler uses a variety of heuristics to determine whether or not
7169 to inline a function. For example, the compiler takes into account
7170 the size of the function being inlined and the amount of inlining
7171 that has already been done in the current function. Therefore,
7172 seemingly insignificant changes in the source program can cause the
7173 warnings produced by @option{-Winline} to appear or disappear.
7175 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
7176 @opindex Wno-invalid-offsetof
7177 @opindex Winvalid-offsetof
7178 Suppress warnings from applying the @code{offsetof} macro to a non-POD
7179 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
7180 to a non-standard-layout type is undefined. In existing C++ implementations,
7181 however, @code{offsetof} typically gives meaningful results.
7182 This flag is for users who are aware that they are
7183 writing nonportable code and who have deliberately chosen to ignore the
7186 The restrictions on @code{offsetof} may be relaxed in a future version
7187 of the C++ standard.
7189 @item -Wint-in-bool-context
7190 @opindex Wint-in-bool-context
7191 @opindex Wno-int-in-bool-context
7192 Warn for suspicious use of integer values where boolean values are expected,
7193 such as conditional expressions (?:) using non-boolean integer constants in
7194 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
7195 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
7196 for all kinds of multiplications regardless of the data type.
7197 This warning is enabled by @option{-Wall}.
7199 @item -Wno-int-to-pointer-cast
7200 @opindex Wno-int-to-pointer-cast
7201 @opindex Wint-to-pointer-cast
7202 Suppress warnings from casts to pointer type of an integer of a
7203 different size. In C++, casting to a pointer type of smaller size is
7204 an error. @option{Wint-to-pointer-cast} is enabled by default.
7207 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
7208 @opindex Wno-pointer-to-int-cast
7209 @opindex Wpointer-to-int-cast
7210 Suppress warnings from casts from a pointer to an integer type of a
7214 @opindex Winvalid-pch
7215 @opindex Wno-invalid-pch
7216 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
7217 the search path but cannot be used.
7221 @opindex Wno-long-long
7222 Warn if @code{long long} type is used. This is enabled by either
7223 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
7224 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
7226 @item -Wvariadic-macros
7227 @opindex Wvariadic-macros
7228 @opindex Wno-variadic-macros
7229 Warn if variadic macros are used in ISO C90 mode, or if the GNU
7230 alternate syntax is used in ISO C99 mode. This is enabled by either
7231 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
7232 messages, use @option{-Wno-variadic-macros}.
7236 @opindex Wno-varargs
7237 Warn upon questionable usage of the macros used to handle variable
7238 arguments like @code{va_start}. This is default. To inhibit the
7239 warning messages, use @option{-Wno-varargs}.
7241 @item -Wvector-operation-performance
7242 @opindex Wvector-operation-performance
7243 @opindex Wno-vector-operation-performance
7244 Warn if vector operation is not implemented via SIMD capabilities of the
7245 architecture. Mainly useful for the performance tuning.
7246 Vector operation can be implemented @code{piecewise}, which means that the
7247 scalar operation is performed on every vector element;
7248 @code{in parallel}, which means that the vector operation is implemented
7249 using scalars of wider type, which normally is more performance efficient;
7250 and @code{as a single scalar}, which means that vector fits into a
7253 @item -Wno-virtual-move-assign
7254 @opindex Wvirtual-move-assign
7255 @opindex Wno-virtual-move-assign
7256 Suppress warnings about inheriting from a virtual base with a
7257 non-trivial C++11 move assignment operator. This is dangerous because
7258 if the virtual base is reachable along more than one path, it is
7259 moved multiple times, which can mean both objects end up in the
7260 moved-from state. If the move assignment operator is written to avoid
7261 moving from a moved-from object, this warning can be disabled.
7266 Warn if a variable-length array is used in the code.
7267 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
7268 the variable-length array.
7270 @item -Wvla-larger-than=@var{byte-size}
7271 @opindex Wvla-larger-than=
7272 @opindex Wno-vla-larger-than
7273 If this option is used, the compiler will warn for declarations of
7274 variable-length arrays whose size is either unbounded, or bounded
7275 by an argument that allows the array size to exceed @var{byte-size}
7276 bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size}
7277 works, but with variable-length arrays.
7279 Note that GCC may optimize small variable-length arrays of a known
7280 value into plain arrays, so this warning may not get triggered for
7283 @option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but
7284 is typically only effective when @option{-ftree-vrp} is active (default
7285 for @option{-O2} and above).
7287 See also @option{-Walloca-larger-than=@var{byte-size}}.
7289 @item -Wno-vla-larger-than
7290 @opindex Wno-vla-larger-than
7291 Disable @option{-Wvla-larger-than=} warnings. The option is equivalent
7292 to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger.
7294 @item -Wvolatile-register-var
7295 @opindex Wvolatile-register-var
7296 @opindex Wno-volatile-register-var
7297 Warn if a register variable is declared volatile. The volatile
7298 modifier does not inhibit all optimizations that may eliminate reads
7299 and/or writes to register variables. This warning is enabled by
7302 @item -Wdisabled-optimization
7303 @opindex Wdisabled-optimization
7304 @opindex Wno-disabled-optimization
7305 Warn if a requested optimization pass is disabled. This warning does
7306 not generally indicate that there is anything wrong with your code; it
7307 merely indicates that GCC's optimizers are unable to handle the code
7308 effectively. Often, the problem is that your code is too big or too
7309 complex; GCC refuses to optimize programs when the optimization
7310 itself is likely to take inordinate amounts of time.
7312 @item -Wpointer-sign @r{(C and Objective-C only)}
7313 @opindex Wpointer-sign
7314 @opindex Wno-pointer-sign
7315 Warn for pointer argument passing or assignment with different signedness.
7316 This option is only supported for C and Objective-C@. It is implied by
7317 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7318 @option{-Wno-pointer-sign}.
7320 @item -Wstack-protector
7321 @opindex Wstack-protector
7322 @opindex Wno-stack-protector
7323 This option is only active when @option{-fstack-protector} is active. It
7324 warns about functions that are not protected against stack smashing.
7326 @item -Woverlength-strings
7327 @opindex Woverlength-strings
7328 @opindex Wno-overlength-strings
7329 Warn about string constants that are longer than the ``minimum
7330 maximum'' length specified in the C standard. Modern compilers
7331 generally allow string constants that are much longer than the
7332 standard's minimum limit, but very portable programs should avoid
7333 using longer strings.
7335 The limit applies @emph{after} string constant concatenation, and does
7336 not count the trailing NUL@. In C90, the limit was 509 characters; in
7337 C99, it was raised to 4095. C++98 does not specify a normative
7338 minimum maximum, so we do not diagnose overlength strings in C++@.
7340 This option is implied by @option{-Wpedantic}, and can be disabled with
7341 @option{-Wno-overlength-strings}.
7343 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7344 @opindex Wunsuffixed-float-constants
7345 @opindex Wno-unsuffixed-float-constants
7347 Issue a warning for any floating constant that does not have
7348 a suffix. When used together with @option{-Wsystem-headers} it
7349 warns about such constants in system header files. This can be useful
7350 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7351 from the decimal floating-point extension to C99.
7353 @item -Wno-designated-init @r{(C and Objective-C only)}
7354 Suppress warnings when a positional initializer is used to initialize
7355 a structure that has been marked with the @code{designated_init}
7359 Issue a warning when HSAIL cannot be emitted for the compiled function or
7364 @node Debugging Options
7365 @section Options for Debugging Your Program
7366 @cindex options, debugging
7367 @cindex debugging information options
7369 To tell GCC to emit extra information for use by a debugger, in almost
7370 all cases you need only to add @option{-g} to your other options.
7372 GCC allows you to use @option{-g} with
7373 @option{-O}. The shortcuts taken by optimized code may occasionally
7374 be surprising: some variables you declared may not exist
7375 at all; flow of control may briefly move where you did not expect it;
7376 some statements may not be executed because they compute constant
7377 results or their values are already at hand; some statements may
7378 execute in different places because they have been moved out of loops.
7379 Nevertheless it is possible to debug optimized output. This makes
7380 it reasonable to use the optimizer for programs that might have bugs.
7382 If you are not using some other optimization option, consider
7383 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7384 With no @option{-O} option at all, some compiler passes that collect
7385 information useful for debugging do not run at all, so that
7386 @option{-Og} may result in a better debugging experience.
7391 Produce debugging information in the operating system's native format
7392 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7395 On most systems that use stabs format, @option{-g} enables use of extra
7396 debugging information that only GDB can use; this extra information
7397 makes debugging work better in GDB but probably makes other debuggers
7399 refuse to read the program. If you want to control for certain whether
7400 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7401 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7405 Produce debugging information for use by GDB@. This means to use the
7406 most expressive format available (DWARF, stabs, or the native format
7407 if neither of those are supported), including GDB extensions if at all
7411 @itemx -gdwarf-@var{version}
7413 Produce debugging information in DWARF format (if that is supported).
7414 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7415 for most targets is 4. DWARF Version 5 is only experimental.
7417 Note that with DWARF Version 2, some ports require and always
7418 use some non-conflicting DWARF 3 extensions in the unwind tables.
7420 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7421 for maximum benefit.
7423 GCC no longer supports DWARF Version 1, which is substantially
7424 different than Version 2 and later. For historical reasons, some
7425 other DWARF-related options such as
7426 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7427 in their names, but apply to all currently-supported versions of DWARF.
7431 Produce debugging information in stabs format (if that is supported),
7432 without GDB extensions. This is the format used by DBX on most BSD
7433 systems. On MIPS, Alpha and System V Release 4 systems this option
7434 produces stabs debugging output that is not understood by DBX@.
7435 On System V Release 4 systems this option requires the GNU assembler.
7439 Produce debugging information in stabs format (if that is supported),
7440 using GNU extensions understood only by the GNU debugger (GDB)@. The
7441 use of these extensions is likely to make other debuggers crash or
7442 refuse to read the program.
7446 Produce debugging information in XCOFF format (if that is supported).
7447 This is the format used by the DBX debugger on IBM RS/6000 systems.
7451 Produce debugging information in XCOFF format (if that is supported),
7452 using GNU extensions understood only by the GNU debugger (GDB)@. The
7453 use of these extensions is likely to make other debuggers crash or
7454 refuse to read the program, and may cause assemblers other than the GNU
7455 assembler (GAS) to fail with an error.
7459 Produce debugging information in Alpha/VMS debug format (if that is
7460 supported). This is the format used by DEBUG on Alpha/VMS systems.
7463 @itemx -ggdb@var{level}
7464 @itemx -gstabs@var{level}
7465 @itemx -gxcoff@var{level}
7466 @itemx -gvms@var{level}
7467 Request debugging information and also use @var{level} to specify how
7468 much information. The default level is 2.
7470 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7473 Level 1 produces minimal information, enough for making backtraces in
7474 parts of the program that you don't plan to debug. This includes
7475 descriptions of functions and external variables, and line number
7476 tables, but no information about local variables.
7478 Level 3 includes extra information, such as all the macro definitions
7479 present in the program. Some debuggers support macro expansion when
7480 you use @option{-g3}.
7482 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7483 confusion with @option{-gdwarf-@var{level}}.
7484 Instead use an additional @option{-g@var{level}} option to change the
7485 debug level for DWARF.
7487 @item -feliminate-unused-debug-symbols
7488 @opindex feliminate-unused-debug-symbols
7489 Produce debugging information in stabs format (if that is supported),
7490 for only symbols that are actually used.
7492 @item -femit-class-debug-always
7493 @opindex femit-class-debug-always
7494 Instead of emitting debugging information for a C++ class in only one
7495 object file, emit it in all object files using the class. This option
7496 should be used only with debuggers that are unable to handle the way GCC
7497 normally emits debugging information for classes because using this
7498 option increases the size of debugging information by as much as a
7501 @item -fno-merge-debug-strings
7502 @opindex fmerge-debug-strings
7503 @opindex fno-merge-debug-strings
7504 Direct the linker to not merge together strings in the debugging
7505 information that are identical in different object files. Merging is
7506 not supported by all assemblers or linkers. Merging decreases the size
7507 of the debug information in the output file at the cost of increasing
7508 link processing time. Merging is enabled by default.
7510 @item -fdebug-prefix-map=@var{old}=@var{new}
7511 @opindex fdebug-prefix-map
7512 When compiling files residing in directory @file{@var{old}}, record
7513 debugging information describing them as if the files resided in
7514 directory @file{@var{new}} instead. This can be used to replace a
7515 build-time path with an install-time path in the debug info. It can
7516 also be used to change an absolute path to a relative path by using
7517 @file{.} for @var{new}. This can give more reproducible builds, which
7518 are location independent, but may require an extra command to tell GDB
7519 where to find the source files. See also @option{-ffile-prefix-map}.
7521 @item -fvar-tracking
7522 @opindex fvar-tracking
7523 Run variable tracking pass. It computes where variables are stored at each
7524 position in code. Better debugging information is then generated
7525 (if the debugging information format supports this information).
7527 It is enabled by default when compiling with optimization (@option{-Os},
7528 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7529 the debug info format supports it.
7531 @item -fvar-tracking-assignments
7532 @opindex fvar-tracking-assignments
7533 @opindex fno-var-tracking-assignments
7534 Annotate assignments to user variables early in the compilation and
7535 attempt to carry the annotations over throughout the compilation all the
7536 way to the end, in an attempt to improve debug information while
7537 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7539 It can be enabled even if var-tracking is disabled, in which case
7540 annotations are created and maintained, but discarded at the end.
7541 By default, this flag is enabled together with @option{-fvar-tracking},
7542 except when selective scheduling is enabled.
7545 @opindex gsplit-dwarf
7546 Separate as much DWARF debugging information as possible into a
7547 separate output file with the extension @file{.dwo}. This option allows
7548 the build system to avoid linking files with debug information. To
7549 be useful, this option requires a debugger capable of reading @file{.dwo}
7552 @item -gdescribe-dies
7553 @opindex gdescribe-dies
7554 Add description attributes to some DWARF DIEs that have no name attribute,
7555 such as artificial variables, external references and call site
7560 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7562 @item -ggnu-pubnames
7563 @opindex ggnu-pubnames
7564 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7565 suitable for conversion into a GDB@ index. This option is only useful
7566 with a linker that can produce GDB@ index version 7.
7568 @item -fdebug-types-section
7569 @opindex fdebug-types-section
7570 @opindex fno-debug-types-section
7571 When using DWARF Version 4 or higher, type DIEs can be put into
7572 their own @code{.debug_types} section instead of making them part of the
7573 @code{.debug_info} section. It is more efficient to put them in a separate
7574 comdat section since the linker can then remove duplicates.
7575 But not all DWARF consumers support @code{.debug_types} sections yet
7576 and on some objects @code{.debug_types} produces larger instead of smaller
7577 debugging information.
7579 @item -grecord-gcc-switches
7580 @itemx -gno-record-gcc-switches
7581 @opindex grecord-gcc-switches
7582 @opindex gno-record-gcc-switches
7583 This switch causes the command-line options used to invoke the
7584 compiler that may affect code generation to be appended to the
7585 DW_AT_producer attribute in DWARF debugging information. The options
7586 are concatenated with spaces separating them from each other and from
7587 the compiler version.
7588 It is enabled by default.
7589 See also @option{-frecord-gcc-switches} for another
7590 way of storing compiler options into the object file.
7592 @item -gstrict-dwarf
7593 @opindex gstrict-dwarf
7594 Disallow using extensions of later DWARF standard version than selected
7595 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
7596 DWARF extensions from later standard versions is allowed.
7598 @item -gno-strict-dwarf
7599 @opindex gno-strict-dwarf
7600 Allow using extensions of later DWARF standard version than selected with
7601 @option{-gdwarf-@var{version}}.
7603 @item -gas-loc-support
7604 @opindex gas-loc-support
7605 Inform the compiler that the assembler supports @code{.loc} directives.
7606 It may then use them for the assembler to generate DWARF2+ line number
7609 This is generally desirable, because assembler-generated line-number
7610 tables are a lot more compact than those the compiler can generate
7613 This option will be enabled by default if, at GCC configure time, the
7614 assembler was found to support such directives.
7616 @item -gno-as-loc-support
7617 @opindex gno-as-loc-support
7618 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
7619 line number tables are to be generated.
7621 @item gas-locview-support
7622 @opindex gas-locview-support
7623 Inform the compiler that the assembler supports @code{view} assignment
7624 and reset assertion checking in @code{.loc} directives.
7626 This option will be enabled by default if, at GCC configure time, the
7627 assembler was found to support them.
7629 @item gno-as-locview-support
7630 Force GCC to assign view numbers internally, if
7631 @option{-gvariable-location-views} are explicitly requested.
7634 @itemx -gno-column-info
7635 @opindex gcolumn-info
7636 @opindex gno-column-info
7637 Emit location column information into DWARF debugging information, rather
7638 than just file and line.
7639 This option is enabled by default.
7641 @item -gstatement-frontiers
7642 @itemx -gno-statement-frontiers
7643 @opindex gstatement-frontiers
7644 @opindex gno-statement-frontiers
7645 This option causes GCC to create markers in the internal representation
7646 at the beginning of statements, and to keep them roughly in place
7647 throughout compilation, using them to guide the output of @code{is_stmt}
7648 markers in the line number table. This is enabled by default when
7649 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
7650 @dots{}), and outputting DWARF 2 debug information at the normal level.
7652 @item -gvariable-location-views
7653 @itemx -gvariable-location-views=incompat5
7654 @itemx -gno-variable-location-views
7655 @opindex gvariable-location-views
7656 @opindex gvariable-location-views=incompat5
7657 @opindex gno-variable-location-views
7658 Augment variable location lists with progressive view numbers implied
7659 from the line number table. This enables debug information consumers to
7660 inspect state at certain points of the program, even if no instructions
7661 associated with the corresponding source locations are present at that
7662 point. If the assembler lacks support for view numbers in line number
7663 tables, this will cause the compiler to emit the line number table,
7664 which generally makes them somewhat less compact. The augmented line
7665 number tables and location lists are fully backward-compatible, so they
7666 can be consumed by debug information consumers that are not aware of
7667 these augmentations, but they won't derive any benefit from them either.
7669 This is enabled by default when outputting DWARF 2 debug information at
7670 the normal level, as long as there is assembler support,
7671 @option{-fvar-tracking-assignments} is enabled and
7672 @option{-gstrict-dwarf} is not. When assembler support is not
7673 available, this may still be enabled, but it will force GCC to output
7674 internal line number tables, and if
7675 @option{-ginternal-reset-location-views} is not enabled, that will most
7676 certainly lead to silently mismatching location views.
7678 There is a proposed representation for view numbers that is not backward
7679 compatible with the location list format introduced in DWARF 5, that can
7680 be enabled with @option{-gvariable-location-views=incompat5}. This
7681 option may be removed in the future, is only provided as a reference
7682 implementation of the proposed representation. Debug information
7683 consumers are not expected to support this extended format, and they
7684 would be rendered unable to decode location lists using it.
7686 @item -ginternal-reset-location-views
7687 @itemx -gnointernal-reset-location-views
7688 @opindex ginternal-reset-location-views
7689 @opindex gno-internal-reset-location-views
7690 Attempt to determine location views that can be omitted from location
7691 view lists. This requires the compiler to have very accurate insn
7692 length estimates, which isn't always the case, and it may cause
7693 incorrect view lists to be generated silently when using an assembler
7694 that does not support location view lists. The GNU assembler will flag
7695 any such error as a @code{view number mismatch}. This is only enabled
7696 on ports that define a reliable estimation function.
7698 @item -ginline-points
7699 @itemx -gno-inline-points
7700 @opindex ginline-points
7701 @opindex gno-inline-points
7702 Generate extended debug information for inlined functions. Location
7703 view tracking markers are inserted at inlined entry points, so that
7704 address and view numbers can be computed and output in debug
7705 information. This can be enabled independently of location views, in
7706 which case the view numbers won't be output, but it can only be enabled
7707 along with statement frontiers, and it is only enabled by default if
7708 location views are enabled.
7710 @item -gz@r{[}=@var{type}@r{]}
7712 Produce compressed debug sections in DWARF format, if that is supported.
7713 If @var{type} is not given, the default type depends on the capabilities
7714 of the assembler and linker used. @var{type} may be one of
7715 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
7716 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
7717 compression in traditional GNU format). If the linker doesn't support
7718 writing compressed debug sections, the option is rejected. Otherwise,
7719 if the assembler does not support them, @option{-gz} is silently ignored
7720 when producing object files.
7722 @item -femit-struct-debug-baseonly
7723 @opindex femit-struct-debug-baseonly
7724 Emit debug information for struct-like types
7725 only when the base name of the compilation source file
7726 matches the base name of file in which the struct is defined.
7728 This option substantially reduces the size of debugging information,
7729 but at significant potential loss in type information to the debugger.
7730 See @option{-femit-struct-debug-reduced} for a less aggressive option.
7731 See @option{-femit-struct-debug-detailed} for more detailed control.
7733 This option works only with DWARF debug output.
7735 @item -femit-struct-debug-reduced
7736 @opindex femit-struct-debug-reduced
7737 Emit debug information for struct-like types
7738 only when the base name of the compilation source file
7739 matches the base name of file in which the type is defined,
7740 unless the struct is a template or defined in a system header.
7742 This option significantly reduces the size of debugging information,
7743 with some potential loss in type information to the debugger.
7744 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
7745 See @option{-femit-struct-debug-detailed} for more detailed control.
7747 This option works only with DWARF debug output.
7749 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
7750 @opindex femit-struct-debug-detailed
7751 Specify the struct-like types
7752 for which the compiler generates debug information.
7753 The intent is to reduce duplicate struct debug information
7754 between different object files within the same program.
7756 This option is a detailed version of
7757 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
7758 which serves for most needs.
7760 A specification has the syntax@*
7761 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
7763 The optional first word limits the specification to
7764 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
7765 A struct type is used directly when it is the type of a variable, member.
7766 Indirect uses arise through pointers to structs.
7767 That is, when use of an incomplete struct is valid, the use is indirect.
7769 @samp{struct one direct; struct two * indirect;}.
7771 The optional second word limits the specification to
7772 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
7773 Generic structs are a bit complicated to explain.
7774 For C++, these are non-explicit specializations of template classes,
7775 or non-template classes within the above.
7776 Other programming languages have generics,
7777 but @option{-femit-struct-debug-detailed} does not yet implement them.
7779 The third word specifies the source files for those
7780 structs for which the compiler should emit debug information.
7781 The values @samp{none} and @samp{any} have the normal meaning.
7782 The value @samp{base} means that
7783 the base of name of the file in which the type declaration appears
7784 must match the base of the name of the main compilation file.
7785 In practice, this means that when compiling @file{foo.c}, debug information
7786 is generated for types declared in that file and @file{foo.h},
7787 but not other header files.
7788 The value @samp{sys} means those types satisfying @samp{base}
7789 or declared in system or compiler headers.
7791 You may need to experiment to determine the best settings for your application.
7793 The default is @option{-femit-struct-debug-detailed=all}.
7795 This option works only with DWARF debug output.
7797 @item -fno-dwarf2-cfi-asm
7798 @opindex fdwarf2-cfi-asm
7799 @opindex fno-dwarf2-cfi-asm
7800 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
7801 instead of using GAS @code{.cfi_*} directives.
7803 @item -fno-eliminate-unused-debug-types
7804 @opindex feliminate-unused-debug-types
7805 @opindex fno-eliminate-unused-debug-types
7806 Normally, when producing DWARF output, GCC avoids producing debug symbol
7807 output for types that are nowhere used in the source file being compiled.
7808 Sometimes it is useful to have GCC emit debugging
7809 information for all types declared in a compilation
7810 unit, regardless of whether or not they are actually used
7811 in that compilation unit, for example
7812 if, in the debugger, you want to cast a value to a type that is
7813 not actually used in your program (but is declared). More often,
7814 however, this results in a significant amount of wasted space.
7817 @node Optimize Options
7818 @section Options That Control Optimization
7819 @cindex optimize options
7820 @cindex options, optimization
7822 These options control various sorts of optimizations.
7824 Without any optimization option, the compiler's goal is to reduce the
7825 cost of compilation and to make debugging produce the expected
7826 results. Statements are independent: if you stop the program with a
7827 breakpoint between statements, you can then assign a new value to any
7828 variable or change the program counter to any other statement in the
7829 function and get exactly the results you expect from the source
7832 Turning on optimization flags makes the compiler attempt to improve
7833 the performance and/or code size at the expense of compilation time
7834 and possibly the ability to debug the program.
7836 The compiler performs optimization based on the knowledge it has of the
7837 program. Compiling multiple files at once to a single output file mode allows
7838 the compiler to use information gained from all of the files when compiling
7841 Not all optimizations are controlled directly by a flag. Only
7842 optimizations that have a flag are listed in this section.
7844 Most optimizations are only enabled if an @option{-O} level is set on
7845 the command line. Otherwise they are disabled, even if individual
7846 optimization flags are specified.
7848 Depending on the target and how GCC was configured, a slightly different
7849 set of optimizations may be enabled at each @option{-O} level than
7850 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7851 to find out the exact set of optimizations that are enabled at each level.
7852 @xref{Overall Options}, for examples.
7859 Optimize. Optimizing compilation takes somewhat more time, and a lot
7860 more memory for a large function.
7862 With @option{-O}, the compiler tries to reduce code size and execution
7863 time, without performing any optimizations that take a great deal of
7866 @option{-O} turns on the following optimization flags:
7869 -fbranch-count-reg @gol
7870 -fcombine-stack-adjustments @gol
7872 -fcprop-registers @gol
7875 -fdelayed-branch @gol
7877 -fforward-propagate @gol
7878 -fguess-branch-probability @gol
7879 -fif-conversion2 @gol
7880 -fif-conversion @gol
7881 -finline-functions-called-once @gol
7882 -fipa-pure-const @gol
7884 -fipa-reference @gol
7885 -fipa-reference-addressable @gol
7886 -fmerge-constants @gol
7887 -fmove-loop-invariants @gol
7888 -fomit-frame-pointer @gol
7889 -freorder-blocks @gol
7891 -fshrink-wrap-separate @gol
7892 -fsplit-wide-types @gol
7898 -ftree-coalesce-vars @gol
7899 -ftree-copy-prop @gol
7901 -ftree-dominator-opts @gol
7903 -ftree-forwprop @gol
7906 -ftree-scev-cprop @gol
7916 Optimize even more. GCC performs nearly all supported optimizations
7917 that do not involve a space-speed tradeoff.
7918 As compared to @option{-O}, this option increases both compilation time
7919 and the performance of the generated code.
7921 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7922 also turns on the following optimization flags:
7923 @gccoptlist{-fthread-jumps @gol
7924 -falign-functions -falign-jumps @gol
7925 -falign-loops -falign-labels @gol
7928 -fcse-follow-jumps -fcse-skip-blocks @gol
7929 -fdelete-null-pointer-checks @gol
7930 -fdevirtualize -fdevirtualize-speculatively @gol
7931 -fexpensive-optimizations @gol
7932 -fgcse -fgcse-lm @gol
7933 -fhoist-adjacent-loads @gol
7934 -finline-small-functions @gol
7935 -findirect-inlining @gol
7941 -fisolate-erroneous-paths-dereference @gol
7943 -foptimize-sibling-calls @gol
7944 -foptimize-strlen @gol
7945 -fpartial-inlining @gol
7947 -freorder-blocks-algorithm=stc @gol
7948 -freorder-blocks-and-partition -freorder-functions @gol
7949 -frerun-cse-after-loop @gol
7950 -fsched-interblock -fsched-spec @gol
7951 -fschedule-insns -fschedule-insns2 @gol
7952 -fstore-merging @gol
7953 -fstrict-aliasing @gol
7954 -ftree-builtin-call-dce @gol
7955 -ftree-switch-conversion -ftree-tail-merge @gol
7956 -fcode-hoisting @gol
7961 Please note the warning under @option{-fgcse} about
7962 invoking @option{-O2} on programs that use computed gotos.
7966 Optimize yet more. @option{-O3} turns on all optimizations specified
7967 by @option{-O2} and also turns on the following optimization flags:
7968 @gccoptlist{-finline-functions @gol
7969 -funswitch-loops @gol
7970 -fpredictive-commoning @gol
7971 -fgcse-after-reload @gol
7972 -ftree-loop-vectorize @gol
7973 -ftree-loop-distribution @gol
7974 -ftree-loop-distribute-patterns @gol
7975 -floop-interchange @gol
7976 -floop-unroll-and-jam @gol
7978 -ftree-slp-vectorize @gol
7979 -fvect-cost-model @gol
7980 -ftree-partial-pre @gol
7986 Reduce compilation time and make debugging produce the expected
7987 results. This is the default.
7991 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7992 do not typically increase code size. It also performs further
7993 optimizations designed to reduce code size.
7995 @option{-Os} disables the following optimization flags:
7996 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7997 -falign-labels -fprefetch-loop-arrays}
8001 Disregard strict standards compliance. @option{-Ofast} enables all
8002 @option{-O3} optimizations. It also enables optimizations that are not
8003 valid for all standard-compliant programs.
8004 It turns on @option{-ffast-math} and the Fortran-specific
8005 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
8006 specified, and @option{-fno-protect-parens}.
8010 Optimize debugging experience. @option{-Og} enables optimizations
8011 that do not interfere with debugging. It should be the optimization
8012 level of choice for the standard edit-compile-debug cycle, offering
8013 a reasonable level of optimization while maintaining fast compilation
8014 and a good debugging experience.
8017 If you use multiple @option{-O} options, with or without level numbers,
8018 the last such option is the one that is effective.
8020 Options of the form @option{-f@var{flag}} specify machine-independent
8021 flags. Most flags have both positive and negative forms; the negative
8022 form of @option{-ffoo} is @option{-fno-foo}. In the table
8023 below, only one of the forms is listed---the one you typically
8024 use. You can figure out the other form by either removing @samp{no-}
8027 The following options control specific optimizations. They are either
8028 activated by @option{-O} options or are related to ones that are. You
8029 can use the following flags in the rare cases when ``fine-tuning'' of
8030 optimizations to be performed is desired.
8033 @item -fno-defer-pop
8034 @opindex fno-defer-pop
8035 Always pop the arguments to each function call as soon as that function
8036 returns. For machines that must pop arguments after a function call,
8037 the compiler normally lets arguments accumulate on the stack for several
8038 function calls and pops them all at once.
8040 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8042 @item -fforward-propagate
8043 @opindex fforward-propagate
8044 Perform a forward propagation pass on RTL@. The pass tries to combine two
8045 instructions and checks if the result can be simplified. If loop unrolling
8046 is active, two passes are performed and the second is scheduled after
8049 This option is enabled by default at optimization levels @option{-O},
8050 @option{-O2}, @option{-O3}, @option{-Os}.
8052 @item -ffp-contract=@var{style}
8053 @opindex ffp-contract
8054 @option{-ffp-contract=off} disables floating-point expression contraction.
8055 @option{-ffp-contract=fast} enables floating-point expression contraction
8056 such as forming of fused multiply-add operations if the target has
8057 native support for them.
8058 @option{-ffp-contract=on} enables floating-point expression contraction
8059 if allowed by the language standard. This is currently not implemented
8060 and treated equal to @option{-ffp-contract=off}.
8062 The default is @option{-ffp-contract=fast}.
8064 @item -fomit-frame-pointer
8065 @opindex fomit-frame-pointer
8066 Omit the frame pointer in functions that don't need one. This avoids the
8067 instructions to save, set up and restore the frame pointer; on many targets
8068 it also makes an extra register available.
8070 On some targets this flag has no effect because the standard calling sequence
8071 always uses a frame pointer, so it cannot be omitted.
8073 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
8074 is used in all functions. Several targets always omit the frame pointer in
8077 Enabled by default at @option{-O} and higher.
8079 @item -foptimize-sibling-calls
8080 @opindex foptimize-sibling-calls
8081 Optimize sibling and tail recursive calls.
8083 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8085 @item -foptimize-strlen
8086 @opindex foptimize-strlen
8087 Optimize various standard C string functions (e.g.@: @code{strlen},
8088 @code{strchr} or @code{strcpy}) and
8089 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8091 Enabled at levels @option{-O2}, @option{-O3}.
8095 Do not expand any functions inline apart from those marked with
8096 the @code{always_inline} attribute. This is the default when not
8099 Single functions can be exempted from inlining by marking them
8100 with the @code{noinline} attribute.
8102 @item -finline-small-functions
8103 @opindex finline-small-functions
8104 Integrate functions into their callers when their body is smaller than expected
8105 function call code (so overall size of program gets smaller). The compiler
8106 heuristically decides which functions are simple enough to be worth integrating
8107 in this way. This inlining applies to all functions, even those not declared
8110 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8112 @item -findirect-inlining
8113 @opindex findirect-inlining
8114 Inline also indirect calls that are discovered to be known at compile
8115 time thanks to previous inlining. This option has any effect only
8116 when inlining itself is turned on by the @option{-finline-functions}
8117 or @option{-finline-small-functions} options.
8119 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8121 @item -finline-functions
8122 @opindex finline-functions
8123 Consider all functions for inlining, even if they are not declared inline.
8124 The compiler heuristically decides which functions are worth integrating
8127 If all calls to a given function are integrated, and the function is
8128 declared @code{static}, then the function is normally not output as
8129 assembler code in its own right.
8131 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8133 @item -finline-functions-called-once
8134 @opindex finline-functions-called-once
8135 Consider all @code{static} functions called once for inlining into their
8136 caller even if they are not marked @code{inline}. If a call to a given
8137 function is integrated, then the function is not output as assembler code
8140 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
8142 @item -fearly-inlining
8143 @opindex fearly-inlining
8144 Inline functions marked by @code{always_inline} and functions whose body seems
8145 smaller than the function call overhead early before doing
8146 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8147 makes profiling significantly cheaper and usually inlining faster on programs
8148 having large chains of nested wrapper functions.
8154 Perform interprocedural scalar replacement of aggregates, removal of
8155 unused parameters and replacement of parameters passed by reference
8156 by parameters passed by value.
8158 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8160 @item -finline-limit=@var{n}
8161 @opindex finline-limit
8162 By default, GCC limits the size of functions that can be inlined. This flag
8163 allows coarse control of this limit. @var{n} is the size of functions that
8164 can be inlined in number of pseudo instructions.
8166 Inlining is actually controlled by a number of parameters, which may be
8167 specified individually by using @option{--param @var{name}=@var{value}}.
8168 The @option{-finline-limit=@var{n}} option sets some of these parameters
8172 @item max-inline-insns-single
8173 is set to @var{n}/2.
8174 @item max-inline-insns-auto
8175 is set to @var{n}/2.
8178 See below for a documentation of the individual
8179 parameters controlling inlining and for the defaults of these parameters.
8181 @emph{Note:} there may be no value to @option{-finline-limit} that results
8182 in default behavior.
8184 @emph{Note:} pseudo instruction represents, in this particular context, an
8185 abstract measurement of function's size. In no way does it represent a count
8186 of assembly instructions and as such its exact meaning might change from one
8187 release to an another.
8189 @item -fno-keep-inline-dllexport
8190 @opindex fno-keep-inline-dllexport
8191 This is a more fine-grained version of @option{-fkeep-inline-functions},
8192 which applies only to functions that are declared using the @code{dllexport}
8193 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
8196 @item -fkeep-inline-functions
8197 @opindex fkeep-inline-functions
8198 In C, emit @code{static} functions that are declared @code{inline}
8199 into the object file, even if the function has been inlined into all
8200 of its callers. This switch does not affect functions using the
8201 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8202 inline functions into the object file.
8204 @item -fkeep-static-functions
8205 @opindex fkeep-static-functions
8206 Emit @code{static} functions into the object file, even if the function
8209 @item -fkeep-static-consts
8210 @opindex fkeep-static-consts
8211 Emit variables declared @code{static const} when optimization isn't turned
8212 on, even if the variables aren't referenced.
8214 GCC enables this option by default. If you want to force the compiler to
8215 check if a variable is referenced, regardless of whether or not
8216 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8218 @item -fmerge-constants
8219 @opindex fmerge-constants
8220 Attempt to merge identical constants (string constants and floating-point
8221 constants) across compilation units.
8223 This option is the default for optimized compilation if the assembler and
8224 linker support it. Use @option{-fno-merge-constants} to inhibit this
8227 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8229 @item -fmerge-all-constants
8230 @opindex fmerge-all-constants
8231 Attempt to merge identical constants and identical variables.
8233 This option implies @option{-fmerge-constants}. In addition to
8234 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8235 arrays or initialized constant variables with integral or floating-point
8236 types. Languages like C or C++ require each variable, including multiple
8237 instances of the same variable in recursive calls, to have distinct locations,
8238 so using this option results in non-conforming
8241 @item -fmodulo-sched
8242 @opindex fmodulo-sched
8243 Perform swing modulo scheduling immediately before the first scheduling
8244 pass. This pass looks at innermost loops and reorders their
8245 instructions by overlapping different iterations.
8247 @item -fmodulo-sched-allow-regmoves
8248 @opindex fmodulo-sched-allow-regmoves
8249 Perform more aggressive SMS-based modulo scheduling with register moves
8250 allowed. By setting this flag certain anti-dependences edges are
8251 deleted, which triggers the generation of reg-moves based on the
8252 life-range analysis. This option is effective only with
8253 @option{-fmodulo-sched} enabled.
8255 @item -fno-branch-count-reg
8256 @opindex fno-branch-count-reg
8257 Avoid running a pass scanning for opportunities to use ``decrement and
8258 branch'' instructions on a count register instead of generating sequences
8259 of instructions that decrement a register, compare it against zero, and
8260 then branch based upon the result. This option is only meaningful on
8261 architectures that support such instructions, which include x86, PowerPC,
8262 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
8263 doesn't remove the decrement and branch instructions from the generated
8264 instruction stream introduced by other optimization passes.
8266 Enabled by default at @option{-O1} and higher.
8268 The default is @option{-fbranch-count-reg}.
8270 @item -fno-function-cse
8271 @opindex fno-function-cse
8272 Do not put function addresses in registers; make each instruction that
8273 calls a constant function contain the function's address explicitly.
8275 This option results in less efficient code, but some strange hacks
8276 that alter the assembler output may be confused by the optimizations
8277 performed when this option is not used.
8279 The default is @option{-ffunction-cse}
8281 @item -fno-zero-initialized-in-bss
8282 @opindex fno-zero-initialized-in-bss
8283 If the target supports a BSS section, GCC by default puts variables that
8284 are initialized to zero into BSS@. This can save space in the resulting
8287 This option turns off this behavior because some programs explicitly
8288 rely on variables going to the data section---e.g., so that the
8289 resulting executable can find the beginning of that section and/or make
8290 assumptions based on that.
8292 The default is @option{-fzero-initialized-in-bss}.
8294 @item -fthread-jumps
8295 @opindex fthread-jumps
8296 Perform optimizations that check to see if a jump branches to a
8297 location where another comparison subsumed by the first is found. If
8298 so, the first branch is redirected to either the destination of the
8299 second branch or a point immediately following it, depending on whether
8300 the condition is known to be true or false.
8302 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8304 @item -fsplit-wide-types
8305 @opindex fsplit-wide-types
8306 When using a type that occupies multiple registers, such as @code{long
8307 long} on a 32-bit system, split the registers apart and allocate them
8308 independently. This normally generates better code for those types,
8309 but may make debugging more difficult.
8311 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8314 @item -fcse-follow-jumps
8315 @opindex fcse-follow-jumps
8316 In common subexpression elimination (CSE), scan through jump instructions
8317 when the target of the jump is not reached by any other path. For
8318 example, when CSE encounters an @code{if} statement with an
8319 @code{else} clause, CSE follows the jump when the condition
8322 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8324 @item -fcse-skip-blocks
8325 @opindex fcse-skip-blocks
8326 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8327 follow jumps that conditionally skip over blocks. When CSE
8328 encounters a simple @code{if} statement with no else clause,
8329 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8330 body of the @code{if}.
8332 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8334 @item -frerun-cse-after-loop
8335 @opindex frerun-cse-after-loop
8336 Re-run common subexpression elimination after loop optimizations are
8339 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8343 Perform a global common subexpression elimination pass.
8344 This pass also performs global constant and copy propagation.
8346 @emph{Note:} When compiling a program using computed gotos, a GCC
8347 extension, you may get better run-time performance if you disable
8348 the global common subexpression elimination pass by adding
8349 @option{-fno-gcse} to the command line.
8351 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8355 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8356 attempts to move loads that are only killed by stores into themselves. This
8357 allows a loop containing a load/store sequence to be changed to a load outside
8358 the loop, and a copy/store within the loop.
8360 Enabled by default when @option{-fgcse} is enabled.
8364 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8365 global common subexpression elimination. This pass attempts to move
8366 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8367 loops containing a load/store sequence can be changed to a load before
8368 the loop and a store after the loop.
8370 Not enabled at any optimization level.
8374 When @option{-fgcse-las} is enabled, the global common subexpression
8375 elimination pass eliminates redundant loads that come after stores to the
8376 same memory location (both partial and full redundancies).
8378 Not enabled at any optimization level.
8380 @item -fgcse-after-reload
8381 @opindex fgcse-after-reload
8382 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8383 pass is performed after reload. The purpose of this pass is to clean up
8386 @item -faggressive-loop-optimizations
8387 @opindex faggressive-loop-optimizations
8388 This option tells the loop optimizer to use language constraints to
8389 derive bounds for the number of iterations of a loop. This assumes that
8390 loop code does not invoke undefined behavior by for example causing signed
8391 integer overflows or out-of-bound array accesses. The bounds for the
8392 number of iterations of a loop are used to guide loop unrolling and peeling
8393 and loop exit test optimizations.
8394 This option is enabled by default.
8396 @item -funconstrained-commons
8397 @opindex funconstrained-commons
8398 This option tells the compiler that variables declared in common blocks
8399 (e.g.@: Fortran) may later be overridden with longer trailing arrays. This
8400 prevents certain optimizations that depend on knowing the array bounds.
8402 @item -fcrossjumping
8403 @opindex fcrossjumping
8404 Perform cross-jumping transformation.
8405 This transformation unifies equivalent code and saves code size. The
8406 resulting code may or may not perform better than without cross-jumping.
8408 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8410 @item -fauto-inc-dec
8411 @opindex fauto-inc-dec
8412 Combine increments or decrements of addresses with memory accesses.
8413 This pass is always skipped on architectures that do not have
8414 instructions to support this. Enabled by default at @option{-O} and
8415 higher on architectures that support this.
8419 Perform dead code elimination (DCE) on RTL@.
8420 Enabled by default at @option{-O} and higher.
8424 Perform dead store elimination (DSE) on RTL@.
8425 Enabled by default at @option{-O} and higher.
8427 @item -fif-conversion
8428 @opindex fif-conversion
8429 Attempt to transform conditional jumps into branch-less equivalents. This
8430 includes use of conditional moves, min, max, set flags and abs instructions, and
8431 some tricks doable by standard arithmetics. The use of conditional execution
8432 on chips where it is available is controlled by @option{-fif-conversion2}.
8434 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8436 @item -fif-conversion2
8437 @opindex fif-conversion2
8438 Use conditional execution (where available) to transform conditional jumps into
8439 branch-less equivalents.
8441 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8443 @item -fdeclone-ctor-dtor
8444 @opindex fdeclone-ctor-dtor
8445 The C++ ABI requires multiple entry points for constructors and
8446 destructors: one for a base subobject, one for a complete object, and
8447 one for a virtual destructor that calls operator delete afterwards.
8448 For a hierarchy with virtual bases, the base and complete variants are
8449 clones, which means two copies of the function. With this option, the
8450 base and complete variants are changed to be thunks that call a common
8453 Enabled by @option{-Os}.
8455 @item -fdelete-null-pointer-checks
8456 @opindex fdelete-null-pointer-checks
8457 Assume that programs cannot safely dereference null pointers, and that
8458 no code or data element resides at address zero.
8459 This option enables simple constant
8460 folding optimizations at all optimization levels. In addition, other
8461 optimization passes in GCC use this flag to control global dataflow
8462 analyses that eliminate useless checks for null pointers; these assume
8463 that a memory access to address zero always results in a trap, so
8464 that if a pointer is checked after it has already been dereferenced,
8467 Note however that in some environments this assumption is not true.
8468 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8469 for programs that depend on that behavior.
8471 This option is enabled by default on most targets. On Nios II ELF, it
8472 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8474 Passes that use the dataflow information
8475 are enabled independently at different optimization levels.
8477 @item -fdevirtualize
8478 @opindex fdevirtualize
8479 Attempt to convert calls to virtual functions to direct calls. This
8480 is done both within a procedure and interprocedurally as part of
8481 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8482 propagation (@option{-fipa-cp}).
8483 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8485 @item -fdevirtualize-speculatively
8486 @opindex fdevirtualize-speculatively
8487 Attempt to convert calls to virtual functions to speculative direct calls.
8488 Based on the analysis of the type inheritance graph, determine for a given call
8489 the set of likely targets. If the set is small, preferably of size 1, change
8490 the call into a conditional deciding between direct and indirect calls. The
8491 speculative calls enable more optimizations, such as inlining. When they seem
8492 useless after further optimization, they are converted back into original form.
8494 @item -fdevirtualize-at-ltrans
8495 @opindex fdevirtualize-at-ltrans
8496 Stream extra information needed for aggressive devirtualization when running
8497 the link-time optimizer in local transformation mode.
8498 This option enables more devirtualization but
8499 significantly increases the size of streamed data. For this reason it is
8500 disabled by default.
8502 @item -fexpensive-optimizations
8503 @opindex fexpensive-optimizations
8504 Perform a number of minor optimizations that are relatively expensive.
8506 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8510 Attempt to remove redundant extension instructions. This is especially
8511 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8512 registers after writing to their lower 32-bit half.
8514 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8515 @option{-O3}, @option{-Os}.
8517 @item -fno-lifetime-dse
8518 @opindex fno-lifetime-dse
8519 In C++ the value of an object is only affected by changes within its
8520 lifetime: when the constructor begins, the object has an indeterminate
8521 value, and any changes during the lifetime of the object are dead when
8522 the object is destroyed. Normally dead store elimination will take
8523 advantage of this; if your code relies on the value of the object
8524 storage persisting beyond the lifetime of the object, you can use this
8525 flag to disable this optimization. To preserve stores before the
8526 constructor starts (e.g.@: because your operator new clears the object
8527 storage) but still treat the object as dead after the destructor you,
8528 can use @option{-flifetime-dse=1}. The default behavior can be
8529 explicitly selected with @option{-flifetime-dse=2}.
8530 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8532 @item -flive-range-shrinkage
8533 @opindex flive-range-shrinkage
8534 Attempt to decrease register pressure through register live range
8535 shrinkage. This is helpful for fast processors with small or moderate
8538 @item -fira-algorithm=@var{algorithm}
8539 @opindex fira-algorithm
8540 Use the specified coloring algorithm for the integrated register
8541 allocator. The @var{algorithm} argument can be @samp{priority}, which
8542 specifies Chow's priority coloring, or @samp{CB}, which specifies
8543 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8544 for all architectures, but for those targets that do support it, it is
8545 the default because it generates better code.
8547 @item -fira-region=@var{region}
8548 @opindex fira-region
8549 Use specified regions for the integrated register allocator. The
8550 @var{region} argument should be one of the following:
8555 Use all loops as register allocation regions.
8556 This can give the best results for machines with a small and/or
8557 irregular register set.
8560 Use all loops except for loops with small register pressure
8561 as the regions. This value usually gives
8562 the best results in most cases and for most architectures,
8563 and is enabled by default when compiling with optimization for speed
8564 (@option{-O}, @option{-O2}, @dots{}).
8567 Use all functions as a single region.
8568 This typically results in the smallest code size, and is enabled by default for
8569 @option{-Os} or @option{-O0}.
8573 @item -fira-hoist-pressure
8574 @opindex fira-hoist-pressure
8575 Use IRA to evaluate register pressure in the code hoisting pass for
8576 decisions to hoist expressions. This option usually results in smaller
8577 code, but it can slow the compiler down.
8579 This option is enabled at level @option{-Os} for all targets.
8581 @item -fira-loop-pressure
8582 @opindex fira-loop-pressure
8583 Use IRA to evaluate register pressure in loops for decisions to move
8584 loop invariants. This option usually results in generation
8585 of faster and smaller code on machines with large register files (>= 32
8586 registers), but it can slow the compiler down.
8588 This option is enabled at level @option{-O3} for some targets.
8590 @item -fno-ira-share-save-slots
8591 @opindex fno-ira-share-save-slots
8592 Disable sharing of stack slots used for saving call-used hard
8593 registers living through a call. Each hard register gets a
8594 separate stack slot, and as a result function stack frames are
8597 @item -fno-ira-share-spill-slots
8598 @opindex fno-ira-share-spill-slots
8599 Disable sharing of stack slots allocated for pseudo-registers. Each
8600 pseudo-register that does not get a hard register gets a separate
8601 stack slot, and as a result function stack frames are larger.
8605 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8606 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8607 values if it is profitable.
8609 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8611 @item -fdelayed-branch
8612 @opindex fdelayed-branch
8613 If supported for the target machine, attempt to reorder instructions
8614 to exploit instruction slots available after delayed branch
8617 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os},
8618 but not at @option{-Og}.
8620 @item -fschedule-insns
8621 @opindex fschedule-insns
8622 If supported for the target machine, attempt to reorder instructions to
8623 eliminate execution stalls due to required data being unavailable. This
8624 helps machines that have slow floating point or memory load instructions
8625 by allowing other instructions to be issued until the result of the load
8626 or floating-point instruction is required.
8628 Enabled at levels @option{-O2}, @option{-O3}.
8630 @item -fschedule-insns2
8631 @opindex fschedule-insns2
8632 Similar to @option{-fschedule-insns}, but requests an additional pass of
8633 instruction scheduling after register allocation has been done. This is
8634 especially useful on machines with a relatively small number of
8635 registers and where memory load instructions take more than one cycle.
8637 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8639 @item -fno-sched-interblock
8640 @opindex fno-sched-interblock
8641 Disable instruction scheduling across basic blocks, which
8642 is normally enabled when scheduling before register allocation, i.e.@:
8643 with @option{-fschedule-insns} or at @option{-O2} or higher.
8645 @item -fno-sched-spec
8646 @opindex fno-sched-spec
8647 Disable speculative motion of non-load instructions, which
8648 is normally enabled when scheduling before register allocation, i.e.@:
8649 with @option{-fschedule-insns} or at @option{-O2} or higher.
8651 @item -fsched-pressure
8652 @opindex fsched-pressure
8653 Enable register pressure sensitive insn scheduling before register
8654 allocation. This only makes sense when scheduling before register
8655 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8656 @option{-O2} or higher. Usage of this option can improve the
8657 generated code and decrease its size by preventing register pressure
8658 increase above the number of available hard registers and subsequent
8659 spills in register allocation.
8661 @item -fsched-spec-load
8662 @opindex fsched-spec-load
8663 Allow speculative motion of some load instructions. This only makes
8664 sense when scheduling before register allocation, i.e.@: with
8665 @option{-fschedule-insns} or at @option{-O2} or higher.
8667 @item -fsched-spec-load-dangerous
8668 @opindex fsched-spec-load-dangerous
8669 Allow speculative motion of more load instructions. This only makes
8670 sense when scheduling before register allocation, i.e.@: with
8671 @option{-fschedule-insns} or at @option{-O2} or higher.
8673 @item -fsched-stalled-insns
8674 @itemx -fsched-stalled-insns=@var{n}
8675 @opindex fsched-stalled-insns
8676 Define how many insns (if any) can be moved prematurely from the queue
8677 of stalled insns into the ready list during the second scheduling pass.
8678 @option{-fno-sched-stalled-insns} means that no insns are moved
8679 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8680 on how many queued insns can be moved prematurely.
8681 @option{-fsched-stalled-insns} without a value is equivalent to
8682 @option{-fsched-stalled-insns=1}.
8684 @item -fsched-stalled-insns-dep
8685 @itemx -fsched-stalled-insns-dep=@var{n}
8686 @opindex fsched-stalled-insns-dep
8687 Define how many insn groups (cycles) are examined for a dependency
8688 on a stalled insn that is a candidate for premature removal from the queue
8689 of stalled insns. This has an effect only during the second scheduling pass,
8690 and only if @option{-fsched-stalled-insns} is used.
8691 @option{-fno-sched-stalled-insns-dep} is equivalent to
8692 @option{-fsched-stalled-insns-dep=0}.
8693 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8694 @option{-fsched-stalled-insns-dep=1}.
8696 @item -fsched2-use-superblocks
8697 @opindex fsched2-use-superblocks
8698 When scheduling after register allocation, use superblock scheduling.
8699 This allows motion across basic block boundaries,
8700 resulting in faster schedules. This option is experimental, as not all machine
8701 descriptions used by GCC model the CPU closely enough to avoid unreliable
8702 results from the algorithm.
8704 This only makes sense when scheduling after register allocation, i.e.@: with
8705 @option{-fschedule-insns2} or at @option{-O2} or higher.
8707 @item -fsched-group-heuristic
8708 @opindex fsched-group-heuristic
8709 Enable the group heuristic in the scheduler. This heuristic favors
8710 the instruction that belongs to a schedule group. This is enabled
8711 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8712 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8714 @item -fsched-critical-path-heuristic
8715 @opindex fsched-critical-path-heuristic
8716 Enable the critical-path heuristic in the scheduler. This heuristic favors
8717 instructions on the critical path. This is enabled by default when
8718 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8719 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8721 @item -fsched-spec-insn-heuristic
8722 @opindex fsched-spec-insn-heuristic
8723 Enable the speculative instruction heuristic in the scheduler. This
8724 heuristic favors speculative instructions with greater dependency weakness.
8725 This is enabled by default when scheduling is enabled, i.e.@:
8726 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8727 or at @option{-O2} or higher.
8729 @item -fsched-rank-heuristic
8730 @opindex fsched-rank-heuristic
8731 Enable the rank heuristic in the scheduler. This heuristic favors
8732 the instruction belonging to a basic block with greater size or frequency.
8733 This is enabled by default when scheduling is enabled, i.e.@:
8734 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8735 at @option{-O2} or higher.
8737 @item -fsched-last-insn-heuristic
8738 @opindex fsched-last-insn-heuristic
8739 Enable the last-instruction heuristic in the scheduler. This heuristic
8740 favors the instruction that is less dependent on the last instruction
8741 scheduled. This is enabled by default when scheduling is enabled,
8742 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8743 at @option{-O2} or higher.
8745 @item -fsched-dep-count-heuristic
8746 @opindex fsched-dep-count-heuristic
8747 Enable the dependent-count heuristic in the scheduler. This heuristic
8748 favors the instruction that has more instructions depending on it.
8749 This is enabled by default when scheduling is enabled, i.e.@:
8750 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8751 at @option{-O2} or higher.
8753 @item -freschedule-modulo-scheduled-loops
8754 @opindex freschedule-modulo-scheduled-loops
8755 Modulo scheduling is performed before traditional scheduling. If a loop
8756 is modulo scheduled, later scheduling passes may change its schedule.
8757 Use this option to control that behavior.
8759 @item -fselective-scheduling
8760 @opindex fselective-scheduling
8761 Schedule instructions using selective scheduling algorithm. Selective
8762 scheduling runs instead of the first scheduler pass.
8764 @item -fselective-scheduling2
8765 @opindex fselective-scheduling2
8766 Schedule instructions using selective scheduling algorithm. Selective
8767 scheduling runs instead of the second scheduler pass.
8769 @item -fsel-sched-pipelining
8770 @opindex fsel-sched-pipelining
8771 Enable software pipelining of innermost loops during selective scheduling.
8772 This option has no effect unless one of @option{-fselective-scheduling} or
8773 @option{-fselective-scheduling2} is turned on.
8775 @item -fsel-sched-pipelining-outer-loops
8776 @opindex fsel-sched-pipelining-outer-loops
8777 When pipelining loops during selective scheduling, also pipeline outer loops.
8778 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8780 @item -fsemantic-interposition
8781 @opindex fsemantic-interposition
8782 Some object formats, like ELF, allow interposing of symbols by the
8784 This means that for symbols exported from the DSO, the compiler cannot perform
8785 interprocedural propagation, inlining and other optimizations in anticipation
8786 that the function or variable in question may change. While this feature is
8787 useful, for example, to rewrite memory allocation functions by a debugging
8788 implementation, it is expensive in the terms of code quality.
8789 With @option{-fno-semantic-interposition} the compiler assumes that
8790 if interposition happens for functions the overwriting function will have
8791 precisely the same semantics (and side effects).
8792 Similarly if interposition happens
8793 for variables, the constructor of the variable will be the same. The flag
8794 has no effect for functions explicitly declared inline
8795 (where it is never allowed for interposition to change semantics)
8796 and for symbols explicitly declared weak.
8799 @opindex fshrink-wrap
8800 Emit function prologues only before parts of the function that need it,
8801 rather than at the top of the function. This flag is enabled by default at
8802 @option{-O} and higher.
8804 @item -fshrink-wrap-separate
8805 @opindex fshrink-wrap-separate
8806 Shrink-wrap separate parts of the prologue and epilogue separately, so that
8807 those parts are only executed when needed.
8808 This option is on by default, but has no effect unless @option{-fshrink-wrap}
8809 is also turned on and the target supports this.
8811 @item -fcaller-saves
8812 @opindex fcaller-saves
8813 Enable allocation of values to registers that are clobbered by
8814 function calls, by emitting extra instructions to save and restore the
8815 registers around such calls. Such allocation is done only when it
8816 seems to result in better code.
8818 This option is always enabled by default on certain machines, usually
8819 those which have no call-preserved registers to use instead.
8821 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8823 @item -fcombine-stack-adjustments
8824 @opindex fcombine-stack-adjustments
8825 Tracks stack adjustments (pushes and pops) and stack memory references
8826 and then tries to find ways to combine them.
8828 Enabled by default at @option{-O1} and higher.
8832 Use caller save registers for allocation if those registers are not used by
8833 any called function. In that case it is not necessary to save and restore
8834 them around calls. This is only possible if called functions are part of
8835 same compilation unit as current function and they are compiled before it.
8837 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
8838 is disabled if generated code will be instrumented for profiling
8839 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
8840 exactly (this happens on targets that do not expose prologues
8841 and epilogues in RTL).
8843 @item -fconserve-stack
8844 @opindex fconserve-stack
8845 Attempt to minimize stack usage. The compiler attempts to use less
8846 stack space, even if that makes the program slower. This option
8847 implies setting the @option{large-stack-frame} parameter to 100
8848 and the @option{large-stack-frame-growth} parameter to 400.
8850 @item -ftree-reassoc
8851 @opindex ftree-reassoc
8852 Perform reassociation on trees. This flag is enabled by default
8853 at @option{-O} and higher.
8855 @item -fcode-hoisting
8856 @opindex fcode-hoisting
8857 Perform code hoisting. Code hoisting tries to move the
8858 evaluation of expressions executed on all paths to the function exit
8859 as early as possible. This is especially useful as a code size
8860 optimization, but it often helps for code speed as well.
8861 This flag is enabled by default at @option{-O2} and higher.
8865 Perform partial redundancy elimination (PRE) on trees. This flag is
8866 enabled by default at @option{-O2} and @option{-O3}.
8868 @item -ftree-partial-pre
8869 @opindex ftree-partial-pre
8870 Make partial redundancy elimination (PRE) more aggressive. This flag is
8871 enabled by default at @option{-O3}.
8873 @item -ftree-forwprop
8874 @opindex ftree-forwprop
8875 Perform forward propagation on trees. This flag is enabled by default
8876 at @option{-O} and higher.
8880 Perform full redundancy elimination (FRE) on trees. The difference
8881 between FRE and PRE is that FRE only considers expressions
8882 that are computed on all paths leading to the redundant computation.
8883 This analysis is faster than PRE, though it exposes fewer redundancies.
8884 This flag is enabled by default at @option{-O} and higher.
8886 @item -ftree-phiprop
8887 @opindex ftree-phiprop
8888 Perform hoisting of loads from conditional pointers on trees. This
8889 pass is enabled by default at @option{-O} and higher.
8891 @item -fhoist-adjacent-loads
8892 @opindex fhoist-adjacent-loads
8893 Speculatively hoist loads from both branches of an if-then-else if the
8894 loads are from adjacent locations in the same structure and the target
8895 architecture has a conditional move instruction. This flag is enabled
8896 by default at @option{-O2} and higher.
8898 @item -ftree-copy-prop
8899 @opindex ftree-copy-prop
8900 Perform copy propagation on trees. This pass eliminates unnecessary
8901 copy operations. This flag is enabled by default at @option{-O} and
8904 @item -fipa-pure-const
8905 @opindex fipa-pure-const
8906 Discover which functions are pure or constant.
8907 Enabled by default at @option{-O} and higher.
8909 @item -fipa-reference
8910 @opindex fipa-reference
8911 Discover which static variables do not escape the
8913 Enabled by default at @option{-O} and higher.
8915 @item -fipa-reference-addressable
8916 @opindex fipa-reference-addressable
8917 Discover read-only, write-only and non-addressable static variables.
8918 Enabled by default at @option{-O} and higher.
8922 Perform interprocedural pointer analysis and interprocedural modification
8923 and reference analysis. This option can cause excessive memory and
8924 compile-time usage on large compilation units. It is not enabled by
8925 default at any optimization level.
8928 @opindex fipa-profile
8929 Perform interprocedural profile propagation. The functions called only from
8930 cold functions are marked as cold. Also functions executed once (such as
8931 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8932 functions and loop less parts of functions executed once are then optimized for
8934 Enabled by default at @option{-O} and higher.
8938 Perform interprocedural constant propagation.
8939 This optimization analyzes the program to determine when values passed
8940 to functions are constants and then optimizes accordingly.
8941 This optimization can substantially increase performance
8942 if the application has constants passed to functions.
8943 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8945 @item -fipa-cp-clone
8946 @opindex fipa-cp-clone
8947 Perform function cloning to make interprocedural constant propagation stronger.
8948 When enabled, interprocedural constant propagation performs function cloning
8949 when externally visible function can be called with constant arguments.
8950 Because this optimization can create multiple copies of functions,
8951 it may significantly increase code size
8952 (see @option{--param ipcp-unit-growth=@var{value}}).
8953 This flag is enabled by default at @option{-O3}.
8956 @opindex fipa-bit-cp
8957 When enabled, perform interprocedural bitwise constant
8958 propagation. This flag is enabled by default at @option{-O2}. It
8959 requires that @option{-fipa-cp} is enabled.
8963 When enabled, perform interprocedural propagation of value
8964 ranges. This flag is enabled by default at @option{-O2}. It requires
8965 that @option{-fipa-cp} is enabled.
8969 Perform Identical Code Folding for functions and read-only variables.
8970 The optimization reduces code size and may disturb unwind stacks by replacing
8971 a function by equivalent one with a different name. The optimization works
8972 more effectively with link-time optimization enabled.
8974 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8975 works on different levels and thus the optimizations are not same - there are
8976 equivalences that are found only by GCC and equivalences found only by Gold.
8978 This flag is enabled by default at @option{-O2} and @option{-Os}.
8980 @item -fisolate-erroneous-paths-dereference
8981 @opindex fisolate-erroneous-paths-dereference
8982 Detect paths that trigger erroneous or undefined behavior due to
8983 dereferencing a null pointer. Isolate those paths from the main control
8984 flow and turn the statement with erroneous or undefined behavior into a trap.
8985 This flag is enabled by default at @option{-O2} and higher and depends on
8986 @option{-fdelete-null-pointer-checks} also being enabled.
8988 @item -fisolate-erroneous-paths-attribute
8989 @opindex fisolate-erroneous-paths-attribute
8990 Detect paths that trigger erroneous or undefined behavior due to a null value
8991 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8992 attribute. Isolate those paths from the main control flow and turn the
8993 statement with erroneous or undefined behavior into a trap. This is not
8994 currently enabled, but may be enabled by @option{-O2} in the future.
8998 Perform forward store motion on trees. This flag is
8999 enabled by default at @option{-O} and higher.
9001 @item -ftree-bit-ccp
9002 @opindex ftree-bit-ccp
9003 Perform sparse conditional bit constant propagation on trees and propagate
9004 pointer alignment information.
9005 This pass only operates on local scalar variables and is enabled by default
9006 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
9010 Perform sparse conditional constant propagation (CCP) on trees. This
9011 pass only operates on local scalar variables and is enabled by default
9012 at @option{-O} and higher.
9014 @item -fssa-backprop
9015 @opindex fssa-backprop
9016 Propagate information about uses of a value up the definition chain
9017 in order to simplify the definitions. For example, this pass strips
9018 sign operations if the sign of a value never matters. The flag is
9019 enabled by default at @option{-O} and higher.
9022 @opindex fssa-phiopt
9023 Perform pattern matching on SSA PHI nodes to optimize conditional
9024 code. This pass is enabled by default at @option{-O} and higher.
9026 @item -ftree-switch-conversion
9027 @opindex ftree-switch-conversion
9028 Perform conversion of simple initializations in a switch to
9029 initializations from a scalar array. This flag is enabled by default
9030 at @option{-O2} and higher.
9032 @item -ftree-tail-merge
9033 @opindex ftree-tail-merge
9034 Look for identical code sequences. When found, replace one with a jump to the
9035 other. This optimization is known as tail merging or cross jumping. This flag
9036 is enabled by default at @option{-O2} and higher. The compilation time
9038 be limited using @option{max-tail-merge-comparisons} parameter and
9039 @option{max-tail-merge-iterations} parameter.
9043 Perform dead code elimination (DCE) on trees. This flag is enabled by
9044 default at @option{-O} and higher.
9046 @item -ftree-builtin-call-dce
9047 @opindex ftree-builtin-call-dce
9048 Perform conditional dead code elimination (DCE) for calls to built-in functions
9049 that may set @code{errno} but are otherwise free of side effects. This flag is
9050 enabled by default at @option{-O2} and higher if @option{-Os} is not also
9053 @item -ftree-dominator-opts
9054 @opindex ftree-dominator-opts
9055 Perform a variety of simple scalar cleanups (constant/copy
9056 propagation, redundancy elimination, range propagation and expression
9057 simplification) based on a dominator tree traversal. This also
9058 performs jump threading (to reduce jumps to jumps). This flag is
9059 enabled by default at @option{-O} and higher.
9063 Perform dead store elimination (DSE) on trees. A dead store is a store into
9064 a memory location that is later overwritten by another store without
9065 any intervening loads. In this case the earlier store can be deleted. This
9066 flag is enabled by default at @option{-O} and higher.
9070 Perform loop header copying on trees. This is beneficial since it increases
9071 effectiveness of code motion optimizations. It also saves one jump. This flag
9072 is enabled by default at @option{-O} and higher. It is not enabled
9073 for @option{-Os}, since it usually increases code size.
9075 @item -ftree-loop-optimize
9076 @opindex ftree-loop-optimize
9077 Perform loop optimizations on trees. This flag is enabled by default
9078 at @option{-O} and higher.
9080 @item -ftree-loop-linear
9081 @itemx -floop-strip-mine
9083 @opindex ftree-loop-linear
9084 @opindex floop-strip-mine
9085 @opindex floop-block
9086 Perform loop nest optimizations. Same as
9087 @option{-floop-nest-optimize}. To use this code transformation, GCC has
9088 to be configured with @option{--with-isl} to enable the Graphite loop
9089 transformation infrastructure.
9091 @item -fgraphite-identity
9092 @opindex fgraphite-identity
9093 Enable the identity transformation for graphite. For every SCoP we generate
9094 the polyhedral representation and transform it back to gimple. Using
9095 @option{-fgraphite-identity} we can check the costs or benefits of the
9096 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
9097 are also performed by the code generator isl, like index splitting and
9098 dead code elimination in loops.
9100 @item -floop-nest-optimize
9101 @opindex floop-nest-optimize
9102 Enable the isl based loop nest optimizer. This is a generic loop nest
9103 optimizer based on the Pluto optimization algorithms. It calculates a loop
9104 structure optimized for data-locality and parallelism. This option
9107 @item -floop-parallelize-all
9108 @opindex floop-parallelize-all
9109 Use the Graphite data dependence analysis to identify loops that can
9110 be parallelized. Parallelize all the loops that can be analyzed to
9111 not contain loop carried dependences without checking that it is
9112 profitable to parallelize the loops.
9114 @item -ftree-coalesce-vars
9115 @opindex ftree-coalesce-vars
9116 While transforming the program out of the SSA representation, attempt to
9117 reduce copying by coalescing versions of different user-defined
9118 variables, instead of just compiler temporaries. This may severely
9119 limit the ability to debug an optimized program compiled with
9120 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9121 prevents SSA coalescing of user variables. This option is enabled by
9122 default if optimization is enabled, and it does very little otherwise.
9124 @item -ftree-loop-if-convert
9125 @opindex ftree-loop-if-convert
9126 Attempt to transform conditional jumps in the innermost loops to
9127 branch-less equivalents. The intent is to remove control-flow from
9128 the innermost loops in order to improve the ability of the
9129 vectorization pass to handle these loops. This is enabled by default
9130 if vectorization is enabled.
9132 @item -ftree-loop-distribution
9133 @opindex ftree-loop-distribution
9134 Perform loop distribution. This flag can improve cache performance on
9135 big loop bodies and allow further loop optimizations, like
9136 parallelization or vectorization, to take place. For example, the loop
9153 @item -ftree-loop-distribute-patterns
9154 @opindex ftree-loop-distribute-patterns
9155 Perform loop distribution of patterns that can be code generated with
9156 calls to a library. This flag is enabled by default at @option{-O3}.
9158 This pass distributes the initialization loops and generates a call to
9159 memset zero. For example, the loop
9175 and the initialization loop is transformed into a call to memset zero.
9177 @item -floop-interchange
9178 @opindex floop-interchange
9179 Perform loop interchange outside of graphite. This flag can improve cache
9180 performance on loop nest and allow further loop optimizations, like
9181 vectorization, to take place. For example, the loop
9183 for (int i = 0; i < N; i++)
9184 for (int j = 0; j < N; j++)
9185 for (int k = 0; k < N; k++)
9186 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9190 for (int i = 0; i < N; i++)
9191 for (int k = 0; k < N; k++)
9192 for (int j = 0; j < N; j++)
9193 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9195 This flag is enabled by default at @option{-O3}.
9197 @item -floop-unroll-and-jam
9198 @opindex floop-unroll-and-jam
9199 Apply unroll and jam transformations on feasible loops. In a loop
9200 nest this unrolls the outer loop by some factor and fuses the resulting
9201 multiple inner loops. This flag is enabled by default at @option{-O3}.
9203 @item -ftree-loop-im
9204 @opindex ftree-loop-im
9205 Perform loop invariant motion on trees. This pass moves only invariants that
9206 are hard to handle at RTL level (function calls, operations that expand to
9207 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9208 operands of conditions that are invariant out of the loop, so that we can use
9209 just trivial invariantness analysis in loop unswitching. The pass also includes
9212 @item -ftree-loop-ivcanon
9213 @opindex ftree-loop-ivcanon
9214 Create a canonical counter for number of iterations in loops for which
9215 determining number of iterations requires complicated analysis. Later
9216 optimizations then may determine the number easily. Useful especially
9217 in connection with unrolling.
9219 @item -ftree-scev-cprop
9220 @opindex ftree-scev-cprop
9221 Perform final value replacement. If a variable is modified in a loop
9222 in such a way that its value when exiting the loop can be determined using
9223 only its initial value and the number of loop iterations, replace uses of
9224 the final value by such a computation, provided it is sufficiently cheap.
9225 This reduces data dependencies and may allow further simplifications.
9226 Enabled by default at @option{-O} and higher.
9230 Perform induction variable optimizations (strength reduction, induction
9231 variable merging and induction variable elimination) on trees.
9233 @item -ftree-parallelize-loops=n
9234 @opindex ftree-parallelize-loops
9235 Parallelize loops, i.e., split their iteration space to run in n threads.
9236 This is only possible for loops whose iterations are independent
9237 and can be arbitrarily reordered. The optimization is only
9238 profitable on multiprocessor machines, for loops that are CPU-intensive,
9239 rather than constrained e.g.@: by memory bandwidth. This option
9240 implies @option{-pthread}, and thus is only supported on targets
9241 that have support for @option{-pthread}.
9245 Perform function-local points-to analysis on trees. This flag is
9246 enabled by default at @option{-O} and higher.
9250 Perform scalar replacement of aggregates. This pass replaces structure
9251 references with scalars to prevent committing structures to memory too
9252 early. This flag is enabled by default at @option{-O} and higher.
9254 @item -fstore-merging
9255 @opindex fstore-merging
9256 Perform merging of narrow stores to consecutive memory addresses. This pass
9257 merges contiguous stores of immediate values narrower than a word into fewer
9258 wider stores to reduce the number of instructions. This is enabled by default
9259 at @option{-O2} and higher as well as @option{-Os}.
9263 Perform temporary expression replacement during the SSA->normal phase. Single
9264 use/single def temporaries are replaced at their use location with their
9265 defining expression. This results in non-GIMPLE code, but gives the expanders
9266 much more complex trees to work on resulting in better RTL generation. This is
9267 enabled by default at @option{-O} and higher.
9271 Perform straight-line strength reduction on trees. This recognizes related
9272 expressions involving multiplications and replaces them by less expensive
9273 calculations when possible. This is enabled by default at @option{-O} and
9276 @item -ftree-vectorize
9277 @opindex ftree-vectorize
9278 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9279 and @option{-ftree-slp-vectorize} if not explicitly specified.
9281 @item -ftree-loop-vectorize
9282 @opindex ftree-loop-vectorize
9283 Perform loop vectorization on trees. This flag is enabled by default at
9284 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9286 @item -ftree-slp-vectorize
9287 @opindex ftree-slp-vectorize
9288 Perform basic block vectorization on trees. This flag is enabled by default at
9289 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9291 @item -fvect-cost-model=@var{model}
9292 @opindex fvect-cost-model
9293 Alter the cost model used for vectorization. The @var{model} argument
9294 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9295 With the @samp{unlimited} model the vectorized code-path is assumed
9296 to be profitable while with the @samp{dynamic} model a runtime check
9297 guards the vectorized code-path to enable it only for iteration
9298 counts that will likely execute faster than when executing the original
9299 scalar loop. The @samp{cheap} model disables vectorization of
9300 loops where doing so would be cost prohibitive for example due to
9301 required runtime checks for data dependence or alignment but otherwise
9302 is equal to the @samp{dynamic} model.
9303 The default cost model depends on other optimization flags and is
9304 either @samp{dynamic} or @samp{cheap}.
9306 @item -fsimd-cost-model=@var{model}
9307 @opindex fsimd-cost-model
9308 Alter the cost model used for vectorization of loops marked with the OpenMP
9309 simd directive. The @var{model} argument should be one of
9310 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9311 have the same meaning as described in @option{-fvect-cost-model} and by
9312 default a cost model defined with @option{-fvect-cost-model} is used.
9316 Perform Value Range Propagation on trees. This is similar to the
9317 constant propagation pass, but instead of values, ranges of values are
9318 propagated. This allows the optimizers to remove unnecessary range
9319 checks like array bound checks and null pointer checks. This is
9320 enabled by default at @option{-O2} and higher. Null pointer check
9321 elimination is only done if @option{-fdelete-null-pointer-checks} is
9325 @opindex fsplit-paths
9326 Split paths leading to loop backedges. This can improve dead code
9327 elimination and common subexpression elimination. This is enabled by
9328 default at @option{-O2} and above.
9330 @item -fsplit-ivs-in-unroller
9331 @opindex fsplit-ivs-in-unroller
9332 Enables expression of values of induction variables in later iterations
9333 of the unrolled loop using the value in the first iteration. This breaks
9334 long dependency chains, thus improving efficiency of the scheduling passes.
9336 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9337 same effect. However, that is not reliable in cases where the loop body
9338 is more complicated than a single basic block. It also does not work at all
9339 on some architectures due to restrictions in the CSE pass.
9341 This optimization is enabled by default.
9343 @item -fvariable-expansion-in-unroller
9344 @opindex fvariable-expansion-in-unroller
9345 With this option, the compiler creates multiple copies of some
9346 local variables when unrolling a loop, which can result in superior code.
9348 @item -fpartial-inlining
9349 @opindex fpartial-inlining
9350 Inline parts of functions. This option has any effect only
9351 when inlining itself is turned on by the @option{-finline-functions}
9352 or @option{-finline-small-functions} options.
9354 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9356 @item -fpredictive-commoning
9357 @opindex fpredictive-commoning
9358 Perform predictive commoning optimization, i.e., reusing computations
9359 (especially memory loads and stores) performed in previous
9360 iterations of loops.
9362 This option is enabled at level @option{-O3}.
9364 @item -fprefetch-loop-arrays
9365 @opindex fprefetch-loop-arrays
9366 If supported by the target machine, generate instructions to prefetch
9367 memory to improve the performance of loops that access large arrays.
9369 This option may generate better or worse code; results are highly
9370 dependent on the structure of loops within the source code.
9372 Disabled at level @option{-Os}.
9374 @item -fno-printf-return-value
9375 @opindex fno-printf-return-value
9376 Do not substitute constants for known return value of formatted output
9377 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9378 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9379 transformation allows GCC to optimize or even eliminate branches based
9380 on the known return value of these functions called with arguments that
9381 are either constant, or whose values are known to be in a range that
9382 makes determining the exact return value possible. For example, when
9383 @option{-fprintf-return-value} is in effect, both the branch and the
9384 body of the @code{if} statement (but not the call to @code{snprint})
9385 can be optimized away when @code{i} is a 32-bit or smaller integer
9386 because the return value is guaranteed to be at most 8.
9390 if (snprintf (buf, "%08x", i) >= sizeof buf)
9394 The @option{-fprintf-return-value} option relies on other optimizations
9395 and yields best results with @option{-O2} and above. It works in tandem
9396 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9397 options. The @option{-fprintf-return-value} option is enabled by default.
9400 @itemx -fno-peephole2
9401 @opindex fno-peephole
9402 @opindex fno-peephole2
9403 Disable any machine-specific peephole optimizations. The difference
9404 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9405 are implemented in the compiler; some targets use one, some use the
9406 other, a few use both.
9408 @option{-fpeephole} is enabled by default.
9409 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9411 @item -fno-guess-branch-probability
9412 @opindex fno-guess-branch-probability
9413 Do not guess branch probabilities using heuristics.
9415 GCC uses heuristics to guess branch probabilities if they are
9416 not provided by profiling feedback (@option{-fprofile-arcs}). These
9417 heuristics are based on the control flow graph. If some branch probabilities
9418 are specified by @code{__builtin_expect}, then the heuristics are
9419 used to guess branch probabilities for the rest of the control flow graph,
9420 taking the @code{__builtin_expect} info into account. The interactions
9421 between the heuristics and @code{__builtin_expect} can be complex, and in
9422 some cases, it may be useful to disable the heuristics so that the effects
9423 of @code{__builtin_expect} are easier to understand.
9425 It is also possible to specify expected probability of the expression
9426 with @code{__builtin_expect_with_probability} built-in function.
9428 The default is @option{-fguess-branch-probability} at levels
9429 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9431 @item -freorder-blocks
9432 @opindex freorder-blocks
9433 Reorder basic blocks in the compiled function in order to reduce number of
9434 taken branches and improve code locality.
9436 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9438 @item -freorder-blocks-algorithm=@var{algorithm}
9439 @opindex freorder-blocks-algorithm
9440 Use the specified algorithm for basic block reordering. The
9441 @var{algorithm} argument can be @samp{simple}, which does not increase
9442 code size (except sometimes due to secondary effects like alignment),
9443 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9444 put all often executed code together, minimizing the number of branches
9445 executed by making extra copies of code.
9447 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9448 @samp{stc} at levels @option{-O2}, @option{-O3}.
9450 @item -freorder-blocks-and-partition
9451 @opindex freorder-blocks-and-partition
9452 In addition to reordering basic blocks in the compiled function, in order
9453 to reduce number of taken branches, partitions hot and cold basic blocks
9454 into separate sections of the assembly and @file{.o} files, to improve
9455 paging and cache locality performance.
9457 This optimization is automatically turned off in the presence of
9458 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
9459 section attribute and on any architecture that does not support named
9460 sections. When @option{-fsplit-stack} is used this option is not
9461 enabled by default (to avoid linker errors), but may be enabled
9462 explicitly (if using a working linker).
9464 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
9466 @item -freorder-functions
9467 @opindex freorder-functions
9468 Reorder functions in the object file in order to
9469 improve code locality. This is implemented by using special
9470 subsections @code{.text.hot} for most frequently executed functions and
9471 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9472 the linker so object file format must support named sections and linker must
9473 place them in a reasonable way.
9475 Also profile feedback must be available to make this option effective. See
9476 @option{-fprofile-arcs} for details.
9478 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9480 @item -fstrict-aliasing
9481 @opindex fstrict-aliasing
9482 Allow the compiler to assume the strictest aliasing rules applicable to
9483 the language being compiled. For C (and C++), this activates
9484 optimizations based on the type of expressions. In particular, an
9485 object of one type is assumed never to reside at the same address as an
9486 object of a different type, unless the types are almost the same. For
9487 example, an @code{unsigned int} can alias an @code{int}, but not a
9488 @code{void*} or a @code{double}. A character type may alias any other
9491 @anchor{Type-punning}Pay special attention to code like this:
9504 The practice of reading from a different union member than the one most
9505 recently written to (called ``type-punning'') is common. Even with
9506 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9507 is accessed through the union type. So, the code above works as
9508 expected. @xref{Structures unions enumerations and bit-fields
9509 implementation}. However, this code might not:
9520 Similarly, access by taking the address, casting the resulting pointer
9521 and dereferencing the result has undefined behavior, even if the cast
9522 uses a union type, e.g.:
9526 return ((union a_union *) &d)->i;
9530 The @option{-fstrict-aliasing} option is enabled at levels
9531 @option{-O2}, @option{-O3}, @option{-Os}.
9533 @item -falign-functions
9534 @itemx -falign-functions=@var{n}
9535 @itemx -falign-functions=@var{n}:@var{m}
9536 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}
9537 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2}
9538 @opindex falign-functions
9539 Align the start of functions to the next power-of-two greater than
9540 @var{n}, skipping up to @var{m}-1 bytes. This ensures that at least
9541 the first @var{m} bytes of the function can be fetched by the CPU
9542 without crossing an @var{n}-byte alignment boundary.
9544 If @var{m} is not specified, it defaults to @var{n}.
9546 Examples: @option{-falign-functions=32} aligns functions to the next
9547 32-byte boundary, @option{-falign-functions=24} aligns to the next
9548 32-byte boundary only if this can be done by skipping 23 bytes or less,
9549 @option{-falign-functions=32:7} aligns to the next
9550 32-byte boundary only if this can be done by skipping 6 bytes or less.
9552 The second pair of @var{n2}:@var{m2} values allows you to specify
9553 a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to
9554 the next 64-byte boundary if this can be done by skipping 6 bytes or less,
9555 otherwise aligns to the next 32-byte boundary if this can be done
9556 by skipping 2 bytes or less.
9557 If @var{m2} is not specified, it defaults to @var{n2}.
9559 Some assemblers only support this flag when @var{n} is a power of two;
9560 in that case, it is rounded up.
9562 @option{-fno-align-functions} and @option{-falign-functions=1} are
9563 equivalent and mean that functions are not aligned.
9565 If @var{n} is not specified or is zero, use a machine-dependent default.
9566 The maximum allowed @var{n} option value is 65536.
9568 Enabled at levels @option{-O2}, @option{-O3}.
9570 @item -flimit-function-alignment
9571 If this option is enabled, the compiler tries to avoid unnecessarily
9572 overaligning functions. It attempts to instruct the assembler to align
9573 by the amount specified by @option{-falign-functions}, but not to
9574 skip more bytes than the size of the function.
9576 @item -falign-labels
9577 @itemx -falign-labels=@var{n}
9578 @itemx -falign-labels=@var{n}:@var{m}
9579 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}
9580 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2}
9581 @opindex falign-labels
9582 Align all branch targets to a power-of-two boundary.
9584 Parameters of this option are analogous to the @option{-falign-functions} option.
9585 @option{-fno-align-labels} and @option{-falign-labels=1} are
9586 equivalent and mean that labels are not aligned.
9588 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9589 are greater than this value, then their values are used instead.
9591 If @var{n} is not specified or is zero, use a machine-dependent default
9592 which is very likely to be @samp{1}, meaning no alignment.
9593 The maximum allowed @var{n} option value is 65536.
9595 Enabled at levels @option{-O2}, @option{-O3}.
9598 @itemx -falign-loops=@var{n}
9599 @itemx -falign-loops=@var{n}:@var{m}
9600 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}
9601 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2}
9602 @opindex falign-loops
9603 Align loops to a power-of-two boundary. If the loops are executed
9604 many times, this makes up for any execution of the dummy padding
9607 Parameters of this option are analogous to the @option{-falign-functions} option.
9608 @option{-fno-align-loops} and @option{-falign-loops=1} are
9609 equivalent and mean that loops are not aligned.
9610 The maximum allowed @var{n} option value is 65536.
9612 If @var{n} is not specified or is zero, use a machine-dependent default.
9614 Enabled at levels @option{-O2}, @option{-O3}.
9617 @itemx -falign-jumps=@var{n}
9618 @itemx -falign-jumps=@var{n}:@var{m}
9619 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}
9620 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2}
9621 @opindex falign-jumps
9622 Align branch targets to a power-of-two boundary, for branch targets
9623 where the targets can only be reached by jumping. In this case,
9624 no dummy operations need be executed.
9626 Parameters of this option are analogous to the @option{-falign-functions} option.
9627 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9628 equivalent and mean that loops are not aligned.
9630 If @var{n} is not specified or is zero, use a machine-dependent default.
9631 The maximum allowed @var{n} option value is 65536.
9633 Enabled at levels @option{-O2}, @option{-O3}.
9635 @item -funit-at-a-time
9636 @opindex funit-at-a-time
9637 This option is left for compatibility reasons. @option{-funit-at-a-time}
9638 has no effect, while @option{-fno-unit-at-a-time} implies
9639 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9643 @item -fno-toplevel-reorder
9644 @opindex fno-toplevel-reorder
9645 Do not reorder top-level functions, variables, and @code{asm}
9646 statements. Output them in the same order that they appear in the
9647 input file. When this option is used, unreferenced static variables
9648 are not removed. This option is intended to support existing code
9649 that relies on a particular ordering. For new code, it is better to
9650 use attributes when possible.
9652 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9653 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9658 Constructs webs as commonly used for register allocation purposes and assign
9659 each web individual pseudo register. This allows the register allocation pass
9660 to operate on pseudos directly, but also strengthens several other optimization
9661 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9662 however, make debugging impossible, since variables no longer stay in a
9665 Enabled by default with @option{-funroll-loops}.
9667 @item -fwhole-program
9668 @opindex fwhole-program
9669 Assume that the current compilation unit represents the whole program being
9670 compiled. All public functions and variables with the exception of @code{main}
9671 and those merged by attribute @code{externally_visible} become static functions
9672 and in effect are optimized more aggressively by interprocedural optimizers.
9674 This option should not be used in combination with @option{-flto}.
9675 Instead relying on a linker plugin should provide safer and more precise
9678 @item -flto[=@var{n}]
9680 This option runs the standard link-time optimizer. When invoked
9681 with source code, it generates GIMPLE (one of GCC's internal
9682 representations) and writes it to special ELF sections in the object
9683 file. When the object files are linked together, all the function
9684 bodies are read from these ELF sections and instantiated as if they
9685 had been part of the same translation unit.
9687 To use the link-time optimizer, @option{-flto} and optimization
9688 options should be specified at compile time and during the final link.
9689 It is recommended that you compile all the files participating in the
9690 same link with the same options and also specify those options at
9695 gcc -c -O2 -flto foo.c
9696 gcc -c -O2 -flto bar.c
9697 gcc -o myprog -flto -O2 foo.o bar.o
9700 The first two invocations to GCC save a bytecode representation
9701 of GIMPLE into special ELF sections inside @file{foo.o} and
9702 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9703 @file{foo.o} and @file{bar.o}, merges the two files into a single
9704 internal image, and compiles the result as usual. Since both
9705 @file{foo.o} and @file{bar.o} are merged into a single image, this
9706 causes all the interprocedural analyses and optimizations in GCC to
9707 work across the two files as if they were a single one. This means,
9708 for example, that the inliner is able to inline functions in
9709 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9711 Another (simpler) way to enable link-time optimization is:
9714 gcc -o myprog -flto -O2 foo.c bar.c
9717 The above generates bytecode for @file{foo.c} and @file{bar.c},
9718 merges them together into a single GIMPLE representation and optimizes
9719 them as usual to produce @file{myprog}.
9721 The only important thing to keep in mind is that to enable link-time
9722 optimizations you need to use the GCC driver to perform the link step.
9723 GCC then automatically performs link-time optimization if any of the
9724 objects involved were compiled with the @option{-flto} command-line option.
9726 should specify the optimization options to be used for link-time
9727 optimization though GCC tries to be clever at guessing an
9728 optimization level to use from the options used at compile time
9729 if you fail to specify one at link time. You can always override
9730 the automatic decision to do link-time optimization
9731 by passing @option{-fno-lto} to the link command.
9733 To make whole program optimization effective, it is necessary to make
9734 certain whole program assumptions. The compiler needs to know
9735 what functions and variables can be accessed by libraries and runtime
9736 outside of the link-time optimized unit. When supported by the linker,
9737 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9738 to the compiler about used and externally visible symbols. When
9739 the linker plugin is not available, @option{-fwhole-program} should be
9740 used to allow the compiler to make these assumptions, which leads
9741 to more aggressive optimization decisions.
9743 When @option{-fuse-linker-plugin} is not enabled, when a file is
9744 compiled with @option{-flto}, the generated object file is larger than
9745 a regular object file because it contains GIMPLE bytecodes and the usual
9746 final code (see @option{-ffat-lto-objects}. This means that
9747 object files with LTO information can be linked as normal object
9748 files; if @option{-fno-lto} is passed to the linker, no
9749 interprocedural optimizations are applied. Note that when
9750 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
9751 but you cannot perform a regular, non-LTO link on them.
9753 Additionally, the optimization flags used to compile individual files
9754 are not necessarily related to those used at link time. For instance,
9757 gcc -c -O0 -ffat-lto-objects -flto foo.c
9758 gcc -c -O0 -ffat-lto-objects -flto bar.c
9759 gcc -o myprog -O3 foo.o bar.o
9762 This produces individual object files with unoptimized assembler
9763 code, but the resulting binary @file{myprog} is optimized at
9764 @option{-O3}. If, instead, the final binary is generated with
9765 @option{-fno-lto}, then @file{myprog} is not optimized.
9767 When producing the final binary, GCC only
9768 applies link-time optimizations to those files that contain bytecode.
9769 Therefore, you can mix and match object files and libraries with
9770 GIMPLE bytecodes and final object code. GCC automatically selects
9771 which files to optimize in LTO mode and which files to link without
9774 There are some code generation flags preserved by GCC when
9775 generating bytecodes, as they need to be used during the final link
9776 stage. Generally options specified at link time override those
9777 specified at compile time.
9779 If you do not specify an optimization level option @option{-O} at
9780 link time, then GCC uses the highest optimization level
9781 used when compiling the object files.
9783 Currently, the following options and their settings are taken from
9784 the first object file that explicitly specifies them:
9785 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9786 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9787 and all the @option{-m} target flags.
9789 Certain ABI-changing flags are required to match in all compilation units,
9790 and trying to override this at link time with a conflicting value
9791 is ignored. This includes options such as @option{-freg-struct-return}
9792 and @option{-fpcc-struct-return}.
9794 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9795 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9796 are passed through to the link stage and merged conservatively for
9797 conflicting translation units. Specifically
9798 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9799 precedence; and for example @option{-ffp-contract=off} takes precedence
9800 over @option{-ffp-contract=fast}. You can override them at link time.
9802 If LTO encounters objects with C linkage declared with incompatible
9803 types in separate translation units to be linked together (undefined
9804 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9805 issued. The behavior is still undefined at run time. Similar
9806 diagnostics may be raised for other languages.
9808 Another feature of LTO is that it is possible to apply interprocedural
9809 optimizations on files written in different languages:
9814 gfortran -c -flto baz.f90
9815 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9818 Notice that the final link is done with @command{g++} to get the C++
9819 runtime libraries and @option{-lgfortran} is added to get the Fortran
9820 runtime libraries. In general, when mixing languages in LTO mode, you
9821 should use the same link command options as when mixing languages in a
9822 regular (non-LTO) compilation.
9824 If object files containing GIMPLE bytecode are stored in a library archive, say
9825 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9826 are using a linker with plugin support. To create static libraries suitable
9827 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9828 and @command{ranlib};
9829 to show the symbols of object files with GIMPLE bytecode, use
9830 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9831 and @command{nm} have been compiled with plugin support. At link time, use the
9832 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9833 the LTO optimization process:
9836 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9839 With the linker plugin enabled, the linker extracts the needed
9840 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9841 to make them part of the aggregated GIMPLE image to be optimized.
9843 If you are not using a linker with plugin support and/or do not
9844 enable the linker plugin, then the objects inside @file{libfoo.a}
9845 are extracted and linked as usual, but they do not participate
9846 in the LTO optimization process. In order to make a static library suitable
9847 for both LTO optimization and usual linkage, compile its object files with
9848 @option{-flto} @option{-ffat-lto-objects}.
9850 Link-time optimizations do not require the presence of the whole program to
9851 operate. If the program does not require any symbols to be exported, it is
9852 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9853 the interprocedural optimizers to use more aggressive assumptions which may
9854 lead to improved optimization opportunities.
9855 Use of @option{-fwhole-program} is not needed when linker plugin is
9856 active (see @option{-fuse-linker-plugin}).
9858 The current implementation of LTO makes no
9859 attempt to generate bytecode that is portable between different
9860 types of hosts. The bytecode files are versioned and there is a
9861 strict version check, so bytecode files generated in one version of
9862 GCC do not work with an older or newer version of GCC.
9864 Link-time optimization does not work well with generation of debugging
9865 information on systems other than those using a combination of ELF and
9868 If you specify the optional @var{n}, the optimization and code
9869 generation done at link time is executed in parallel using @var{n}
9870 parallel jobs by utilizing an installed @command{make} program. The
9871 environment variable @env{MAKE} may be used to override the program
9872 used. The default value for @var{n} is 1.
9874 You can also specify @option{-flto=jobserver} to use GNU make's
9875 job server mode to determine the number of parallel jobs. This
9876 is useful when the Makefile calling GCC is already executing in parallel.
9877 You must prepend a @samp{+} to the command recipe in the parent Makefile
9878 for this to work. This option likely only works if @env{MAKE} is
9881 @item -flto-partition=@var{alg}
9882 @opindex flto-partition
9883 Specify the partitioning algorithm used by the link-time optimizer.
9884 The value is either @samp{1to1} to specify a partitioning mirroring
9885 the original source files or @samp{balanced} to specify partitioning
9886 into equally sized chunks (whenever possible) or @samp{max} to create
9887 new partition for every symbol where possible. Specifying @samp{none}
9888 as an algorithm disables partitioning and streaming completely.
9889 The default value is @samp{balanced}. While @samp{1to1} can be used
9890 as an workaround for various code ordering issues, the @samp{max}
9891 partitioning is intended for internal testing only.
9892 The value @samp{one} specifies that exactly one partition should be
9893 used while the value @samp{none} bypasses partitioning and executes
9894 the link-time optimization step directly from the WPA phase.
9896 @item -flto-odr-type-merging
9897 @opindex flto-odr-type-merging
9898 Enable streaming of mangled types names of C++ types and their unification
9899 at link time. This increases size of LTO object files, but enables
9900 diagnostics about One Definition Rule violations.
9902 @item -flto-compression-level=@var{n}
9903 @opindex flto-compression-level
9904 This option specifies the level of compression used for intermediate
9905 language written to LTO object files, and is only meaningful in
9906 conjunction with LTO mode (@option{-flto}). Valid
9907 values are 0 (no compression) to 9 (maximum compression). Values
9908 outside this range are clamped to either 0 or 9. If the option is not
9909 given, a default balanced compression setting is used.
9911 @item -fuse-linker-plugin
9912 @opindex fuse-linker-plugin
9913 Enables the use of a linker plugin during link-time optimization. This
9914 option relies on plugin support in the linker, which is available in gold
9915 or in GNU ld 2.21 or newer.
9917 This option enables the extraction of object files with GIMPLE bytecode out
9918 of library archives. This improves the quality of optimization by exposing
9919 more code to the link-time optimizer. This information specifies what
9920 symbols can be accessed externally (by non-LTO object or during dynamic
9921 linking). Resulting code quality improvements on binaries (and shared
9922 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9923 See @option{-flto} for a description of the effect of this flag and how to
9926 This option is enabled by default when LTO support in GCC is enabled
9927 and GCC was configured for use with
9928 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9930 @item -ffat-lto-objects
9931 @opindex ffat-lto-objects
9932 Fat LTO objects are object files that contain both the intermediate language
9933 and the object code. This makes them usable for both LTO linking and normal
9934 linking. This option is effective only when compiling with @option{-flto}
9935 and is ignored at link time.
9937 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9938 requires the complete toolchain to be aware of LTO. It requires a linker with
9939 linker plugin support for basic functionality. Additionally,
9940 @command{nm}, @command{ar} and @command{ranlib}
9941 need to support linker plugins to allow a full-featured build environment
9942 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9943 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9944 to these tools. With non fat LTO makefiles need to be modified to use them.
9946 Note that modern binutils provide plugin auto-load mechanism.
9947 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
9948 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
9949 @command{gcc-ranlib}).
9951 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9954 @item -fcompare-elim
9955 @opindex fcompare-elim
9956 After register allocation and post-register allocation instruction splitting,
9957 identify arithmetic instructions that compute processor flags similar to a
9958 comparison operation based on that arithmetic. If possible, eliminate the
9959 explicit comparison operation.
9961 This pass only applies to certain targets that cannot explicitly represent
9962 the comparison operation before register allocation is complete.
9964 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9966 @item -fcprop-registers
9967 @opindex fcprop-registers
9968 After register allocation and post-register allocation instruction splitting,
9969 perform a copy-propagation pass to try to reduce scheduling dependencies
9970 and occasionally eliminate the copy.
9972 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9974 @item -fprofile-correction
9975 @opindex fprofile-correction
9976 Profiles collected using an instrumented binary for multi-threaded programs may
9977 be inconsistent due to missed counter updates. When this option is specified,
9978 GCC uses heuristics to correct or smooth out such inconsistencies. By
9979 default, GCC emits an error message when an inconsistent profile is detected.
9982 @itemx -fprofile-use=@var{path}
9983 @opindex fprofile-use
9984 Enable profile feedback-directed optimizations,
9985 and the following optimizations
9986 which are generally profitable only with profile feedback available:
9987 @option{-fbranch-probabilities}, @option{-fvpt},
9988 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9989 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9991 Before you can use this option, you must first generate profiling information.
9992 @xref{Instrumentation Options}, for information about the
9993 @option{-fprofile-generate} option.
9995 By default, GCC emits an error message if the feedback profiles do not
9996 match the source code. This error can be turned into a warning by using
9997 @option{-Wno-error=coverage-mismatch}. Note this may result in poorly
9998 optimized code. Additionally, by default, GCC also emits a warning message if
9999 the feedback profiles do not exist (See @option{-Wmissing-profile}).
10001 If @var{path} is specified, GCC looks at the @var{path} to find
10002 the profile feedback data files. See @option{-fprofile-dir}.
10004 @item -fauto-profile
10005 @itemx -fauto-profile=@var{path}
10006 @opindex fauto-profile
10007 Enable sampling-based feedback-directed optimizations,
10008 and the following optimizations
10009 which are generally profitable only with profile feedback available:
10010 @option{-fbranch-probabilities}, @option{-fvpt},
10011 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
10012 @option{-ftree-vectorize},
10013 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
10014 @option{-fpredictive-commoning}, @option{-funswitch-loops},
10015 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
10017 @var{path} is the name of a file containing AutoFDO profile information.
10018 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
10020 Producing an AutoFDO profile data file requires running your program
10021 with the @command{perf} utility on a supported GNU/Linux target system.
10022 For more information, see @uref{https://perf.wiki.kernel.org/}.
10026 perf record -e br_inst_retired:near_taken -b -o perf.data \
10030 Then use the @command{create_gcov} tool to convert the raw profile data
10031 to a format that can be used by GCC.@ You must also supply the
10032 unstripped binary for your program to this tool.
10033 See @uref{https://github.com/google/autofdo}.
10037 create_gcov --binary=your_program.unstripped --profile=perf.data \
10038 --gcov=profile.afdo
10042 The following options control compiler behavior regarding floating-point
10043 arithmetic. These options trade off between speed and
10044 correctness. All must be specifically enabled.
10047 @item -ffloat-store
10048 @opindex ffloat-store
10049 Do not store floating-point variables in registers, and inhibit other
10050 options that might change whether a floating-point value is taken from a
10051 register or memory.
10053 @cindex floating-point precision
10054 This option prevents undesirable excess precision on machines such as
10055 the 68000 where the floating registers (of the 68881) keep more
10056 precision than a @code{double} is supposed to have. Similarly for the
10057 x86 architecture. For most programs, the excess precision does only
10058 good, but a few programs rely on the precise definition of IEEE floating
10059 point. Use @option{-ffloat-store} for such programs, after modifying
10060 them to store all pertinent intermediate computations into variables.
10062 @item -fexcess-precision=@var{style}
10063 @opindex fexcess-precision
10064 This option allows further control over excess precision on machines
10065 where floating-point operations occur in a format with more precision or
10066 range than the IEEE standard and interchange floating-point types. By
10067 default, @option{-fexcess-precision=fast} is in effect; this means that
10068 operations may be carried out in a wider precision than the types specified
10069 in the source if that would result in faster code, and it is unpredictable
10070 when rounding to the types specified in the source code takes place.
10071 When compiling C, if @option{-fexcess-precision=standard} is specified then
10072 excess precision follows the rules specified in ISO C99; in particular,
10073 both casts and assignments cause values to be rounded to their
10074 semantic types (whereas @option{-ffloat-store} only affects
10075 assignments). This option is enabled by default for C if a strict
10076 conformance option such as @option{-std=c99} is used.
10077 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
10078 regardless of whether a strict conformance option is used.
10081 @option{-fexcess-precision=standard} is not implemented for languages
10082 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
10083 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
10084 semantics apply without excess precision, and in the latter, rounding
10088 @opindex ffast-math
10089 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
10090 @option{-ffinite-math-only}, @option{-fno-rounding-math},
10091 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
10092 @option{-fexcess-precision=fast}.
10094 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
10096 This option is not turned on by any @option{-O} option besides
10097 @option{-Ofast} since it can result in incorrect output for programs
10098 that depend on an exact implementation of IEEE or ISO rules/specifications
10099 for math functions. It may, however, yield faster code for programs
10100 that do not require the guarantees of these specifications.
10102 @item -fno-math-errno
10103 @opindex fno-math-errno
10104 Do not set @code{errno} after calling math functions that are executed
10105 with a single instruction, e.g., @code{sqrt}. A program that relies on
10106 IEEE exceptions for math error handling may want to use this flag
10107 for speed while maintaining IEEE arithmetic compatibility.
10109 This option is not turned on by any @option{-O} option since
10110 it can result in incorrect output for programs that depend on
10111 an exact implementation of IEEE or ISO rules/specifications for
10112 math functions. It may, however, yield faster code for programs
10113 that do not require the guarantees of these specifications.
10115 The default is @option{-fmath-errno}.
10117 On Darwin systems, the math library never sets @code{errno}. There is
10118 therefore no reason for the compiler to consider the possibility that
10119 it might, and @option{-fno-math-errno} is the default.
10121 @item -funsafe-math-optimizations
10122 @opindex funsafe-math-optimizations
10124 Allow optimizations for floating-point arithmetic that (a) assume
10125 that arguments and results are valid and (b) may violate IEEE or
10126 ANSI standards. When used at link time, it may include libraries
10127 or startup files that change the default FPU control word or other
10128 similar optimizations.
10130 This option is not turned on by any @option{-O} option since
10131 it can result in incorrect output for programs that depend on
10132 an exact implementation of IEEE or ISO rules/specifications for
10133 math functions. It may, however, yield faster code for programs
10134 that do not require the guarantees of these specifications.
10135 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10136 @option{-fassociative-math} and @option{-freciprocal-math}.
10138 The default is @option{-fno-unsafe-math-optimizations}.
10140 @item -fassociative-math
10141 @opindex fassociative-math
10143 Allow re-association of operands in series of floating-point operations.
10144 This violates the ISO C and C++ language standard by possibly changing
10145 computation result. NOTE: re-ordering may change the sign of zero as
10146 well as ignore NaNs and inhibit or create underflow or overflow (and
10147 thus cannot be used on code that relies on rounding behavior like
10148 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10149 and thus may not be used when ordered comparisons are required.
10150 This option requires that both @option{-fno-signed-zeros} and
10151 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10152 much sense with @option{-frounding-math}. For Fortran the option
10153 is automatically enabled when both @option{-fno-signed-zeros} and
10154 @option{-fno-trapping-math} are in effect.
10156 The default is @option{-fno-associative-math}.
10158 @item -freciprocal-math
10159 @opindex freciprocal-math
10161 Allow the reciprocal of a value to be used instead of dividing by
10162 the value if this enables optimizations. For example @code{x / y}
10163 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10164 is subject to common subexpression elimination. Note that this loses
10165 precision and increases the number of flops operating on the value.
10167 The default is @option{-fno-reciprocal-math}.
10169 @item -ffinite-math-only
10170 @opindex ffinite-math-only
10171 Allow optimizations for floating-point arithmetic that assume
10172 that arguments and results are not NaNs or +-Infs.
10174 This option is not turned on by any @option{-O} option since
10175 it can result in incorrect output for programs that depend on
10176 an exact implementation of IEEE or ISO rules/specifications for
10177 math functions. It may, however, yield faster code for programs
10178 that do not require the guarantees of these specifications.
10180 The default is @option{-fno-finite-math-only}.
10182 @item -fno-signed-zeros
10183 @opindex fno-signed-zeros
10184 Allow optimizations for floating-point arithmetic that ignore the
10185 signedness of zero. IEEE arithmetic specifies the behavior of
10186 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10187 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10188 This option implies that the sign of a zero result isn't significant.
10190 The default is @option{-fsigned-zeros}.
10192 @item -fno-trapping-math
10193 @opindex fno-trapping-math
10194 Compile code assuming that floating-point operations cannot generate
10195 user-visible traps. These traps include division by zero, overflow,
10196 underflow, inexact result and invalid operation. This option requires
10197 that @option{-fno-signaling-nans} be in effect. Setting this option may
10198 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10200 This option should never be turned on by any @option{-O} option since
10201 it can result in incorrect output for programs that depend on
10202 an exact implementation of IEEE or ISO rules/specifications for
10205 The default is @option{-ftrapping-math}.
10207 @item -frounding-math
10208 @opindex frounding-math
10209 Disable transformations and optimizations that assume default floating-point
10210 rounding behavior. This is round-to-zero for all floating point
10211 to integer conversions, and round-to-nearest for all other arithmetic
10212 truncations. This option should be specified for programs that change
10213 the FP rounding mode dynamically, or that may be executed with a
10214 non-default rounding mode. This option disables constant folding of
10215 floating-point expressions at compile time (which may be affected by
10216 rounding mode) and arithmetic transformations that are unsafe in the
10217 presence of sign-dependent rounding modes.
10219 The default is @option{-fno-rounding-math}.
10221 This option is experimental and does not currently guarantee to
10222 disable all GCC optimizations that are affected by rounding mode.
10223 Future versions of GCC may provide finer control of this setting
10224 using C99's @code{FENV_ACCESS} pragma. This command-line option
10225 will be used to specify the default state for @code{FENV_ACCESS}.
10227 @item -fsignaling-nans
10228 @opindex fsignaling-nans
10229 Compile code assuming that IEEE signaling NaNs may generate user-visible
10230 traps during floating-point operations. Setting this option disables
10231 optimizations that may change the number of exceptions visible with
10232 signaling NaNs. This option implies @option{-ftrapping-math}.
10234 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10237 The default is @option{-fno-signaling-nans}.
10239 This option is experimental and does not currently guarantee to
10240 disable all GCC optimizations that affect signaling NaN behavior.
10242 @item -fno-fp-int-builtin-inexact
10243 @opindex fno-fp-int-builtin-inexact
10244 Do not allow the built-in functions @code{ceil}, @code{floor},
10245 @code{round} and @code{trunc}, and their @code{float} and @code{long
10246 double} variants, to generate code that raises the ``inexact''
10247 floating-point exception for noninteger arguments. ISO C99 and C11
10248 allow these functions to raise the ``inexact'' exception, but ISO/IEC
10249 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
10250 functions to do so.
10252 The default is @option{-ffp-int-builtin-inexact}, allowing the
10253 exception to be raised. This option does nothing unless
10254 @option{-ftrapping-math} is in effect.
10256 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
10257 generate a call to a library function then the ``inexact'' exception
10258 may be raised if the library implementation does not follow TS 18661.
10260 @item -fsingle-precision-constant
10261 @opindex fsingle-precision-constant
10262 Treat floating-point constants as single precision instead of
10263 implicitly converting them to double-precision constants.
10265 @item -fcx-limited-range
10266 @opindex fcx-limited-range
10267 When enabled, this option states that a range reduction step is not
10268 needed when performing complex division. Also, there is no checking
10269 whether the result of a complex multiplication or division is @code{NaN
10270 + I*NaN}, with an attempt to rescue the situation in that case. The
10271 default is @option{-fno-cx-limited-range}, but is enabled by
10272 @option{-ffast-math}.
10274 This option controls the default setting of the ISO C99
10275 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10278 @item -fcx-fortran-rules
10279 @opindex fcx-fortran-rules
10280 Complex multiplication and division follow Fortran rules. Range
10281 reduction is done as part of complex division, but there is no checking
10282 whether the result of a complex multiplication or division is @code{NaN
10283 + I*NaN}, with an attempt to rescue the situation in that case.
10285 The default is @option{-fno-cx-fortran-rules}.
10289 The following options control optimizations that may improve
10290 performance, but are not enabled by any @option{-O} options. This
10291 section includes experimental options that may produce broken code.
10294 @item -fbranch-probabilities
10295 @opindex fbranch-probabilities
10296 After running a program compiled with @option{-fprofile-arcs}
10297 (@pxref{Instrumentation Options}),
10298 you can compile it a second time using
10299 @option{-fbranch-probabilities}, to improve optimizations based on
10300 the number of times each branch was taken. When a program
10301 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10302 counts to a file called @file{@var{sourcename}.gcda} for each source
10303 file. The information in this data file is very dependent on the
10304 structure of the generated code, so you must use the same source code
10305 and the same optimization options for both compilations.
10307 With @option{-fbranch-probabilities}, GCC puts a
10308 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10309 These can be used to improve optimization. Currently, they are only
10310 used in one place: in @file{reorg.c}, instead of guessing which path a
10311 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10312 exactly determine which path is taken more often.
10314 @item -fprofile-values
10315 @opindex fprofile-values
10316 If combined with @option{-fprofile-arcs}, it adds code so that some
10317 data about values of expressions in the program is gathered.
10319 With @option{-fbranch-probabilities}, it reads back the data gathered
10320 from profiling values of expressions for usage in optimizations.
10322 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
10324 @item -fprofile-reorder-functions
10325 @opindex fprofile-reorder-functions
10326 Function reordering based on profile instrumentation collects
10327 first time of execution of a function and orders these functions
10328 in ascending order.
10330 Enabled with @option{-fprofile-use}.
10334 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10335 to add code to gather information about values of expressions.
10337 With @option{-fbranch-probabilities}, it reads back the data gathered
10338 and actually performs the optimizations based on them.
10339 Currently the optimizations include specialization of division operations
10340 using the knowledge about the value of the denominator.
10342 @item -frename-registers
10343 @opindex frename-registers
10344 Attempt to avoid false dependencies in scheduled code by making use
10345 of registers left over after register allocation. This optimization
10346 most benefits processors with lots of registers. Depending on the
10347 debug information format adopted by the target, however, it can
10348 make debugging impossible, since variables no longer stay in
10349 a ``home register''.
10351 Enabled by default with @option{-funroll-loops}.
10353 @item -fschedule-fusion
10354 @opindex fschedule-fusion
10355 Performs a target dependent pass over the instruction stream to schedule
10356 instructions of same type together because target machine can execute them
10357 more efficiently if they are adjacent to each other in the instruction flow.
10359 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10363 Perform tail duplication to enlarge superblock size. This transformation
10364 simplifies the control flow of the function allowing other optimizations to do
10367 Enabled with @option{-fprofile-use}.
10369 @item -funroll-loops
10370 @opindex funroll-loops
10371 Unroll loops whose number of iterations can be determined at compile time or
10372 upon entry to the loop. @option{-funroll-loops} implies
10373 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10374 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10375 a small constant number of iterations). This option makes code larger, and may
10376 or may not make it run faster.
10378 Enabled with @option{-fprofile-use}.
10380 @item -funroll-all-loops
10381 @opindex funroll-all-loops
10382 Unroll all loops, even if their number of iterations is uncertain when
10383 the loop is entered. This usually makes programs run more slowly.
10384 @option{-funroll-all-loops} implies the same options as
10385 @option{-funroll-loops}.
10388 @opindex fpeel-loops
10389 Peels loops for which there is enough information that they do not
10390 roll much (from profile feedback or static analysis). It also turns on
10391 complete loop peeling (i.e.@: complete removal of loops with small constant
10392 number of iterations).
10394 Enabled with @option{-O3} and/or @option{-fprofile-use}.
10396 @item -fmove-loop-invariants
10397 @opindex fmove-loop-invariants
10398 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10399 at level @option{-O1}
10401 @item -fsplit-loops
10402 @opindex fsplit-loops
10403 Split a loop into two if it contains a condition that's always true
10404 for one side of the iteration space and false for the other.
10406 @item -funswitch-loops
10407 @opindex funswitch-loops
10408 Move branches with loop invariant conditions out of the loop, with duplicates
10409 of the loop on both branches (modified according to result of the condition).
10411 @item -ffunction-sections
10412 @itemx -fdata-sections
10413 @opindex ffunction-sections
10414 @opindex fdata-sections
10415 Place each function or data item into its own section in the output
10416 file if the target supports arbitrary sections. The name of the
10417 function or the name of the data item determines the section's name
10418 in the output file.
10420 Use these options on systems where the linker can perform optimizations to
10421 improve locality of reference in the instruction space. Most systems using the
10422 ELF object format have linkers with such optimizations. On AIX, the linker
10423 rearranges sections (CSECTs) based on the call graph. The performance impact
10426 Together with a linker garbage collection (linker @option{--gc-sections}
10427 option) these options may lead to smaller statically-linked executables (after
10430 On ELF/DWARF systems these options do not degenerate the quality of the debug
10431 information. There could be issues with other object files/debug info formats.
10433 Only use these options when there are significant benefits from doing so. When
10434 you specify these options, the assembler and linker create larger object and
10435 executable files and are also slower. These options affect code generation.
10436 They prevent optimizations by the compiler and assembler using relative
10437 locations inside a translation unit since the locations are unknown until
10438 link time. An example of such an optimization is relaxing calls to short call
10441 @item -fbranch-target-load-optimize
10442 @opindex fbranch-target-load-optimize
10443 Perform branch target register load optimization before prologue / epilogue
10445 The use of target registers can typically be exposed only during reload,
10446 thus hoisting loads out of loops and doing inter-block scheduling needs
10447 a separate optimization pass.
10449 @item -fbranch-target-load-optimize2
10450 @opindex fbranch-target-load-optimize2
10451 Perform branch target register load optimization after prologue / epilogue
10454 @item -fbtr-bb-exclusive
10455 @opindex fbtr-bb-exclusive
10456 When performing branch target register load optimization, don't reuse
10457 branch target registers within any basic block.
10460 @opindex fstdarg-opt
10461 Optimize the prologue of variadic argument functions with respect to usage of
10464 @item -fsection-anchors
10465 @opindex fsection-anchors
10466 Try to reduce the number of symbolic address calculations by using
10467 shared ``anchor'' symbols to address nearby objects. This transformation
10468 can help to reduce the number of GOT entries and GOT accesses on some
10471 For example, the implementation of the following function @code{foo}:
10474 static int a, b, c;
10475 int foo (void) @{ return a + b + c; @}
10479 usually calculates the addresses of all three variables, but if you
10480 compile it with @option{-fsection-anchors}, it accesses the variables
10481 from a common anchor point instead. The effect is similar to the
10482 following pseudocode (which isn't valid C):
10487 register int *xr = &x;
10488 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10492 Not all targets support this option.
10494 @item --param @var{name}=@var{value}
10496 In some places, GCC uses various constants to control the amount of
10497 optimization that is done. For example, GCC does not inline functions
10498 that contain more than a certain number of instructions. You can
10499 control some of these constants on the command line using the
10500 @option{--param} option.
10502 The names of specific parameters, and the meaning of the values, are
10503 tied to the internals of the compiler, and are subject to change
10504 without notice in future releases.
10506 In order to get minimal, maximal and default value of a parameter,
10507 one can use @option{--help=param -Q} options.
10509 In each case, the @var{value} is an integer. The allowable choices for
10513 @item predictable-branch-outcome
10514 When branch is predicted to be taken with probability lower than this threshold
10515 (in percent), then it is considered well predictable.
10517 @item max-rtl-if-conversion-insns
10518 RTL if-conversion tries to remove conditional branches around a block and
10519 replace them with conditionally executed instructions. This parameter
10520 gives the maximum number of instructions in a block which should be
10521 considered for if-conversion. The compiler will
10522 also use other heuristics to decide whether if-conversion is likely to be
10525 @item max-rtl-if-conversion-predictable-cost
10526 @itemx max-rtl-if-conversion-unpredictable-cost
10527 RTL if-conversion will try to remove conditional branches around a block
10528 and replace them with conditionally executed instructions. These parameters
10529 give the maximum permissible cost for the sequence that would be generated
10530 by if-conversion depending on whether the branch is statically determined
10531 to be predictable or not. The units for this parameter are the same as
10532 those for the GCC internal seq_cost metric. The compiler will try to
10533 provide a reasonable default for this parameter using the BRANCH_COST
10536 @item max-crossjump-edges
10537 The maximum number of incoming edges to consider for cross-jumping.
10538 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10539 the number of edges incoming to each block. Increasing values mean
10540 more aggressive optimization, making the compilation time increase with
10541 probably small improvement in executable size.
10543 @item min-crossjump-insns
10544 The minimum number of instructions that must be matched at the end
10545 of two blocks before cross-jumping is performed on them. This
10546 value is ignored in the case where all instructions in the block being
10547 cross-jumped from are matched.
10549 @item max-grow-copy-bb-insns
10550 The maximum code size expansion factor when copying basic blocks
10551 instead of jumping. The expansion is relative to a jump instruction.
10553 @item max-goto-duplication-insns
10554 The maximum number of instructions to duplicate to a block that jumps
10555 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10556 passes, GCC factors computed gotos early in the compilation process,
10557 and unfactors them as late as possible. Only computed jumps at the
10558 end of a basic blocks with no more than max-goto-duplication-insns are
10561 @item max-delay-slot-insn-search
10562 The maximum number of instructions to consider when looking for an
10563 instruction to fill a delay slot. If more than this arbitrary number of
10564 instructions are searched, the time savings from filling the delay slot
10565 are minimal, so stop searching. Increasing values mean more
10566 aggressive optimization, making the compilation time increase with probably
10567 small improvement in execution time.
10569 @item max-delay-slot-live-search
10570 When trying to fill delay slots, the maximum number of instructions to
10571 consider when searching for a block with valid live register
10572 information. Increasing this arbitrarily chosen value means more
10573 aggressive optimization, increasing the compilation time. This parameter
10574 should be removed when the delay slot code is rewritten to maintain the
10575 control-flow graph.
10577 @item max-gcse-memory
10578 The approximate maximum amount of memory that can be allocated in
10579 order to perform the global common subexpression elimination
10580 optimization. If more memory than specified is required, the
10581 optimization is not done.
10583 @item max-gcse-insertion-ratio
10584 If the ratio of expression insertions to deletions is larger than this value
10585 for any expression, then RTL PRE inserts or removes the expression and thus
10586 leaves partially redundant computations in the instruction stream.
10588 @item max-pending-list-length
10589 The maximum number of pending dependencies scheduling allows
10590 before flushing the current state and starting over. Large functions
10591 with few branches or calls can create excessively large lists which
10592 needlessly consume memory and resources.
10594 @item max-modulo-backtrack-attempts
10595 The maximum number of backtrack attempts the scheduler should make
10596 when modulo scheduling a loop. Larger values can exponentially increase
10599 @item max-inline-insns-single
10600 Several parameters control the tree inliner used in GCC@.
10601 This number sets the maximum number of instructions (counted in GCC's
10602 internal representation) in a single function that the tree inliner
10603 considers for inlining. This only affects functions declared
10604 inline and methods implemented in a class declaration (C++).
10606 @item max-inline-insns-auto
10607 When you use @option{-finline-functions} (included in @option{-O3}),
10608 a lot of functions that would otherwise not be considered for inlining
10609 by the compiler are investigated. To those functions, a different
10610 (more restrictive) limit compared to functions declared inline can
10613 @item inline-min-speedup
10614 When estimated performance improvement of caller + callee runtime exceeds this
10615 threshold (in percent), the function can be inlined regardless of the limit on
10616 @option{--param max-inline-insns-single} and @option{--param
10617 max-inline-insns-auto}.
10619 @item large-function-insns
10620 The limit specifying really large functions. For functions larger than this
10621 limit after inlining, inlining is constrained by
10622 @option{--param large-function-growth}. This parameter is useful primarily
10623 to avoid extreme compilation time caused by non-linear algorithms used by the
10626 @item large-function-growth
10627 Specifies maximal growth of large function caused by inlining in percents.
10628 For example, parameter value 100 limits large function growth to 2.0 times
10631 @item large-unit-insns
10632 The limit specifying large translation unit. Growth caused by inlining of
10633 units larger than this limit is limited by @option{--param inline-unit-growth}.
10634 For small units this might be too tight.
10635 For example, consider a unit consisting of function A
10636 that is inline and B that just calls A three times. If B is small relative to
10637 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10638 large units consisting of small inlineable functions, however, the overall unit
10639 growth limit is needed to avoid exponential explosion of code size. Thus for
10640 smaller units, the size is increased to @option{--param large-unit-insns}
10641 before applying @option{--param inline-unit-growth}.
10643 @item inline-unit-growth
10644 Specifies maximal overall growth of the compilation unit caused by inlining.
10645 For example, parameter value 20 limits unit growth to 1.2 times the original
10646 size. Cold functions (either marked cold via an attribute or by profile
10647 feedback) are not accounted into the unit size.
10649 @item ipcp-unit-growth
10650 Specifies maximal overall growth of the compilation unit caused by
10651 interprocedural constant propagation. For example, parameter value 10 limits
10652 unit growth to 1.1 times the original size.
10654 @item large-stack-frame
10655 The limit specifying large stack frames. While inlining the algorithm is trying
10656 to not grow past this limit too much.
10658 @item large-stack-frame-growth
10659 Specifies maximal growth of large stack frames caused by inlining in percents.
10660 For example, parameter value 1000 limits large stack frame growth to 11 times
10663 @item max-inline-insns-recursive
10664 @itemx max-inline-insns-recursive-auto
10665 Specifies the maximum number of instructions an out-of-line copy of a
10666 self-recursive inline
10667 function can grow into by performing recursive inlining.
10669 @option{--param max-inline-insns-recursive} applies to functions
10671 For functions not declared inline, recursive inlining
10672 happens only when @option{-finline-functions} (included in @option{-O3}) is
10673 enabled; @option{--param max-inline-insns-recursive-auto} applies instead.
10675 @item max-inline-recursive-depth
10676 @itemx max-inline-recursive-depth-auto
10677 Specifies the maximum recursion depth used for recursive inlining.
10679 @option{--param max-inline-recursive-depth} applies to functions
10680 declared inline. For functions not declared inline, recursive inlining
10681 happens only when @option{-finline-functions} (included in @option{-O3}) is
10682 enabled; @option{--param max-inline-recursive-depth-auto} applies instead.
10684 @item min-inline-recursive-probability
10685 Recursive inlining is profitable only for function having deep recursion
10686 in average and can hurt for function having little recursion depth by
10687 increasing the prologue size or complexity of function body to other
10690 When profile feedback is available (see @option{-fprofile-generate}) the actual
10691 recursion depth can be guessed from the probability that function recurses
10692 via a given call expression. This parameter limits inlining only to call
10693 expressions whose probability exceeds the given threshold (in percents).
10695 @item early-inlining-insns
10696 Specify growth that the early inliner can make. In effect it increases
10697 the amount of inlining for code having a large abstraction penalty.
10699 @item max-early-inliner-iterations
10700 Limit of iterations of the early inliner. This basically bounds
10701 the number of nested indirect calls the early inliner can resolve.
10702 Deeper chains are still handled by late inlining.
10704 @item comdat-sharing-probability
10705 Probability (in percent) that C++ inline function with comdat visibility
10706 are shared across multiple compilation units.
10708 @item profile-func-internal-id
10709 A parameter to control whether to use function internal id in profile
10710 database lookup. If the value is 0, the compiler uses an id that
10711 is based on function assembler name and filename, which makes old profile
10712 data more tolerant to source changes such as function reordering etc.
10714 @item min-vect-loop-bound
10715 The minimum number of iterations under which loops are not vectorized
10716 when @option{-ftree-vectorize} is used. The number of iterations after
10717 vectorization needs to be greater than the value specified by this option
10718 to allow vectorization.
10720 @item gcse-cost-distance-ratio
10721 Scaling factor in calculation of maximum distance an expression
10722 can be moved by GCSE optimizations. This is currently supported only in the
10723 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10724 is with simple expressions, i.e., the expressions that have cost
10725 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10726 hoisting of simple expressions.
10728 @item gcse-unrestricted-cost
10729 Cost, roughly measured as the cost of a single typical machine
10730 instruction, at which GCSE optimizations do not constrain
10731 the distance an expression can travel. This is currently
10732 supported only in the code hoisting pass. The lesser the cost,
10733 the more aggressive code hoisting is. Specifying 0
10734 allows all expressions to travel unrestricted distances.
10736 @item max-hoist-depth
10737 The depth of search in the dominator tree for expressions to hoist.
10738 This is used to avoid quadratic behavior in hoisting algorithm.
10739 The value of 0 does not limit on the search, but may slow down compilation
10742 @item max-tail-merge-comparisons
10743 The maximum amount of similar bbs to compare a bb with. This is used to
10744 avoid quadratic behavior in tree tail merging.
10746 @item max-tail-merge-iterations
10747 The maximum amount of iterations of the pass over the function. This is used to
10748 limit compilation time in tree tail merging.
10750 @item store-merging-allow-unaligned
10751 Allow the store merging pass to introduce unaligned stores if it is legal to
10754 @item max-stores-to-merge
10755 The maximum number of stores to attempt to merge into wider stores in the store
10758 @item max-unrolled-insns
10759 The maximum number of instructions that a loop may have to be unrolled.
10760 If a loop is unrolled, this parameter also determines how many times
10761 the loop code is unrolled.
10763 @item max-average-unrolled-insns
10764 The maximum number of instructions biased by probabilities of their execution
10765 that a loop may have to be unrolled. If a loop is unrolled,
10766 this parameter also determines how many times the loop code is unrolled.
10768 @item max-unroll-times
10769 The maximum number of unrollings of a single loop.
10771 @item max-peeled-insns
10772 The maximum number of instructions that a loop may have to be peeled.
10773 If a loop is peeled, this parameter also determines how many times
10774 the loop code is peeled.
10776 @item max-peel-times
10777 The maximum number of peelings of a single loop.
10779 @item max-peel-branches
10780 The maximum number of branches on the hot path through the peeled sequence.
10782 @item max-completely-peeled-insns
10783 The maximum number of insns of a completely peeled loop.
10785 @item max-completely-peel-times
10786 The maximum number of iterations of a loop to be suitable for complete peeling.
10788 @item max-completely-peel-loop-nest-depth
10789 The maximum depth of a loop nest suitable for complete peeling.
10791 @item max-unswitch-insns
10792 The maximum number of insns of an unswitched loop.
10794 @item max-unswitch-level
10795 The maximum number of branches unswitched in a single loop.
10797 @item lim-expensive
10798 The minimum cost of an expensive expression in the loop invariant motion.
10800 @item iv-consider-all-candidates-bound
10801 Bound on number of candidates for induction variables, below which
10802 all candidates are considered for each use in induction variable
10803 optimizations. If there are more candidates than this,
10804 only the most relevant ones are considered to avoid quadratic time complexity.
10806 @item iv-max-considered-uses
10807 The induction variable optimizations give up on loops that contain more
10808 induction variable uses.
10810 @item iv-always-prune-cand-set-bound
10811 If the number of candidates in the set is smaller than this value,
10812 always try to remove unnecessary ivs from the set
10813 when adding a new one.
10815 @item avg-loop-niter
10816 Average number of iterations of a loop.
10818 @item dse-max-object-size
10819 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
10820 Larger values may result in larger compilation times.
10822 @item dse-max-alias-queries-per-store
10823 Maximum number of queries into the alias oracle per store.
10824 Larger values result in larger compilation times and may result in more
10825 removed dead stores.
10827 @item scev-max-expr-size
10828 Bound on size of expressions used in the scalar evolutions analyzer.
10829 Large expressions slow the analyzer.
10831 @item scev-max-expr-complexity
10832 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10833 Complex expressions slow the analyzer.
10835 @item max-tree-if-conversion-phi-args
10836 Maximum number of arguments in a PHI supported by TREE if conversion
10837 unless the loop is marked with simd pragma.
10839 @item vect-max-version-for-alignment-checks
10840 The maximum number of run-time checks that can be performed when
10841 doing loop versioning for alignment in the vectorizer.
10843 @item vect-max-version-for-alias-checks
10844 The maximum number of run-time checks that can be performed when
10845 doing loop versioning for alias in the vectorizer.
10847 @item vect-max-peeling-for-alignment
10848 The maximum number of loop peels to enhance access alignment
10849 for vectorizer. Value -1 means no limit.
10851 @item max-iterations-to-track
10852 The maximum number of iterations of a loop the brute-force algorithm
10853 for analysis of the number of iterations of the loop tries to evaluate.
10855 @item hot-bb-count-ws-permille
10856 A basic block profile count is considered hot if it contributes to
10857 the given permillage (i.e.@: 0...1000) of the entire profiled execution.
10859 @item hot-bb-frequency-fraction
10860 Select fraction of the entry block frequency of executions of basic block in
10861 function given basic block needs to have to be considered hot.
10863 @item max-predicted-iterations
10864 The maximum number of loop iterations we predict statically. This is useful
10865 in cases where a function contains a single loop with known bound and
10866 another loop with unknown bound.
10867 The known number of iterations is predicted correctly, while
10868 the unknown number of iterations average to roughly 10. This means that the
10869 loop without bounds appears artificially cold relative to the other one.
10871 @item builtin-expect-probability
10872 Control the probability of the expression having the specified value. This
10873 parameter takes a percentage (i.e.@: 0 ... 100) as input.
10875 @item builtin-string-cmp-inline-length
10876 The maximum length of a constant string for a builtin string cmp call
10877 eligible for inlining.
10879 @item align-threshold
10881 Select fraction of the maximal frequency of executions of a basic block in
10882 a function to align the basic block.
10884 @item align-loop-iterations
10886 A loop expected to iterate at least the selected number of iterations is
10889 @item tracer-dynamic-coverage
10890 @itemx tracer-dynamic-coverage-feedback
10892 This value is used to limit superblock formation once the given percentage of
10893 executed instructions is covered. This limits unnecessary code size
10896 The @option{tracer-dynamic-coverage-feedback} parameter
10897 is used only when profile
10898 feedback is available. The real profiles (as opposed to statically estimated
10899 ones) are much less balanced allowing the threshold to be larger value.
10901 @item tracer-max-code-growth
10902 Stop tail duplication once code growth has reached given percentage. This is
10903 a rather artificial limit, as most of the duplicates are eliminated later in
10904 cross jumping, so it may be set to much higher values than is the desired code
10907 @item tracer-min-branch-ratio
10909 Stop reverse growth when the reverse probability of best edge is less than this
10910 threshold (in percent).
10912 @item tracer-min-branch-probability
10913 @itemx tracer-min-branch-probability-feedback
10915 Stop forward growth if the best edge has probability lower than this
10918 Similarly to @option{tracer-dynamic-coverage} two parameters are
10919 provided. @option{tracer-min-branch-probability-feedback} is used for
10920 compilation with profile feedback and @option{tracer-min-branch-probability}
10921 compilation without. The value for compilation with profile feedback
10922 needs to be more conservative (higher) in order to make tracer
10925 @item stack-clash-protection-guard-size
10926 Specify the size of the operating system provided stack guard as
10927 2 raised to @var{num} bytes. Higher values may reduce the
10928 number of explicit probes, but a value larger than the operating system
10929 provided guard will leave code vulnerable to stack clash style attacks.
10931 @item stack-clash-protection-probe-interval
10932 Stack clash protection involves probing stack space as it is allocated. This
10933 param controls the maximum distance between probes into the stack as 2 raised
10934 to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value
10935 larger than the operating system provided guard will leave code vulnerable to
10936 stack clash style attacks.
10938 @item max-cse-path-length
10940 The maximum number of basic blocks on path that CSE considers.
10942 @item max-cse-insns
10943 The maximum number of instructions CSE processes before flushing.
10945 @item ggc-min-expand
10947 GCC uses a garbage collector to manage its own memory allocation. This
10948 parameter specifies the minimum percentage by which the garbage
10949 collector's heap should be allowed to expand between collections.
10950 Tuning this may improve compilation speed; it has no effect on code
10953 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10954 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10955 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10956 GCC is not able to calculate RAM on a particular platform, the lower
10957 bound of 30% is used. Setting this parameter and
10958 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10959 every opportunity. This is extremely slow, but can be useful for
10962 @item ggc-min-heapsize
10964 Minimum size of the garbage collector's heap before it begins bothering
10965 to collect garbage. The first collection occurs after the heap expands
10966 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10967 tuning this may improve compilation speed, and has no effect on code
10970 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10971 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10972 with a lower bound of 4096 (four megabytes) and an upper bound of
10973 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10974 particular platform, the lower bound is used. Setting this parameter
10975 very large effectively disables garbage collection. Setting this
10976 parameter and @option{ggc-min-expand} to zero causes a full collection
10977 to occur at every opportunity.
10979 @item max-reload-search-insns
10980 The maximum number of instruction reload should look backward for equivalent
10981 register. Increasing values mean more aggressive optimization, making the
10982 compilation time increase with probably slightly better performance.
10984 @item max-cselib-memory-locations
10985 The maximum number of memory locations cselib should take into account.
10986 Increasing values mean more aggressive optimization, making the compilation time
10987 increase with probably slightly better performance.
10989 @item max-sched-ready-insns
10990 The maximum number of instructions ready to be issued the scheduler should
10991 consider at any given time during the first scheduling pass. Increasing
10992 values mean more thorough searches, making the compilation time increase
10993 with probably little benefit.
10995 @item max-sched-region-blocks
10996 The maximum number of blocks in a region to be considered for
10997 interblock scheduling.
10999 @item max-pipeline-region-blocks
11000 The maximum number of blocks in a region to be considered for
11001 pipelining in the selective scheduler.
11003 @item max-sched-region-insns
11004 The maximum number of insns in a region to be considered for
11005 interblock scheduling.
11007 @item max-pipeline-region-insns
11008 The maximum number of insns in a region to be considered for
11009 pipelining in the selective scheduler.
11011 @item min-spec-prob
11012 The minimum probability (in percents) of reaching a source block
11013 for interblock speculative scheduling.
11015 @item max-sched-extend-regions-iters
11016 The maximum number of iterations through CFG to extend regions.
11017 A value of 0 disables region extensions.
11019 @item max-sched-insn-conflict-delay
11020 The maximum conflict delay for an insn to be considered for speculative motion.
11022 @item sched-spec-prob-cutoff
11023 The minimal probability of speculation success (in percents), so that
11024 speculative insns are scheduled.
11026 @item sched-state-edge-prob-cutoff
11027 The minimum probability an edge must have for the scheduler to save its
11030 @item sched-mem-true-dep-cost
11031 Minimal distance (in CPU cycles) between store and load targeting same
11034 @item selsched-max-lookahead
11035 The maximum size of the lookahead window of selective scheduling. It is a
11036 depth of search for available instructions.
11038 @item selsched-max-sched-times
11039 The maximum number of times that an instruction is scheduled during
11040 selective scheduling. This is the limit on the number of iterations
11041 through which the instruction may be pipelined.
11043 @item selsched-insns-to-rename
11044 The maximum number of best instructions in the ready list that are considered
11045 for renaming in the selective scheduler.
11048 The minimum value of stage count that swing modulo scheduler
11051 @item max-last-value-rtl
11052 The maximum size measured as number of RTLs that can be recorded in an expression
11053 in combiner for a pseudo register as last known value of that register.
11055 @item max-combine-insns
11056 The maximum number of instructions the RTL combiner tries to combine.
11058 @item integer-share-limit
11059 Small integer constants can use a shared data structure, reducing the
11060 compiler's memory usage and increasing its speed. This sets the maximum
11061 value of a shared integer constant.
11063 @item ssp-buffer-size
11064 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
11065 protection when @option{-fstack-protection} is used.
11067 @item min-size-for-stack-sharing
11068 The minimum size of variables taking part in stack slot sharing when not
11071 @item max-jump-thread-duplication-stmts
11072 Maximum number of statements allowed in a block that needs to be
11073 duplicated when threading jumps.
11075 @item max-fields-for-field-sensitive
11076 Maximum number of fields in a structure treated in
11077 a field sensitive manner during pointer analysis.
11079 @item prefetch-latency
11080 Estimate on average number of instructions that are executed before
11081 prefetch finishes. The distance prefetched ahead is proportional
11082 to this constant. Increasing this number may also lead to less
11083 streams being prefetched (see @option{simultaneous-prefetches}).
11085 @item simultaneous-prefetches
11086 Maximum number of prefetches that can run at the same time.
11088 @item l1-cache-line-size
11089 The size of cache line in L1 cache, in bytes.
11091 @item l1-cache-size
11092 The size of L1 cache, in kilobytes.
11094 @item l2-cache-size
11095 The size of L2 cache, in kilobytes.
11097 @item prefetch-dynamic-strides
11098 Whether the loop array prefetch pass should issue software prefetch hints
11099 for strides that are non-constant. In some cases this may be
11100 beneficial, though the fact the stride is non-constant may make it
11101 hard to predict when there is clear benefit to issuing these hints.
11103 Set to 1 if the prefetch hints should be issued for non-constant
11104 strides. Set to 0 if prefetch hints should be issued only for strides that
11105 are known to be constant and below @option{prefetch-minimum-stride}.
11107 @item prefetch-minimum-stride
11108 Minimum constant stride, in bytes, to start using prefetch hints for. If
11109 the stride is less than this threshold, prefetch hints will not be issued.
11111 This setting is useful for processors that have hardware prefetchers, in
11112 which case there may be conflicts between the hardware prefetchers and
11113 the software prefetchers. If the hardware prefetchers have a maximum
11114 stride they can handle, it should be used here to improve the use of
11115 software prefetchers.
11117 A value of -1 means we don't have a threshold and therefore
11118 prefetch hints can be issued for any constant stride.
11120 This setting is only useful for strides that are known and constant.
11122 @item loop-interchange-max-num-stmts
11123 The maximum number of stmts in a loop to be interchanged.
11125 @item loop-interchange-stride-ratio
11126 The minimum ratio between stride of two loops for interchange to be profitable.
11128 @item min-insn-to-prefetch-ratio
11129 The minimum ratio between the number of instructions and the
11130 number of prefetches to enable prefetching in a loop.
11132 @item prefetch-min-insn-to-mem-ratio
11133 The minimum ratio between the number of instructions and the
11134 number of memory references to enable prefetching in a loop.
11136 @item use-canonical-types
11137 Whether the compiler should use the ``canonical'' type system.
11138 Should always be 1, which uses a more efficient internal
11139 mechanism for comparing types in C++ and Objective-C++. However, if
11140 bugs in the canonical type system are causing compilation failures,
11141 set this value to 0 to disable canonical types.
11143 @item switch-conversion-max-branch-ratio
11144 Switch initialization conversion refuses to create arrays that are
11145 bigger than @option{switch-conversion-max-branch-ratio} times the number of
11146 branches in the switch.
11148 @item max-partial-antic-length
11149 Maximum length of the partial antic set computed during the tree
11150 partial redundancy elimination optimization (@option{-ftree-pre}) when
11151 optimizing at @option{-O3} and above. For some sorts of source code
11152 the enhanced partial redundancy elimination optimization can run away,
11153 consuming all of the memory available on the host machine. This
11154 parameter sets a limit on the length of the sets that are computed,
11155 which prevents the runaway behavior. Setting a value of 0 for
11156 this parameter allows an unlimited set length.
11158 @item rpo-vn-max-loop-depth
11159 Maximum loop depth that is value-numbered optimistically.
11160 When the limit hits the innermost
11161 @var{rpo-vn-max-loop-depth} loops and the outermost loop in the
11162 loop nest are value-numbered optimistically and the remaining ones not.
11164 @item sccvn-max-alias-queries-per-access
11165 Maximum number of alias-oracle queries we perform when looking for
11166 redundancies for loads and stores. If this limit is hit the search
11167 is aborted and the load or store is not considered redundant. The
11168 number of queries is algorithmically limited to the number of
11169 stores on all paths from the load to the function entry.
11171 @item ira-max-loops-num
11172 IRA uses regional register allocation by default. If a function
11173 contains more loops than the number given by this parameter, only at most
11174 the given number of the most frequently-executed loops form regions
11175 for regional register allocation.
11177 @item ira-max-conflict-table-size
11178 Although IRA uses a sophisticated algorithm to compress the conflict
11179 table, the table can still require excessive amounts of memory for
11180 huge functions. If the conflict table for a function could be more
11181 than the size in MB given by this parameter, the register allocator
11182 instead uses a faster, simpler, and lower-quality
11183 algorithm that does not require building a pseudo-register conflict table.
11185 @item ira-loop-reserved-regs
11186 IRA can be used to evaluate more accurate register pressure in loops
11187 for decisions to move loop invariants (see @option{-O3}). The number
11188 of available registers reserved for some other purposes is given
11189 by this parameter. Default of the parameter
11190 is the best found from numerous experiments.
11192 @item lra-inheritance-ebb-probability-cutoff
11193 LRA tries to reuse values reloaded in registers in subsequent insns.
11194 This optimization is called inheritance. EBB is used as a region to
11195 do this optimization. The parameter defines a minimal fall-through
11196 edge probability in percentage used to add BB to inheritance EBB in
11197 LRA. The default value was chosen
11198 from numerous runs of SPEC2000 on x86-64.
11200 @item loop-invariant-max-bbs-in-loop
11201 Loop invariant motion can be very expensive, both in compilation time and
11202 in amount of needed compile-time memory, with very large loops. Loops
11203 with more basic blocks than this parameter won't have loop invariant
11204 motion optimization performed on them.
11206 @item loop-max-datarefs-for-datadeps
11207 Building data dependencies is expensive for very large loops. This
11208 parameter limits the number of data references in loops that are
11209 considered for data dependence analysis. These large loops are no
11210 handled by the optimizations using loop data dependencies.
11212 @item max-vartrack-size
11213 Sets a maximum number of hash table slots to use during variable
11214 tracking dataflow analysis of any function. If this limit is exceeded
11215 with variable tracking at assignments enabled, analysis for that
11216 function is retried without it, after removing all debug insns from
11217 the function. If the limit is exceeded even without debug insns, var
11218 tracking analysis is completely disabled for the function. Setting
11219 the parameter to zero makes it unlimited.
11221 @item max-vartrack-expr-depth
11222 Sets a maximum number of recursion levels when attempting to map
11223 variable names or debug temporaries to value expressions. This trades
11224 compilation time for more complete debug information. If this is set too
11225 low, value expressions that are available and could be represented in
11226 debug information may end up not being used; setting this higher may
11227 enable the compiler to find more complex debug expressions, but compile
11228 time and memory use may grow.
11230 @item max-debug-marker-count
11231 Sets a threshold on the number of debug markers (e.g.@: begin stmt
11232 markers) to avoid complexity explosion at inlining or expanding to RTL.
11233 If a function has more such gimple stmts than the set limit, such stmts
11234 will be dropped from the inlined copy of a function, and from its RTL
11237 @item min-nondebug-insn-uid
11238 Use uids starting at this parameter for nondebug insns. The range below
11239 the parameter is reserved exclusively for debug insns created by
11240 @option{-fvar-tracking-assignments}, but debug insns may get
11241 (non-overlapping) uids above it if the reserved range is exhausted.
11243 @item ipa-sra-ptr-growth-factor
11244 IPA-SRA replaces a pointer to an aggregate with one or more new
11245 parameters only when their cumulative size is less or equal to
11246 @option{ipa-sra-ptr-growth-factor} times the size of the original
11249 @item sra-max-scalarization-size-Ospeed
11250 @itemx sra-max-scalarization-size-Osize
11251 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11252 replace scalar parts of aggregates with uses of independent scalar
11253 variables. These parameters control the maximum size, in storage units,
11254 of aggregate which is considered for replacement when compiling for
11256 (@option{sra-max-scalarization-size-Ospeed}) or size
11257 (@option{sra-max-scalarization-size-Osize}) respectively.
11259 @item tm-max-aggregate-size
11260 When making copies of thread-local variables in a transaction, this
11261 parameter specifies the size in bytes after which variables are
11262 saved with the logging functions as opposed to save/restore code
11263 sequence pairs. This option only applies when using
11266 @item graphite-max-nb-scop-params
11267 To avoid exponential effects in the Graphite loop transforms, the
11268 number of parameters in a Static Control Part (SCoP) is bounded.
11269 A value of zero can be used to lift
11270 the bound. A variable whose value is unknown at compilation time and
11271 defined outside a SCoP is a parameter of the SCoP.
11273 @item loop-block-tile-size
11274 Loop blocking or strip mining transforms, enabled with
11275 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11276 loop in the loop nest by a given number of iterations. The strip
11277 length can be changed using the @option{loop-block-tile-size}
11280 @item ipa-cp-value-list-size
11281 IPA-CP attempts to track all possible values and types passed to a function's
11282 parameter in order to propagate them and perform devirtualization.
11283 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11284 stores per one formal parameter of a function.
11286 @item ipa-cp-eval-threshold
11287 IPA-CP calculates its own score of cloning profitability heuristics
11288 and performs those cloning opportunities with scores that exceed
11289 @option{ipa-cp-eval-threshold}.
11291 @item ipa-cp-recursion-penalty
11292 Percentage penalty the recursive functions will receive when they
11293 are evaluated for cloning.
11295 @item ipa-cp-single-call-penalty
11296 Percentage penalty functions containing a single call to another
11297 function will receive when they are evaluated for cloning.
11299 @item ipa-max-agg-items
11300 IPA-CP is also capable to propagate a number of scalar values passed
11301 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11302 number of such values per one parameter.
11304 @item ipa-cp-loop-hint-bonus
11305 When IPA-CP determines that a cloning candidate would make the number
11306 of iterations of a loop known, it adds a bonus of
11307 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11310 @item ipa-cp-array-index-hint-bonus
11311 When IPA-CP determines that a cloning candidate would make the index of
11312 an array access known, it adds a bonus of
11313 @option{ipa-cp-array-index-hint-bonus} to the profitability
11314 score of the candidate.
11316 @item ipa-max-aa-steps
11317 During its analysis of function bodies, IPA-CP employs alias analysis
11318 in order to track values pointed to by function parameters. In order
11319 not spend too much time analyzing huge functions, it gives up and
11320 consider all memory clobbered after examining
11321 @option{ipa-max-aa-steps} statements modifying memory.
11323 @item lto-partitions
11324 Specify desired number of partitions produced during WHOPR compilation.
11325 The number of partitions should exceed the number of CPUs used for compilation.
11327 @item lto-min-partition
11328 Size of minimal partition for WHOPR (in estimated instructions).
11329 This prevents expenses of splitting very small programs into too many
11332 @item lto-max-partition
11333 Size of max partition for WHOPR (in estimated instructions).
11334 to provide an upper bound for individual size of partition.
11335 Meant to be used only with balanced partitioning.
11337 @item cxx-max-namespaces-for-diagnostic-help
11338 The maximum number of namespaces to consult for suggestions when C++
11339 name lookup fails for an identifier.
11341 @item sink-frequency-threshold
11342 The maximum relative execution frequency (in percents) of the target block
11343 relative to a statement's original block to allow statement sinking of a
11344 statement. Larger numbers result in more aggressive statement sinking.
11345 A small positive adjustment is applied for
11346 statements with memory operands as those are even more profitable so sink.
11348 @item max-stores-to-sink
11349 The maximum number of conditional store pairs that can be sunk. Set to 0
11350 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11351 (@option{-ftree-loop-if-convert}) is disabled.
11353 @item allow-store-data-races
11354 Allow optimizers to introduce new data races on stores.
11355 Set to 1 to allow, otherwise to 0.
11357 @item case-values-threshold
11358 The smallest number of different values for which it is best to use a
11359 jump-table instead of a tree of conditional branches. If the value is
11360 0, use the default for the machine.
11362 @item tree-reassoc-width
11363 Set the maximum number of instructions executed in parallel in
11364 reassociated tree. This parameter overrides target dependent
11365 heuristics used by default if has non zero value.
11367 @item sched-pressure-algorithm
11368 Choose between the two available implementations of
11369 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11370 and is the more likely to prevent instructions from being reordered.
11371 Algorithm 2 was designed to be a compromise between the relatively
11372 conservative approach taken by algorithm 1 and the rather aggressive
11373 approach taken by the default scheduler. It relies more heavily on
11374 having a regular register file and accurate register pressure classes.
11375 See @file{haifa-sched.c} in the GCC sources for more details.
11377 The default choice depends on the target.
11379 @item max-slsr-cand-scan
11380 Set the maximum number of existing candidates that are considered when
11381 seeking a basis for a new straight-line strength reduction candidate.
11384 Enable buffer overflow detection for global objects. This kind
11385 of protection is enabled by default if you are using
11386 @option{-fsanitize=address} option.
11387 To disable global objects protection use @option{--param asan-globals=0}.
11390 Enable buffer overflow detection for stack objects. This kind of
11391 protection is enabled by default when using @option{-fsanitize=address}.
11392 To disable stack protection use @option{--param asan-stack=0} option.
11394 @item asan-instrument-reads
11395 Enable buffer overflow detection for memory reads. This kind of
11396 protection is enabled by default when using @option{-fsanitize=address}.
11397 To disable memory reads protection use
11398 @option{--param asan-instrument-reads=0}.
11400 @item asan-instrument-writes
11401 Enable buffer overflow detection for memory writes. This kind of
11402 protection is enabled by default when using @option{-fsanitize=address}.
11403 To disable memory writes protection use
11404 @option{--param asan-instrument-writes=0} option.
11406 @item asan-memintrin
11407 Enable detection for built-in functions. This kind of protection
11408 is enabled by default when using @option{-fsanitize=address}.
11409 To disable built-in functions protection use
11410 @option{--param asan-memintrin=0}.
11412 @item asan-use-after-return
11413 Enable detection of use-after-return. This kind of protection
11414 is enabled by default when using the @option{-fsanitize=address} option.
11415 To disable it use @option{--param asan-use-after-return=0}.
11417 Note: By default the check is disabled at run time. To enable it,
11418 add @code{detect_stack_use_after_return=1} to the environment variable
11419 @env{ASAN_OPTIONS}.
11421 @item asan-instrumentation-with-call-threshold
11422 If number of memory accesses in function being instrumented
11423 is greater or equal to this number, use callbacks instead of inline checks.
11424 E.g. to disable inline code use
11425 @option{--param asan-instrumentation-with-call-threshold=0}.
11427 @item use-after-scope-direct-emission-threshold
11428 If the size of a local variable in bytes is smaller or equal to this
11429 number, directly poison (or unpoison) shadow memory instead of using
11430 run-time callbacks.
11432 @item max-fsm-thread-path-insns
11433 Maximum number of instructions to copy when duplicating blocks on a
11434 finite state automaton jump thread path.
11436 @item max-fsm-thread-length
11437 Maximum number of basic blocks on a finite state automaton jump thread
11440 @item max-fsm-thread-paths
11441 Maximum number of new jump thread paths to create for a finite state
11444 @item parloops-chunk-size
11445 Chunk size of omp schedule for loops parallelized by parloops.
11447 @item parloops-schedule
11448 Schedule type of omp schedule for loops parallelized by parloops (static,
11449 dynamic, guided, auto, runtime).
11451 @item parloops-min-per-thread
11452 The minimum number of iterations per thread of an innermost parallelized
11453 loop for which the parallelized variant is preferred over the single threaded
11454 one. Note that for a parallelized loop nest the
11455 minimum number of iterations of the outermost loop per thread is two.
11457 @item max-ssa-name-query-depth
11458 Maximum depth of recursion when querying properties of SSA names in things
11459 like fold routines. One level of recursion corresponds to following a
11462 @item hsa-gen-debug-stores
11463 Enable emission of special debug stores within HSA kernels which are
11464 then read and reported by libgomp plugin. Generation of these stores
11465 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
11468 @item max-speculative-devirt-maydefs
11469 The maximum number of may-defs we analyze when looking for a must-def
11470 specifying the dynamic type of an object that invokes a virtual call
11471 we may be able to devirtualize speculatively.
11473 @item max-vrp-switch-assertions
11474 The maximum number of assertions to add along the default edge of a switch
11475 statement during VRP.
11477 @item unroll-jam-min-percent
11478 The minimum percentage of memory references that must be optimized
11479 away for the unroll-and-jam transformation to be considered profitable.
11481 @item unroll-jam-max-unroll
11482 The maximum number of times the outer loop should be unrolled by
11483 the unroll-and-jam transformation.
11485 @item max-rtl-if-conversion-unpredictable-cost
11486 Maximum permissible cost for the sequence that would be generated
11487 by the RTL if-conversion pass for a branch that is considered unpredictable.
11489 @item max-variable-expansions-in-unroller
11490 If @option{-fvariable-expansion-in-unroller} is used, the maximum number
11491 of times that an individual variable will be expanded during loop unrolling.
11493 @item tracer-min-branch-probability-feedback
11494 Stop forward growth if the probability of best edge is less than
11495 this threshold (in percent). Used when profile feedback is available.
11497 @item partial-inlining-entry-probability
11498 Maximum probability of the entry BB of split region
11499 (in percent relative to entry BB of the function)
11500 to make partial inlining happen.
11502 @item max-tracked-strlens
11503 Maximum number of strings for which strlen optimization pass will
11504 track string lengths.
11506 @item gcse-after-reload-partial-fraction
11507 The threshold ratio for performing partial redundancy
11508 elimination after reload.
11510 @item gcse-after-reload-critical-fraction
11511 The threshold ratio of critical edges execution count that
11512 permit performing redundancy elimination after reload.
11514 @item max-loop-header-insns
11515 The maximum number of insns in loop header duplicated
11516 by the copy loop headers pass.
11518 @item vect-epilogues-nomask
11519 Enable loop epilogue vectorization using smaller vector size.
11521 @item slp-max-insns-in-bb
11522 Maximum number of instructions in basic block to be
11523 considered for SLP vectorization.
11525 @item avoid-fma-max-bits
11526 Maximum number of bits for which we avoid creating FMAs.
11528 @item sms-loop-average-count-threshold
11529 A threshold on the average loop count considered by the swing modulo scheduler.
11531 @item sms-dfa-history
11532 The number of cycles the swing modulo scheduler considers when checking
11533 conflicts using DFA.
11535 @item hot-bb-count-fraction
11536 Select fraction of the maximal count of repetitions of basic block
11537 in program given basic block needs
11538 to have to be considered hot (used in non-LTO mode)
11540 @item max-inline-insns-recursive-auto
11541 The maximum number of instructions non-inline function
11542 can grow to via recursive inlining.
11544 @item graphite-allow-codegen-errors
11545 Whether codegen errors should be ICEs when @option{-fchecking}.
11547 @item sms-max-ii-factor
11548 A factor for tuning the upper bound that swing modulo scheduler
11549 uses for scheduling a loop.
11551 @item lra-max-considered-reload-pseudos
11552 The max number of reload pseudos which are considered during
11553 spilling a non-reload pseudo.
11555 @item max-pow-sqrt-depth
11556 Maximum depth of sqrt chains to use when synthesizing exponentiation
11557 by a real constant.
11559 @item max-dse-active-local-stores
11560 Maximum number of active local stores in RTL dead store elimination.
11562 @item asan-instrument-allocas
11563 Enable asan allocas/VLAs protection.
11565 @item max-iterations-computation-cost
11566 Bound on the cost of an expression to compute the number of iterations.
11568 @item max-isl-operations
11569 Maximum number of isl operations, 0 means unlimited.
11571 @item graphite-max-arrays-per-scop
11572 Maximum number of arrays per scop.
11574 @item max-vartrack-reverse-op-size
11575 Max. size of loc list for which reverse ops should be added.
11577 @item unlikely-bb-count-fraction
11578 The minimum fraction of profile runs a given basic block execution count
11579 must be not to be considered unlikely.
11581 @item tracer-dynamic-coverage-feedback
11582 The percentage of function, weighted by execution frequency,
11583 that must be covered by trace formation.
11584 Used when profile feedback is available.
11586 @item max-inline-recursive-depth-auto
11587 The maximum depth of recursive inlining for non-inline functions.
11589 @item fsm-scale-path-stmts
11590 Scale factor to apply to the number of statements in a threading path
11591 when comparing to the number of (scaled) blocks.
11593 @item fsm-maximum-phi-arguments
11594 Maximum number of arguments a PHI may have before the FSM threader
11595 will not try to thread through its block.
11597 @item uninit-control-dep-attempts
11598 Maximum number of nested calls to search for control dependencies
11599 during uninitialized variable analysis.
11601 @item indir-call-topn-profile
11602 Track top N target addresses in indirect-call profile.
11604 @item max-once-peeled-insns
11605 The maximum number of insns of a peeled loop that rolls only once.
11607 @item sra-max-scalarization-size-Osize
11608 Maximum size, in storage units, of an aggregate
11609 which should be considered for scalarization when compiling for size.
11611 @item fsm-scale-path-blocks
11612 Scale factor to apply to the number of blocks in a threading path
11613 when comparing to the number of (scaled) statements.
11615 @item sched-autopref-queue-depth
11616 Hardware autoprefetcher scheduler model control flag.
11617 Number of lookahead cycles the model looks into; at '
11618 ' only enable instruction sorting heuristic.
11624 @node Instrumentation Options
11625 @section Program Instrumentation Options
11626 @cindex instrumentation options
11627 @cindex program instrumentation options
11628 @cindex run-time error checking options
11629 @cindex profiling options
11630 @cindex options, program instrumentation
11631 @cindex options, run-time error checking
11632 @cindex options, profiling
11634 GCC supports a number of command-line options that control adding
11635 run-time instrumentation to the code it normally generates.
11636 For example, one purpose of instrumentation is collect profiling
11637 statistics for use in finding program hot spots, code coverage
11638 analysis, or profile-guided optimizations.
11639 Another class of program instrumentation is adding run-time checking
11640 to detect programming errors like invalid pointer
11641 dereferences or out-of-bounds array accesses, as well as deliberately
11642 hostile attacks such as stack smashing or C++ vtable hijacking.
11643 There is also a general hook which can be used to implement other
11644 forms of tracing or function-level instrumentation for debug or
11645 program analysis purposes.
11648 @cindex @command{prof}
11651 Generate extra code to write profile information suitable for the
11652 analysis program @command{prof}. You must use this option when compiling
11653 the source files you want data about, and you must also use it when
11656 @cindex @command{gprof}
11659 Generate extra code to write profile information suitable for the
11660 analysis program @command{gprof}. You must use this option when compiling
11661 the source files you want data about, and you must also use it when
11664 @item -fprofile-arcs
11665 @opindex fprofile-arcs
11666 Add code so that program flow @dfn{arcs} are instrumented. During
11667 execution the program records how many times each branch and call is
11668 executed and how many times it is taken or returns. On targets that support
11669 constructors with priority support, profiling properly handles constructors,
11670 destructors and C++ constructors (and destructors) of classes which are used
11671 as a type of a global variable.
11674 program exits it saves this data to a file called
11675 @file{@var{auxname}.gcda} for each source file. The data may be used for
11676 profile-directed optimizations (@option{-fbranch-probabilities}), or for
11677 test coverage analysis (@option{-ftest-coverage}). Each object file's
11678 @var{auxname} is generated from the name of the output file, if
11679 explicitly specified and it is not the final executable, otherwise it is
11680 the basename of the source file. In both cases any suffix is removed
11681 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
11682 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
11683 @xref{Cross-profiling}.
11685 @cindex @command{gcov}
11689 This option is used to compile and link code instrumented for coverage
11690 analysis. The option is a synonym for @option{-fprofile-arcs}
11691 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
11692 linking). See the documentation for those options for more details.
11697 Compile the source files with @option{-fprofile-arcs} plus optimization
11698 and code generation options. For test coverage analysis, use the
11699 additional @option{-ftest-coverage} option. You do not need to profile
11700 every source file in a program.
11703 Compile the source files additionally with @option{-fprofile-abs-path}
11704 to create absolute path names in the @file{.gcno} files. This allows
11705 @command{gcov} to find the correct sources in projects where compilations
11706 occur with different working directories.
11709 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
11710 (the latter implies the former).
11713 Run the program on a representative workload to generate the arc profile
11714 information. This may be repeated any number of times. You can run
11715 concurrent instances of your program, and provided that the file system
11716 supports locking, the data files will be correctly updated. Unless
11717 a strict ISO C dialect option is in effect, @code{fork} calls are
11718 detected and correctly handled without double counting.
11721 For profile-directed optimizations, compile the source files again with
11722 the same optimization and code generation options plus
11723 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
11724 Control Optimization}).
11727 For test coverage analysis, use @command{gcov} to produce human readable
11728 information from the @file{.gcno} and @file{.gcda} files. Refer to the
11729 @command{gcov} documentation for further information.
11733 With @option{-fprofile-arcs}, for each function of your program GCC
11734 creates a program flow graph, then finds a spanning tree for the graph.
11735 Only arcs that are not on the spanning tree have to be instrumented: the
11736 compiler adds code to count the number of times that these arcs are
11737 executed. When an arc is the only exit or only entrance to a block, the
11738 instrumentation code can be added to the block; otherwise, a new basic
11739 block must be created to hold the instrumentation code.
11742 @item -ftest-coverage
11743 @opindex ftest-coverage
11744 Produce a notes file that the @command{gcov} code-coverage utility
11745 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
11746 show program coverage. Each source file's note file is called
11747 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
11748 above for a description of @var{auxname} and instructions on how to
11749 generate test coverage data. Coverage data matches the source files
11750 more closely if you do not optimize.
11752 @item -fprofile-abs-path
11753 @opindex fprofile-abs-path
11754 Automatically convert relative source file names to absolute path names
11755 in the @file{.gcno} files. This allows @command{gcov} to find the correct
11756 sources in projects where compilations occur with different working
11759 @item -fprofile-dir=@var{path}
11760 @opindex fprofile-dir
11762 Set the directory to search for the profile data files in to @var{path}.
11763 This option affects only the profile data generated by
11764 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
11765 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
11766 and its related options. Both absolute and relative paths can be used.
11767 By default, GCC uses the current directory as @var{path}, thus the
11768 profile data file appears in the same directory as the object file.
11769 In order to prevent the file name clashing, if the object file name is
11770 not an absolute path, we mangle the absolute path of the
11771 @file{@var{sourcename}.gcda} file and use it as the file name of a
11774 When an executable is run in a massive parallel environment, it is recommended
11775 to save profile to different folders. That can be done with variables
11776 in @var{path} that are exported during run-time:
11784 value of environment variable @var{VAR}
11788 @item -fprofile-generate
11789 @itemx -fprofile-generate=@var{path}
11790 @opindex fprofile-generate
11792 Enable options usually used for instrumenting application to produce
11793 profile useful for later recompilation with profile feedback based
11794 optimization. You must use @option{-fprofile-generate} both when
11795 compiling and when linking your program.
11797 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
11799 If @var{path} is specified, GCC looks at the @var{path} to find
11800 the profile feedback data files. See @option{-fprofile-dir}.
11802 To optimize the program based on the collected profile information, use
11803 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
11805 @item -fprofile-update=@var{method}
11806 @opindex fprofile-update
11808 Alter the update method for an application instrumented for profile
11809 feedback based optimization. The @var{method} argument should be one of
11810 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
11811 The first one is useful for single-threaded applications,
11812 while the second one prevents profile corruption by emitting thread-safe code.
11814 @strong{Warning:} When an application does not properly join all threads
11815 (or creates an detached thread), a profile file can be still corrupted.
11817 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
11818 when supported by a target, or to @samp{single} otherwise. The GCC driver
11819 automatically selects @samp{prefer-atomic} when @option{-pthread}
11820 is present in the command line.
11822 @item -fsanitize=address
11823 @opindex fsanitize=address
11824 Enable AddressSanitizer, a fast memory error detector.
11825 Memory access instructions are instrumented to detect
11826 out-of-bounds and use-after-free bugs.
11827 The option enables @option{-fsanitize-address-use-after-scope}.
11828 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
11829 more details. The run-time behavior can be influenced using the
11830 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
11831 the available options are shown at startup of the instrumented program. See
11832 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
11833 for a list of supported options.
11834 The option cannot be combined with @option{-fsanitize=thread}.
11836 @item -fsanitize=kernel-address
11837 @opindex fsanitize=kernel-address
11838 Enable AddressSanitizer for Linux kernel.
11839 See @uref{https://github.com/google/kasan/wiki} for more details.
11841 @item -fsanitize=pointer-compare
11842 @opindex fsanitize=pointer-compare
11843 Instrument comparison operation (<, <=, >, >=) with pointer operands.
11844 The option must be combined with either @option{-fsanitize=kernel-address} or
11845 @option{-fsanitize=address}
11846 The option cannot be combined with @option{-fsanitize=thread}.
11847 Note: By default the check is disabled at run time. To enable it,
11848 add @code{detect_invalid_pointer_pairs=2} to the environment variable
11849 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
11850 invalid operation only when both pointers are non-null.
11852 @item -fsanitize=pointer-subtract
11853 @opindex fsanitize=pointer-subtract
11854 Instrument subtraction with pointer operands.
11855 The option must be combined with either @option{-fsanitize=kernel-address} or
11856 @option{-fsanitize=address}
11857 The option cannot be combined with @option{-fsanitize=thread}.
11858 Note: By default the check is disabled at run time. To enable it,
11859 add @code{detect_invalid_pointer_pairs=2} to the environment variable
11860 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
11861 invalid operation only when both pointers are non-null.
11863 @item -fsanitize=thread
11864 @opindex fsanitize=thread
11865 Enable ThreadSanitizer, a fast data race detector.
11866 Memory access instructions are instrumented to detect
11867 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
11868 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
11869 environment variable; see
11870 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
11872 The option cannot be combined with @option{-fsanitize=address},
11873 @option{-fsanitize=leak}.
11875 Note that sanitized atomic builtins cannot throw exceptions when
11876 operating on invalid memory addresses with non-call exceptions
11877 (@option{-fnon-call-exceptions}).
11879 @item -fsanitize=leak
11880 @opindex fsanitize=leak
11881 Enable LeakSanitizer, a memory leak detector.
11882 This option only matters for linking of executables and
11883 the executable is linked against a library that overrides @code{malloc}
11884 and other allocator functions. See
11885 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
11886 details. The run-time behavior can be influenced using the
11887 @env{LSAN_OPTIONS} environment variable.
11888 The option cannot be combined with @option{-fsanitize=thread}.
11890 @item -fsanitize=undefined
11891 @opindex fsanitize=undefined
11892 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
11893 Various computations are instrumented to detect undefined behavior
11894 at runtime. Current suboptions are:
11898 @item -fsanitize=shift
11899 @opindex fsanitize=shift
11900 This option enables checking that the result of a shift operation is
11901 not undefined. Note that what exactly is considered undefined differs
11902 slightly between C and C++, as well as between ISO C90 and C99, etc.
11903 This option has two suboptions, @option{-fsanitize=shift-base} and
11904 @option{-fsanitize=shift-exponent}.
11906 @item -fsanitize=shift-exponent
11907 @opindex fsanitize=shift-exponent
11908 This option enables checking that the second argument of a shift operation
11909 is not negative and is smaller than the precision of the promoted first
11912 @item -fsanitize=shift-base
11913 @opindex fsanitize=shift-base
11914 If the second argument of a shift operation is within range, check that the
11915 result of a shift operation is not undefined. Note that what exactly is
11916 considered undefined differs slightly between C and C++, as well as between
11917 ISO C90 and C99, etc.
11919 @item -fsanitize=integer-divide-by-zero
11920 @opindex fsanitize=integer-divide-by-zero
11921 Detect integer division by zero as well as @code{INT_MIN / -1} division.
11923 @item -fsanitize=unreachable
11924 @opindex fsanitize=unreachable
11925 With this option, the compiler turns the @code{__builtin_unreachable}
11926 call into a diagnostics message call instead. When reaching the
11927 @code{__builtin_unreachable} call, the behavior is undefined.
11929 @item -fsanitize=vla-bound
11930 @opindex fsanitize=vla-bound
11931 This option instructs the compiler to check that the size of a variable
11932 length array is positive.
11934 @item -fsanitize=null
11935 @opindex fsanitize=null
11936 This option enables pointer checking. Particularly, the application
11937 built with this option turned on will issue an error message when it
11938 tries to dereference a NULL pointer, or if a reference (possibly an
11939 rvalue reference) is bound to a NULL pointer, or if a method is invoked
11940 on an object pointed by a NULL pointer.
11942 @item -fsanitize=return
11943 @opindex fsanitize=return
11944 This option enables return statement checking. Programs
11945 built with this option turned on will issue an error message
11946 when the end of a non-void function is reached without actually
11947 returning a value. This option works in C++ only.
11949 @item -fsanitize=signed-integer-overflow
11950 @opindex fsanitize=signed-integer-overflow
11951 This option enables signed integer overflow checking. We check that
11952 the result of @code{+}, @code{*}, and both unary and binary @code{-}
11953 does not overflow in the signed arithmetics. Note, integer promotion
11954 rules must be taken into account. That is, the following is not an
11957 signed char a = SCHAR_MAX;
11961 @item -fsanitize=bounds
11962 @opindex fsanitize=bounds
11963 This option enables instrumentation of array bounds. Various out of bounds
11964 accesses are detected. Flexible array members, flexible array member-like
11965 arrays, and initializers of variables with static storage are not instrumented.
11967 @item -fsanitize=bounds-strict
11968 @opindex fsanitize=bounds-strict
11969 This option enables strict instrumentation of array bounds. Most out of bounds
11970 accesses are detected, including flexible array members and flexible array
11971 member-like arrays. Initializers of variables with static storage are not
11974 @item -fsanitize=alignment
11975 @opindex fsanitize=alignment
11977 This option enables checking of alignment of pointers when they are
11978 dereferenced, or when a reference is bound to insufficiently aligned target,
11979 or when a method or constructor is invoked on insufficiently aligned object.
11981 @item -fsanitize=object-size
11982 @opindex fsanitize=object-size
11983 This option enables instrumentation of memory references using the
11984 @code{__builtin_object_size} function. Various out of bounds pointer
11985 accesses are detected.
11987 @item -fsanitize=float-divide-by-zero
11988 @opindex fsanitize=float-divide-by-zero
11989 Detect floating-point division by zero. Unlike other similar options,
11990 @option{-fsanitize=float-divide-by-zero} is not enabled by
11991 @option{-fsanitize=undefined}, since floating-point division by zero can
11992 be a legitimate way of obtaining infinities and NaNs.
11994 @item -fsanitize=float-cast-overflow
11995 @opindex fsanitize=float-cast-overflow
11996 This option enables floating-point type to integer conversion checking.
11997 We check that the result of the conversion does not overflow.
11998 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
11999 not enabled by @option{-fsanitize=undefined}.
12000 This option does not work well with @code{FE_INVALID} exceptions enabled.
12002 @item -fsanitize=nonnull-attribute
12003 @opindex fsanitize=nonnull-attribute
12005 This option enables instrumentation of calls, checking whether null values
12006 are not passed to arguments marked as requiring a non-null value by the
12007 @code{nonnull} function attribute.
12009 @item -fsanitize=returns-nonnull-attribute
12010 @opindex fsanitize=returns-nonnull-attribute
12012 This option enables instrumentation of return statements in functions
12013 marked with @code{returns_nonnull} function attribute, to detect returning
12014 of null values from such functions.
12016 @item -fsanitize=bool
12017 @opindex fsanitize=bool
12019 This option enables instrumentation of loads from bool. If a value other
12020 than 0/1 is loaded, a run-time error is issued.
12022 @item -fsanitize=enum
12023 @opindex fsanitize=enum
12025 This option enables instrumentation of loads from an enum type. If
12026 a value outside the range of values for the enum type is loaded,
12027 a run-time error is issued.
12029 @item -fsanitize=vptr
12030 @opindex fsanitize=vptr
12032 This option enables instrumentation of C++ member function calls, member
12033 accesses and some conversions between pointers to base and derived classes,
12034 to verify the referenced object has the correct dynamic type.
12036 @item -fsanitize=pointer-overflow
12037 @opindex fsanitize=pointer-overflow
12039 This option enables instrumentation of pointer arithmetics. If the pointer
12040 arithmetics overflows, a run-time error is issued.
12042 @item -fsanitize=builtin
12043 @opindex fsanitize=builtin
12045 This option enables instrumentation of arguments to selected builtin
12046 functions. If an invalid value is passed to such arguments, a run-time
12047 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
12048 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
12053 While @option{-ftrapv} causes traps for signed overflows to be emitted,
12054 @option{-fsanitize=undefined} gives a diagnostic message.
12055 This currently works only for the C family of languages.
12057 @item -fno-sanitize=all
12058 @opindex fno-sanitize=all
12060 This option disables all previously enabled sanitizers.
12061 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
12064 @item -fasan-shadow-offset=@var{number}
12065 @opindex fasan-shadow-offset
12066 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
12067 It is useful for experimenting with different shadow memory layouts in
12068 Kernel AddressSanitizer.
12070 @item -fsanitize-sections=@var{s1},@var{s2},...
12071 @opindex fsanitize-sections
12072 Sanitize global variables in selected user-defined sections. @var{si} may
12075 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
12076 @opindex fsanitize-recover
12077 @opindex fno-sanitize-recover
12078 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
12079 mentioned in comma-separated list of @var{opts}. Enabling this option
12080 for a sanitizer component causes it to attempt to continue
12081 running the program as if no error happened. This means multiple
12082 runtime errors can be reported in a single program run, and the exit
12083 code of the program may indicate success even when errors
12084 have been reported. The @option{-fno-sanitize-recover=} option
12085 can be used to alter
12086 this behavior: only the first detected error is reported
12087 and program then exits with a non-zero exit code.
12089 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
12090 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
12091 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
12092 @option{-fsanitize=bounds-strict},
12093 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
12094 For these sanitizers error recovery is turned on by default,
12095 except @option{-fsanitize=address}, for which this feature is experimental.
12096 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
12097 accepted, the former enables recovery for all sanitizers that support it,
12098 the latter disables recovery for all sanitizers that support it.
12100 Even if a recovery mode is turned on the compiler side, it needs to be also
12101 enabled on the runtime library side, otherwise the failures are still fatal.
12102 The runtime library defaults to @code{halt_on_error=0} for
12103 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
12104 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
12105 setting the @code{halt_on_error} flag in the corresponding environment variable.
12107 Syntax without an explicit @var{opts} parameter is deprecated. It is
12108 equivalent to specifying an @var{opts} list of:
12111 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
12114 @item -fsanitize-address-use-after-scope
12115 @opindex fsanitize-address-use-after-scope
12116 Enable sanitization of local variables to detect use-after-scope bugs.
12117 The option sets @option{-fstack-reuse} to @samp{none}.
12119 @item -fsanitize-undefined-trap-on-error
12120 @opindex fsanitize-undefined-trap-on-error
12121 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
12122 report undefined behavior using @code{__builtin_trap} rather than
12123 a @code{libubsan} library routine. The advantage of this is that the
12124 @code{libubsan} library is not needed and is not linked in, so this
12125 is usable even in freestanding environments.
12127 @item -fsanitize-coverage=trace-pc
12128 @opindex fsanitize-coverage=trace-pc
12129 Enable coverage-guided fuzzing code instrumentation.
12130 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
12132 @item -fsanitize-coverage=trace-cmp
12133 @opindex fsanitize-coverage=trace-cmp
12134 Enable dataflow guided fuzzing code instrumentation.
12135 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
12136 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
12137 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
12138 variable or @code{__sanitizer_cov_trace_const_cmp1},
12139 @code{__sanitizer_cov_trace_const_cmp2},
12140 @code{__sanitizer_cov_trace_const_cmp4} or
12141 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
12142 operand constant, @code{__sanitizer_cov_trace_cmpf} or
12143 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
12144 @code{__sanitizer_cov_trace_switch} for switch statements.
12146 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
12147 @opindex fcf-protection
12148 Enable code instrumentation of control-flow transfers to increase
12149 program security by checking that target addresses of control-flow
12150 transfer instructions (such as indirect function call, function return,
12151 indirect jump) are valid. This prevents diverting the flow of control
12152 to an unexpected target. This is intended to protect against such
12153 threats as Return-oriented Programming (ROP), and similarly
12154 call/jmp-oriented programming (COP/JOP).
12156 The value @code{branch} tells the compiler to implement checking of
12157 validity of control-flow transfer at the point of indirect branch
12158 instructions, i.e.@: call/jmp instructions. The value @code{return}
12159 implements checking of validity at the point of returning from a
12160 function. The value @code{full} is an alias for specifying both
12161 @code{branch} and @code{return}. The value @code{none} turns off
12164 The macro @code{__CET__} is defined when @option{-fcf-protection} is
12165 used. The first bit of @code{__CET__} is set to 1 for the value
12166 @code{branch} and the second bit of @code{__CET__} is set to 1 for
12169 You can also use the @code{nocf_check} attribute to identify
12170 which functions and calls should be skipped from instrumentation
12171 (@pxref{Function Attributes}).
12173 Currently the x86 GNU/Linux target provides an implementation based
12174 on Intel Control-flow Enforcement Technology (CET).
12176 @item -fstack-protector
12177 @opindex fstack-protector
12178 Emit extra code to check for buffer overflows, such as stack smashing
12179 attacks. This is done by adding a guard variable to functions with
12180 vulnerable objects. This includes functions that call @code{alloca}, and
12181 functions with buffers larger than 8 bytes. The guards are initialized
12182 when a function is entered and then checked when the function exits.
12183 If a guard check fails, an error message is printed and the program exits.
12185 @item -fstack-protector-all
12186 @opindex fstack-protector-all
12187 Like @option{-fstack-protector} except that all functions are protected.
12189 @item -fstack-protector-strong
12190 @opindex fstack-protector-strong
12191 Like @option{-fstack-protector} but includes additional functions to
12192 be protected --- those that have local array definitions, or have
12193 references to local frame addresses.
12195 @item -fstack-protector-explicit
12196 @opindex fstack-protector-explicit
12197 Like @option{-fstack-protector} but only protects those functions which
12198 have the @code{stack_protect} attribute.
12200 @item -fstack-check
12201 @opindex fstack-check
12202 Generate code to verify that you do not go beyond the boundary of the
12203 stack. You should specify this flag if you are running in an
12204 environment with multiple threads, but you only rarely need to specify it in
12205 a single-threaded environment since stack overflow is automatically
12206 detected on nearly all systems if there is only one stack.
12208 Note that this switch does not actually cause checking to be done; the
12209 operating system or the language runtime must do that. The switch causes
12210 generation of code to ensure that they see the stack being extended.
12212 You can additionally specify a string parameter: @samp{no} means no
12213 checking, @samp{generic} means force the use of old-style checking,
12214 @samp{specific} means use the best checking method and is equivalent
12215 to bare @option{-fstack-check}.
12217 Old-style checking is a generic mechanism that requires no specific
12218 target support in the compiler but comes with the following drawbacks:
12222 Modified allocation strategy for large objects: they are always
12223 allocated dynamically if their size exceeds a fixed threshold. Note this
12224 may change the semantics of some code.
12227 Fixed limit on the size of the static frame of functions: when it is
12228 topped by a particular function, stack checking is not reliable and
12229 a warning is issued by the compiler.
12232 Inefficiency: because of both the modified allocation strategy and the
12233 generic implementation, code performance is hampered.
12236 Note that old-style stack checking is also the fallback method for
12237 @samp{specific} if no target support has been added in the compiler.
12239 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
12240 and stack overflows. @samp{specific} is an excellent choice when compiling
12241 Ada code. It is not generally sufficient to protect against stack-clash
12242 attacks. To protect against those you want @samp{-fstack-clash-protection}.
12244 @item -fstack-clash-protection
12245 @opindex fstack-clash-protection
12246 Generate code to prevent stack clash style attacks. When this option is
12247 enabled, the compiler will only allocate one page of stack space at a time
12248 and each page is accessed immediately after allocation. Thus, it prevents
12249 allocations from jumping over any stack guard page provided by the
12252 Most targets do not fully support stack clash protection. However, on
12253 those targets @option{-fstack-clash-protection} will protect dynamic stack
12254 allocations. @option{-fstack-clash-protection} may also provide limited
12255 protection for static stack allocations if the target supports
12256 @option{-fstack-check=specific}.
12258 @item -fstack-limit-register=@var{reg}
12259 @itemx -fstack-limit-symbol=@var{sym}
12260 @itemx -fno-stack-limit
12261 @opindex fstack-limit-register
12262 @opindex fstack-limit-symbol
12263 @opindex fno-stack-limit
12264 Generate code to ensure that the stack does not grow beyond a certain value,
12265 either the value of a register or the address of a symbol. If a larger
12266 stack is required, a signal is raised at run time. For most targets,
12267 the signal is raised before the stack overruns the boundary, so
12268 it is possible to catch the signal without taking special precautions.
12270 For instance, if the stack starts at absolute address @samp{0x80000000}
12271 and grows downwards, you can use the flags
12272 @option{-fstack-limit-symbol=__stack_limit} and
12273 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12274 of 128KB@. Note that this may only work with the GNU linker.
12276 You can locally override stack limit checking by using the
12277 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
12279 @item -fsplit-stack
12280 @opindex fsplit-stack
12281 Generate code to automatically split the stack before it overflows.
12282 The resulting program has a discontiguous stack which can only
12283 overflow if the program is unable to allocate any more memory. This
12284 is most useful when running threaded programs, as it is no longer
12285 necessary to calculate a good stack size to use for each thread. This
12286 is currently only implemented for the x86 targets running
12289 When code compiled with @option{-fsplit-stack} calls code compiled
12290 without @option{-fsplit-stack}, there may not be much stack space
12291 available for the latter code to run. If compiling all code,
12292 including library code, with @option{-fsplit-stack} is not an option,
12293 then the linker can fix up these calls so that the code compiled
12294 without @option{-fsplit-stack} always has a large stack. Support for
12295 this is implemented in the gold linker in GNU binutils release 2.21
12298 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
12299 @opindex fvtable-verify
12300 This option is only available when compiling C++ code.
12301 It turns on (or off, if using @option{-fvtable-verify=none}) the security
12302 feature that verifies at run time, for every virtual call, that
12303 the vtable pointer through which the call is made is valid for the type of
12304 the object, and has not been corrupted or overwritten. If an invalid vtable
12305 pointer is detected at run time, an error is reported and execution of the
12306 program is immediately halted.
12308 This option causes run-time data structures to be built at program startup,
12309 which are used for verifying the vtable pointers.
12310 The options @samp{std} and @samp{preinit}
12311 control the timing of when these data structures are built. In both cases the
12312 data structures are built before execution reaches @code{main}. Using
12313 @option{-fvtable-verify=std} causes the data structures to be built after
12314 shared libraries have been loaded and initialized.
12315 @option{-fvtable-verify=preinit} causes them to be built before shared
12316 libraries have been loaded and initialized.
12318 If this option appears multiple times in the command line with different
12319 values specified, @samp{none} takes highest priority over both @samp{std} and
12320 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
12323 @opindex fvtv-debug
12324 When used in conjunction with @option{-fvtable-verify=std} or
12325 @option{-fvtable-verify=preinit}, causes debug versions of the
12326 runtime functions for the vtable verification feature to be called.
12327 This flag also causes the compiler to log information about which
12328 vtable pointers it finds for each class.
12329 This information is written to a file named @file{vtv_set_ptr_data.log}
12330 in the directory named by the environment variable @env{VTV_LOGS_DIR}
12331 if that is defined or the current working directory otherwise.
12333 Note: This feature @emph{appends} data to the log file. If you want a fresh log
12334 file, be sure to delete any existing one.
12337 @opindex fvtv-counts
12338 This is a debugging flag. When used in conjunction with
12339 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
12340 causes the compiler to keep track of the total number of virtual calls
12341 it encounters and the number of verifications it inserts. It also
12342 counts the number of calls to certain run-time library functions
12343 that it inserts and logs this information for each compilation unit.
12344 The compiler writes this information to a file named
12345 @file{vtv_count_data.log} in the directory named by the environment
12346 variable @env{VTV_LOGS_DIR} if that is defined or the current working
12347 directory otherwise. It also counts the size of the vtable pointer sets
12348 for each class, and writes this information to @file{vtv_class_set_sizes.log}
12349 in the same directory.
12351 Note: This feature @emph{appends} data to the log files. To get fresh log
12352 files, be sure to delete any existing ones.
12354 @item -finstrument-functions
12355 @opindex finstrument-functions
12356 Generate instrumentation calls for entry and exit to functions. Just
12357 after function entry and just before function exit, the following
12358 profiling functions are called with the address of the current
12359 function and its call site. (On some platforms,
12360 @code{__builtin_return_address} does not work beyond the current
12361 function, so the call site information may not be available to the
12362 profiling functions otherwise.)
12365 void __cyg_profile_func_enter (void *this_fn,
12367 void __cyg_profile_func_exit (void *this_fn,
12371 The first argument is the address of the start of the current function,
12372 which may be looked up exactly in the symbol table.
12374 This instrumentation is also done for functions expanded inline in other
12375 functions. The profiling calls indicate where, conceptually, the
12376 inline function is entered and exited. This means that addressable
12377 versions of such functions must be available. If all your uses of a
12378 function are expanded inline, this may mean an additional expansion of
12379 code size. If you use @code{extern inline} in your C code, an
12380 addressable version of such functions must be provided. (This is
12381 normally the case anyway, but if you get lucky and the optimizer always
12382 expands the functions inline, you might have gotten away without
12383 providing static copies.)
12385 A function may be given the attribute @code{no_instrument_function}, in
12386 which case this instrumentation is not done. This can be used, for
12387 example, for the profiling functions listed above, high-priority
12388 interrupt routines, and any functions from which the profiling functions
12389 cannot safely be called (perhaps signal handlers, if the profiling
12390 routines generate output or allocate memory).
12392 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
12393 @opindex finstrument-functions-exclude-file-list
12395 Set the list of functions that are excluded from instrumentation (see
12396 the description of @option{-finstrument-functions}). If the file that
12397 contains a function definition matches with one of @var{file}, then
12398 that function is not instrumented. The match is done on substrings:
12399 if the @var{file} parameter is a substring of the file name, it is
12400 considered to be a match.
12405 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
12409 excludes any inline function defined in files whose pathnames
12410 contain @file{/bits/stl} or @file{include/sys}.
12412 If, for some reason, you want to include letter @samp{,} in one of
12413 @var{sym}, write @samp{\,}. For example,
12414 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
12415 (note the single quote surrounding the option).
12417 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
12418 @opindex finstrument-functions-exclude-function-list
12420 This is similar to @option{-finstrument-functions-exclude-file-list},
12421 but this option sets the list of function names to be excluded from
12422 instrumentation. The function name to be matched is its user-visible
12423 name, such as @code{vector<int> blah(const vector<int> &)}, not the
12424 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
12425 match is done on substrings: if the @var{sym} parameter is a substring
12426 of the function name, it is considered to be a match. For C99 and C++
12427 extended identifiers, the function name must be given in UTF-8, not
12428 using universal character names.
12430 @item -fpatchable-function-entry=@var{N}[,@var{M}]
12431 @opindex fpatchable-function-entry
12432 Generate @var{N} NOPs right at the beginning
12433 of each function, with the function entry point before the @var{M}th NOP.
12434 If @var{M} is omitted, it defaults to @code{0} so the
12435 function entry points to the address just at the first NOP.
12436 The NOP instructions reserve extra space which can be used to patch in
12437 any desired instrumentation at run time, provided that the code segment
12438 is writable. The amount of space is controllable indirectly via
12439 the number of NOPs; the NOP instruction used corresponds to the instruction
12440 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
12441 is target-specific and may also depend on the architecture variant and/or
12442 other compilation options.
12444 For run-time identification, the starting addresses of these areas,
12445 which correspond to their respective function entries minus @var{M},
12446 are additionally collected in the @code{__patchable_function_entries}
12447 section of the resulting binary.
12449 Note that the value of @code{__attribute__ ((patchable_function_entry
12450 (N,M)))} takes precedence over command-line option
12451 @option{-fpatchable-function-entry=N,M}. This can be used to increase
12452 the area size or to remove it completely on a single function.
12453 If @code{N=0}, no pad location is recorded.
12455 The NOP instructions are inserted at---and maybe before, depending on
12456 @var{M}---the function entry address, even before the prologue.
12461 @node Preprocessor Options
12462 @section Options Controlling the Preprocessor
12463 @cindex preprocessor options
12464 @cindex options, preprocessor
12466 These options control the C preprocessor, which is run on each C source
12467 file before actual compilation.
12469 If you use the @option{-E} option, nothing is done except preprocessing.
12470 Some of these options make sense only together with @option{-E} because
12471 they cause the preprocessor output to be unsuitable for actual
12474 In addition to the options listed here, there are a number of options
12475 to control search paths for include files documented in
12476 @ref{Directory Options}.
12477 Options to control preprocessor diagnostics are listed in
12478 @ref{Warning Options}.
12481 @include cppopts.texi
12483 @item -Wp,@var{option}
12485 You can use @option{-Wp,@var{option}} to bypass the compiler driver
12486 and pass @var{option} directly through to the preprocessor. If
12487 @var{option} contains commas, it is split into multiple options at the
12488 commas. However, many options are modified, translated or interpreted
12489 by the compiler driver before being passed to the preprocessor, and
12490 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
12491 interface is undocumented and subject to change, so whenever possible
12492 you should avoid using @option{-Wp} and let the driver handle the
12495 @item -Xpreprocessor @var{option}
12496 @opindex Xpreprocessor
12497 Pass @var{option} as an option to the preprocessor. You can use this to
12498 supply system-specific preprocessor options that GCC does not
12501 If you want to pass an option that takes an argument, you must use
12502 @option{-Xpreprocessor} twice, once for the option and once for the argument.
12504 @item -no-integrated-cpp
12505 @opindex no-integrated-cpp
12506 Perform preprocessing as a separate pass before compilation.
12507 By default, GCC performs preprocessing as an integrated part of
12508 input tokenization and parsing.
12509 If this option is provided, the appropriate language front end
12510 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
12511 and Objective-C, respectively) is instead invoked twice,
12512 once for preprocessing only and once for actual compilation
12513 of the preprocessed input.
12514 This option may be useful in conjunction with the @option{-B} or
12515 @option{-wrapper} options to specify an alternate preprocessor or
12516 perform additional processing of the program source between
12517 normal preprocessing and compilation.
12521 @node Assembler Options
12522 @section Passing Options to the Assembler
12524 @c prevent bad page break with this line
12525 You can pass options to the assembler.
12528 @item -Wa,@var{option}
12530 Pass @var{option} as an option to the assembler. If @var{option}
12531 contains commas, it is split into multiple options at the commas.
12533 @item -Xassembler @var{option}
12534 @opindex Xassembler
12535 Pass @var{option} as an option to the assembler. You can use this to
12536 supply system-specific assembler options that GCC does not
12539 If you want to pass an option that takes an argument, you must use
12540 @option{-Xassembler} twice, once for the option and once for the argument.
12545 @section Options for Linking
12546 @cindex link options
12547 @cindex options, linking
12549 These options come into play when the compiler links object files into
12550 an executable output file. They are meaningless if the compiler is
12551 not doing a link step.
12555 @item @var{object-file-name}
12556 A file name that does not end in a special recognized suffix is
12557 considered to name an object file or library. (Object files are
12558 distinguished from libraries by the linker according to the file
12559 contents.) If linking is done, these object files are used as input
12568 If any of these options is used, then the linker is not run, and
12569 object file names should not be used as arguments. @xref{Overall
12572 @item -flinker-output=@var{type}
12573 @opindex flinker-output
12574 This option controls the code generation of the link time optimizer. By
12575 default the linker output is determined by the linker plugin automatically. For
12576 debugging the compiler and in the case of incremental linking to non-lto object
12577 file is desired, it may be useful to control the type manually.
12579 If @var{type} is @samp{exec} the code generation is configured to produce static
12580 binary. In this case @option{-fpic} and @option{-fpie} are both disabled.
12582 If @var{type} is @samp{dyn} the code generation is configured to produce shared
12583 library. In this case @option{-fpic} or @option{-fPIC} is preserved, but not
12584 enabled automatically. This makes it possible to build shared libraries without
12585 position independent code on architectures this is possible, i.e.@: on x86.
12587 If @var{type} is @samp{pie} the code generation is configured to produce
12588 @option{-fpie} executable. This result in similar optimizations as @samp{exec}
12589 except that @option{-fpie} is not disabled if specified at compilation time.
12591 If @var{type} is @samp{rel} the compiler assumes that incremental linking is
12592 done. The sections containing intermediate code for link-time optimization are
12593 merged, pre-optimized, and output to the resulting object file. In addition, if
12594 @option{-ffat-lto-objects} is specified the binary code is produced for future
12595 non-lto linking. The object file produced by incremental linking will be smaller
12596 than a static library produced from the same object files. At link-time the
12597 result of incremental linking will also load faster to compiler than a static
12598 library assuming that majority of objects in the library are used.
12600 Finally @samp{nolto-rel} configure compiler to for incremental linking where
12601 code generation is forced, final binary is produced and the intermediate code
12602 for later link-time optimization is stripped. When multiple object files are
12603 linked together the resulting code will be optimized better than with link time
12604 optimizations disabled (for example, the cross-module inlining will happen),
12605 most of benefits of whole program optimizations are however lost.
12607 During the incremental link (by @option{-r}) the linker plugin will default to
12608 @option{rel}. With current interfaces to GNU Binutils it is however not
12609 possible to link incrementally LTO objects and non-LTO objects into a single
12610 mixed object file. In the case any of object files in incremental link can not
12611 be used for link-time optimization the linker plugin will output warning and
12612 use @samp{nolto-rel}. To maintain the whole program optimization it is
12613 recommended to link such objects into static library instead. Alternatively it
12614 is possible to use H.J. Lu's binutils with support for mixed objects.
12617 @opindex fuse-ld=bfd
12618 Use the @command{bfd} linker instead of the default linker.
12620 @item -fuse-ld=gold
12621 @opindex fuse-ld=gold
12622 Use the @command{gold} linker instead of the default linker.
12625 @opindex fuse-ld=lld
12626 Use the LLVM @command{lld} linker instead of the default linker.
12629 @item -l@var{library}
12630 @itemx -l @var{library}
12632 Search the library named @var{library} when linking. (The second
12633 alternative with the library as a separate argument is only for
12634 POSIX compliance and is not recommended.)
12636 It makes a difference where in the command you write this option; the
12637 linker searches and processes libraries and object files in the order they
12638 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
12639 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
12640 to functions in @samp{z}, those functions may not be loaded.
12642 The linker searches a standard list of directories for the library,
12643 which is actually a file named @file{lib@var{library}.a}. The linker
12644 then uses this file as if it had been specified precisely by name.
12646 The directories searched include several standard system directories
12647 plus any that you specify with @option{-L}.
12649 Normally the files found this way are library files---archive files
12650 whose members are object files. The linker handles an archive file by
12651 scanning through it for members which define symbols that have so far
12652 been referenced but not defined. But if the file that is found is an
12653 ordinary object file, it is linked in the usual fashion. The only
12654 difference between using an @option{-l} option and specifying a file name
12655 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
12656 and searches several directories.
12660 You need this special case of the @option{-l} option in order to
12661 link an Objective-C or Objective-C++ program.
12663 @item -nostartfiles
12664 @opindex nostartfiles
12665 Do not use the standard system startup files when linking.
12666 The standard system libraries are used normally, unless @option{-nostdlib},
12667 @option{-nolibc}, or @option{-nodefaultlibs} is used.
12669 @item -nodefaultlibs
12670 @opindex nodefaultlibs
12671 Do not use the standard system libraries when linking.
12672 Only the libraries you specify are passed to the linker, and options
12673 specifying linkage of the system libraries, such as @option{-static-libgcc}
12674 or @option{-shared-libgcc}, are ignored.
12675 The standard startup files are used normally, unless @option{-nostartfiles}
12678 The compiler may generate calls to @code{memcmp},
12679 @code{memset}, @code{memcpy} and @code{memmove}.
12680 These entries are usually resolved by entries in
12681 libc. These entry points should be supplied through some other
12682 mechanism when this option is specified.
12686 Do not use the C library or system libraries tightly coupled with it when
12687 linking. Still link with the startup files, @file{libgcc} or toolchain
12688 provided language support libraries such as @file{libgnat}, @file{libgfortran}
12689 or @file{libstdc++} unless options preventing their inclusion are used as
12690 well. This typically removes @option{-lc} from the link command line, as well
12691 as system libraries that normally go with it and become meaningless when
12692 absence of a C library is assumed, for example @option{-lpthread} or
12693 @option{-lm} in some configurations. This is intended for bare-board
12694 targets when there is indeed no C library available.
12698 Do not use the standard system startup files or libraries when linking.
12699 No startup files and only the libraries you specify are passed to
12700 the linker, and options specifying linkage of the system libraries, such as
12701 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
12703 The compiler may generate calls to @code{memcmp}, @code{memset},
12704 @code{memcpy} and @code{memmove}.
12705 These entries are usually resolved by entries in
12706 libc. These entry points should be supplied through some other
12707 mechanism when this option is specified.
12709 @cindex @option{-lgcc}, use with @option{-nostdlib}
12710 @cindex @option{-nostdlib} and unresolved references
12711 @cindex unresolved references and @option{-nostdlib}
12712 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
12713 @cindex @option{-nodefaultlibs} and unresolved references
12714 @cindex unresolved references and @option{-nodefaultlibs}
12715 One of the standard libraries bypassed by @option{-nostdlib} and
12716 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
12717 which GCC uses to overcome shortcomings of particular machines, or special
12718 needs for some languages.
12719 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
12720 Collection (GCC) Internals},
12721 for more discussion of @file{libgcc.a}.)
12722 In most cases, you need @file{libgcc.a} even when you want to avoid
12723 other standard libraries. In other words, when you specify @option{-nostdlib}
12724 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
12725 This ensures that you have no unresolved references to internal GCC
12726 library subroutines.
12727 (An example of such an internal subroutine is @code{__main}, used to ensure C++
12728 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
12729 GNU Compiler Collection (GCC) Internals}.)
12731 @item -e @var{entry}
12732 @itemx --entry=@var{entry}
12736 Specify that the program entry point is @var{entry}. The argument is
12737 interpreted by the linker; the GNU linker accepts either a symbol name
12742 Produce a dynamically linked position independent executable on targets
12743 that support it. For predictable results, you must also specify the same
12744 set of options used for compilation (@option{-fpie}, @option{-fPIE},
12745 or model suboptions) when you specify this linker option.
12749 Don't produce a dynamically linked position independent executable.
12752 @opindex static-pie
12753 Produce a static position independent executable on targets that support
12754 it. A static position independent executable is similar to a static
12755 executable, but can be loaded at any address without a dynamic linker.
12756 For predictable results, you must also specify the same set of options
12757 used for compilation (@option{-fpie}, @option{-fPIE}, or model
12758 suboptions) when you specify this linker option.
12762 Link with the POSIX threads library. This option is supported on
12763 GNU/Linux targets, most other Unix derivatives, and also on
12764 x86 Cygwin and MinGW targets. On some targets this option also sets
12765 flags for the preprocessor, so it should be used consistently for both
12766 compilation and linking.
12770 Produce a relocatable object as output. This is also known as partial
12775 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
12776 that support it. This instructs the linker to add all symbols, not
12777 only used ones, to the dynamic symbol table. This option is needed
12778 for some uses of @code{dlopen} or to allow obtaining backtraces
12779 from within a program.
12783 Remove all symbol table and relocation information from the executable.
12787 On systems that support dynamic linking, this overrides @option{-pie}
12788 and prevents linking with the shared libraries. On other systems, this
12789 option has no effect.
12793 Produce a shared object which can then be linked with other objects to
12794 form an executable. Not all systems support this option. For predictable
12795 results, you must also specify the same set of options used for compilation
12796 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
12797 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
12798 needs to build supplementary stub code for constructors to work. On
12799 multi-libbed systems, @samp{gcc -shared} must select the correct support
12800 libraries to link against. Failing to supply the correct flags may lead
12801 to subtle defects. Supplying them in cases where they are not necessary
12804 @item -shared-libgcc
12805 @itemx -static-libgcc
12806 @opindex shared-libgcc
12807 @opindex static-libgcc
12808 On systems that provide @file{libgcc} as a shared library, these options
12809 force the use of either the shared or static version, respectively.
12810 If no shared version of @file{libgcc} was built when the compiler was
12811 configured, these options have no effect.
12813 There are several situations in which an application should use the
12814 shared @file{libgcc} instead of the static version. The most common
12815 of these is when the application wishes to throw and catch exceptions
12816 across different shared libraries. In that case, each of the libraries
12817 as well as the application itself should use the shared @file{libgcc}.
12819 Therefore, the G++ driver automatically adds @option{-shared-libgcc}
12820 whenever you build a shared library or a main executable, because C++
12821 programs typically use exceptions, so this is the right thing to do.
12823 If, instead, you use the GCC driver to create shared libraries, you may
12824 find that they are not always linked with the shared @file{libgcc}.
12825 If GCC finds, at its configuration time, that you have a non-GNU linker
12826 or a GNU linker that does not support option @option{--eh-frame-hdr},
12827 it links the shared version of @file{libgcc} into shared libraries
12828 by default. Otherwise, it takes advantage of the linker and optimizes
12829 away the linking with the shared version of @file{libgcc}, linking with
12830 the static version of libgcc by default. This allows exceptions to
12831 propagate through such shared libraries, without incurring relocation
12832 costs at library load time.
12834 However, if a library or main executable is supposed to throw or catch
12835 exceptions, you must link it using the G++ driver, or using the option
12836 @option{-shared-libgcc}, such that it is linked with the shared
12839 @item -static-libasan
12840 @opindex static-libasan
12841 When the @option{-fsanitize=address} option is used to link a program,
12842 the GCC driver automatically links against @option{libasan}. If
12843 @file{libasan} is available as a shared library, and the @option{-static}
12844 option is not used, then this links against the shared version of
12845 @file{libasan}. The @option{-static-libasan} option directs the GCC
12846 driver to link @file{libasan} statically, without necessarily linking
12847 other libraries statically.
12849 @item -static-libtsan
12850 @opindex static-libtsan
12851 When the @option{-fsanitize=thread} option is used to link a program,
12852 the GCC driver automatically links against @option{libtsan}. If
12853 @file{libtsan} is available as a shared library, and the @option{-static}
12854 option is not used, then this links against the shared version of
12855 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
12856 driver to link @file{libtsan} statically, without necessarily linking
12857 other libraries statically.
12859 @item -static-liblsan
12860 @opindex static-liblsan
12861 When the @option{-fsanitize=leak} option is used to link a program,
12862 the GCC driver automatically links against @option{liblsan}. If
12863 @file{liblsan} is available as a shared library, and the @option{-static}
12864 option is not used, then this links against the shared version of
12865 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
12866 driver to link @file{liblsan} statically, without necessarily linking
12867 other libraries statically.
12869 @item -static-libubsan
12870 @opindex static-libubsan
12871 When the @option{-fsanitize=undefined} option is used to link a program,
12872 the GCC driver automatically links against @option{libubsan}. If
12873 @file{libubsan} is available as a shared library, and the @option{-static}
12874 option is not used, then this links against the shared version of
12875 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
12876 driver to link @file{libubsan} statically, without necessarily linking
12877 other libraries statically.
12879 @item -static-libstdc++
12880 @opindex static-libstdc++
12881 When the @command{g++} program is used to link a C++ program, it
12882 normally automatically links against @option{libstdc++}. If
12883 @file{libstdc++} is available as a shared library, and the
12884 @option{-static} option is not used, then this links against the
12885 shared version of @file{libstdc++}. That is normally fine. However, it
12886 is sometimes useful to freeze the version of @file{libstdc++} used by
12887 the program without going all the way to a fully static link. The
12888 @option{-static-libstdc++} option directs the @command{g++} driver to
12889 link @file{libstdc++} statically, without necessarily linking other
12890 libraries statically.
12894 Bind references to global symbols when building a shared object. Warn
12895 about any unresolved references (unless overridden by the link editor
12896 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
12899 @item -T @var{script}
12901 @cindex linker script
12902 Use @var{script} as the linker script. This option is supported by most
12903 systems using the GNU linker. On some targets, such as bare-board
12904 targets without an operating system, the @option{-T} option may be required
12905 when linking to avoid references to undefined symbols.
12907 @item -Xlinker @var{option}
12909 Pass @var{option} as an option to the linker. You can use this to
12910 supply system-specific linker options that GCC does not recognize.
12912 If you want to pass an option that takes a separate argument, you must use
12913 @option{-Xlinker} twice, once for the option and once for the argument.
12914 For example, to pass @option{-assert definitions}, you must write
12915 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
12916 @option{-Xlinker "-assert definitions"}, because this passes the entire
12917 string as a single argument, which is not what the linker expects.
12919 When using the GNU linker, it is usually more convenient to pass
12920 arguments to linker options using the @option{@var{option}=@var{value}}
12921 syntax than as separate arguments. For example, you can specify
12922 @option{-Xlinker -Map=output.map} rather than
12923 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
12924 this syntax for command-line options.
12926 @item -Wl,@var{option}
12928 Pass @var{option} as an option to the linker. If @var{option} contains
12929 commas, it is split into multiple options at the commas. You can use this
12930 syntax to pass an argument to the option.
12931 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
12932 linker. When using the GNU linker, you can also get the same effect with
12933 @option{-Wl,-Map=output.map}.
12935 @item -u @var{symbol}
12937 Pretend the symbol @var{symbol} is undefined, to force linking of
12938 library modules to define it. You can use @option{-u} multiple times with
12939 different symbols to force loading of additional library modules.
12941 @item -z @var{keyword}
12943 @option{-z} is passed directly on to the linker along with the keyword
12944 @var{keyword}. See the section in the documentation of your linker for
12945 permitted values and their meanings.
12948 @node Directory Options
12949 @section Options for Directory Search
12950 @cindex directory options
12951 @cindex options, directory search
12952 @cindex search path
12954 These options specify directories to search for header files, for
12955 libraries and for parts of the compiler:
12958 @include cppdiropts.texi
12960 @item -iplugindir=@var{dir}
12961 @opindex iplugindir=
12962 Set the directory to search for plugins that are passed
12963 by @option{-fplugin=@var{name}} instead of
12964 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
12965 to be used by the user, but only passed by the driver.
12969 Add directory @var{dir} to the list of directories to be searched
12972 @item -B@var{prefix}
12974 This option specifies where to find the executables, libraries,
12975 include files, and data files of the compiler itself.
12977 The compiler driver program runs one or more of the subprograms
12978 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
12979 @var{prefix} as a prefix for each program it tries to run, both with and
12980 without @samp{@var{machine}/@var{version}/} for the corresponding target
12981 machine and compiler version.
12983 For each subprogram to be run, the compiler driver first tries the
12984 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
12985 is not specified, the driver tries two standard prefixes,
12986 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
12987 those results in a file name that is found, the unmodified program
12988 name is searched for using the directories specified in your
12989 @env{PATH} environment variable.
12991 The compiler checks to see if the path provided by @option{-B}
12992 refers to a directory, and if necessary it adds a directory
12993 separator character at the end of the path.
12995 @option{-B} prefixes that effectively specify directory names also apply
12996 to libraries in the linker, because the compiler translates these
12997 options into @option{-L} options for the linker. They also apply to
12998 include files in the preprocessor, because the compiler translates these
12999 options into @option{-isystem} options for the preprocessor. In this case,
13000 the compiler appends @samp{include} to the prefix.
13002 The runtime support file @file{libgcc.a} can also be searched for using
13003 the @option{-B} prefix, if needed. If it is not found there, the two
13004 standard prefixes above are tried, and that is all. The file is left
13005 out of the link if it is not found by those means.
13007 Another way to specify a prefix much like the @option{-B} prefix is to use
13008 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
13011 As a special kludge, if the path provided by @option{-B} is
13012 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
13013 9, then it is replaced by @file{[dir/]include}. This is to help
13014 with boot-strapping the compiler.
13016 @item -no-canonical-prefixes
13017 @opindex no-canonical-prefixes
13018 Do not expand any symbolic links, resolve references to @samp{/../}
13019 or @samp{/./}, or make the path absolute when generating a relative
13022 @item --sysroot=@var{dir}
13024 Use @var{dir} as the logical root directory for headers and libraries.
13025 For example, if the compiler normally searches for headers in
13026 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
13027 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
13029 If you use both this option and the @option{-isysroot} option, then
13030 the @option{--sysroot} option applies to libraries, but the
13031 @option{-isysroot} option applies to header files.
13033 The GNU linker (beginning with version 2.16) has the necessary support
13034 for this option. If your linker does not support this option, the
13035 header file aspect of @option{--sysroot} still works, but the
13036 library aspect does not.
13038 @item --no-sysroot-suffix
13039 @opindex no-sysroot-suffix
13040 For some targets, a suffix is added to the root directory specified
13041 with @option{--sysroot}, depending on the other options used, so that
13042 headers may for example be found in
13043 @file{@var{dir}/@var{suffix}/usr/include} instead of
13044 @file{@var{dir}/usr/include}. This option disables the addition of
13049 @node Code Gen Options
13050 @section Options for Code Generation Conventions
13051 @cindex code generation conventions
13052 @cindex options, code generation
13053 @cindex run-time options
13055 These machine-independent options control the interface conventions
13056 used in code generation.
13058 Most of them have both positive and negative forms; the negative form
13059 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
13060 one of the forms is listed---the one that is not the default. You
13061 can figure out the other form by either removing @samp{no-} or adding
13065 @item -fstack-reuse=@var{reuse-level}
13066 @opindex fstack_reuse
13067 This option controls stack space reuse for user declared local/auto variables
13068 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
13069 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
13070 local variables and temporaries, @samp{named_vars} enables the reuse only for
13071 user defined local variables with names, and @samp{none} disables stack reuse
13072 completely. The default value is @samp{all}. The option is needed when the
13073 program extends the lifetime of a scoped local variable or a compiler generated
13074 temporary beyond the end point defined by the language. When a lifetime of
13075 a variable ends, and if the variable lives in memory, the optimizing compiler
13076 has the freedom to reuse its stack space with other temporaries or scoped
13077 local variables whose live range does not overlap with it. Legacy code extending
13078 local lifetime is likely to break with the stack reuse optimization.
13097 if (*p == 10) // out of scope use of local1
13108 A(int k) : i(k), j(k) @{ @}
13115 void foo(const A& ar)
13122 foo(A(10)); // temp object's lifetime ends when foo returns
13128 ap->i+= 10; // ap references out of scope temp whose space
13129 // is reused with a. What is the value of ap->i?
13134 The lifetime of a compiler generated temporary is well defined by the C++
13135 standard. When a lifetime of a temporary ends, and if the temporary lives
13136 in memory, the optimizing compiler has the freedom to reuse its stack
13137 space with other temporaries or scoped local variables whose live range
13138 does not overlap with it. However some of the legacy code relies on
13139 the behavior of older compilers in which temporaries' stack space is
13140 not reused, the aggressive stack reuse can lead to runtime errors. This
13141 option is used to control the temporary stack reuse optimization.
13145 This option generates traps for signed overflow on addition, subtraction,
13146 multiplication operations.
13147 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13148 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13149 @option{-fwrapv} being effective. Note that only active options override, so
13150 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13151 results in @option{-ftrapv} being effective.
13155 This option instructs the compiler to assume that signed arithmetic
13156 overflow of addition, subtraction and multiplication wraps around
13157 using twos-complement representation. This flag enables some optimizations
13158 and disables others.
13159 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13160 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13161 @option{-fwrapv} being effective. Note that only active options override, so
13162 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13163 results in @option{-ftrapv} being effective.
13165 @item -fwrapv-pointer
13166 @opindex fwrapv-pointer
13167 This option instructs the compiler to assume that pointer arithmetic
13168 overflow on addition and subtraction wraps around using twos-complement
13169 representation. This flag disables some optimizations which assume
13170 pointer overflow is invalid.
13172 @item -fstrict-overflow
13173 @opindex fstrict-overflow
13174 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
13175 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
13178 @opindex fexceptions
13179 Enable exception handling. Generates extra code needed to propagate
13180 exceptions. For some targets, this implies GCC generates frame
13181 unwind information for all functions, which can produce significant data
13182 size overhead, although it does not affect execution. If you do not
13183 specify this option, GCC enables it by default for languages like
13184 C++ that normally require exception handling, and disables it for
13185 languages like C that do not normally require it. However, you may need
13186 to enable this option when compiling C code that needs to interoperate
13187 properly with exception handlers written in C++. You may also wish to
13188 disable this option if you are compiling older C++ programs that don't
13189 use exception handling.
13191 @item -fnon-call-exceptions
13192 @opindex fnon-call-exceptions
13193 Generate code that allows trapping instructions to throw exceptions.
13194 Note that this requires platform-specific runtime support that does
13195 not exist everywhere. Moreover, it only allows @emph{trapping}
13196 instructions to throw exceptions, i.e.@: memory references or floating-point
13197 instructions. It does not allow exceptions to be thrown from
13198 arbitrary signal handlers such as @code{SIGALRM}.
13200 @item -fdelete-dead-exceptions
13201 @opindex fdelete-dead-exceptions
13202 Consider that instructions that may throw exceptions but don't otherwise
13203 contribute to the execution of the program can be optimized away.
13204 This option is enabled by default for the Ada front end, as permitted by
13205 the Ada language specification.
13206 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
13208 @item -funwind-tables
13209 @opindex funwind-tables
13210 Similar to @option{-fexceptions}, except that it just generates any needed
13211 static data, but does not affect the generated code in any other way.
13212 You normally do not need to enable this option; instead, a language processor
13213 that needs this handling enables it on your behalf.
13215 @item -fasynchronous-unwind-tables
13216 @opindex fasynchronous-unwind-tables
13217 Generate unwind table in DWARF format, if supported by target machine. The
13218 table is exact at each instruction boundary, so it can be used for stack
13219 unwinding from asynchronous events (such as debugger or garbage collector).
13221 @item -fno-gnu-unique
13222 @opindex fno-gnu-unique
13223 On systems with recent GNU assembler and C library, the C++ compiler
13224 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
13225 of template static data members and static local variables in inline
13226 functions are unique even in the presence of @code{RTLD_LOCAL}; this
13227 is necessary to avoid problems with a library used by two different
13228 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
13229 therefore disagreeing with the other one about the binding of the
13230 symbol. But this causes @code{dlclose} to be ignored for affected
13231 DSOs; if your program relies on reinitialization of a DSO via
13232 @code{dlclose} and @code{dlopen}, you can use
13233 @option{-fno-gnu-unique}.
13235 @item -fpcc-struct-return
13236 @opindex fpcc-struct-return
13237 Return ``short'' @code{struct} and @code{union} values in memory like
13238 longer ones, rather than in registers. This convention is less
13239 efficient, but it has the advantage of allowing intercallability between
13240 GCC-compiled files and files compiled with other compilers, particularly
13241 the Portable C Compiler (pcc).
13243 The precise convention for returning structures in memory depends
13244 on the target configuration macros.
13246 Short structures and unions are those whose size and alignment match
13247 that of some integer type.
13249 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13250 switch is not binary compatible with code compiled with the
13251 @option{-freg-struct-return} switch.
13252 Use it to conform to a non-default application binary interface.
13254 @item -freg-struct-return
13255 @opindex freg-struct-return
13256 Return @code{struct} and @code{union} values in registers when possible.
13257 This is more efficient for small structures than
13258 @option{-fpcc-struct-return}.
13260 If you specify neither @option{-fpcc-struct-return} nor
13261 @option{-freg-struct-return}, GCC defaults to whichever convention is
13262 standard for the target. If there is no standard convention, GCC
13263 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13264 the principal compiler. In those cases, we can choose the standard, and
13265 we chose the more efficient register return alternative.
13267 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13268 switch is not binary compatible with code compiled with the
13269 @option{-fpcc-struct-return} switch.
13270 Use it to conform to a non-default application binary interface.
13272 @item -fshort-enums
13273 @opindex fshort-enums
13274 Allocate to an @code{enum} type only as many bytes as it needs for the
13275 declared range of possible values. Specifically, the @code{enum} type
13276 is equivalent to the smallest integer type that has enough room.
13278 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13279 code that is not binary compatible with code generated without that switch.
13280 Use it to conform to a non-default application binary interface.
13282 @item -fshort-wchar
13283 @opindex fshort-wchar
13284 Override the underlying type for @code{wchar_t} to be @code{short
13285 unsigned int} instead of the default for the target. This option is
13286 useful for building programs to run under WINE@.
13288 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13289 code that is not binary compatible with code generated without that switch.
13290 Use it to conform to a non-default application binary interface.
13293 @opindex fno-common
13294 @cindex tentative definitions
13295 In C code, this option controls the placement of global variables
13296 defined without an initializer, known as @dfn{tentative definitions}
13297 in the C standard. Tentative definitions are distinct from declarations
13298 of a variable with the @code{extern} keyword, which do not allocate storage.
13300 Unix C compilers have traditionally allocated storage for
13301 uninitialized global variables in a common block. This allows the
13302 linker to resolve all tentative definitions of the same variable
13303 in different compilation units to the same object, or to a non-tentative
13305 This is the behavior specified by @option{-fcommon}, and is the default for
13306 GCC on most targets.
13307 On the other hand, this behavior is not required by ISO
13308 C, and on some targets may carry a speed or code size penalty on
13309 variable references.
13311 The @option{-fno-common} option specifies that the compiler should instead
13312 place uninitialized global variables in the BSS section of the object file.
13313 This inhibits the merging of tentative definitions by the linker so
13314 you get a multiple-definition error if the same
13315 variable is defined in more than one compilation unit.
13316 Compiling with @option{-fno-common} is useful on targets for which
13317 it provides better performance, or if you wish to verify that the
13318 program will work on other systems that always treat uninitialized
13319 variable definitions this way.
13323 Ignore the @code{#ident} directive.
13325 @item -finhibit-size-directive
13326 @opindex finhibit-size-directive
13327 Don't output a @code{.size} assembler directive, or anything else that
13328 would cause trouble if the function is split in the middle, and the
13329 two halves are placed at locations far apart in memory. This option is
13330 used when compiling @file{crtstuff.c}; you should not need to use it
13333 @item -fverbose-asm
13334 @opindex fverbose-asm
13335 Put extra commentary information in the generated assembly code to
13336 make it more readable. This option is generally only of use to those
13337 who actually need to read the generated assembly code (perhaps while
13338 debugging the compiler itself).
13340 @option{-fno-verbose-asm}, the default, causes the
13341 extra information to be omitted and is useful when comparing two assembler
13344 The added comments include:
13349 information on the compiler version and command-line options,
13352 the source code lines associated with the assembly instructions,
13353 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
13356 hints on which high-level expressions correspond to
13357 the various assembly instruction operands.
13361 For example, given this C source file:
13369 for (i = 0; i < n; i++)
13376 compiling to (x86_64) assembly via @option{-S} and emitting the result
13377 direct to stdout via @option{-o} @option{-}
13380 gcc -S test.c -fverbose-asm -Os -o -
13383 gives output similar to this:
13387 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
13394 .type test, @@function
13398 # test.c:4: int total = 0;
13399 xorl %eax, %eax # <retval>
13400 # test.c:6: for (i = 0; i < n; i++)
13401 xorl %edx, %edx # i
13403 # test.c:6: for (i = 0; i < n; i++)
13404 cmpl %edi, %edx # n, i
13406 # test.c:7: total += i * i;
13407 movl %edx, %ecx # i, tmp92
13408 imull %edx, %ecx # i, tmp92
13409 # test.c:6: for (i = 0; i < n; i++)
13411 # test.c:7: total += i * i;
13412 addl %ecx, %eax # tmp92, <retval>
13420 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
13421 .section .note.GNU-stack,"",@@progbits
13424 The comments are intended for humans rather than machines and hence the
13425 precise format of the comments is subject to change.
13427 @item -frecord-gcc-switches
13428 @opindex frecord-gcc-switches
13429 This switch causes the command line used to invoke the
13430 compiler to be recorded into the object file that is being created.
13431 This switch is only implemented on some targets and the exact format
13432 of the recording is target and binary file format dependent, but it
13433 usually takes the form of a section containing ASCII text. This
13434 switch is related to the @option{-fverbose-asm} switch, but that
13435 switch only records information in the assembler output file as
13436 comments, so it never reaches the object file.
13437 See also @option{-grecord-gcc-switches} for another
13438 way of storing compiler options into the object file.
13442 @cindex global offset table
13444 Generate position-independent code (PIC) suitable for use in a shared
13445 library, if supported for the target machine. Such code accesses all
13446 constant addresses through a global offset table (GOT)@. The dynamic
13447 loader resolves the GOT entries when the program starts (the dynamic
13448 loader is not part of GCC; it is part of the operating system). If
13449 the GOT size for the linked executable exceeds a machine-specific
13450 maximum size, you get an error message from the linker indicating that
13451 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13452 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
13453 on the m68k and RS/6000. The x86 has no such limit.)
13455 Position-independent code requires special support, and therefore works
13456 only on certain machines. For the x86, GCC supports PIC for System V
13457 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13458 position-independent.
13460 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13465 If supported for the target machine, emit position-independent code,
13466 suitable for dynamic linking and avoiding any limit on the size of the
13467 global offset table. This option makes a difference on AArch64, m68k,
13468 PowerPC and SPARC@.
13470 Position-independent code requires special support, and therefore works
13471 only on certain machines.
13473 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13480 These options are similar to @option{-fpic} and @option{-fPIC}, but the
13481 generated position-independent code can be only linked into executables.
13482 Usually these options are used to compile code that will be linked using
13483 the @option{-pie} GCC option.
13485 @option{-fpie} and @option{-fPIE} both define the macros
13486 @code{__pie__} and @code{__PIE__}. The macros have the value 1
13487 for @option{-fpie} and 2 for @option{-fPIE}.
13491 Do not use the PLT for external function calls in position-independent code.
13492 Instead, load the callee address at call sites from the GOT and branch to it.
13493 This leads to more efficient code by eliminating PLT stubs and exposing
13494 GOT loads to optimizations. On architectures such as 32-bit x86 where
13495 PLT stubs expect the GOT pointer in a specific register, this gives more
13496 register allocation freedom to the compiler.
13497 Lazy binding requires use of the PLT;
13498 with @option{-fno-plt} all external symbols are resolved at load time.
13500 Alternatively, the function attribute @code{noplt} can be used to avoid calls
13501 through the PLT for specific external functions.
13503 In position-dependent code, a few targets also convert calls to
13504 functions that are marked to not use the PLT to use the GOT instead.
13506 @item -fno-jump-tables
13507 @opindex fno-jump-tables
13508 Do not use jump tables for switch statements even where it would be
13509 more efficient than other code generation strategies. This option is
13510 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13511 building code that forms part of a dynamic linker and cannot
13512 reference the address of a jump table. On some targets, jump tables
13513 do not require a GOT and this option is not needed.
13515 @item -ffixed-@var{reg}
13517 Treat the register named @var{reg} as a fixed register; generated code
13518 should never refer to it (except perhaps as a stack pointer, frame
13519 pointer or in some other fixed role).
13521 @var{reg} must be the name of a register. The register names accepted
13522 are machine-specific and are defined in the @code{REGISTER_NAMES}
13523 macro in the machine description macro file.
13525 This flag does not have a negative form, because it specifies a
13528 @item -fcall-used-@var{reg}
13529 @opindex fcall-used
13530 Treat the register named @var{reg} as an allocable register that is
13531 clobbered by function calls. It may be allocated for temporaries or
13532 variables that do not live across a call. Functions compiled this way
13533 do not save and restore the register @var{reg}.
13535 It is an error to use this flag with the frame pointer or stack pointer.
13536 Use of this flag for other registers that have fixed pervasive roles in
13537 the machine's execution model produces disastrous results.
13539 This flag does not have a negative form, because it specifies a
13542 @item -fcall-saved-@var{reg}
13543 @opindex fcall-saved
13544 Treat the register named @var{reg} as an allocable register saved by
13545 functions. It may be allocated even for temporaries or variables that
13546 live across a call. Functions compiled this way save and restore
13547 the register @var{reg} if they use it.
13549 It is an error to use this flag with the frame pointer or stack pointer.
13550 Use of this flag for other registers that have fixed pervasive roles in
13551 the machine's execution model produces disastrous results.
13553 A different sort of disaster results from the use of this flag for
13554 a register in which function values may be returned.
13556 This flag does not have a negative form, because it specifies a
13559 @item -fpack-struct[=@var{n}]
13560 @opindex fpack-struct
13561 Without a value specified, pack all structure members together without
13562 holes. When a value is specified (which must be a small power of two), pack
13563 structure members according to this value, representing the maximum
13564 alignment (that is, objects with default alignment requirements larger than
13565 this are output potentially unaligned at the next fitting location.
13567 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13568 code that is not binary compatible with code generated without that switch.
13569 Additionally, it makes the code suboptimal.
13570 Use it to conform to a non-default application binary interface.
13572 @item -fleading-underscore
13573 @opindex fleading-underscore
13574 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13575 change the way C symbols are represented in the object file. One use
13576 is to help link with legacy assembly code.
13578 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13579 generate code that is not binary compatible with code generated without that
13580 switch. Use it to conform to a non-default application binary interface.
13581 Not all targets provide complete support for this switch.
13583 @item -ftls-model=@var{model}
13584 @opindex ftls-model
13585 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13586 The @var{model} argument should be one of @samp{global-dynamic},
13587 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
13588 Note that the choice is subject to optimization: the compiler may use
13589 a more efficient model for symbols not visible outside of the translation
13590 unit, or if @option{-fpic} is not given on the command line.
13592 The default without @option{-fpic} is @samp{initial-exec}; with
13593 @option{-fpic} the default is @samp{global-dynamic}.
13595 @item -ftrampolines
13596 @opindex ftrampolines
13597 For targets that normally need trampolines for nested functions, always
13598 generate them instead of using descriptors. Otherwise, for targets that
13599 do not need them, like for example HP-PA or IA-64, do nothing.
13601 A trampoline is a small piece of code that is created at run time on the
13602 stack when the address of a nested function is taken, and is used to call
13603 the nested function indirectly. Therefore, it requires the stack to be
13604 made executable in order for the program to work properly.
13606 @option{-fno-trampolines} is enabled by default on a language by language
13607 basis to let the compiler avoid generating them, if it computes that this
13608 is safe, and replace them with descriptors. Descriptors are made up of data
13609 only, but the generated code must be prepared to deal with them. As of this
13610 writing, @option{-fno-trampolines} is enabled by default only for Ada.
13612 Moreover, code compiled with @option{-ftrampolines} and code compiled with
13613 @option{-fno-trampolines} are not binary compatible if nested functions are
13614 present. This option must therefore be used on a program-wide basis and be
13615 manipulated with extreme care.
13617 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
13618 @opindex fvisibility
13619 Set the default ELF image symbol visibility to the specified option---all
13620 symbols are marked with this unless overridden within the code.
13621 Using this feature can very substantially improve linking and
13622 load times of shared object libraries, produce more optimized
13623 code, provide near-perfect API export and prevent symbol clashes.
13624 It is @strong{strongly} recommended that you use this in any shared objects
13627 Despite the nomenclature, @samp{default} always means public; i.e.,
13628 available to be linked against from outside the shared object.
13629 @samp{protected} and @samp{internal} are pretty useless in real-world
13630 usage so the only other commonly used option is @samp{hidden}.
13631 The default if @option{-fvisibility} isn't specified is
13632 @samp{default}, i.e., make every symbol public.
13634 A good explanation of the benefits offered by ensuring ELF
13635 symbols have the correct visibility is given by ``How To Write
13636 Shared Libraries'' by Ulrich Drepper (which can be found at
13637 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
13638 solution made possible by this option to marking things hidden when
13639 the default is public is to make the default hidden and mark things
13640 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
13641 and @code{__attribute__ ((visibility("default")))} instead of
13642 @code{__declspec(dllexport)} you get almost identical semantics with
13643 identical syntax. This is a great boon to those working with
13644 cross-platform projects.
13646 For those adding visibility support to existing code, you may find
13647 @code{#pragma GCC visibility} of use. This works by you enclosing
13648 the declarations you wish to set visibility for with (for example)
13649 @code{#pragma GCC visibility push(hidden)} and
13650 @code{#pragma GCC visibility pop}.
13651 Bear in mind that symbol visibility should be viewed @strong{as
13652 part of the API interface contract} and thus all new code should
13653 always specify visibility when it is not the default; i.e., declarations
13654 only for use within the local DSO should @strong{always} be marked explicitly
13655 as hidden as so to avoid PLT indirection overheads---making this
13656 abundantly clear also aids readability and self-documentation of the code.
13657 Note that due to ISO C++ specification requirements, @code{operator new} and
13658 @code{operator delete} must always be of default visibility.
13660 Be aware that headers from outside your project, in particular system
13661 headers and headers from any other library you use, may not be
13662 expecting to be compiled with visibility other than the default. You
13663 may need to explicitly say @code{#pragma GCC visibility push(default)}
13664 before including any such headers.
13666 @code{extern} declarations are not affected by @option{-fvisibility}, so
13667 a lot of code can be recompiled with @option{-fvisibility=hidden} with
13668 no modifications. However, this means that calls to @code{extern}
13669 functions with no explicit visibility use the PLT, so it is more
13670 effective to use @code{__attribute ((visibility))} and/or
13671 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
13672 declarations should be treated as hidden.
13674 Note that @option{-fvisibility} does affect C++ vague linkage
13675 entities. This means that, for instance, an exception class that is
13676 be thrown between DSOs must be explicitly marked with default
13677 visibility so that the @samp{type_info} nodes are unified between
13680 An overview of these techniques, their benefits and how to use them
13681 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
13683 @item -fstrict-volatile-bitfields
13684 @opindex fstrict-volatile-bitfields
13685 This option should be used if accesses to volatile bit-fields (or other
13686 structure fields, although the compiler usually honors those types
13687 anyway) should use a single access of the width of the
13688 field's type, aligned to a natural alignment if possible. For
13689 example, targets with memory-mapped peripheral registers might require
13690 all such accesses to be 16 bits wide; with this flag you can
13691 declare all peripheral bit-fields as @code{unsigned short} (assuming short
13692 is 16 bits on these targets) to force GCC to use 16-bit accesses
13693 instead of, perhaps, a more efficient 32-bit access.
13695 If this option is disabled, the compiler uses the most efficient
13696 instruction. In the previous example, that might be a 32-bit load
13697 instruction, even though that accesses bytes that do not contain
13698 any portion of the bit-field, or memory-mapped registers unrelated to
13699 the one being updated.
13701 In some cases, such as when the @code{packed} attribute is applied to a
13702 structure field, it may not be possible to access the field with a single
13703 read or write that is correctly aligned for the target machine. In this
13704 case GCC falls back to generating multiple accesses rather than code that
13705 will fault or truncate the result at run time.
13707 Note: Due to restrictions of the C/C++11 memory model, write accesses are
13708 not allowed to touch non bit-field members. It is therefore recommended
13709 to define all bits of the field's type as bit-field members.
13711 The default value of this option is determined by the application binary
13712 interface for the target processor.
13714 @item -fsync-libcalls
13715 @opindex fsync-libcalls
13716 This option controls whether any out-of-line instance of the @code{__sync}
13717 family of functions may be used to implement the C++11 @code{__atomic}
13718 family of functions.
13720 The default value of this option is enabled, thus the only useful form
13721 of the option is @option{-fno-sync-libcalls}. This option is used in
13722 the implementation of the @file{libatomic} runtime library.
13726 @node Developer Options
13727 @section GCC Developer Options
13728 @cindex developer options
13729 @cindex debugging GCC
13730 @cindex debug dump options
13731 @cindex dump options
13732 @cindex compilation statistics
13734 This section describes command-line options that are primarily of
13735 interest to GCC developers, including options to support compiler
13736 testing and investigation of compiler bugs and compile-time
13737 performance problems. This includes options that produce debug dumps
13738 at various points in the compilation; that print statistics such as
13739 memory use and execution time; and that print information about GCC's
13740 configuration, such as where it searches for libraries. You should
13741 rarely need to use any of these options for ordinary compilation and
13744 Many developer options that cause GCC to dump output to a file take an
13745 optional @samp{=@var{filename}} suffix. You can specify @samp{stdout}
13746 or @samp{-} to dump to standard output, and @samp{stderr} for standard
13749 If @samp{=@var{filename}} is omitted, a default dump file name is
13750 constructed by concatenating the base dump file name, a pass number,
13751 phase letter, and pass name. The base dump file name is the name of
13752 output file produced by the compiler if explicitly specified and not
13753 an executable; otherwise it is the source file name.
13754 The pass number is determined by the order passes are registered with
13755 the compiler's pass manager.
13756 This is generally the same as the order of execution, but passes
13757 registered by plugins, target-specific passes, or passes that are
13758 otherwise registered late are numbered higher than the pass named
13759 @samp{final}, even if they are executed earlier. The phase letter is
13760 one of @samp{i} (inter-procedural analysis), @samp{l}
13761 (language-specific), @samp{r} (RTL), or @samp{t} (tree).
13762 The files are created in the directory of the output file.
13766 @item -d@var{letters}
13767 @itemx -fdump-rtl-@var{pass}
13768 @itemx -fdump-rtl-@var{pass}=@var{filename}
13770 @opindex fdump-rtl-@var{pass}
13771 Says to make debugging dumps during compilation at times specified by
13772 @var{letters}. This is used for debugging the RTL-based passes of the
13775 Some @option{-d@var{letters}} switches have different meaning when
13776 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
13777 for information about preprocessor-specific dump options.
13779 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
13780 @option{-d} option @var{letters}. Here are the possible
13781 letters for use in @var{pass} and @var{letters}, and their meanings:
13785 @item -fdump-rtl-alignments
13786 @opindex fdump-rtl-alignments
13787 Dump after branch alignments have been computed.
13789 @item -fdump-rtl-asmcons
13790 @opindex fdump-rtl-asmcons
13791 Dump after fixing rtl statements that have unsatisfied in/out constraints.
13793 @item -fdump-rtl-auto_inc_dec
13794 @opindex fdump-rtl-auto_inc_dec
13795 Dump after auto-inc-dec discovery. This pass is only run on
13796 architectures that have auto inc or auto dec instructions.
13798 @item -fdump-rtl-barriers
13799 @opindex fdump-rtl-barriers
13800 Dump after cleaning up the barrier instructions.
13802 @item -fdump-rtl-bbpart
13803 @opindex fdump-rtl-bbpart
13804 Dump after partitioning hot and cold basic blocks.
13806 @item -fdump-rtl-bbro
13807 @opindex fdump-rtl-bbro
13808 Dump after block reordering.
13810 @item -fdump-rtl-btl1
13811 @itemx -fdump-rtl-btl2
13812 @opindex fdump-rtl-btl2
13813 @opindex fdump-rtl-btl2
13814 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
13815 after the two branch
13816 target load optimization passes.
13818 @item -fdump-rtl-bypass
13819 @opindex fdump-rtl-bypass
13820 Dump after jump bypassing and control flow optimizations.
13822 @item -fdump-rtl-combine
13823 @opindex fdump-rtl-combine
13824 Dump after the RTL instruction combination pass.
13826 @item -fdump-rtl-compgotos
13827 @opindex fdump-rtl-compgotos
13828 Dump after duplicating the computed gotos.
13830 @item -fdump-rtl-ce1
13831 @itemx -fdump-rtl-ce2
13832 @itemx -fdump-rtl-ce3
13833 @opindex fdump-rtl-ce1
13834 @opindex fdump-rtl-ce2
13835 @opindex fdump-rtl-ce3
13836 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
13837 @option{-fdump-rtl-ce3} enable dumping after the three
13838 if conversion passes.
13840 @item -fdump-rtl-cprop_hardreg
13841 @opindex fdump-rtl-cprop_hardreg
13842 Dump after hard register copy propagation.
13844 @item -fdump-rtl-csa
13845 @opindex fdump-rtl-csa
13846 Dump after combining stack adjustments.
13848 @item -fdump-rtl-cse1
13849 @itemx -fdump-rtl-cse2
13850 @opindex fdump-rtl-cse1
13851 @opindex fdump-rtl-cse2
13852 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
13853 the two common subexpression elimination passes.
13855 @item -fdump-rtl-dce
13856 @opindex fdump-rtl-dce
13857 Dump after the standalone dead code elimination passes.
13859 @item -fdump-rtl-dbr
13860 @opindex fdump-rtl-dbr
13861 Dump after delayed branch scheduling.
13863 @item -fdump-rtl-dce1
13864 @itemx -fdump-rtl-dce2
13865 @opindex fdump-rtl-dce1
13866 @opindex fdump-rtl-dce2
13867 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
13868 the two dead store elimination passes.
13870 @item -fdump-rtl-eh
13871 @opindex fdump-rtl-eh
13872 Dump after finalization of EH handling code.
13874 @item -fdump-rtl-eh_ranges
13875 @opindex fdump-rtl-eh_ranges
13876 Dump after conversion of EH handling range regions.
13878 @item -fdump-rtl-expand
13879 @opindex fdump-rtl-expand
13880 Dump after RTL generation.
13882 @item -fdump-rtl-fwprop1
13883 @itemx -fdump-rtl-fwprop2
13884 @opindex fdump-rtl-fwprop1
13885 @opindex fdump-rtl-fwprop2
13886 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
13887 dumping after the two forward propagation passes.
13889 @item -fdump-rtl-gcse1
13890 @itemx -fdump-rtl-gcse2
13891 @opindex fdump-rtl-gcse1
13892 @opindex fdump-rtl-gcse2
13893 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
13894 after global common subexpression elimination.
13896 @item -fdump-rtl-init-regs
13897 @opindex fdump-rtl-init-regs
13898 Dump after the initialization of the registers.
13900 @item -fdump-rtl-initvals
13901 @opindex fdump-rtl-initvals
13902 Dump after the computation of the initial value sets.
13904 @item -fdump-rtl-into_cfglayout
13905 @opindex fdump-rtl-into_cfglayout
13906 Dump after converting to cfglayout mode.
13908 @item -fdump-rtl-ira
13909 @opindex fdump-rtl-ira
13910 Dump after iterated register allocation.
13912 @item -fdump-rtl-jump
13913 @opindex fdump-rtl-jump
13914 Dump after the second jump optimization.
13916 @item -fdump-rtl-loop2
13917 @opindex fdump-rtl-loop2
13918 @option{-fdump-rtl-loop2} enables dumping after the rtl
13919 loop optimization passes.
13921 @item -fdump-rtl-mach
13922 @opindex fdump-rtl-mach
13923 Dump after performing the machine dependent reorganization pass, if that
13926 @item -fdump-rtl-mode_sw
13927 @opindex fdump-rtl-mode_sw
13928 Dump after removing redundant mode switches.
13930 @item -fdump-rtl-rnreg
13931 @opindex fdump-rtl-rnreg
13932 Dump after register renumbering.
13934 @item -fdump-rtl-outof_cfglayout
13935 @opindex fdump-rtl-outof_cfglayout
13936 Dump after converting from cfglayout mode.
13938 @item -fdump-rtl-peephole2
13939 @opindex fdump-rtl-peephole2
13940 Dump after the peephole pass.
13942 @item -fdump-rtl-postreload
13943 @opindex fdump-rtl-postreload
13944 Dump after post-reload optimizations.
13946 @item -fdump-rtl-pro_and_epilogue
13947 @opindex fdump-rtl-pro_and_epilogue
13948 Dump after generating the function prologues and epilogues.
13950 @item -fdump-rtl-sched1
13951 @itemx -fdump-rtl-sched2
13952 @opindex fdump-rtl-sched1
13953 @opindex fdump-rtl-sched2
13954 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
13955 after the basic block scheduling passes.
13957 @item -fdump-rtl-ree
13958 @opindex fdump-rtl-ree
13959 Dump after sign/zero extension elimination.
13961 @item -fdump-rtl-seqabstr
13962 @opindex fdump-rtl-seqabstr
13963 Dump after common sequence discovery.
13965 @item -fdump-rtl-shorten
13966 @opindex fdump-rtl-shorten
13967 Dump after shortening branches.
13969 @item -fdump-rtl-sibling
13970 @opindex fdump-rtl-sibling
13971 Dump after sibling call optimizations.
13973 @item -fdump-rtl-split1
13974 @itemx -fdump-rtl-split2
13975 @itemx -fdump-rtl-split3
13976 @itemx -fdump-rtl-split4
13977 @itemx -fdump-rtl-split5
13978 @opindex fdump-rtl-split1
13979 @opindex fdump-rtl-split2
13980 @opindex fdump-rtl-split3
13981 @opindex fdump-rtl-split4
13982 @opindex fdump-rtl-split5
13983 These options enable dumping after five rounds of
13984 instruction splitting.
13986 @item -fdump-rtl-sms
13987 @opindex fdump-rtl-sms
13988 Dump after modulo scheduling. This pass is only run on some
13991 @item -fdump-rtl-stack
13992 @opindex fdump-rtl-stack
13993 Dump after conversion from GCC's ``flat register file'' registers to the
13994 x87's stack-like registers. This pass is only run on x86 variants.
13996 @item -fdump-rtl-subreg1
13997 @itemx -fdump-rtl-subreg2
13998 @opindex fdump-rtl-subreg1
13999 @opindex fdump-rtl-subreg2
14000 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
14001 the two subreg expansion passes.
14003 @item -fdump-rtl-unshare
14004 @opindex fdump-rtl-unshare
14005 Dump after all rtl has been unshared.
14007 @item -fdump-rtl-vartrack
14008 @opindex fdump-rtl-vartrack
14009 Dump after variable tracking.
14011 @item -fdump-rtl-vregs
14012 @opindex fdump-rtl-vregs
14013 Dump after converting virtual registers to hard registers.
14015 @item -fdump-rtl-web
14016 @opindex fdump-rtl-web
14017 Dump after live range splitting.
14019 @item -fdump-rtl-regclass
14020 @itemx -fdump-rtl-subregs_of_mode_init
14021 @itemx -fdump-rtl-subregs_of_mode_finish
14022 @itemx -fdump-rtl-dfinit
14023 @itemx -fdump-rtl-dfinish
14024 @opindex fdump-rtl-regclass
14025 @opindex fdump-rtl-subregs_of_mode_init
14026 @opindex fdump-rtl-subregs_of_mode_finish
14027 @opindex fdump-rtl-dfinit
14028 @opindex fdump-rtl-dfinish
14029 These dumps are defined but always produce empty files.
14032 @itemx -fdump-rtl-all
14034 @opindex fdump-rtl-all
14035 Produce all the dumps listed above.
14039 Annotate the assembler output with miscellaneous debugging information.
14043 Dump all macro definitions, at the end of preprocessing, in addition to
14048 Produce a core dump whenever an error occurs.
14052 Annotate the assembler output with a comment indicating which
14053 pattern and alternative is used. The length and cost of each instruction are
14058 Dump the RTL in the assembler output as a comment before each instruction.
14059 Also turns on @option{-dp} annotation.
14063 Just generate RTL for a function instead of compiling it. Usually used
14064 with @option{-fdump-rtl-expand}.
14068 @opindex fdump-debug
14069 Dump debugging information generated during the debug
14072 @item -fdump-earlydebug
14073 @opindex fdump-earlydebug
14074 Dump debugging information generated during the early debug
14077 @item -fdump-noaddr
14078 @opindex fdump-noaddr
14079 When doing debugging dumps, suppress address output. This makes it more
14080 feasible to use diff on debugging dumps for compiler invocations with
14081 different compiler binaries and/or different
14082 text / bss / data / heap / stack / dso start locations.
14085 @opindex freport-bug
14086 Collect and dump debug information into a temporary file if an
14087 internal compiler error (ICE) occurs.
14089 @item -fdump-unnumbered
14090 @opindex fdump-unnumbered
14091 When doing debugging dumps, suppress instruction numbers and address output.
14092 This makes it more feasible to use diff on debugging dumps for compiler
14093 invocations with different options, in particular with and without
14096 @item -fdump-unnumbered-links
14097 @opindex fdump-unnumbered-links
14098 When doing debugging dumps (see @option{-d} option above), suppress
14099 instruction numbers for the links to the previous and next instructions
14102 @item -fdump-ipa-@var{switch}
14103 @itemx -fdump-ipa-@var{switch}-@var{options}
14105 Control the dumping at various stages of inter-procedural analysis
14106 language tree to a file. The file name is generated by appending a
14107 switch specific suffix to the source file name, and the file is created
14108 in the same directory as the output file. The following dumps are
14113 Enables all inter-procedural analysis dumps.
14116 Dumps information about call-graph optimization, unused function removal,
14117 and inlining decisions.
14120 Dump after function inlining.
14124 Additionally, the options @option{-optimized}, @option{-missed},
14125 @option{-note}, and @option{-all} can be provided, with the same meaning
14126 as for @option{-fopt-info}, defaulting to @option{-optimized}.
14128 For example, @option{-fdump-ipa-inline-optimized-missed} will emit
14129 information on callsites that were inlined, along with callsites
14130 that were not inlined.
14132 By default, the dump will contain messages about successful
14133 optimizations (equivalent to @option{-optimized}) together with
14134 low-level details about the analysis.
14136 @item -fdump-lang-all
14137 @itemx -fdump-lang-@var{switch}
14138 @itemx -fdump-lang-@var{switch}-@var{options}
14139 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
14140 @opindex fdump-lang-all
14141 @opindex fdump-lang
14142 Control the dumping of language-specific information. The @var{options}
14143 and @var{filename} portions behave as described in the
14144 @option{-fdump-tree} option. The following @var{switch} values are
14150 Enable all language-specific dumps.
14153 Dump class hierarchy information. Virtual table information is emitted
14154 unless '@option{slim}' is specified. This option is applicable to C++ only.
14157 Dump the raw internal tree data. This option is applicable to C++ only.
14161 @item -fdump-passes
14162 @opindex fdump-passes
14163 Print on @file{stderr} the list of optimization passes that are turned
14164 on and off by the current command-line options.
14166 @item -fdump-statistics-@var{option}
14167 @opindex fdump-statistics
14168 Enable and control dumping of pass statistics in a separate file. The
14169 file name is generated by appending a suffix ending in
14170 @samp{.statistics} to the source file name, and the file is created in
14171 the same directory as the output file. If the @samp{-@var{option}}
14172 form is used, @samp{-stats} causes counters to be summed over the
14173 whole compilation unit while @samp{-details} dumps every event as
14174 the passes generate them. The default with no option is to sum
14175 counters for each function compiled.
14177 @item -fdump-tree-all
14178 @itemx -fdump-tree-@var{switch}
14179 @itemx -fdump-tree-@var{switch}-@var{options}
14180 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
14181 @opindex fdump-tree-all
14182 @opindex fdump-tree
14183 Control the dumping at various stages of processing the intermediate
14184 language tree to a file. If the @samp{-@var{options}}
14185 form is used, @var{options} is a list of @samp{-} separated options
14186 which control the details of the dump. Not all options are applicable
14187 to all dumps; those that are not meaningful are ignored. The
14188 following options are available
14192 Print the address of each node. Usually this is not meaningful as it
14193 changes according to the environment and source file. Its primary use
14194 is for tying up a dump file with a debug environment.
14196 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
14197 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
14198 use working backward from mangled names in the assembly file.
14200 When dumping front-end intermediate representations, inhibit dumping
14201 of members of a scope or body of a function merely because that scope
14202 has been reached. Only dump such items when they are directly reachable
14203 by some other path.
14205 When dumping pretty-printed trees, this option inhibits dumping the
14206 bodies of control structures.
14208 When dumping RTL, print the RTL in slim (condensed) form instead of
14209 the default LISP-like representation.
14211 Print a raw representation of the tree. By default, trees are
14212 pretty-printed into a C-like representation.
14214 Enable more detailed dumps (not honored by every dump option). Also
14215 include information from the optimization passes.
14217 Enable dumping various statistics about the pass (not honored by every dump
14220 Enable showing basic block boundaries (disabled in raw dumps).
14222 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
14223 dump a representation of the control flow graph suitable for viewing with
14224 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
14225 the file is pretty-printed as a subgraph, so that GraphViz can render them
14226 all in a single plot.
14228 This option currently only works for RTL dumps, and the RTL is always
14229 dumped in slim form.
14231 Enable showing virtual operands for every statement.
14233 Enable showing line numbers for statements.
14235 Enable showing the unique ID (@code{DECL_UID}) for each variable.
14237 Enable showing the tree dump for each statement.
14239 Enable showing the EH region number holding each statement.
14241 Enable showing scalar evolution analysis details.
14243 Enable showing optimization information (only available in certain
14246 Enable showing missed optimization information (only available in certain
14249 Enable other detailed optimization information (only available in
14252 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
14253 and @option{lineno}.
14255 Turn on all optimization options, i.e., @option{optimized},
14256 @option{missed}, and @option{note}.
14259 To determine what tree dumps are available or find the dump for a pass
14260 of interest follow the steps below.
14264 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
14265 look for a code that corresponds to the pass you are interested in.
14266 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
14267 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
14268 The number at the end distinguishes distinct invocations of the same pass.
14270 To enable the creation of the dump file, append the pass code to
14271 the @option{-fdump-} option prefix and invoke GCC with it. For example,
14272 to enable the dump from the Early Value Range Propagation pass, invoke
14273 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
14274 specify the name of the dump file. If you don't specify one, GCC
14275 creates as described below.
14277 Find the pass dump in a file whose name is composed of three components
14278 separated by a period: the name of the source file GCC was invoked to
14279 compile, a numeric suffix indicating the pass number followed by the
14280 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
14281 and finally the pass code. For example, the Early VRP pass dump might
14282 be in a file named @file{myfile.c.038t.evrp} in the current working
14283 directory. Note that the numeric codes are not stable and may change
14284 from one version of GCC to another.
14288 @itemx -fopt-info-@var{options}
14289 @itemx -fopt-info-@var{options}=@var{filename}
14291 Controls optimization dumps from various optimization passes. If the
14292 @samp{-@var{options}} form is used, @var{options} is a list of
14293 @samp{-} separated option keywords to select the dump details and
14296 The @var{options} can be divided into three groups:
14299 options describing what kinds of messages should be emitted,
14301 options describing the verbosity of the dump, and
14303 options describing which optimizations should be included.
14305 The options from each group can be freely mixed as they are
14306 non-overlapping. However, in case of any conflicts,
14307 the later options override the earlier options on the command
14310 The following options control which kinds of messages should be emitted:
14314 Print information when an optimization is successfully applied. It is
14315 up to a pass to decide which information is relevant. For example, the
14316 vectorizer passes print the source location of loops which are
14317 successfully vectorized.
14319 Print information about missed optimizations. Individual passes
14320 control which information to include in the output.
14322 Print verbose information about optimizations, such as certain
14323 transformations, more detailed messages about decisions etc.
14325 Print detailed optimization information. This includes
14326 @samp{optimized}, @samp{missed}, and @samp{note}.
14329 The following option controls the dump verbosity:
14333 By default, only ``high-level'' messages are emitted. This option enables
14334 additional, more detailed, messages, which are likely to only be of interest
14338 One or more of the following option keywords can be used to describe a
14339 group of optimizations:
14343 Enable dumps from all interprocedural optimizations.
14345 Enable dumps from all loop optimizations.
14347 Enable dumps from all inlining optimizations.
14349 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
14351 Enable dumps from all vectorization optimizations.
14353 Enable dumps from all optimizations. This is a superset of
14354 the optimization groups listed above.
14357 If @var{options} is
14358 omitted, it defaults to @samp{optimized-optall}, which means to dump messages
14359 about successful optimizations from all the passes, omitting messages
14360 that are treated as ``internals''.
14362 If the @var{filename} is provided, then the dumps from all the
14363 applicable optimizations are concatenated into the @var{filename}.
14364 Otherwise the dump is output onto @file{stderr}. Though multiple
14365 @option{-fopt-info} options are accepted, only one of them can include
14366 a @var{filename}. If other filenames are provided then all but the
14367 first such option are ignored.
14369 Note that the output @var{filename} is overwritten
14370 in case of multiple translation units. If a combined output from
14371 multiple translation units is desired, @file{stderr} should be used
14374 In the following example, the optimization info is output to
14383 gcc -O3 -fopt-info-missed=missed.all
14387 outputs missed optimization report from all the passes into
14388 @file{missed.all}, and this one:
14391 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
14395 prints information about missed optimization opportunities from
14396 vectorization passes on @file{stderr}.
14397 Note that @option{-fopt-info-vec-missed} is equivalent to
14398 @option{-fopt-info-missed-vec}. The order of the optimization group
14399 names and message types listed after @option{-fopt-info} does not matter.
14401 As another example,
14403 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
14407 outputs information about missed optimizations as well as
14408 optimized locations from all the inlining passes into
14414 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
14418 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
14419 in conflict since only one output file is allowed. In this case, only
14420 the first option takes effect and the subsequent options are
14421 ignored. Thus only @file{vec.miss} is produced which contains
14422 dumps from the vectorizer about missed opportunities.
14424 @item -fsave-optimization-record
14425 @opindex fsave-optimization-record
14426 Write a SRCFILE.opt-record.json file detailing what optimizations
14427 were performed, for those optimizations that support @option{-fopt-info}.
14429 This option is experimental and the format of the data within the JSON
14430 file is subject to change.
14432 It is roughly equivalent to a machine-readable version of
14433 @option{-fopt-info-all}, as a collection of messages with source file,
14434 line number and column number, with the following additional data for
14440 the execution count of the code being optimized, along with metadata about
14441 whether this was from actual profile data, or just an estimate, allowing
14442 consumers to prioritize messages by code hotness,
14445 the function name of the code being optimized, where applicable,
14448 the ``inlining chain'' for the code being optimized, so that when
14449 a function is inlined into several different places (which might
14450 themselves be inlined), the reader can distinguish between the copies,
14453 objects identifying those parts of the message that refer to expressions,
14454 statements or symbol-table nodes, which of these categories they are, and,
14455 when available, their source code location,
14458 the GCC pass that emitted the message, and
14461 the location in GCC's own code from which the message was emitted
14465 Additionally, some messages are logically nested within other
14466 messages, reflecting implementation details of the optimization
14469 @item -fsched-verbose=@var{n}
14470 @opindex fsched-verbose
14471 On targets that use instruction scheduling, this option controls the
14472 amount of debugging output the scheduler prints to the dump files.
14474 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
14475 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
14476 For @var{n} greater than one, it also output basic block probabilities,
14477 detailed ready list information and unit/insn info. For @var{n} greater
14478 than two, it includes RTL at abort point, control-flow and regions info.
14479 And for @var{n} over four, @option{-fsched-verbose} also includes
14484 @item -fenable-@var{kind}-@var{pass}
14485 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
14489 This is a set of options that are used to explicitly disable/enable
14490 optimization passes. These options are intended for use for debugging GCC.
14491 Compiler users should use regular options for enabling/disabling
14496 @item -fdisable-ipa-@var{pass}
14497 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
14498 statically invoked in the compiler multiple times, the pass name should be
14499 appended with a sequential number starting from 1.
14501 @item -fdisable-rtl-@var{pass}
14502 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
14503 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
14504 statically invoked in the compiler multiple times, the pass name should be
14505 appended with a sequential number starting from 1. @var{range-list} is a
14506 comma-separated list of function ranges or assembler names. Each range is a number
14507 pair separated by a colon. The range is inclusive in both ends. If the range
14508 is trivial, the number pair can be simplified as a single number. If the
14509 function's call graph node's @var{uid} falls within one of the specified ranges,
14510 the @var{pass} is disabled for that function. The @var{uid} is shown in the
14511 function header of a dump file, and the pass names can be dumped by using
14512 option @option{-fdump-passes}.
14514 @item -fdisable-tree-@var{pass}
14515 @itemx -fdisable-tree-@var{pass}=@var{range-list}
14516 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
14519 @item -fenable-ipa-@var{pass}
14520 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
14521 statically invoked in the compiler multiple times, the pass name should be
14522 appended with a sequential number starting from 1.
14524 @item -fenable-rtl-@var{pass}
14525 @itemx -fenable-rtl-@var{pass}=@var{range-list}
14526 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
14527 description and examples.
14529 @item -fenable-tree-@var{pass}
14530 @itemx -fenable-tree-@var{pass}=@var{range-list}
14531 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
14532 of option arguments.
14536 Here are some examples showing uses of these options.
14540 # disable ccp1 for all functions
14541 -fdisable-tree-ccp1
14542 # disable complete unroll for function whose cgraph node uid is 1
14543 -fenable-tree-cunroll=1
14544 # disable gcse2 for functions at the following ranges [1,1],
14545 # [300,400], and [400,1000]
14546 # disable gcse2 for functions foo and foo2
14547 -fdisable-rtl-gcse2=foo,foo2
14548 # disable early inlining
14549 -fdisable-tree-einline
14550 # disable ipa inlining
14551 -fdisable-ipa-inline
14552 # enable tree full unroll
14553 -fenable-tree-unroll
14558 @itemx -fchecking=@var{n}
14560 @opindex fno-checking
14561 Enable internal consistency checking. The default depends on
14562 the compiler configuration. @option{-fchecking=2} enables further
14563 internal consistency checking that might affect code generation.
14565 @item -frandom-seed=@var{string}
14566 @opindex frandom-seed
14567 This option provides a seed that GCC uses in place of
14568 random numbers in generating certain symbol names
14569 that have to be different in every compiled file. It is also used to
14570 place unique stamps in coverage data files and the object files that
14571 produce them. You can use the @option{-frandom-seed} option to produce
14572 reproducibly identical object files.
14574 The @var{string} can either be a number (decimal, octal or hex) or an
14575 arbitrary string (in which case it's converted to a number by
14578 The @var{string} should be different for every file you compile.
14581 @itemx -save-temps=cwd
14582 @opindex save-temps
14583 Store the usual ``temporary'' intermediate files permanently; place them
14584 in the current directory and name them based on the source file. Thus,
14585 compiling @file{foo.c} with @option{-c -save-temps} produces files
14586 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
14587 preprocessed @file{foo.i} output file even though the compiler now
14588 normally uses an integrated preprocessor.
14590 When used in combination with the @option{-x} command-line option,
14591 @option{-save-temps} is sensible enough to avoid over writing an
14592 input source file with the same extension as an intermediate file.
14593 The corresponding intermediate file may be obtained by renaming the
14594 source file before using @option{-save-temps}.
14596 If you invoke GCC in parallel, compiling several different source
14597 files that share a common base name in different subdirectories or the
14598 same source file compiled for multiple output destinations, it is
14599 likely that the different parallel compilers will interfere with each
14600 other, and overwrite the temporary files. For instance:
14603 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
14604 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
14607 may result in @file{foo.i} and @file{foo.o} being written to
14608 simultaneously by both compilers.
14610 @item -save-temps=obj
14611 @opindex save-temps=obj
14612 Store the usual ``temporary'' intermediate files permanently. If the
14613 @option{-o} option is used, the temporary files are based on the
14614 object file. If the @option{-o} option is not used, the
14615 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
14620 gcc -save-temps=obj -c foo.c
14621 gcc -save-temps=obj -c bar.c -o dir/xbar.o
14622 gcc -save-temps=obj foobar.c -o dir2/yfoobar
14626 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
14627 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
14628 @file{dir2/yfoobar.o}.
14630 @item -time@r{[}=@var{file}@r{]}
14632 Report the CPU time taken by each subprocess in the compilation
14633 sequence. For C source files, this is the compiler proper and assembler
14634 (plus the linker if linking is done).
14636 Without the specification of an output file, the output looks like this:
14643 The first number on each line is the ``user time'', that is time spent
14644 executing the program itself. The second number is ``system time'',
14645 time spent executing operating system routines on behalf of the program.
14646 Both numbers are in seconds.
14648 With the specification of an output file, the output is appended to the
14649 named file, and it looks like this:
14652 0.12 0.01 cc1 @var{options}
14653 0.00 0.01 as @var{options}
14656 The ``user time'' and the ``system time'' are moved before the program
14657 name, and the options passed to the program are displayed, so that one
14658 can later tell what file was being compiled, and with which options.
14660 @item -fdump-final-insns@r{[}=@var{file}@r{]}
14661 @opindex fdump-final-insns
14662 Dump the final internal representation (RTL) to @var{file}. If the
14663 optional argument is omitted (or if @var{file} is @code{.}), the name
14664 of the dump file is determined by appending @code{.gkd} to the
14665 compilation output file name.
14667 @item -fcompare-debug@r{[}=@var{opts}@r{]}
14668 @opindex fcompare-debug
14669 @opindex fno-compare-debug
14670 If no error occurs during compilation, run the compiler a second time,
14671 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
14672 passed to the second compilation. Dump the final internal
14673 representation in both compilations, and print an error if they differ.
14675 If the equal sign is omitted, the default @option{-gtoggle} is used.
14677 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
14678 and nonzero, implicitly enables @option{-fcompare-debug}. If
14679 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
14680 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
14683 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
14684 is equivalent to @option{-fno-compare-debug}, which disables the dumping
14685 of the final representation and the second compilation, preventing even
14686 @env{GCC_COMPARE_DEBUG} from taking effect.
14688 To verify full coverage during @option{-fcompare-debug} testing, set
14689 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
14690 which GCC rejects as an invalid option in any actual compilation
14691 (rather than preprocessing, assembly or linking). To get just a
14692 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
14693 not overridden} will do.
14695 @item -fcompare-debug-second
14696 @opindex fcompare-debug-second
14697 This option is implicitly passed to the compiler for the second
14698 compilation requested by @option{-fcompare-debug}, along with options to
14699 silence warnings, and omitting other options that would cause the compiler
14700 to produce output to files or to standard output as a side effect. Dump
14701 files and preserved temporary files are renamed so as to contain the
14702 @code{.gk} additional extension during the second compilation, to avoid
14703 overwriting those generated by the first.
14705 When this option is passed to the compiler driver, it causes the
14706 @emph{first} compilation to be skipped, which makes it useful for little
14707 other than debugging the compiler proper.
14711 Turn off generation of debug info, if leaving out this option
14712 generates it, or turn it on at level 2 otherwise. The position of this
14713 argument in the command line does not matter; it takes effect after all
14714 other options are processed, and it does so only once, no matter how
14715 many times it is given. This is mainly intended to be used with
14716 @option{-fcompare-debug}.
14718 @item -fvar-tracking-assignments-toggle
14719 @opindex fvar-tracking-assignments-toggle
14720 @opindex fno-var-tracking-assignments-toggle
14721 Toggle @option{-fvar-tracking-assignments}, in the same way that
14722 @option{-gtoggle} toggles @option{-g}.
14726 Makes the compiler print out each function name as it is compiled, and
14727 print some statistics about each pass when it finishes.
14729 @item -ftime-report
14730 @opindex ftime-report
14731 Makes the compiler print some statistics about the time consumed by each
14732 pass when it finishes.
14734 @item -ftime-report-details
14735 @opindex ftime-report-details
14736 Record the time consumed by infrastructure parts separately for each pass.
14738 @item -fira-verbose=@var{n}
14739 @opindex fira-verbose
14740 Control the verbosity of the dump file for the integrated register allocator.
14741 The default value is 5. If the value @var{n} is greater or equal to 10,
14742 the dump output is sent to stderr using the same format as @var{n} minus 10.
14745 @opindex flto-report
14746 Prints a report with internal details on the workings of the link-time
14747 optimizer. The contents of this report vary from version to version.
14748 It is meant to be useful to GCC developers when processing object
14749 files in LTO mode (via @option{-flto}).
14751 Disabled by default.
14753 @item -flto-report-wpa
14754 @opindex flto-report-wpa
14755 Like @option{-flto-report}, but only print for the WPA phase of Link
14759 @opindex fmem-report
14760 Makes the compiler print some statistics about permanent memory
14761 allocation when it finishes.
14763 @item -fmem-report-wpa
14764 @opindex fmem-report-wpa
14765 Makes the compiler print some statistics about permanent memory
14766 allocation for the WPA phase only.
14768 @item -fpre-ipa-mem-report
14769 @opindex fpre-ipa-mem-report
14770 @item -fpost-ipa-mem-report
14771 @opindex fpost-ipa-mem-report
14772 Makes the compiler print some statistics about permanent memory
14773 allocation before or after interprocedural optimization.
14775 @item -fprofile-report
14776 @opindex fprofile-report
14777 Makes the compiler print some statistics about consistency of the
14778 (estimated) profile and effect of individual passes.
14780 @item -fstack-usage
14781 @opindex fstack-usage
14782 Makes the compiler output stack usage information for the program, on a
14783 per-function basis. The filename for the dump is made by appending
14784 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
14785 the output file, if explicitly specified and it is not an executable,
14786 otherwise it is the basename of the source file. An entry is made up
14791 The name of the function.
14795 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
14798 The qualifier @code{static} means that the function manipulates the stack
14799 statically: a fixed number of bytes are allocated for the frame on function
14800 entry and released on function exit; no stack adjustments are otherwise made
14801 in the function. The second field is this fixed number of bytes.
14803 The qualifier @code{dynamic} means that the function manipulates the stack
14804 dynamically: in addition to the static allocation described above, stack
14805 adjustments are made in the body of the function, for example to push/pop
14806 arguments around function calls. If the qualifier @code{bounded} is also
14807 present, the amount of these adjustments is bounded at compile time and
14808 the second field is an upper bound of the total amount of stack used by
14809 the function. If it is not present, the amount of these adjustments is
14810 not bounded at compile time and the second field only represents the
14815 Emit statistics about front-end processing at the end of the compilation.
14816 This option is supported only by the C++ front end, and
14817 the information is generally only useful to the G++ development team.
14819 @item -fdbg-cnt-list
14820 @opindex fdbg-cnt-list
14821 Print the name and the counter upper bound for all debug counters.
14824 @item -fdbg-cnt=@var{counter-value-list}
14826 Set the internal debug counter lower and upper bound. @var{counter-value-list}
14827 is a comma-separated list of @var{name}:@var{lower_bound}:@var{upper_bound}
14828 tuples which sets the lower and the upper bound of each debug
14829 counter @var{name}. The @var{lower_bound} is optional and is zero
14830 initialized if not set.
14831 All debug counters have the initial upper bound of @code{UINT_MAX};
14832 thus @code{dbg_cnt} returns true always unless the upper bound
14833 is set by this option.
14834 For example, with @option{-fdbg-cnt=dce:2:4,tail_call:10},
14835 @code{dbg_cnt(dce)} returns true only for third and fourth invocation.
14836 For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations.
14838 @item -print-file-name=@var{library}
14839 @opindex print-file-name
14840 Print the full absolute name of the library file @var{library} that
14841 would be used when linking---and don't do anything else. With this
14842 option, GCC does not compile or link anything; it just prints the
14845 @item -print-multi-directory
14846 @opindex print-multi-directory
14847 Print the directory name corresponding to the multilib selected by any
14848 other switches present in the command line. This directory is supposed
14849 to exist in @env{GCC_EXEC_PREFIX}.
14851 @item -print-multi-lib
14852 @opindex print-multi-lib
14853 Print the mapping from multilib directory names to compiler switches
14854 that enable them. The directory name is separated from the switches by
14855 @samp{;}, and each switch starts with an @samp{@@} instead of the
14856 @samp{-}, without spaces between multiple switches. This is supposed to
14857 ease shell processing.
14859 @item -print-multi-os-directory
14860 @opindex print-multi-os-directory
14861 Print the path to OS libraries for the selected
14862 multilib, relative to some @file{lib} subdirectory. If OS libraries are
14863 present in the @file{lib} subdirectory and no multilibs are used, this is
14864 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
14865 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
14866 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
14867 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
14869 @item -print-multiarch
14870 @opindex print-multiarch
14871 Print the path to OS libraries for the selected multiarch,
14872 relative to some @file{lib} subdirectory.
14874 @item -print-prog-name=@var{program}
14875 @opindex print-prog-name
14876 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
14878 @item -print-libgcc-file-name
14879 @opindex print-libgcc-file-name
14880 Same as @option{-print-file-name=libgcc.a}.
14882 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
14883 but you do want to link with @file{libgcc.a}. You can do:
14886 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
14889 @item -print-search-dirs
14890 @opindex print-search-dirs
14891 Print the name of the configured installation directory and a list of
14892 program and library directories @command{gcc} searches---and don't do anything else.
14894 This is useful when @command{gcc} prints the error message
14895 @samp{installation problem, cannot exec cpp0: No such file or directory}.
14896 To resolve this you either need to put @file{cpp0} and the other compiler
14897 components where @command{gcc} expects to find them, or you can set the environment
14898 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
14899 Don't forget the trailing @samp{/}.
14900 @xref{Environment Variables}.
14902 @item -print-sysroot
14903 @opindex print-sysroot
14904 Print the target sysroot directory that is used during
14905 compilation. This is the target sysroot specified either at configure
14906 time or using the @option{--sysroot} option, possibly with an extra
14907 suffix that depends on compilation options. If no target sysroot is
14908 specified, the option prints nothing.
14910 @item -print-sysroot-headers-suffix
14911 @opindex print-sysroot-headers-suffix
14912 Print the suffix added to the target sysroot when searching for
14913 headers, or give an error if the compiler is not configured with such
14914 a suffix---and don't do anything else.
14917 @opindex dumpmachine
14918 Print the compiler's target machine (for example,
14919 @samp{i686-pc-linux-gnu})---and don't do anything else.
14922 @opindex dumpversion
14923 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
14924 anything else. This is the compiler version used in filesystem paths and
14925 specs. Depending on how the compiler has been configured it can be just
14926 a single number (major version), two numbers separated by a dot (major and
14927 minor version) or three numbers separated by dots (major, minor and patchlevel
14930 @item -dumpfullversion
14931 @opindex dumpfullversion
14932 Print the full compiler version---and don't do anything else. The output is
14933 always three numbers separated by dots, major, minor and patchlevel version.
14937 Print the compiler's built-in specs---and don't do anything else. (This
14938 is used when GCC itself is being built.) @xref{Spec Files}.
14941 @node Submodel Options
14942 @section Machine-Dependent Options
14943 @cindex submodel options
14944 @cindex specifying hardware config
14945 @cindex hardware models and configurations, specifying
14946 @cindex target-dependent options
14947 @cindex machine-dependent options
14949 Each target machine supported by GCC can have its own options---for
14950 example, to allow you to compile for a particular processor variant or
14951 ABI, or to control optimizations specific to that machine. By
14952 convention, the names of machine-specific options start with
14955 Some configurations of the compiler also support additional target-specific
14956 options, usually for compatibility with other compilers on the same
14959 @c This list is ordered alphanumerically by subsection name.
14960 @c It should be the same order and spelling as these options are listed
14961 @c in Machine Dependent Options
14964 * AArch64 Options::
14965 * Adapteva Epiphany Options::
14969 * Blackfin Options::
14975 * DEC Alpha Options::
14979 * GNU/Linux Options::
14989 * MicroBlaze Options::
14992 * MN10300 Options::
14996 * Nios II Options::
14997 * Nvidia PTX Options::
14998 * OpenRISC Options::
15000 * picoChip Options::
15001 * PowerPC Options::
15002 * PowerPC SPE Options::
15005 * RS/6000 and PowerPC Options::
15007 * S/390 and zSeries Options::
15010 * Solaris 2 Options::
15013 * System V Options::
15014 * TILE-Gx Options::
15015 * TILEPro Options::
15020 * VxWorks Options::
15022 * x86 Windows Options::
15023 * Xstormy16 Options::
15025 * zSeries Options::
15028 @node AArch64 Options
15029 @subsection AArch64 Options
15030 @cindex AArch64 Options
15032 These options are defined for AArch64 implementations:
15036 @item -mabi=@var{name}
15038 Generate code for the specified data model. Permissible values
15039 are @samp{ilp32} for SysV-like data model where int, long int and pointers
15040 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
15041 but long int and pointers are 64 bits.
15043 The default depends on the specific target configuration. Note that
15044 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
15045 entire program with the same ABI, and link with a compatible set of libraries.
15048 @opindex mbig-endian
15049 Generate big-endian code. This is the default when GCC is configured for an
15050 @samp{aarch64_be-*-*} target.
15052 @item -mgeneral-regs-only
15053 @opindex mgeneral-regs-only
15054 Generate code which uses only the general-purpose registers. This will prevent
15055 the compiler from using floating-point and Advanced SIMD registers but will not
15056 impose any restrictions on the assembler.
15058 @item -mlittle-endian
15059 @opindex mlittle-endian
15060 Generate little-endian code. This is the default when GCC is configured for an
15061 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
15063 @item -mcmodel=tiny
15064 @opindex mcmodel=tiny
15065 Generate code for the tiny code model. The program and its statically defined
15066 symbols must be within 1MB of each other. Programs can be statically or
15067 dynamically linked.
15069 @item -mcmodel=small
15070 @opindex mcmodel=small
15071 Generate code for the small code model. The program and its statically defined
15072 symbols must be within 4GB of each other. Programs can be statically or
15073 dynamically linked. This is the default code model.
15075 @item -mcmodel=large
15076 @opindex mcmodel=large
15077 Generate code for the large code model. This makes no assumptions about
15078 addresses and sizes of sections. Programs can be statically linked only.
15080 @item -mstrict-align
15081 @itemx -mno-strict-align
15082 @opindex mstrict-align
15083 @opindex mno-strict-align
15084 Avoid or allow generating memory accesses that may not be aligned on a natural
15085 object boundary as described in the architecture specification.
15087 @item -momit-leaf-frame-pointer
15088 @itemx -mno-omit-leaf-frame-pointer
15089 @opindex momit-leaf-frame-pointer
15090 @opindex mno-omit-leaf-frame-pointer
15091 Omit or keep the frame pointer in leaf functions. The former behavior is the
15094 @item -mtls-dialect=desc
15095 @opindex mtls-dialect=desc
15096 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
15097 of TLS variables. This is the default.
15099 @item -mtls-dialect=traditional
15100 @opindex mtls-dialect=traditional
15101 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
15104 @item -mtls-size=@var{size}
15106 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
15107 This option requires binutils 2.26 or newer.
15109 @item -mfix-cortex-a53-835769
15110 @itemx -mno-fix-cortex-a53-835769
15111 @opindex mfix-cortex-a53-835769
15112 @opindex mno-fix-cortex-a53-835769
15113 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
15114 This involves inserting a NOP instruction between memory instructions and
15115 64-bit integer multiply-accumulate instructions.
15117 @item -mfix-cortex-a53-843419
15118 @itemx -mno-fix-cortex-a53-843419
15119 @opindex mfix-cortex-a53-843419
15120 @opindex mno-fix-cortex-a53-843419
15121 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
15122 This erratum workaround is made at link time and this will only pass the
15123 corresponding flag to the linker.
15125 @item -mlow-precision-recip-sqrt
15126 @itemx -mno-low-precision-recip-sqrt
15127 @opindex mlow-precision-recip-sqrt
15128 @opindex mno-low-precision-recip-sqrt
15129 Enable or disable the reciprocal square root approximation.
15130 This option only has an effect if @option{-ffast-math} or
15131 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15132 precision of reciprocal square root results to about 16 bits for
15133 single precision and to 32 bits for double precision.
15135 @item -mlow-precision-sqrt
15136 @itemx -mno-low-precision-sqrt
15137 @opindex mlow-precision-sqrt
15138 @opindex mno-low-precision-sqrt
15139 Enable or disable the square root approximation.
15140 This option only has an effect if @option{-ffast-math} or
15141 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15142 precision of square root results to about 16 bits for
15143 single precision and to 32 bits for double precision.
15144 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
15146 @item -mlow-precision-div
15147 @itemx -mno-low-precision-div
15148 @opindex mlow-precision-div
15149 @opindex mno-low-precision-div
15150 Enable or disable the division approximation.
15151 This option only has an effect if @option{-ffast-math} or
15152 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15153 precision of division results to about 16 bits for
15154 single precision and to 32 bits for double precision.
15156 @item -mtrack-speculation
15157 @itemx -mno-track-speculation
15158 Enable or disable generation of additional code to track speculative
15159 execution through conditional branches. The tracking state can then
15160 be used by the compiler when expanding calls to
15161 @code{__builtin_speculation_safe_copy} to permit a more efficient code
15162 sequence to be generated.
15164 @item -march=@var{name}
15166 Specify the name of the target architecture and, optionally, one or
15167 more feature modifiers. This option has the form
15168 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
15170 The permissible values for @var{arch} are @samp{armv8-a},
15171 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a} or @samp{armv8.4-a}
15174 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
15175 support for the ARMv8.4-A architecture extensions.
15177 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
15178 support for the ARMv8.3-A architecture extensions.
15180 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
15181 support for the ARMv8.2-A architecture extensions.
15183 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
15184 support for the ARMv8.1-A architecture extension. In particular, it
15185 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
15187 The value @samp{native} is available on native AArch64 GNU/Linux and
15188 causes the compiler to pick the architecture of the host system. This
15189 option has no effect if the compiler is unable to recognize the
15190 architecture of the host system,
15192 The permissible values for @var{feature} are listed in the sub-section
15193 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15194 Feature Modifiers}. Where conflicting feature modifiers are
15195 specified, the right-most feature is used.
15197 GCC uses @var{name} to determine what kind of instructions it can emit
15198 when generating assembly code. If @option{-march} is specified
15199 without either of @option{-mtune} or @option{-mcpu} also being
15200 specified, the code is tuned to perform well across a range of target
15201 processors implementing the target architecture.
15203 @item -mtune=@var{name}
15205 Specify the name of the target processor for which GCC should tune the
15206 performance of the code. Permissible values for this option are:
15207 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15208 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15209 @samp{cortex-a76}, @samp{ares}, @samp{exynos-m1}, @samp{falkor},
15210 @samp{qdf24xx}, @samp{saphira}, @samp{phecda}, @samp{xgene1}, @samp{vulcan},
15211 @samp{thunderx}, @samp{thunderxt88}, @samp{thunderxt88p1}, @samp{thunderxt81},
15212 @samp{tsv110}, @samp{thunderxt83}, @samp{thunderx2t99},
15213 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15214 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15215 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}
15218 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15219 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15220 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC
15221 should tune for a big.LITTLE system.
15223 Additionally on native AArch64 GNU/Linux systems the value
15224 @samp{native} tunes performance to the host system. This option has no effect
15225 if the compiler is unable to recognize the processor of the host system.
15227 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
15228 are specified, the code is tuned to perform well across a range
15229 of target processors.
15231 This option cannot be suffixed by feature modifiers.
15233 @item -mcpu=@var{name}
15235 Specify the name of the target processor, optionally suffixed by one
15236 or more feature modifiers. This option has the form
15237 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
15238 the permissible values for @var{cpu} are the same as those available
15239 for @option{-mtune}. The permissible values for @var{feature} are
15240 documented in the sub-section on
15241 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15242 Feature Modifiers}. Where conflicting feature modifiers are
15243 specified, the right-most feature is used.
15245 GCC uses @var{name} to determine what kind of instructions it can emit when
15246 generating assembly code (as if by @option{-march}) and to determine
15247 the target processor for which to tune for performance (as if
15248 by @option{-mtune}). Where this option is used in conjunction
15249 with @option{-march} or @option{-mtune}, those options take precedence
15250 over the appropriate part of this option.
15252 @item -moverride=@var{string}
15254 Override tuning decisions made by the back-end in response to a
15255 @option{-mtune=} switch. The syntax, semantics, and accepted values
15256 for @var{string} in this option are not guaranteed to be consistent
15259 This option is only intended to be useful when developing GCC.
15261 @item -mverbose-cost-dump
15262 @opindex mverbose-cost-dump
15263 Enable verbose cost model dumping in the debug dump files. This option is
15264 provided for use in debugging the compiler.
15266 @item -mpc-relative-literal-loads
15267 @itemx -mno-pc-relative-literal-loads
15268 @opindex mpc-relative-literal-loads
15269 @opindex mno-pc-relative-literal-loads
15270 Enable or disable PC-relative literal loads. With this option literal pools are
15271 accessed using a single instruction and emitted after each function. This
15272 limits the maximum size of functions to 1MB. This is enabled by default for
15273 @option{-mcmodel=tiny}.
15275 @item -msign-return-address=@var{scope}
15276 @opindex msign-return-address
15277 Select the function scope on which return address signing will be applied.
15278 Permissible values are @samp{none}, which disables return address signing,
15279 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
15280 functions, and @samp{all}, which enables pointer signing for all functions. The
15281 default value is @samp{none}.
15283 @item -msve-vector-bits=@var{bits}
15284 @opindex msve-vector-bits
15285 Specify the number of bits in an SVE vector register. This option only has
15286 an effect when SVE is enabled.
15288 GCC supports two forms of SVE code generation: ``vector-length
15289 agnostic'' output that works with any size of vector register and
15290 ``vector-length specific'' output that only works when the vector
15291 registers are a particular size. Replacing @var{bits} with
15292 @samp{scalable} selects vector-length agnostic output while
15293 replacing it with a number selects vector-length specific output.
15294 The possible lengths in the latter case are: 128, 256, 512, 1024
15295 and 2048. @samp{scalable} is the default.
15297 At present, @samp{-msve-vector-bits=128} produces the same output
15298 as @samp{-msve-vector-bits=scalable}.
15302 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
15303 @anchor{aarch64-feature-modifiers}
15304 @cindex @option{-march} feature modifiers
15305 @cindex @option{-mcpu} feature modifiers
15306 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
15307 the following and their inverses @option{no@var{feature}}:
15311 Enable CRC extension. This is on by default for
15312 @option{-march=armv8.1-a}.
15314 Enable Crypto extension. This also enables Advanced SIMD and floating-point
15317 Enable floating-point instructions. This is on by default for all possible
15318 values for options @option{-march} and @option{-mcpu}.
15320 Enable Advanced SIMD instructions. This also enables floating-point
15321 instructions. This is on by default for all possible values for options
15322 @option{-march} and @option{-mcpu}.
15324 Enable Scalable Vector Extension instructions. This also enables Advanced
15325 SIMD and floating-point instructions.
15327 Enable Large System Extension instructions. This is on by default for
15328 @option{-march=armv8.1-a}.
15330 Enable Round Double Multiply Accumulate instructions. This is on by default
15331 for @option{-march=armv8.1-a}.
15333 Enable FP16 extension. This also enables floating-point instructions.
15335 Enable FP16 fmla extension. This also enables FP16 extensions and
15336 floating-point instructions. This option is enabled by default for @option{-march=armv8.4-a}. Use of this option with architectures prior to Armv8.2-A is not supported.
15339 Enable the RcPc extension. This does not change code generation from GCC,
15340 but is passed on to the assembler, enabling inline asm statements to use
15341 instructions from the RcPc extension.
15343 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
15345 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
15348 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
15350 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
15351 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
15353 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
15354 Use of this option with architectures prior to Armv8.2-A is not supported.
15356 Enable the Statistical Profiling extension. This option is only to enable the
15357 extension at the assembler level and does not affect code generation.
15361 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
15362 which implies @option{fp}.
15363 Conversely, @option{nofp} implies @option{nosimd}, which implies
15364 @option{nocrypto}, @option{noaes} and @option{nosha2}.
15366 @node Adapteva Epiphany Options
15367 @subsection Adapteva Epiphany Options
15369 These @samp{-m} options are defined for Adapteva Epiphany:
15372 @item -mhalf-reg-file
15373 @opindex mhalf-reg-file
15374 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
15375 That allows code to run on hardware variants that lack these registers.
15377 @item -mprefer-short-insn-regs
15378 @opindex mprefer-short-insn-regs
15379 Preferentially allocate registers that allow short instruction generation.
15380 This can result in increased instruction count, so this may either reduce or
15381 increase overall code size.
15383 @item -mbranch-cost=@var{num}
15384 @opindex mbranch-cost
15385 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15386 This cost is only a heuristic and is not guaranteed to produce
15387 consistent results across releases.
15391 Enable the generation of conditional moves.
15393 @item -mnops=@var{num}
15395 Emit @var{num} NOPs before every other generated instruction.
15397 @item -mno-soft-cmpsf
15398 @opindex mno-soft-cmpsf
15399 For single-precision floating-point comparisons, emit an @code{fsub} instruction
15400 and test the flags. This is faster than a software comparison, but can
15401 get incorrect results in the presence of NaNs, or when two different small
15402 numbers are compared such that their difference is calculated as zero.
15403 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
15404 software comparisons.
15406 @item -mstack-offset=@var{num}
15407 @opindex mstack-offset
15408 Set the offset between the top of the stack and the stack pointer.
15409 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
15410 can be used by leaf functions without stack allocation.
15411 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
15412 Note also that this option changes the ABI; compiling a program with a
15413 different stack offset than the libraries have been compiled with
15414 generally does not work.
15415 This option can be useful if you want to evaluate if a different stack
15416 offset would give you better code, but to actually use a different stack
15417 offset to build working programs, it is recommended to configure the
15418 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
15420 @item -mno-round-nearest
15421 @opindex mno-round-nearest
15422 Make the scheduler assume that the rounding mode has been set to
15423 truncating. The default is @option{-mround-nearest}.
15426 @opindex mlong-calls
15427 If not otherwise specified by an attribute, assume all calls might be beyond
15428 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
15429 function address into a register before performing a (otherwise direct) call.
15430 This is the default.
15432 @item -mshort-calls
15433 @opindex short-calls
15434 If not otherwise specified by an attribute, assume all direct calls are
15435 in the range of the @code{b} / @code{bl} instructions, so use these instructions
15436 for direct calls. The default is @option{-mlong-calls}.
15440 Assume addresses can be loaded as 16-bit unsigned values. This does not
15441 apply to function addresses for which @option{-mlong-calls} semantics
15444 @item -mfp-mode=@var{mode}
15446 Set the prevailing mode of the floating-point unit.
15447 This determines the floating-point mode that is provided and expected
15448 at function call and return time. Making this mode match the mode you
15449 predominantly need at function start can make your programs smaller and
15450 faster by avoiding unnecessary mode switches.
15452 @var{mode} can be set to one the following values:
15456 Any mode at function entry is valid, and retained or restored when
15457 the function returns, and when it calls other functions.
15458 This mode is useful for compiling libraries or other compilation units
15459 you might want to incorporate into different programs with different
15460 prevailing FPU modes, and the convenience of being able to use a single
15461 object file outweighs the size and speed overhead for any extra
15462 mode switching that might be needed, compared with what would be needed
15463 with a more specific choice of prevailing FPU mode.
15466 This is the mode used for floating-point calculations with
15467 truncating (i.e.@: round towards zero) rounding mode. That includes
15468 conversion from floating point to integer.
15470 @item round-nearest
15471 This is the mode used for floating-point calculations with
15472 round-to-nearest-or-even rounding mode.
15475 This is the mode used to perform integer calculations in the FPU, e.g.@:
15476 integer multiply, or integer multiply-and-accumulate.
15479 The default is @option{-mfp-mode=caller}
15481 @item -mnosplit-lohi
15482 @itemx -mno-postinc
15483 @itemx -mno-postmodify
15484 @opindex mnosplit-lohi
15485 @opindex mno-postinc
15486 @opindex mno-postmodify
15487 Code generation tweaks that disable, respectively, splitting of 32-bit
15488 loads, generation of post-increment addresses, and generation of
15489 post-modify addresses. The defaults are @option{msplit-lohi},
15490 @option{-mpost-inc}, and @option{-mpost-modify}.
15492 @item -mnovect-double
15493 @opindex mno-vect-double
15494 Change the preferred SIMD mode to SImode. The default is
15495 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
15497 @item -max-vect-align=@var{num}
15498 @opindex max-vect-align
15499 The maximum alignment for SIMD vector mode types.
15500 @var{num} may be 4 or 8. The default is 8.
15501 Note that this is an ABI change, even though many library function
15502 interfaces are unaffected if they don't use SIMD vector modes
15503 in places that affect size and/or alignment of relevant types.
15505 @item -msplit-vecmove-early
15506 @opindex msplit-vecmove-early
15507 Split vector moves into single word moves before reload. In theory this
15508 can give better register allocation, but so far the reverse seems to be
15509 generally the case.
15511 @item -m1reg-@var{reg}
15513 Specify a register to hold the constant @minus{}1, which makes loading small negative
15514 constants and certain bitmasks faster.
15515 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
15516 which specify use of that register as a fixed register,
15517 and @samp{none}, which means that no register is used for this
15518 purpose. The default is @option{-m1reg-none}.
15523 @subsection ARC Options
15524 @cindex ARC options
15526 The following options control the architecture variant for which code
15529 @c architecture variants
15532 @item -mbarrel-shifter
15533 @opindex mbarrel-shifter
15534 Generate instructions supported by barrel shifter. This is the default
15535 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
15538 @opindex mjli-alawys
15539 Force to call a function using jli_s instruction. This option is
15540 valid only for ARCv2 architecture.
15542 @item -mcpu=@var{cpu}
15544 Set architecture type, register usage, and instruction scheduling
15545 parameters for @var{cpu}. There are also shortcut alias options
15546 available for backward compatibility and convenience. Supported
15547 values for @var{cpu} are
15553 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
15557 Compile for ARC601. Alias: @option{-mARC601}.
15562 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
15563 This is the default when configured with @option{--with-cpu=arc700}@.
15566 Compile for ARC EM.
15569 Compile for ARC HS.
15572 Compile for ARC EM CPU with no hardware extensions.
15575 Compile for ARC EM4 CPU.
15578 Compile for ARC EM4 DMIPS CPU.
15581 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
15585 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
15586 double assist instructions.
15589 Compile for ARC HS CPU with no hardware extensions except the atomic
15593 Compile for ARC HS34 CPU.
15596 Compile for ARC HS38 CPU.
15599 Compile for ARC HS38 CPU with all hardware extensions on.
15602 Compile for ARC 600 CPU with @code{norm} instructions enabled.
15604 @item arc600_mul32x16
15605 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
15606 instructions enabled.
15609 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
15610 instructions enabled.
15613 Compile for ARC 601 CPU with @code{norm} instructions enabled.
15615 @item arc601_mul32x16
15616 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
15617 instructions enabled.
15620 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
15621 instructions enabled.
15624 Compile for ARC 700 on NPS400 chip.
15627 Compile for ARC EM minimalist configuration featuring reduced register
15634 @itemx -mdpfp-compact
15635 @opindex mdpfp-compact
15636 Generate double-precision FPX instructions, tuned for the compact
15640 @opindex mdpfp-fast
15641 Generate double-precision FPX instructions, tuned for the fast
15644 @item -mno-dpfp-lrsr
15645 @opindex mno-dpfp-lrsr
15646 Disable @code{lr} and @code{sr} instructions from using FPX extension
15651 Generate extended arithmetic instructions. Currently only
15652 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
15653 supported. This is always enabled for @option{-mcpu=ARC700}.
15657 Do not generate @code{mpy}-family instructions for ARC700. This option is
15662 Generate 32x16-bit multiply and multiply-accumulate instructions.
15666 Generate @code{mul64} and @code{mulu64} instructions.
15667 Only valid for @option{-mcpu=ARC600}.
15671 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
15676 @itemx -mspfp-compact
15677 @opindex mspfp-compact
15678 Generate single-precision FPX instructions, tuned for the compact
15682 @opindex mspfp-fast
15683 Generate single-precision FPX instructions, tuned for the fast
15688 Enable generation of ARC SIMD instructions via target-specific
15689 builtins. Only valid for @option{-mcpu=ARC700}.
15692 @opindex msoft-float
15693 This option ignored; it is provided for compatibility purposes only.
15694 Software floating-point code is emitted by default, and this default
15695 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
15696 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
15697 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
15701 Generate @code{swap} instructions.
15705 This enables use of the locked load/store conditional extension to implement
15706 atomic memory built-in functions. Not available for ARC 6xx or ARC
15711 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
15713 @item -mcode-density
15714 @opindex mcode-density
15715 Enable code density instructions for ARC EM.
15716 This option is on by default for ARC HS.
15720 Enable double load/store operations for ARC HS cores.
15722 @item -mtp-regno=@var{regno}
15724 Specify thread pointer register number.
15726 @item -mmpy-option=@var{multo}
15727 @opindex mmpy-option
15728 Compile ARCv2 code with a multiplier design option. You can specify
15729 the option using either a string or numeric value for @var{multo}.
15730 @samp{wlh1} is the default value. The recognized values are:
15735 No multiplier available.
15739 16x16 multiplier, fully pipelined.
15740 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
15744 32x32 multiplier, fully
15745 pipelined (1 stage). The following instructions are additionally
15746 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15750 32x32 multiplier, fully pipelined
15751 (2 stages). The following instructions are additionally enabled: @code{mpy},
15752 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15756 Two 16x16 multipliers, blocking,
15757 sequential. The following instructions are additionally enabled: @code{mpy},
15758 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15762 One 16x16 multiplier, blocking,
15763 sequential. The following instructions are additionally enabled: @code{mpy},
15764 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15768 One 32x4 multiplier, blocking,
15769 sequential. The following instructions are additionally enabled: @code{mpy},
15770 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
15774 ARC HS SIMD support.
15778 ARC HS SIMD support.
15782 ARC HS SIMD support.
15786 This option is only available for ARCv2 cores@.
15788 @item -mfpu=@var{fpu}
15790 Enables support for specific floating-point hardware extensions for ARCv2
15791 cores. Supported values for @var{fpu} are:
15796 Enables support for single-precision floating-point hardware
15800 Enables support for double-precision floating-point hardware
15801 extensions. The single-precision floating-point extension is also
15802 enabled. Not available for ARC EM@.
15805 Enables support for double-precision floating-point hardware
15806 extensions using double-precision assist instructions. The single-precision
15807 floating-point extension is also enabled. This option is
15808 only available for ARC EM@.
15811 Enables support for double-precision floating-point hardware
15812 extensions using double-precision assist instructions.
15813 The single-precision floating-point, square-root, and divide
15814 extensions are also enabled. This option is
15815 only available for ARC EM@.
15818 Enables support for double-precision floating-point hardware
15819 extensions using double-precision assist instructions.
15820 The single-precision floating-point and fused multiply and add
15821 hardware extensions are also enabled. This option is
15822 only available for ARC EM@.
15825 Enables support for double-precision floating-point hardware
15826 extensions using double-precision assist instructions.
15827 All single-precision floating-point hardware extensions are also
15828 enabled. This option is only available for ARC EM@.
15831 Enables support for single-precision floating-point, square-root and divide
15832 hardware extensions@.
15835 Enables support for double-precision floating-point, square-root and divide
15836 hardware extensions. This option
15837 includes option @samp{fpus_div}. Not available for ARC EM@.
15840 Enables support for single-precision floating-point and
15841 fused multiply and add hardware extensions@.
15844 Enables support for double-precision floating-point and
15845 fused multiply and add hardware extensions. This option
15846 includes option @samp{fpus_fma}. Not available for ARC EM@.
15849 Enables support for all single-precision floating-point hardware
15853 Enables support for all single- and double-precision floating-point
15854 hardware extensions. Not available for ARC EM@.
15858 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
15859 @opindex mirq-ctrl-saved
15860 Specifies general-purposes registers that the processor automatically
15861 saves/restores on interrupt entry and exit. @var{register-range} is
15862 specified as two registers separated by a dash. The register range
15863 always starts with @code{r0}, the upper limit is @code{fp} register.
15864 @var{blink} and @var{lp_count} are optional. This option is only
15865 valid for ARC EM and ARC HS cores.
15867 @item -mrgf-banked-regs=@var{number}
15868 @opindex mrgf-banked-regs
15869 Specifies the number of registers replicated in second register bank
15870 on entry to fast interrupt. Fast interrupts are interrupts with the
15871 highest priority level P0. These interrupts save only PC and STATUS32
15872 registers to avoid memory transactions during interrupt entry and exit
15873 sequences. Use this option when you are using fast interrupts in an
15874 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
15876 @item -mlpc-width=@var{width}
15877 @opindex mlpc-width
15878 Specify the width of the @code{lp_count} register. Valid values for
15879 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
15880 fixed to 32 bits. If the width is less than 32, the compiler does not
15881 attempt to transform loops in your program to use the zero-delay loop
15882 mechanism unless it is known that the @code{lp_count} register can
15883 hold the required loop-counter value. Depending on the width
15884 specified, the compiler and run-time library might continue to use the
15885 loop mechanism for various needs. This option defines macro
15886 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
15890 This option instructs the compiler to generate code for a 16-entry
15891 register file. This option defines the @code{__ARC_RF16__}
15892 preprocessor macro.
15894 @item -mbranch-index
15895 @opindex mbranch-index
15896 Enable use of @code{bi} or @code{bih} instructions to implement jump
15901 The following options are passed through to the assembler, and also
15902 define preprocessor macro symbols.
15904 @c Flags used by the assembler, but for which we define preprocessor
15905 @c macro symbols as well.
15908 @opindex mdsp-packa
15909 Passed down to the assembler to enable the DSP Pack A extensions.
15910 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
15915 Passed down to the assembler to enable the dual Viterbi butterfly
15916 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
15917 option is deprecated.
15919 @c ARC700 4.10 extension instruction
15922 Passed down to the assembler to enable the locked load/store
15923 conditional extension. Also sets the preprocessor symbol
15928 Passed down to the assembler. Also sets the preprocessor symbol
15929 @code{__Xxmac_d16}. This option is deprecated.
15933 Passed down to the assembler. Also sets the preprocessor symbol
15934 @code{__Xxmac_24}. This option is deprecated.
15936 @c ARC700 4.10 extension instruction
15939 Passed down to the assembler to enable the 64-bit time-stamp counter
15940 extension instruction. Also sets the preprocessor symbol
15941 @code{__Xrtsc}. This option is deprecated.
15943 @c ARC700 4.10 extension instruction
15946 Passed down to the assembler to enable the swap byte ordering
15947 extension instruction. Also sets the preprocessor symbol
15951 @opindex mtelephony
15952 Passed down to the assembler to enable dual- and single-operand
15953 instructions for telephony. Also sets the preprocessor symbol
15954 @code{__Xtelephony}. This option is deprecated.
15958 Passed down to the assembler to enable the XY memory extension. Also
15959 sets the preprocessor symbol @code{__Xxy}.
15963 The following options control how the assembly code is annotated:
15965 @c Assembly annotation options
15969 Annotate assembler instructions with estimated addresses.
15971 @item -mannotate-align
15972 @opindex mannotate-align
15973 Explain what alignment considerations lead to the decision to make an
15974 instruction short or long.
15978 The following options are passed through to the linker:
15980 @c options passed through to the linker
15984 Passed through to the linker, to specify use of the @code{arclinux} emulation.
15985 This option is enabled by default in tool chains built for
15986 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
15987 when profiling is not requested.
15989 @item -marclinux_prof
15990 @opindex marclinux_prof
15991 Passed through to the linker, to specify use of the
15992 @code{arclinux_prof} emulation. This option is enabled by default in
15993 tool chains built for @w{@code{arc-linux-uclibc}} and
15994 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
15998 The following options control the semantics of generated code:
16000 @c semantically relevant code generation options
16003 @opindex mlong-calls
16004 Generate calls as register indirect calls, thus providing access
16005 to the full 32-bit address range.
16007 @item -mmedium-calls
16008 @opindex mmedium-calls
16009 Don't use less than 25-bit addressing range for calls, which is the
16010 offset available for an unconditional branch-and-link
16011 instruction. Conditional execution of function calls is suppressed, to
16012 allow use of the 25-bit range, rather than the 21-bit range with
16013 conditional branch-and-link. This is the default for tool chains built
16014 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
16018 Put definitions of externally-visible data in a small data section if
16019 that data is no bigger than @var{num} bytes. The default value of
16020 @var{num} is 4 for any ARC configuration, or 8 when we have double
16021 load/store operations.
16025 Do not generate sdata references. This is the default for tool chains
16026 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
16029 @item -mvolatile-cache
16030 @opindex mvolatile-cache
16031 Use ordinarily cached memory accesses for volatile references. This is the
16034 @item -mno-volatile-cache
16035 @opindex mno-volatile-cache
16036 Enable cache bypass for volatile references.
16040 The following options fine tune code generation:
16041 @c code generation tuning options
16044 @opindex malign-call
16045 Do alignment optimizations for call instructions.
16047 @item -mauto-modify-reg
16048 @opindex mauto-modify-reg
16049 Enable the use of pre/post modify with register displacement.
16051 @item -mbbit-peephole
16052 @opindex mbbit-peephole
16053 Enable bbit peephole2.
16057 This option disables a target-specific pass in @file{arc_reorg} to
16058 generate compare-and-branch (@code{br@var{cc}}) instructions.
16059 It has no effect on
16060 generation of these instructions driven by the combiner pass.
16062 @item -mcase-vector-pcrel
16063 @opindex mcase-vector-pcrel
16064 Use PC-relative switch case tables to enable case table shortening.
16065 This is the default for @option{-Os}.
16067 @item -mcompact-casesi
16068 @opindex mcompact-casesi
16069 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
16070 and only available for ARCv1 cores. This option is deprecated.
16072 @item -mno-cond-exec
16073 @opindex mno-cond-exec
16074 Disable the ARCompact-specific pass to generate conditional
16075 execution instructions.
16077 Due to delay slot scheduling and interactions between operand numbers,
16078 literal sizes, instruction lengths, and the support for conditional execution,
16079 the target-independent pass to generate conditional execution is often lacking,
16080 so the ARC port has kept a special pass around that tries to find more
16081 conditional execution generation opportunities after register allocation,
16082 branch shortening, and delay slot scheduling have been done. This pass
16083 generally, but not always, improves performance and code size, at the cost of
16084 extra compilation time, which is why there is an option to switch it off.
16085 If you have a problem with call instructions exceeding their allowable
16086 offset range because they are conditionalized, you should consider using
16087 @option{-mmedium-calls} instead.
16089 @item -mearly-cbranchsi
16090 @opindex mearly-cbranchsi
16091 Enable pre-reload use of the @code{cbranchsi} pattern.
16093 @item -mexpand-adddi
16094 @opindex mexpand-adddi
16095 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
16096 @code{add.f}, @code{adc} etc. This option is deprecated.
16098 @item -mindexed-loads
16099 @opindex mindexed-loads
16100 Enable the use of indexed loads. This can be problematic because some
16101 optimizers then assume that indexed stores exist, which is not
16106 Enable Local Register Allocation. This is still experimental for ARC,
16107 so by default the compiler uses standard reload
16108 (i.e.@: @option{-mno-lra}).
16110 @item -mlra-priority-none
16111 @opindex mlra-priority-none
16112 Don't indicate any priority for target registers.
16114 @item -mlra-priority-compact
16115 @opindex mlra-priority-compact
16116 Indicate target register priority for r0..r3 / r12..r15.
16118 @item -mlra-priority-noncompact
16119 @opindex mlra-priority-noncompact
16120 Reduce target register priority for r0..r3 / r12..r15.
16122 @item -mno-millicode
16123 @opindex mno-millicode
16124 When optimizing for size (using @option{-Os}), prologues and epilogues
16125 that have to save or restore a large number of registers are often
16126 shortened by using call to a special function in libgcc; this is
16127 referred to as a @emph{millicode} call. As these calls can pose
16128 performance issues, and/or cause linking issues when linking in a
16129 nonstandard way, this option is provided to turn off millicode call
16133 @opindex mmixed-code
16134 Tweak register allocation to help 16-bit instruction generation.
16135 This generally has the effect of decreasing the average instruction size
16136 while increasing the instruction count.
16140 Enable @samp{q} instruction alternatives.
16141 This is the default for @option{-Os}.
16145 Enable @samp{Rcq} constraint handling.
16146 Most short code generation depends on this.
16147 This is the default.
16151 Enable @samp{Rcw} constraint handling.
16152 Most ccfsm condexec mostly depends on this.
16153 This is the default.
16155 @item -msize-level=@var{level}
16156 @opindex msize-level
16157 Fine-tune size optimization with regards to instruction lengths and alignment.
16158 The recognized values for @var{level} are:
16161 No size optimization. This level is deprecated and treated like @samp{1}.
16164 Short instructions are used opportunistically.
16167 In addition, alignment of loops and of code after barriers are dropped.
16170 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
16174 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
16175 the behavior when this is not set is equivalent to level @samp{1}.
16177 @item -mtune=@var{cpu}
16179 Set instruction scheduling parameters for @var{cpu}, overriding any implied
16180 by @option{-mcpu=}.
16182 Supported values for @var{cpu} are
16186 Tune for ARC600 CPU.
16189 Tune for ARC601 CPU.
16192 Tune for ARC700 CPU with standard multiplier block.
16195 Tune for ARC700 CPU with XMAC block.
16198 Tune for ARC725D CPU.
16201 Tune for ARC750D CPU.
16205 @item -mmultcost=@var{num}
16207 Cost to assume for a multiply instruction, with @samp{4} being equal to a
16208 normal instruction.
16210 @item -munalign-prob-threshold=@var{probability}
16211 @opindex munalign-prob-threshold
16212 Set probability threshold for unaligning branches.
16213 When tuning for @samp{ARC700} and optimizing for speed, branches without
16214 filled delay slot are preferably emitted unaligned and long, unless
16215 profiling indicates that the probability for the branch to be taken
16216 is below @var{probability}. @xref{Cross-profiling}.
16217 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
16221 The following options are maintained for backward compatibility, but
16222 are now deprecated and will be removed in a future release:
16224 @c Deprecated options
16232 @opindex mbig-endian
16235 Compile code for big-endian targets. Use of these options is now
16236 deprecated. Big-endian code is supported by configuring GCC to build
16237 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
16238 for which big endian is the default.
16240 @item -mlittle-endian
16241 @opindex mlittle-endian
16244 Compile code for little-endian targets. Use of these options is now
16245 deprecated. Little-endian code is supported by configuring GCC to build
16246 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
16247 for which little endian is the default.
16249 @item -mbarrel_shifter
16250 @opindex mbarrel_shifter
16251 Replaced by @option{-mbarrel-shifter}.
16253 @item -mdpfp_compact
16254 @opindex mdpfp_compact
16255 Replaced by @option{-mdpfp-compact}.
16258 @opindex mdpfp_fast
16259 Replaced by @option{-mdpfp-fast}.
16262 @opindex mdsp_packa
16263 Replaced by @option{-mdsp-packa}.
16267 Replaced by @option{-mea}.
16271 Replaced by @option{-mmac-24}.
16275 Replaced by @option{-mmac-d16}.
16277 @item -mspfp_compact
16278 @opindex mspfp_compact
16279 Replaced by @option{-mspfp-compact}.
16282 @opindex mspfp_fast
16283 Replaced by @option{-mspfp-fast}.
16285 @item -mtune=@var{cpu}
16287 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
16288 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
16289 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
16291 @item -multcost=@var{num}
16293 Replaced by @option{-mmultcost}.
16298 @subsection ARM Options
16299 @cindex ARM options
16301 These @samp{-m} options are defined for the ARM port:
16304 @item -mabi=@var{name}
16306 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
16307 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
16310 @opindex mapcs-frame
16311 Generate a stack frame that is compliant with the ARM Procedure Call
16312 Standard for all functions, even if this is not strictly necessary for
16313 correct execution of the code. Specifying @option{-fomit-frame-pointer}
16314 with this option causes the stack frames not to be generated for
16315 leaf functions. The default is @option{-mno-apcs-frame}.
16316 This option is deprecated.
16320 This is a synonym for @option{-mapcs-frame} and is deprecated.
16323 @c not currently implemented
16324 @item -mapcs-stack-check
16325 @opindex mapcs-stack-check
16326 Generate code to check the amount of stack space available upon entry to
16327 every function (that actually uses some stack space). If there is
16328 insufficient space available then either the function
16329 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
16330 called, depending upon the amount of stack space required. The runtime
16331 system is required to provide these functions. The default is
16332 @option{-mno-apcs-stack-check}, since this produces smaller code.
16334 @c not currently implemented
16335 @item -mapcs-reentrant
16336 @opindex mapcs-reentrant
16337 Generate reentrant, position-independent code. The default is
16338 @option{-mno-apcs-reentrant}.
16341 @item -mthumb-interwork
16342 @opindex mthumb-interwork
16343 Generate code that supports calling between the ARM and Thumb
16344 instruction sets. Without this option, on pre-v5 architectures, the
16345 two instruction sets cannot be reliably used inside one program. The
16346 default is @option{-mno-thumb-interwork}, since slightly larger code
16347 is generated when @option{-mthumb-interwork} is specified. In AAPCS
16348 configurations this option is meaningless.
16350 @item -mno-sched-prolog
16351 @opindex mno-sched-prolog
16352 Prevent the reordering of instructions in the function prologue, or the
16353 merging of those instruction with the instructions in the function's
16354 body. This means that all functions start with a recognizable set
16355 of instructions (or in fact one of a choice from a small set of
16356 different function prologues), and this information can be used to
16357 locate the start of functions inside an executable piece of code. The
16358 default is @option{-msched-prolog}.
16360 @item -mfloat-abi=@var{name}
16361 @opindex mfloat-abi
16362 Specifies which floating-point ABI to use. Permissible values
16363 are: @samp{soft}, @samp{softfp} and @samp{hard}.
16365 Specifying @samp{soft} causes GCC to generate output containing
16366 library calls for floating-point operations.
16367 @samp{softfp} allows the generation of code using hardware floating-point
16368 instructions, but still uses the soft-float calling conventions.
16369 @samp{hard} allows generation of floating-point instructions
16370 and uses FPU-specific calling conventions.
16372 The default depends on the specific target configuration. Note that
16373 the hard-float and soft-float ABIs are not link-compatible; you must
16374 compile your entire program with the same ABI, and link with a
16375 compatible set of libraries.
16377 @item -mlittle-endian
16378 @opindex mlittle-endian
16379 Generate code for a processor running in little-endian mode. This is
16380 the default for all standard configurations.
16383 @opindex mbig-endian
16384 Generate code for a processor running in big-endian mode; the default is
16385 to compile code for a little-endian processor.
16390 When linking a big-endian image select between BE8 and BE32 formats.
16391 The option has no effect for little-endian images and is ignored. The
16392 default is dependent on the selected target architecture. For ARMv6
16393 and later architectures the default is BE8, for older architectures
16394 the default is BE32. BE32 format has been deprecated by ARM.
16396 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
16398 This specifies the name of the target ARM architecture. GCC uses this
16399 name to determine what kind of instructions it can emit when generating
16400 assembly code. This option can be used in conjunction with or instead
16401 of the @option{-mcpu=} option.
16403 Permissible names are:
16405 @samp{armv5t}, @samp{armv5te},
16406 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
16407 @samp{armv6z}, @samp{armv6zk},
16408 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
16409 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
16413 @samp{armv6-m}, @samp{armv6s-m},
16414 @samp{armv7-m}, @samp{armv7e-m},
16415 @samp{armv8-m.base}, @samp{armv8-m.main},
16416 @samp{iwmmxt} and @samp{iwmmxt2}.
16418 Additionally, the following architectures, which lack support for the
16419 Thumb execution state, are recognized but support is deprecated: @samp{armv4}.
16421 Many of the architectures support extensions. These can be added by
16422 appending @samp{+@var{extension}} to the architecture name. Extension
16423 options are processed in order and capabilities accumulate. An extension
16424 will also enable any necessary base extensions
16425 upon which it depends. For example, the @samp{+crypto} extension
16426 will always enable the @samp{+simd} extension. The exception to the
16427 additive construction is for extensions that are prefixed with
16428 @samp{+no@dots{}}: these extensions disable the specified option and
16429 any other extensions that may depend on the presence of that
16432 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
16433 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
16434 entirely disabled by the @samp{+nofp} option that follows it.
16436 Most extension names are generically named, but have an effect that is
16437 dependent upon the architecture to which it is applied. For example,
16438 the @samp{+simd} option can be applied to both @samp{armv7-a} and
16439 @samp{armv8-a} architectures, but will enable the original ARMv7-A
16440 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
16441 variant for @samp{armv8-a}.
16443 The table below lists the supported extensions for each architecture.
16444 Architectures not mentioned do not support any extensions.
16457 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
16458 used as an alias for this extension.
16461 Disable the floating-point instructions.
16465 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
16468 The VFPv3 floating-point instructions, with 16 double-precision
16469 registers. The extension @samp{+vfpv3-d16} can be used as an alias
16470 for this extension. Note that floating-point is not supported by the
16471 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
16472 ARMv7-R architectures.
16475 Disable the floating-point instructions.
16481 The VFPv3 floating-point instructions, with 16 double-precision
16482 registers. The extension @samp{+vfpv3-d16} can be used as an alias
16483 for this extension.
16486 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
16487 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
16488 for this extension.
16491 The VFPv3 floating-point instructions, with 32 double-precision
16494 @item +vfpv3-d16-fp16
16495 The VFPv3 floating-point instructions, with 16 double-precision
16496 registers and the half-precision floating-point conversion operations.
16499 The VFPv3 floating-point instructions, with 32 double-precision
16500 registers and the half-precision floating-point conversion operations.
16503 The VFPv4 floating-point instructions, with 16 double-precision
16507 The VFPv4 floating-point instructions, with 32 double-precision
16511 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
16512 the half-precision floating-point conversion operations.
16515 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
16518 Disable the Advanced SIMD instructions (does not disable floating point).
16521 Disable the floating-point and Advanced SIMD instructions.
16525 The extended version of the ARMv7-A architecture with support for
16529 The VFPv4 floating-point instructions, with 16 double-precision registers.
16530 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
16533 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
16534 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
16537 The VFPv3 floating-point instructions, with 16 double-precision
16541 The VFPv3 floating-point instructions, with 32 double-precision
16544 @item +vfpv3-d16-fp16
16545 The VFPv3 floating-point instructions, with 16 double-precision
16546 registers and the half-precision floating-point conversion operations.
16549 The VFPv3 floating-point instructions, with 32 double-precision
16550 registers and the half-precision floating-point conversion operations.
16553 The VFPv4 floating-point instructions, with 16 double-precision
16557 The VFPv4 floating-point instructions, with 32 double-precision
16561 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
16562 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
16565 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
16566 the half-precision floating-point conversion operations.
16569 Disable the Advanced SIMD instructions (does not disable floating point).
16572 Disable the floating-point and Advanced SIMD instructions.
16578 The Cyclic Redundancy Check (CRC) instructions.
16580 The ARMv8-A Advanced SIMD and floating-point instructions.
16582 The cryptographic instructions.
16584 Disable the cryptographic instructions.
16586 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16592 The ARMv8.1-A Advanced SIMD and floating-point instructions.
16595 The cryptographic instructions. This also enables the Advanced SIMD and
16596 floating-point instructions.
16599 Disable the cryptographic instructions.
16602 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16609 The half-precision floating-point data processing instructions.
16610 This also enables the Advanced SIMD and floating-point instructions.
16613 The half-precision floating-point fmla extension. This also enables
16614 the half-precision floating-point extension and Advanced SIMD and
16615 floating-point instructions.
16618 The ARMv8.1-A Advanced SIMD and floating-point instructions.
16621 The cryptographic instructions. This also enables the Advanced SIMD and
16622 floating-point instructions.
16625 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
16628 Disable the cryptographic extension.
16631 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16637 The half-precision floating-point data processing instructions.
16638 This also enables the Advanced SIMD and floating-point instructions as well
16639 as the Dot Product extension and the half-precision floating-point fmla
16643 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
16644 Dot Product extension.
16647 The cryptographic instructions. This also enables the Advanced SIMD and
16648 floating-point instructions as well as the Dot Product extension.
16651 Disable the cryptographic extension.
16654 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16660 The single-precision VFPv3 floating-point instructions. The extension
16661 @samp{+vfpv3xd} can be used as an alias for this extension.
16664 The VFPv3 floating-point instructions with 16 double-precision registers.
16665 The extension +vfpv3-d16 can be used as an alias for this extension.
16668 Disable the floating-point extension.
16671 The ARM-state integer division instructions.
16674 Disable the ARM-state integer division extension.
16680 The single-precision VFPv4 floating-point instructions.
16683 The single-precision FPv5 floating-point instructions.
16686 The single- and double-precision FPv5 floating-point instructions.
16689 Disable the floating-point extensions.
16695 The DSP instructions.
16698 Disable the DSP extension.
16701 The single-precision floating-point instructions.
16704 The single- and double-precision floating-point instructions.
16707 Disable the floating-point extension.
16713 The Cyclic Redundancy Check (CRC) instructions.
16715 The single-precision FPv5 floating-point instructions.
16717 The ARMv8-A Advanced SIMD and floating-point instructions.
16719 The cryptographic instructions.
16721 Disable the cryptographic instructions.
16723 Disable the floating-point, Advanced SIMD and cryptographic instructions.
16728 @option{-march=native} causes the compiler to auto-detect the architecture
16729 of the build computer. At present, this feature is only supported on
16730 GNU/Linux, and not all architectures are recognized. If the auto-detect
16731 is unsuccessful the option has no effect.
16733 @item -mtune=@var{name}
16735 This option specifies the name of the target ARM processor for
16736 which GCC should tune the performance of the code.
16737 For some ARM implementations better performance can be obtained by using
16739 Permissible names are: @samp{arm2}, @samp{arm250},
16740 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
16741 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
16742 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
16743 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
16745 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
16746 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
16747 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
16748 @samp{strongarm1110},
16749 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
16750 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
16751 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
16752 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
16753 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
16754 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
16755 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
16756 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
16757 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
16758 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
16759 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
16760 @samp{cortex-a76}, @samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f},
16761 @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52},
16769 @samp{cortex-m0plus},
16770 @samp{cortex-m1.small-multiply},
16771 @samp{cortex-m0.small-multiply},
16772 @samp{cortex-m0plus.small-multiply},
16774 @samp{marvell-pj4},
16775 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
16776 @samp{fa526}, @samp{fa626},
16777 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
16780 Additionally, this option can specify that GCC should tune the performance
16781 of the code for a big.LITTLE system. Permissible names are:
16782 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
16783 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
16784 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
16785 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}.
16787 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
16788 performance for a blend of processors within architecture @var{arch}.
16789 The aim is to generate code that run well on the current most popular
16790 processors, balancing between optimizations that benefit some CPUs in the
16791 range, and avoiding performance pitfalls of other CPUs. The effects of
16792 this option may change in future GCC versions as CPU models come and go.
16794 @option{-mtune} permits the same extension options as @option{-mcpu}, but
16795 the extension options do not affect the tuning of the generated code.
16797 @option{-mtune=native} causes the compiler to auto-detect the CPU
16798 of the build computer. At present, this feature is only supported on
16799 GNU/Linux, and not all architectures are recognized. If the auto-detect is
16800 unsuccessful the option has no effect.
16802 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
16804 This specifies the name of the target ARM processor. GCC uses this name
16805 to derive the name of the target ARM architecture (as if specified
16806 by @option{-march}) and the ARM processor type for which to tune for
16807 performance (as if specified by @option{-mtune}). Where this option
16808 is used in conjunction with @option{-march} or @option{-mtune},
16809 those options take precedence over the appropriate part of this option.
16811 Many of the supported CPUs implement optional architectural
16812 extensions. Where this is so the architectural extensions are
16813 normally enabled by default. If implementations that lack the
16814 extension exist, then the extension syntax can be used to disable
16815 those extensions that have been omitted. For floating-point and
16816 Advanced SIMD (Neon) instructions, the settings of the options
16817 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
16818 floating-point and Advanced SIMD instructions will only be used if
16819 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
16820 @option{-mfpu} other than @samp{auto} will override the available
16821 floating-point and SIMD extension instructions.
16823 For example, @samp{cortex-a9} can be found in three major
16824 configurations: integer only, with just a floating-point unit or with
16825 floating-point and Advanced SIMD. The default is to enable all the
16826 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
16827 be used to disable just the SIMD or both the SIMD and floating-point
16828 instructions respectively.
16830 Permissible names for this option are the same as those for
16833 The following extension options are common to the listed CPUs:
16837 Disable the DSP instructions on @samp{cortex-m33}.
16840 Disables the floating-point instructions on @samp{arm9e},
16841 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
16842 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
16843 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
16844 @samp{cortex-m4}, @samp{cortex-m7} and @samp{cortex-m33}.
16845 Disables the floating-point and SIMD instructions on
16846 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
16847 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
16848 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
16849 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
16850 @samp{cortex-a53} and @samp{cortex-a55}.
16853 Disables the double-precision component of the floating-point instructions
16854 on @samp{cortex-r5}, @samp{cortex-r52} and @samp{cortex-m7}.
16857 Disables the SIMD (but not floating-point) instructions on
16858 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
16859 and @samp{cortex-a9}.
16862 Enables the cryptographic instructions on @samp{cortex-a32},
16863 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
16864 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
16865 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
16866 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
16867 @samp{cortex-a75.cortex-a55}.
16870 Additionally the @samp{generic-armv7-a} pseudo target defaults to
16871 VFPv3 with 16 double-precision registers. It supports the following
16872 extension options: @samp{vfpv3-d16}, @samp{vfpv3},
16873 @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16}, @samp{vfpv4-d16},
16874 @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3}, @samp{neon-fp16},
16875 @samp{neon-vfpv4}. The meanings are the same as for the extensions to
16876 @option{-march=armv7-a}.
16878 @option{-mcpu=generic-@var{arch}} is also permissible, and is
16879 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
16880 See @option{-mtune} for more information.
16882 @option{-mcpu=native} causes the compiler to auto-detect the CPU
16883 of the build computer. At present, this feature is only supported on
16884 GNU/Linux, and not all architectures are recognized. If the auto-detect
16885 is unsuccessful the option has no effect.
16887 @item -mfpu=@var{name}
16889 This specifies what floating-point hardware (or hardware emulation) is
16890 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
16892 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
16893 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
16894 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
16895 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
16896 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
16897 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
16898 is an alias for @samp{vfpv2}.
16900 The setting @samp{auto} is the default and is special. It causes the
16901 compiler to select the floating-point and Advanced SIMD instructions
16902 based on the settings of @option{-mcpu} and @option{-march}.
16904 If the selected floating-point hardware includes the NEON extension
16905 (e.g.@: @option{-mfpu=neon}), note that floating-point
16906 operations are not generated by GCC's auto-vectorization pass unless
16907 @option{-funsafe-math-optimizations} is also specified. This is
16908 because NEON hardware does not fully implement the IEEE 754 standard for
16909 floating-point arithmetic (in particular denormal values are treated as
16910 zero), so the use of NEON instructions may lead to a loss of precision.
16912 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}).
16914 @item -mfp16-format=@var{name}
16915 @opindex mfp16-format
16916 Specify the format of the @code{__fp16} half-precision floating-point type.
16917 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
16918 the default is @samp{none}, in which case the @code{__fp16} type is not
16919 defined. @xref{Half-Precision}, for more information.
16921 @item -mstructure-size-boundary=@var{n}
16922 @opindex mstructure-size-boundary
16923 The sizes of all structures and unions are rounded up to a multiple
16924 of the number of bits set by this option. Permissible values are 8, 32
16925 and 64. The default value varies for different toolchains. For the COFF
16926 targeted toolchain the default value is 8. A value of 64 is only allowed
16927 if the underlying ABI supports it.
16929 Specifying a larger number can produce faster, more efficient code, but
16930 can also increase the size of the program. Different values are potentially
16931 incompatible. Code compiled with one value cannot necessarily expect to
16932 work with code or libraries compiled with another value, if they exchange
16933 information using structures or unions.
16935 This option is deprecated.
16937 @item -mabort-on-noreturn
16938 @opindex mabort-on-noreturn
16939 Generate a call to the function @code{abort} at the end of a
16940 @code{noreturn} function. It is executed if the function tries to
16944 @itemx -mno-long-calls
16945 @opindex mlong-calls
16946 @opindex mno-long-calls
16947 Tells the compiler to perform function calls by first loading the
16948 address of the function into a register and then performing a subroutine
16949 call on this register. This switch is needed if the target function
16950 lies outside of the 64-megabyte addressing range of the offset-based
16951 version of subroutine call instruction.
16953 Even if this switch is enabled, not all function calls are turned
16954 into long calls. The heuristic is that static functions, functions
16955 that have the @code{short_call} attribute, functions that are inside
16956 the scope of a @code{#pragma no_long_calls} directive, and functions whose
16957 definitions have already been compiled within the current compilation
16958 unit are not turned into long calls. The exceptions to this rule are
16959 that weak function definitions, functions with the @code{long_call}
16960 attribute or the @code{section} attribute, and functions that are within
16961 the scope of a @code{#pragma long_calls} directive are always
16962 turned into long calls.
16964 This feature is not enabled by default. Specifying
16965 @option{-mno-long-calls} restores the default behavior, as does
16966 placing the function calls within the scope of a @code{#pragma
16967 long_calls_off} directive. Note these switches have no effect on how
16968 the compiler generates code to handle function calls via function
16971 @item -msingle-pic-base
16972 @opindex msingle-pic-base
16973 Treat the register used for PIC addressing as read-only, rather than
16974 loading it in the prologue for each function. The runtime system is
16975 responsible for initializing this register with an appropriate value
16976 before execution begins.
16978 @item -mpic-register=@var{reg}
16979 @opindex mpic-register
16980 Specify the register to be used for PIC addressing.
16981 For standard PIC base case, the default is any suitable register
16982 determined by compiler. For single PIC base case, the default is
16983 @samp{R9} if target is EABI based or stack-checking is enabled,
16984 otherwise the default is @samp{R10}.
16986 @item -mpic-data-is-text-relative
16987 @opindex mpic-data-is-text-relative
16988 Assume that the displacement between the text and data segments is fixed
16989 at static link time. This permits using PC-relative addressing
16990 operations to access data known to be in the data segment. For
16991 non-VxWorks RTP targets, this option is enabled by default. When
16992 disabled on such targets, it will enable @option{-msingle-pic-base} by
16995 @item -mpoke-function-name
16996 @opindex mpoke-function-name
16997 Write the name of each function into the text section, directly
16998 preceding the function prologue. The generated code is similar to this:
17002 .ascii "arm_poke_function_name", 0
17005 .word 0xff000000 + (t1 - t0)
17006 arm_poke_function_name
17008 stmfd sp!, @{fp, ip, lr, pc@}
17012 When performing a stack backtrace, code can inspect the value of
17013 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
17014 location @code{pc - 12} and the top 8 bits are set, then we know that
17015 there is a function name embedded immediately preceding this location
17016 and has length @code{((pc[-3]) & 0xff000000)}.
17023 Select between generating code that executes in ARM and Thumb
17024 states. The default for most configurations is to generate code
17025 that executes in ARM state, but the default can be changed by
17026 configuring GCC with the @option{--with-mode=}@var{state}
17029 You can also override the ARM and Thumb mode for each function
17030 by using the @code{target("thumb")} and @code{target("arm")} function attributes
17031 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
17034 @opindex mflip-thumb
17035 Switch ARM/Thumb modes on alternating functions.
17036 This option is provided for regression testing of mixed Thumb/ARM code
17037 generation, and is not intended for ordinary use in compiling code.
17040 @opindex mtpcs-frame
17041 Generate a stack frame that is compliant with the Thumb Procedure Call
17042 Standard for all non-leaf functions. (A leaf function is one that does
17043 not call any other functions.) The default is @option{-mno-tpcs-frame}.
17045 @item -mtpcs-leaf-frame
17046 @opindex mtpcs-leaf-frame
17047 Generate a stack frame that is compliant with the Thumb Procedure Call
17048 Standard for all leaf functions. (A leaf function is one that does
17049 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
17051 @item -mcallee-super-interworking
17052 @opindex mcallee-super-interworking
17053 Gives all externally visible functions in the file being compiled an ARM
17054 instruction set header which switches to Thumb mode before executing the
17055 rest of the function. This allows these functions to be called from
17056 non-interworking code. This option is not valid in AAPCS configurations
17057 because interworking is enabled by default.
17059 @item -mcaller-super-interworking
17060 @opindex mcaller-super-interworking
17061 Allows calls via function pointers (including virtual functions) to
17062 execute correctly regardless of whether the target code has been
17063 compiled for interworking or not. There is a small overhead in the cost
17064 of executing a function pointer if this option is enabled. This option
17065 is not valid in AAPCS configurations because interworking is enabled
17068 @item -mtp=@var{name}
17070 Specify the access model for the thread local storage pointer. The valid
17071 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
17072 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
17073 (supported in the arm6k architecture), and @samp{auto}, which uses the
17074 best available method for the selected processor. The default setting is
17077 @item -mtls-dialect=@var{dialect}
17078 @opindex mtls-dialect
17079 Specify the dialect to use for accessing thread local storage. Two
17080 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
17081 @samp{gnu} dialect selects the original GNU scheme for supporting
17082 local and global dynamic TLS models. The @samp{gnu2} dialect
17083 selects the GNU descriptor scheme, which provides better performance
17084 for shared libraries. The GNU descriptor scheme is compatible with
17085 the original scheme, but does require new assembler, linker and
17086 library support. Initial and local exec TLS models are unaffected by
17087 this option and always use the original scheme.
17089 @item -mword-relocations
17090 @opindex mword-relocations
17091 Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32).
17092 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
17093 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
17094 is specified. This option conflicts with @option{-mslow-flash-data}.
17096 @item -mfix-cortex-m3-ldrd
17097 @opindex mfix-cortex-m3-ldrd
17098 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
17099 with overlapping destination and base registers are used. This option avoids
17100 generating these instructions. This option is enabled by default when
17101 @option{-mcpu=cortex-m3} is specified.
17103 @item -munaligned-access
17104 @itemx -mno-unaligned-access
17105 @opindex munaligned-access
17106 @opindex mno-unaligned-access
17107 Enables (or disables) reading and writing of 16- and 32- bit values
17108 from addresses that are not 16- or 32- bit aligned. By default
17109 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
17110 ARMv8-M Baseline architectures, and enabled for all other
17111 architectures. If unaligned access is not enabled then words in packed
17112 data structures are accessed a byte at a time.
17114 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
17115 generated object file to either true or false, depending upon the
17116 setting of this option. If unaligned access is enabled then the
17117 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
17120 @item -mneon-for-64bits
17121 @opindex mneon-for-64bits
17122 Enables using Neon to handle scalar 64-bits operations. This is
17123 disabled by default since the cost of moving data from core registers
17126 @item -mslow-flash-data
17127 @opindex mslow-flash-data
17128 Assume loading data from flash is slower than fetching instruction.
17129 Therefore literal load is minimized for better performance.
17130 This option is only supported when compiling for ARMv7 M-profile and
17131 off by default. It conflicts with @option{-mword-relocations}.
17133 @item -masm-syntax-unified
17134 @opindex masm-syntax-unified
17135 Assume inline assembler is using unified asm syntax. The default is
17136 currently off which implies divided syntax. This option has no impact
17137 on Thumb2. However, this may change in future releases of GCC.
17138 Divided syntax should be considered deprecated.
17140 @item -mrestrict-it
17141 @opindex mrestrict-it
17142 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
17143 IT blocks can only contain a single 16-bit instruction from a select
17144 set of instructions. This option is on by default for ARMv8-A Thumb mode.
17146 @item -mprint-tune-info
17147 @opindex mprint-tune-info
17148 Print CPU tuning information as comment in assembler file. This is
17149 an option used only for regression testing of the compiler and not
17150 intended for ordinary use in compiling code. This option is disabled
17153 @item -mverbose-cost-dump
17154 @opindex mverbose-cost-dump
17155 Enable verbose cost model dumping in the debug dump files. This option is
17156 provided for use in debugging the compiler.
17159 @opindex mpure-code
17160 Do not allow constant data to be placed in code sections.
17161 Additionally, when compiling for ELF object format give all text sections the
17162 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
17163 is only available when generating non-pic code for M-profile targets with the
17168 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
17169 Development Tools Engineering Specification", which can be found on
17170 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
17174 @subsection AVR Options
17175 @cindex AVR Options
17177 These options are defined for AVR implementations:
17180 @item -mmcu=@var{mcu}
17182 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
17184 The default for this option is@tie{}@samp{avr2}.
17186 GCC supports the following AVR devices and ISAs:
17188 @include avr-mmcu.texi
17193 Assume that all data in static storage can be accessed by LDS / STS
17194 instructions. This option has only an effect on reduced Tiny devices like
17195 ATtiny40. See also the @code{absdata}
17196 @ref{AVR Variable Attributes,variable attribute}.
17198 @item -maccumulate-args
17199 @opindex maccumulate-args
17200 Accumulate outgoing function arguments and acquire/release the needed
17201 stack space for outgoing function arguments once in function
17202 prologue/epilogue. Without this option, outgoing arguments are pushed
17203 before calling a function and popped afterwards.
17205 Popping the arguments after the function call can be expensive on
17206 AVR so that accumulating the stack space might lead to smaller
17207 executables because arguments need not be removed from the
17208 stack after such a function call.
17210 This option can lead to reduced code size for functions that perform
17211 several calls to functions that get their arguments on the stack like
17212 calls to printf-like functions.
17214 @item -mbranch-cost=@var{cost}
17215 @opindex mbranch-cost
17216 Set the branch costs for conditional branch instructions to
17217 @var{cost}. Reasonable values for @var{cost} are small, non-negative
17218 integers. The default branch cost is 0.
17220 @item -mcall-prologues
17221 @opindex mcall-prologues
17222 Functions prologues/epilogues are expanded as calls to appropriate
17223 subroutines. Code size is smaller.
17225 @item -mgas-isr-prologues
17226 @opindex mgas-isr-prologues
17227 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
17228 instruction supported by GNU Binutils.
17229 If this option is on, the feature can still be disabled for individual
17230 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
17231 function attribute. This feature is activated per default
17232 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
17233 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
17237 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
17238 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
17239 and @code{long long} is 4 bytes. Please note that this option does not
17240 conform to the C standards, but it results in smaller code
17243 @item -mmain-is-OS_task
17244 @opindex mmain-is-OS_task
17245 Do not save registers in @code{main}. The effect is the same like
17246 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
17247 to @code{main}. It is activated per default if optimization is on.
17249 @item -mn-flash=@var{num}
17251 Assume that the flash memory has a size of
17252 @var{num} times 64@tie{}KiB.
17254 @item -mno-interrupts
17255 @opindex mno-interrupts
17256 Generated code is not compatible with hardware interrupts.
17257 Code size is smaller.
17261 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
17262 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
17263 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
17264 the assembler's command line and the @option{--relax} option to the
17265 linker's command line.
17267 Jump relaxing is performed by the linker because jump offsets are not
17268 known before code is located. Therefore, the assembler code generated by the
17269 compiler is the same, but the instructions in the executable may
17270 differ from instructions in the assembler code.
17272 Relaxing must be turned on if linker stubs are needed, see the
17273 section on @code{EIND} and linker stubs below.
17277 Assume that the device supports the Read-Modify-Write
17278 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
17280 @item -mshort-calls
17281 @opindex mshort-calls
17283 Assume that @code{RJMP} and @code{RCALL} can target the whole
17286 This option is used internally for multilib selection. It is
17287 not an optimization option, and you don't need to set it by hand.
17291 Treat the stack pointer register as an 8-bit register,
17292 i.e.@: assume the high byte of the stack pointer is zero.
17293 In general, you don't need to set this option by hand.
17295 This option is used internally by the compiler to select and
17296 build multilibs for architectures @code{avr2} and @code{avr25}.
17297 These architectures mix devices with and without @code{SPH}.
17298 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
17299 the compiler driver adds or removes this option from the compiler
17300 proper's command line, because the compiler then knows if the device
17301 or architecture has an 8-bit stack pointer and thus no @code{SPH}
17306 Use address register @code{X} in a way proposed by the hardware. This means
17307 that @code{X} is only used in indirect, post-increment or
17308 pre-decrement addressing.
17310 Without this option, the @code{X} register may be used in the same way
17311 as @code{Y} or @code{Z} which then is emulated by additional
17313 For example, loading a value with @code{X+const} addressing with a
17314 small non-negative @code{const < 64} to a register @var{Rn} is
17318 adiw r26, const ; X += const
17319 ld @var{Rn}, X ; @var{Rn} = *X
17320 sbiw r26, const ; X -= const
17324 @opindex mtiny-stack
17325 Only change the lower 8@tie{}bits of the stack pointer.
17327 @item -mfract-convert-truncate
17328 @opindex mfract-convert-truncate
17329 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
17332 @opindex nodevicelib
17333 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
17335 @item -Waddr-space-convert
17336 @opindex Waddr-space-convert
17337 @opindex Wno-addr-space-convert
17338 Warn about conversions between address spaces in the case where the
17339 resulting address space is not contained in the incoming address space.
17341 @item -Wmisspelled-isr
17342 @opindex Wmisspelled-isr
17343 @opindex Wno-misspelled-isr
17344 Warn if the ISR is misspelled, i.e.@: without __vector prefix.
17345 Enabled by default.
17348 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
17349 @cindex @code{EIND}
17350 Pointers in the implementation are 16@tie{}bits wide.
17351 The address of a function or label is represented as word address so
17352 that indirect jumps and calls can target any code address in the
17353 range of 64@tie{}Ki words.
17355 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
17356 bytes of program memory space, there is a special function register called
17357 @code{EIND} that serves as most significant part of the target address
17358 when @code{EICALL} or @code{EIJMP} instructions are used.
17360 Indirect jumps and calls on these devices are handled as follows by
17361 the compiler and are subject to some limitations:
17366 The compiler never sets @code{EIND}.
17369 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
17370 instructions or might read @code{EIND} directly in order to emulate an
17371 indirect call/jump by means of a @code{RET} instruction.
17374 The compiler assumes that @code{EIND} never changes during the startup
17375 code or during the application. In particular, @code{EIND} is not
17376 saved/restored in function or interrupt service routine
17380 For indirect calls to functions and computed goto, the linker
17381 generates @emph{stubs}. Stubs are jump pads sometimes also called
17382 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
17383 The stub contains a direct jump to the desired address.
17386 Linker relaxation must be turned on so that the linker generates
17387 the stubs correctly in all situations. See the compiler option
17388 @option{-mrelax} and the linker option @option{--relax}.
17389 There are corner cases where the linker is supposed to generate stubs
17390 but aborts without relaxation and without a helpful error message.
17393 The default linker script is arranged for code with @code{EIND = 0}.
17394 If code is supposed to work for a setup with @code{EIND != 0}, a custom
17395 linker script has to be used in order to place the sections whose
17396 name start with @code{.trampolines} into the segment where @code{EIND}
17400 The startup code from libgcc never sets @code{EIND}.
17401 Notice that startup code is a blend of code from libgcc and AVR-LibC.
17402 For the impact of AVR-LibC on @code{EIND}, see the
17403 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
17406 It is legitimate for user-specific startup code to set up @code{EIND}
17407 early, for example by means of initialization code located in
17408 section @code{.init3}. Such code runs prior to general startup code
17409 that initializes RAM and calls constructors, but after the bit
17410 of startup code from AVR-LibC that sets @code{EIND} to the segment
17411 where the vector table is located.
17413 #include <avr/io.h>
17416 __attribute__((section(".init3"),naked,used,no_instrument_function))
17417 init3_set_eind (void)
17419 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
17420 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
17425 The @code{__trampolines_start} symbol is defined in the linker script.
17428 Stubs are generated automatically by the linker if
17429 the following two conditions are met:
17432 @item The address of a label is taken by means of the @code{gs} modifier
17433 (short for @emph{generate stubs}) like so:
17435 LDI r24, lo8(gs(@var{func}))
17436 LDI r25, hi8(gs(@var{func}))
17438 @item The final location of that label is in a code segment
17439 @emph{outside} the segment where the stubs are located.
17443 The compiler emits such @code{gs} modifiers for code labels in the
17444 following situations:
17446 @item Taking address of a function or code label.
17447 @item Computed goto.
17448 @item If prologue-save function is used, see @option{-mcall-prologues}
17449 command-line option.
17450 @item Switch/case dispatch tables. If you do not want such dispatch
17451 tables you can specify the @option{-fno-jump-tables} command-line option.
17452 @item C and C++ constructors/destructors called during startup/shutdown.
17453 @item If the tools hit a @code{gs()} modifier explained above.
17457 Jumping to non-symbolic addresses like so is @emph{not} supported:
17462 /* Call function at word address 0x2 */
17463 return ((int(*)(void)) 0x2)();
17467 Instead, a stub has to be set up, i.e.@: the function has to be called
17468 through a symbol (@code{func_4} in the example):
17473 extern int func_4 (void);
17475 /* Call function at byte address 0x4 */
17480 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
17481 Alternatively, @code{func_4} can be defined in the linker script.
17484 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
17485 @cindex @code{RAMPD}
17486 @cindex @code{RAMPX}
17487 @cindex @code{RAMPY}
17488 @cindex @code{RAMPZ}
17489 Some AVR devices support memories larger than the 64@tie{}KiB range
17490 that can be accessed with 16-bit pointers. To access memory locations
17491 outside this 64@tie{}KiB range, the content of a @code{RAMP}
17492 register is used as high part of the address:
17493 The @code{X}, @code{Y}, @code{Z} address register is concatenated
17494 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
17495 register, respectively, to get a wide address. Similarly,
17496 @code{RAMPD} is used together with direct addressing.
17500 The startup code initializes the @code{RAMP} special function
17501 registers with zero.
17504 If a @ref{AVR Named Address Spaces,named address space} other than
17505 generic or @code{__flash} is used, then @code{RAMPZ} is set
17506 as needed before the operation.
17509 If the device supports RAM larger than 64@tie{}KiB and the compiler
17510 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
17511 is reset to zero after the operation.
17514 If the device comes with a specific @code{RAMP} register, the ISR
17515 prologue/epilogue saves/restores that SFR and initializes it with
17516 zero in case the ISR code might (implicitly) use it.
17519 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
17520 If you use inline assembler to read from locations outside the
17521 16-bit address range and change one of the @code{RAMP} registers,
17522 you must reset it to zero after the access.
17526 @subsubsection AVR Built-in Macros
17528 GCC defines several built-in macros so that the user code can test
17529 for the presence or absence of features. Almost any of the following
17530 built-in macros are deduced from device capabilities and thus
17531 triggered by the @option{-mmcu=} command-line option.
17533 For even more AVR-specific built-in macros see
17534 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
17539 Build-in macro that resolves to a decimal number that identifies the
17540 architecture and depends on the @option{-mmcu=@var{mcu}} option.
17541 Possible values are:
17543 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
17544 @code{4}, @code{5}, @code{51}, @code{6}
17546 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
17547 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
17552 @code{102}, @code{103}, @code{104},
17553 @code{105}, @code{106}, @code{107}
17555 for @var{mcu}=@code{avrtiny},
17556 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
17557 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
17558 If @var{mcu} specifies a device, this built-in macro is set
17559 accordingly. For example, with @option{-mmcu=atmega8} the macro is
17560 defined to @code{4}.
17562 @item __AVR_@var{Device}__
17563 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
17564 the device's name. For example, @option{-mmcu=atmega8} defines the
17565 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
17566 @code{__AVR_ATtiny261A__}, etc.
17568 The built-in macros' names follow
17569 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
17570 the device name as from the AVR user manual. The difference between
17571 @var{Device} in the built-in macro and @var{device} in
17572 @option{-mmcu=@var{device}} is that the latter is always lowercase.
17574 If @var{device} is not a device but only a core architecture like
17575 @samp{avr51}, this macro is not defined.
17577 @item __AVR_DEVICE_NAME__
17578 Setting @option{-mmcu=@var{device}} defines this built-in macro to
17579 the device's name. For example, with @option{-mmcu=atmega8} the macro
17580 is defined to @code{atmega8}.
17582 If @var{device} is not a device but only a core architecture like
17583 @samp{avr51}, this macro is not defined.
17585 @item __AVR_XMEGA__
17586 The device / architecture belongs to the XMEGA family of devices.
17588 @item __AVR_HAVE_ELPM__
17589 The device has the @code{ELPM} instruction.
17591 @item __AVR_HAVE_ELPMX__
17592 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
17593 R@var{n},Z+} instructions.
17595 @item __AVR_HAVE_MOVW__
17596 The device has the @code{MOVW} instruction to perform 16-bit
17597 register-register moves.
17599 @item __AVR_HAVE_LPMX__
17600 The device has the @code{LPM R@var{n},Z} and
17601 @code{LPM R@var{n},Z+} instructions.
17603 @item __AVR_HAVE_MUL__
17604 The device has a hardware multiplier.
17606 @item __AVR_HAVE_JMP_CALL__
17607 The device has the @code{JMP} and @code{CALL} instructions.
17608 This is the case for devices with more than 8@tie{}KiB of program
17611 @item __AVR_HAVE_EIJMP_EICALL__
17612 @itemx __AVR_3_BYTE_PC__
17613 The device has the @code{EIJMP} and @code{EICALL} instructions.
17614 This is the case for devices with more than 128@tie{}KiB of program memory.
17615 This also means that the program counter
17616 (PC) is 3@tie{}bytes wide.
17618 @item __AVR_2_BYTE_PC__
17619 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
17620 with up to 128@tie{}KiB of program memory.
17622 @item __AVR_HAVE_8BIT_SP__
17623 @itemx __AVR_HAVE_16BIT_SP__
17624 The stack pointer (SP) register is treated as 8-bit respectively
17625 16-bit register by the compiler.
17626 The definition of these macros is affected by @option{-mtiny-stack}.
17628 @item __AVR_HAVE_SPH__
17630 The device has the SPH (high part of stack pointer) special function
17631 register or has an 8-bit stack pointer, respectively.
17632 The definition of these macros is affected by @option{-mmcu=} and
17633 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
17636 @item __AVR_HAVE_RAMPD__
17637 @itemx __AVR_HAVE_RAMPX__
17638 @itemx __AVR_HAVE_RAMPY__
17639 @itemx __AVR_HAVE_RAMPZ__
17640 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
17641 @code{RAMPZ} special function register, respectively.
17643 @item __NO_INTERRUPTS__
17644 This macro reflects the @option{-mno-interrupts} command-line option.
17646 @item __AVR_ERRATA_SKIP__
17647 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
17648 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
17649 instructions because of a hardware erratum. Skip instructions are
17650 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
17651 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
17654 @item __AVR_ISA_RMW__
17655 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
17657 @item __AVR_SFR_OFFSET__=@var{offset}
17658 Instructions that can address I/O special function registers directly
17659 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
17660 address as if addressed by an instruction to access RAM like @code{LD}
17661 or @code{STS}. This offset depends on the device architecture and has
17662 to be subtracted from the RAM address in order to get the
17663 respective I/O@tie{}address.
17665 @item __AVR_SHORT_CALLS__
17666 The @option{-mshort-calls} command line option is set.
17668 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
17669 Some devices support reading from flash memory by means of @code{LD*}
17670 instructions. The flash memory is seen in the data address space
17671 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
17672 is not defined, this feature is not available. If defined,
17673 the address space is linear and there is no need to put
17674 @code{.rodata} into RAM. This is handled by the default linker
17675 description file, and is currently available for
17676 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
17677 there is no need to use address spaces like @code{__flash} or
17678 features like attribute @code{progmem} and @code{pgm_read_*}.
17680 @item __WITH_AVRLIBC__
17681 The compiler is configured to be used together with AVR-Libc.
17682 See the @option{--with-avrlibc} configure option.
17686 @node Blackfin Options
17687 @subsection Blackfin Options
17688 @cindex Blackfin Options
17691 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
17693 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
17694 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
17695 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
17696 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
17697 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
17698 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
17699 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
17700 @samp{bf561}, @samp{bf592}.
17702 The optional @var{sirevision} specifies the silicon revision of the target
17703 Blackfin processor. Any workarounds available for the targeted silicon revision
17704 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
17705 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
17706 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
17707 hexadecimal digits representing the major and minor numbers in the silicon
17708 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
17709 is not defined. If @var{sirevision} is @samp{any}, the
17710 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
17711 If this optional @var{sirevision} is not used, GCC assumes the latest known
17712 silicon revision of the targeted Blackfin processor.
17714 GCC defines a preprocessor macro for the specified @var{cpu}.
17715 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
17716 provided by libgloss to be linked in if @option{-msim} is not given.
17718 Without this option, @samp{bf532} is used as the processor by default.
17720 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
17721 only the preprocessor macro is defined.
17725 Specifies that the program will be run on the simulator. This causes
17726 the simulator BSP provided by libgloss to be linked in. This option
17727 has effect only for @samp{bfin-elf} toolchain.
17728 Certain other options, such as @option{-mid-shared-library} and
17729 @option{-mfdpic}, imply @option{-msim}.
17731 @item -momit-leaf-frame-pointer
17732 @opindex momit-leaf-frame-pointer
17733 Don't keep the frame pointer in a register for leaf functions. This
17734 avoids the instructions to save, set up and restore frame pointers and
17735 makes an extra register available in leaf functions.
17737 @item -mspecld-anomaly
17738 @opindex mspecld-anomaly
17739 When enabled, the compiler ensures that the generated code does not
17740 contain speculative loads after jump instructions. If this option is used,
17741 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
17743 @item -mno-specld-anomaly
17744 @opindex mno-specld-anomaly
17745 Don't generate extra code to prevent speculative loads from occurring.
17747 @item -mcsync-anomaly
17748 @opindex mcsync-anomaly
17749 When enabled, the compiler ensures that the generated code does not
17750 contain CSYNC or SSYNC instructions too soon after conditional branches.
17751 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
17753 @item -mno-csync-anomaly
17754 @opindex mno-csync-anomaly
17755 Don't generate extra code to prevent CSYNC or SSYNC instructions from
17756 occurring too soon after a conditional branch.
17760 When enabled, the compiler is free to take advantage of the knowledge that
17761 the entire program fits into the low 64k of memory.
17764 @opindex mno-low-64k
17765 Assume that the program is arbitrarily large. This is the default.
17767 @item -mstack-check-l1
17768 @opindex mstack-check-l1
17769 Do stack checking using information placed into L1 scratchpad memory by the
17772 @item -mid-shared-library
17773 @opindex mid-shared-library
17774 Generate code that supports shared libraries via the library ID method.
17775 This allows for execute in place and shared libraries in an environment
17776 without virtual memory management. This option implies @option{-fPIC}.
17777 With a @samp{bfin-elf} target, this option implies @option{-msim}.
17779 @item -mno-id-shared-library
17780 @opindex mno-id-shared-library
17781 Generate code that doesn't assume ID-based shared libraries are being used.
17782 This is the default.
17784 @item -mleaf-id-shared-library
17785 @opindex mleaf-id-shared-library
17786 Generate code that supports shared libraries via the library ID method,
17787 but assumes that this library or executable won't link against any other
17788 ID shared libraries. That allows the compiler to use faster code for jumps
17791 @item -mno-leaf-id-shared-library
17792 @opindex mno-leaf-id-shared-library
17793 Do not assume that the code being compiled won't link against any ID shared
17794 libraries. Slower code is generated for jump and call insns.
17796 @item -mshared-library-id=n
17797 @opindex mshared-library-id
17798 Specifies the identification number of the ID-based shared library being
17799 compiled. Specifying a value of 0 generates more compact code; specifying
17800 other values forces the allocation of that number to the current
17801 library but is no more space- or time-efficient than omitting this option.
17805 Generate code that allows the data segment to be located in a different
17806 area of memory from the text segment. This allows for execute in place in
17807 an environment without virtual memory management by eliminating relocations
17808 against the text section.
17810 @item -mno-sep-data
17811 @opindex mno-sep-data
17812 Generate code that assumes that the data segment follows the text segment.
17813 This is the default.
17816 @itemx -mno-long-calls
17817 @opindex mlong-calls
17818 @opindex mno-long-calls
17819 Tells the compiler to perform function calls by first loading the
17820 address of the function into a register and then performing a subroutine
17821 call on this register. This switch is needed if the target function
17822 lies outside of the 24-bit addressing range of the offset-based
17823 version of subroutine call instruction.
17825 This feature is not enabled by default. Specifying
17826 @option{-mno-long-calls} restores the default behavior. Note these
17827 switches have no effect on how the compiler generates code to handle
17828 function calls via function pointers.
17832 Link with the fast floating-point library. This library relaxes some of
17833 the IEEE floating-point standard's rules for checking inputs against
17834 Not-a-Number (NAN), in the interest of performance.
17837 @opindex minline-plt
17838 Enable inlining of PLT entries in function calls to functions that are
17839 not known to bind locally. It has no effect without @option{-mfdpic}.
17842 @opindex mmulticore
17843 Build a standalone application for multicore Blackfin processors.
17844 This option causes proper start files and link scripts supporting
17845 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
17846 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
17848 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
17849 selects the one-application-per-core programming model. Without
17850 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
17851 programming model is used. In this model, the main function of Core B
17852 should be named as @code{coreb_main}.
17854 If this option is not used, the single-core application programming
17859 Build a standalone application for Core A of BF561 when using
17860 the one-application-per-core programming model. Proper start files
17861 and link scripts are used to support Core A, and the macro
17862 @code{__BFIN_COREA} is defined.
17863 This option can only be used in conjunction with @option{-mmulticore}.
17867 Build a standalone application for Core B of BF561 when using
17868 the one-application-per-core programming model. Proper start files
17869 and link scripts are used to support Core B, and the macro
17870 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
17871 should be used instead of @code{main}.
17872 This option can only be used in conjunction with @option{-mmulticore}.
17876 Build a standalone application for SDRAM. Proper start files and
17877 link scripts are used to put the application into SDRAM, and the macro
17878 @code{__BFIN_SDRAM} is defined.
17879 The loader should initialize SDRAM before loading the application.
17883 Assume that ICPLBs are enabled at run time. This has an effect on certain
17884 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
17885 are enabled; for standalone applications the default is off.
17889 @subsection C6X Options
17890 @cindex C6X Options
17893 @item -march=@var{name}
17895 This specifies the name of the target architecture. GCC uses this
17896 name to determine what kind of instructions it can emit when generating
17897 assembly code. Permissible names are: @samp{c62x},
17898 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
17901 @opindex mbig-endian
17902 Generate code for a big-endian target.
17904 @item -mlittle-endian
17905 @opindex mlittle-endian
17906 Generate code for a little-endian target. This is the default.
17910 Choose startup files and linker script suitable for the simulator.
17912 @item -msdata=default
17913 @opindex msdata=default
17914 Put small global and static data in the @code{.neardata} section,
17915 which is pointed to by register @code{B14}. Put small uninitialized
17916 global and static data in the @code{.bss} section, which is adjacent
17917 to the @code{.neardata} section. Put small read-only data into the
17918 @code{.rodata} section. The corresponding sections used for large
17919 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
17922 @opindex msdata=all
17923 Put all data, not just small objects, into the sections reserved for
17924 small data, and use addressing relative to the @code{B14} register to
17928 @opindex msdata=none
17929 Make no use of the sections reserved for small data, and use absolute
17930 addresses to access all data. Put all initialized global and static
17931 data in the @code{.fardata} section, and all uninitialized data in the
17932 @code{.far} section. Put all constant data into the @code{.const}
17937 @subsection CRIS Options
17938 @cindex CRIS Options
17940 These options are defined specifically for the CRIS ports.
17943 @item -march=@var{architecture-type}
17944 @itemx -mcpu=@var{architecture-type}
17947 Generate code for the specified architecture. The choices for
17948 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
17949 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
17950 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
17953 @item -mtune=@var{architecture-type}
17955 Tune to @var{architecture-type} everything applicable about the generated
17956 code, except for the ABI and the set of available instructions. The
17957 choices for @var{architecture-type} are the same as for
17958 @option{-march=@var{architecture-type}}.
17960 @item -mmax-stack-frame=@var{n}
17961 @opindex mmax-stack-frame
17962 Warn when the stack frame of a function exceeds @var{n} bytes.
17968 The options @option{-metrax4} and @option{-metrax100} are synonyms for
17969 @option{-march=v3} and @option{-march=v8} respectively.
17971 @item -mmul-bug-workaround
17972 @itemx -mno-mul-bug-workaround
17973 @opindex mmul-bug-workaround
17974 @opindex mno-mul-bug-workaround
17975 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
17976 models where it applies. This option is active by default.
17980 Enable CRIS-specific verbose debug-related information in the assembly
17981 code. This option also has the effect of turning off the @samp{#NO_APP}
17982 formatted-code indicator to the assembler at the beginning of the
17987 Do not use condition-code results from previous instruction; always emit
17988 compare and test instructions before use of condition codes.
17990 @item -mno-side-effects
17991 @opindex mno-side-effects
17992 Do not emit instructions with side effects in addressing modes other than
17995 @item -mstack-align
17996 @itemx -mno-stack-align
17997 @itemx -mdata-align
17998 @itemx -mno-data-align
17999 @itemx -mconst-align
18000 @itemx -mno-const-align
18001 @opindex mstack-align
18002 @opindex mno-stack-align
18003 @opindex mdata-align
18004 @opindex mno-data-align
18005 @opindex mconst-align
18006 @opindex mno-const-align
18007 These options (@samp{no-} options) arrange (eliminate arrangements) for the
18008 stack frame, individual data and constants to be aligned for the maximum
18009 single data access size for the chosen CPU model. The default is to
18010 arrange for 32-bit alignment. ABI details such as structure layout are
18011 not affected by these options.
18019 Similar to the stack- data- and const-align options above, these options
18020 arrange for stack frame, writable data and constants to all be 32-bit,
18021 16-bit or 8-bit aligned. The default is 32-bit alignment.
18023 @item -mno-prologue-epilogue
18024 @itemx -mprologue-epilogue
18025 @opindex mno-prologue-epilogue
18026 @opindex mprologue-epilogue
18027 With @option{-mno-prologue-epilogue}, the normal function prologue and
18028 epilogue which set up the stack frame are omitted and no return
18029 instructions or return sequences are generated in the code. Use this
18030 option only together with visual inspection of the compiled code: no
18031 warnings or errors are generated when call-saved registers must be saved,
18032 or storage for local variables needs to be allocated.
18036 @opindex mno-gotplt
18038 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
18039 instruction sequences that load addresses for functions from the PLT part
18040 of the GOT rather than (traditional on other architectures) calls to the
18041 PLT@. The default is @option{-mgotplt}.
18045 Legacy no-op option only recognized with the cris-axis-elf and
18046 cris-axis-linux-gnu targets.
18050 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
18054 This option, recognized for the cris-axis-elf, arranges
18055 to link with input-output functions from a simulator library. Code,
18056 initialized data and zero-initialized data are allocated consecutively.
18060 Like @option{-sim}, but pass linker options to locate initialized data at
18061 0x40000000 and zero-initialized data at 0x80000000.
18065 @subsection CR16 Options
18066 @cindex CR16 Options
18068 These options are defined specifically for the CR16 ports.
18074 Enable the use of multiply-accumulate instructions. Disabled by default.
18078 @opindex mcr16cplus
18080 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
18085 Links the library libsim.a which is in compatible with simulator. Applicable
18086 to ELF compiler only.
18090 Choose integer type as 32-bit wide.
18094 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
18096 @item -mdata-model=@var{model}
18097 @opindex mdata-model
18098 Choose a data model. The choices for @var{model} are @samp{near},
18099 @samp{far} or @samp{medium}. @samp{medium} is default.
18100 However, @samp{far} is not valid with @option{-mcr16c}, as the
18101 CR16C architecture does not support the far data model.
18104 @node C-SKY Options
18105 @subsection C-SKY Options
18106 @cindex C-SKY Options
18108 GCC supports these options when compiling for C-SKY V2 processors.
18112 @item -march=@var{arch}
18114 Specify the C-SKY target architecture. Valid values for @var{arch} are:
18115 @samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
18116 The default is @samp{ck810}.
18118 @item -mcpu=@var{cpu}
18120 Specify the C-SKY target processor. Valid values for @var{cpu} are:
18121 @samp{ck801}, @samp{ck801t},
18122 @samp{ck802}, @samp{ck802t}, @samp{ck802j},
18123 @samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
18124 @samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
18125 @samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
18126 @samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
18127 @samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
18128 @samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
18129 @samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
18130 @samp{ck803eftr1}, @samp{ck803efhtr1},
18131 @samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
18132 @samp{ck803sef}, @samp{ck803seft},
18133 @samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
18134 @samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
18135 @samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
18136 @samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
18139 @opindex mbig-endian
18142 @itemx -mlittle-endian
18143 @opindex mlittle-endian
18147 Select big- or little-endian code. The default is little-endian.
18150 @opindex mhard-float
18151 @itemx -msoft-float
18152 @opindex msoft-float
18154 Select hardware or software floating-point implementations.
18155 The default is soft float.
18157 @item -mdouble-float
18158 @itemx -mno-double-float
18159 @opindex mdouble-float
18160 When @option{-mhard-float} is in effect, enable generation of
18161 double-precision float instructions. This is the default except
18162 when compiling for CK803.
18167 When @option{-mhard-float} is in effect, enable generation of
18168 @code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
18169 This is the default except when compiling for CK803.
18171 @item -mfpu=@var{fpu}
18173 Select the floating-point processor. This option can only be used with
18174 @option{-mhard-float}.
18175 Values for @var{fpu} are
18176 @samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
18177 @samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
18178 @samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
18183 Enable the extended @code{lrw} instruction. This option defaults to on
18184 for CK801 and off otherwise.
18189 Enable interrupt stack instructions; the default is off.
18191 The @option{-mistack} option is required to handle the
18192 @code{interrupt} and @code{isr} function attributes
18193 (@pxref{C-SKY Function Attributes}).
18197 Enable multiprocessor instructions; the default is off.
18201 Enable coprocessor instructions; the default is off.
18205 Enable coprocessor instructions; the default is off.
18209 Enable C-SKY security instructions; the default is off.
18213 Enable C-SKY trust instructions; the default is off.
18221 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
18222 All of these options default to off.
18227 Generate divide instructions. Default is off.
18232 Generate code for Smart Mode, using only registers numbered 0-7 to allow
18233 use of 16-bit instructions. This option is ignored for CK801 where this
18234 is the required behavior, and it defaults to on for CK802.
18235 For other targets, the default is off.
18237 @item -mhigh-registers
18238 @itemx -mno-high-registers
18239 @opindex mhigh-registers
18240 Generate code using the high registers numbered 16-31. This option
18241 is not supported on CK801, CK802, or CK803, and is enabled by default
18242 for other processors.
18247 Generate code using global anchor symbol addresses.
18250 @itemx -mno-pushpop
18252 Generate code using @code{push} and @code{pop} instructions. This option
18255 @item -mmultiple-stld
18257 @itemx -mno-multiple-stld
18259 @opindex mmultiple-stld
18260 Generate code using @code{stm} and @code{ldm} instructions. This option
18261 isn't supported on CK801 but is enabled by default on other processors.
18264 @itemx -mno-constpool
18265 @opindex mconstpool
18266 Create constant pools in the compiler instead of deferring it to the
18267 assembler. This option is the default and required for correct code
18268 generation on CK801 and CK802, and is optional on other processors.
18271 @item -mno-stack-size
18272 @opindex mstack-size
18273 Emit @code{.stack_size} directives for each function in the assembly
18274 output. This option defaults to off.
18279 Generate code for the C-SKY compiler runtime instead of libgcc. This
18280 option defaults to off.
18282 @item -mbranch-cost=@var{n}
18283 @opindex mbranch-cost=
18284 Set the branch costs to roughly @code{n} instructions. The default is 1.
18286 @item -msched-prolog
18287 @itemx -mno-sched-prolog
18288 @opindex msched-prolog
18289 Permit scheduling of function prologue and epilogue sequences. Using
18290 this option can result in code that is not compliant with the C-SKY V2 ABI
18291 prologue requirements and that cannot be debugged or backtraced.
18292 It is disabled by default.
18296 @node Darwin Options
18297 @subsection Darwin Options
18298 @cindex Darwin options
18300 These options are defined for all architectures running the Darwin operating
18303 FSF GCC on Darwin does not create ``fat'' object files; it creates
18304 an object file for the single architecture that GCC was built to
18305 target. Apple's GCC on Darwin does create ``fat'' files if multiple
18306 @option{-arch} options are used; it does so by running the compiler or
18307 linker multiple times and joining the results together with
18310 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
18311 @samp{i686}) is determined by the flags that specify the ISA
18312 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
18313 @option{-force_cpusubtype_ALL} option can be used to override this.
18315 The Darwin tools vary in their behavior when presented with an ISA
18316 mismatch. The assembler, @file{as}, only permits instructions to
18317 be used that are valid for the subtype of the file it is generating,
18318 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
18319 The linker for shared libraries, @file{/usr/bin/libtool}, fails
18320 and prints an error if asked to create a shared library with a less
18321 restrictive subtype than its input files (for instance, trying to put
18322 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
18323 for executables, @command{ld}, quietly gives the executable the most
18324 restrictive subtype of any of its input files.
18329 Add the framework directory @var{dir} to the head of the list of
18330 directories to be searched for header files. These directories are
18331 interleaved with those specified by @option{-I} options and are
18332 scanned in a left-to-right order.
18334 A framework directory is a directory with frameworks in it. A
18335 framework is a directory with a @file{Headers} and/or
18336 @file{PrivateHeaders} directory contained directly in it that ends
18337 in @file{.framework}. The name of a framework is the name of this
18338 directory excluding the @file{.framework}. Headers associated with
18339 the framework are found in one of those two directories, with
18340 @file{Headers} being searched first. A subframework is a framework
18341 directory that is in a framework's @file{Frameworks} directory.
18342 Includes of subframework headers can only appear in a header of a
18343 framework that contains the subframework, or in a sibling subframework
18344 header. Two subframeworks are siblings if they occur in the same
18345 framework. A subframework should not have the same name as a
18346 framework; a warning is issued if this is violated. Currently a
18347 subframework cannot have subframeworks; in the future, the mechanism
18348 may be extended to support this. The standard frameworks can be found
18349 in @file{/System/Library/Frameworks} and
18350 @file{/Library/Frameworks}. An example include looks like
18351 @code{#include <Framework/header.h>}, where @file{Framework} denotes
18352 the name of the framework and @file{header.h} is found in the
18353 @file{PrivateHeaders} or @file{Headers} directory.
18355 @item -iframework@var{dir}
18356 @opindex iframework
18357 Like @option{-F} except the directory is a treated as a system
18358 directory. The main difference between this @option{-iframework} and
18359 @option{-F} is that with @option{-iframework} the compiler does not
18360 warn about constructs contained within header files found via
18361 @var{dir}. This option is valid only for the C family of languages.
18365 Emit debugging information for symbols that are used. For stabs
18366 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
18367 This is by default ON@.
18371 Emit debugging information for all symbols and types.
18373 @item -mmacosx-version-min=@var{version}
18374 The earliest version of MacOS X that this executable will run on
18375 is @var{version}. Typical values of @var{version} include @code{10.1},
18376 @code{10.2}, and @code{10.3.9}.
18378 If the compiler was built to use the system's headers by default,
18379 then the default for this option is the system version on which the
18380 compiler is running, otherwise the default is to make choices that
18381 are compatible with as many systems and code bases as possible.
18385 Enable kernel development mode. The @option{-mkernel} option sets
18386 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
18387 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
18388 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
18389 applicable. This mode also sets @option{-mno-altivec},
18390 @option{-msoft-float}, @option{-fno-builtin} and
18391 @option{-mlong-branch} for PowerPC targets.
18393 @item -mone-byte-bool
18394 @opindex mone-byte-bool
18395 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
18396 By default @code{sizeof(bool)} is @code{4} when compiling for
18397 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
18398 option has no effect on x86.
18400 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
18401 to generate code that is not binary compatible with code generated
18402 without that switch. Using this switch may require recompiling all
18403 other modules in a program, including system libraries. Use this
18404 switch to conform to a non-default data model.
18406 @item -mfix-and-continue
18407 @itemx -ffix-and-continue
18408 @itemx -findirect-data
18409 @opindex mfix-and-continue
18410 @opindex ffix-and-continue
18411 @opindex findirect-data
18412 Generate code suitable for fast turnaround development, such as to
18413 allow GDB to dynamically load @file{.o} files into already-running
18414 programs. @option{-findirect-data} and @option{-ffix-and-continue}
18415 are provided for backwards compatibility.
18419 Loads all members of static archive libraries.
18420 See man ld(1) for more information.
18422 @item -arch_errors_fatal
18423 @opindex arch_errors_fatal
18424 Cause the errors having to do with files that have the wrong architecture
18427 @item -bind_at_load
18428 @opindex bind_at_load
18429 Causes the output file to be marked such that the dynamic linker will
18430 bind all undefined references when the file is loaded or launched.
18434 Produce a Mach-o bundle format file.
18435 See man ld(1) for more information.
18437 @item -bundle_loader @var{executable}
18438 @opindex bundle_loader
18439 This option specifies the @var{executable} that will load the build
18440 output file being linked. See man ld(1) for more information.
18443 @opindex dynamiclib
18444 When passed this option, GCC produces a dynamic library instead of
18445 an executable when linking, using the Darwin @file{libtool} command.
18447 @item -force_cpusubtype_ALL
18448 @opindex force_cpusubtype_ALL
18449 This causes GCC's output file to have the @samp{ALL} subtype, instead of
18450 one controlled by the @option{-mcpu} or @option{-march} option.
18452 @item -allowable_client @var{client_name}
18453 @itemx -client_name
18454 @itemx -compatibility_version
18455 @itemx -current_version
18457 @itemx -dependency-file
18459 @itemx -dylinker_install_name
18461 @itemx -exported_symbols_list
18464 @itemx -flat_namespace
18465 @itemx -force_flat_namespace
18466 @itemx -headerpad_max_install_names
18469 @itemx -install_name
18470 @itemx -keep_private_externs
18471 @itemx -multi_module
18472 @itemx -multiply_defined
18473 @itemx -multiply_defined_unused
18476 @itemx -no_dead_strip_inits_and_terms
18477 @itemx -nofixprebinding
18478 @itemx -nomultidefs
18480 @itemx -noseglinkedit
18481 @itemx -pagezero_size
18483 @itemx -prebind_all_twolevel_modules
18484 @itemx -private_bundle
18486 @itemx -read_only_relocs
18488 @itemx -sectobjectsymbols
18492 @itemx -sectobjectsymbols
18495 @itemx -segs_read_only_addr
18497 @itemx -segs_read_write_addr
18498 @itemx -seg_addr_table
18499 @itemx -seg_addr_table_filename
18500 @itemx -seglinkedit
18502 @itemx -segs_read_only_addr
18503 @itemx -segs_read_write_addr
18504 @itemx -single_module
18506 @itemx -sub_library
18508 @itemx -sub_umbrella
18509 @itemx -twolevel_namespace
18512 @itemx -unexported_symbols_list
18513 @itemx -weak_reference_mismatches
18514 @itemx -whatsloaded
18515 @opindex allowable_client
18516 @opindex client_name
18517 @opindex compatibility_version
18518 @opindex current_version
18519 @opindex dead_strip
18520 @opindex dependency-file
18521 @opindex dylib_file
18522 @opindex dylinker_install_name
18524 @opindex exported_symbols_list
18526 @opindex flat_namespace
18527 @opindex force_flat_namespace
18528 @opindex headerpad_max_install_names
18529 @opindex image_base
18531 @opindex install_name
18532 @opindex keep_private_externs
18533 @opindex multi_module
18534 @opindex multiply_defined
18535 @opindex multiply_defined_unused
18536 @opindex noall_load
18537 @opindex no_dead_strip_inits_and_terms
18538 @opindex nofixprebinding
18539 @opindex nomultidefs
18541 @opindex noseglinkedit
18542 @opindex pagezero_size
18544 @opindex prebind_all_twolevel_modules
18545 @opindex private_bundle
18546 @opindex read_only_relocs
18548 @opindex sectobjectsymbols
18551 @opindex sectcreate
18552 @opindex sectobjectsymbols
18555 @opindex segs_read_only_addr
18556 @opindex segs_read_write_addr
18557 @opindex seg_addr_table
18558 @opindex seg_addr_table_filename
18559 @opindex seglinkedit
18561 @opindex segs_read_only_addr
18562 @opindex segs_read_write_addr
18563 @opindex single_module
18565 @opindex sub_library
18566 @opindex sub_umbrella
18567 @opindex twolevel_namespace
18570 @opindex unexported_symbols_list
18571 @opindex weak_reference_mismatches
18572 @opindex whatsloaded
18573 These options are passed to the Darwin linker. The Darwin linker man page
18574 describes them in detail.
18577 @node DEC Alpha Options
18578 @subsection DEC Alpha Options
18580 These @samp{-m} options are defined for the DEC Alpha implementations:
18583 @item -mno-soft-float
18584 @itemx -msoft-float
18585 @opindex mno-soft-float
18586 @opindex msoft-float
18587 Use (do not use) the hardware floating-point instructions for
18588 floating-point operations. When @option{-msoft-float} is specified,
18589 functions in @file{libgcc.a} are used to perform floating-point
18590 operations. Unless they are replaced by routines that emulate the
18591 floating-point operations, or compiled in such a way as to call such
18592 emulations routines, these routines issue floating-point
18593 operations. If you are compiling for an Alpha without floating-point
18594 operations, you must ensure that the library is built so as not to call
18597 Note that Alpha implementations without floating-point operations are
18598 required to have floating-point registers.
18601 @itemx -mno-fp-regs
18603 @opindex mno-fp-regs
18604 Generate code that uses (does not use) the floating-point register set.
18605 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
18606 register set is not used, floating-point operands are passed in integer
18607 registers as if they were integers and floating-point results are passed
18608 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
18609 so any function with a floating-point argument or return value called by code
18610 compiled with @option{-mno-fp-regs} must also be compiled with that
18613 A typical use of this option is building a kernel that does not use,
18614 and hence need not save and restore, any floating-point registers.
18618 The Alpha architecture implements floating-point hardware optimized for
18619 maximum performance. It is mostly compliant with the IEEE floating-point
18620 standard. However, for full compliance, software assistance is
18621 required. This option generates code fully IEEE-compliant code
18622 @emph{except} that the @var{inexact-flag} is not maintained (see below).
18623 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
18624 defined during compilation. The resulting code is less efficient but is
18625 able to correctly support denormalized numbers and exceptional IEEE
18626 values such as not-a-number and plus/minus infinity. Other Alpha
18627 compilers call this option @option{-ieee_with_no_inexact}.
18629 @item -mieee-with-inexact
18630 @opindex mieee-with-inexact
18631 This is like @option{-mieee} except the generated code also maintains
18632 the IEEE @var{inexact-flag}. Turning on this option causes the
18633 generated code to implement fully-compliant IEEE math. In addition to
18634 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
18635 macro. On some Alpha implementations the resulting code may execute
18636 significantly slower than the code generated by default. Since there is
18637 very little code that depends on the @var{inexact-flag}, you should
18638 normally not specify this option. Other Alpha compilers call this
18639 option @option{-ieee_with_inexact}.
18641 @item -mfp-trap-mode=@var{trap-mode}
18642 @opindex mfp-trap-mode
18643 This option controls what floating-point related traps are enabled.
18644 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
18645 The trap mode can be set to one of four values:
18649 This is the default (normal) setting. The only traps that are enabled
18650 are the ones that cannot be disabled in software (e.g., division by zero
18654 In addition to the traps enabled by @samp{n}, underflow traps are enabled
18658 Like @samp{u}, but the instructions are marked to be safe for software
18659 completion (see Alpha architecture manual for details).
18662 Like @samp{su}, but inexact traps are enabled as well.
18665 @item -mfp-rounding-mode=@var{rounding-mode}
18666 @opindex mfp-rounding-mode
18667 Selects the IEEE rounding mode. Other Alpha compilers call this option
18668 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
18673 Normal IEEE rounding mode. Floating-point numbers are rounded towards
18674 the nearest machine number or towards the even machine number in case
18678 Round towards minus infinity.
18681 Chopped rounding mode. Floating-point numbers are rounded towards zero.
18684 Dynamic rounding mode. A field in the floating-point control register
18685 (@var{fpcr}, see Alpha architecture reference manual) controls the
18686 rounding mode in effect. The C library initializes this register for
18687 rounding towards plus infinity. Thus, unless your program modifies the
18688 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
18691 @item -mtrap-precision=@var{trap-precision}
18692 @opindex mtrap-precision
18693 In the Alpha architecture, floating-point traps are imprecise. This
18694 means without software assistance it is impossible to recover from a
18695 floating trap and program execution normally needs to be terminated.
18696 GCC can generate code that can assist operating system trap handlers
18697 in determining the exact location that caused a floating-point trap.
18698 Depending on the requirements of an application, different levels of
18699 precisions can be selected:
18703 Program precision. This option is the default and means a trap handler
18704 can only identify which program caused a floating-point exception.
18707 Function precision. The trap handler can determine the function that
18708 caused a floating-point exception.
18711 Instruction precision. The trap handler can determine the exact
18712 instruction that caused a floating-point exception.
18715 Other Alpha compilers provide the equivalent options called
18716 @option{-scope_safe} and @option{-resumption_safe}.
18718 @item -mieee-conformant
18719 @opindex mieee-conformant
18720 This option marks the generated code as IEEE conformant. You must not
18721 use this option unless you also specify @option{-mtrap-precision=i} and either
18722 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
18723 is to emit the line @samp{.eflag 48} in the function prologue of the
18724 generated assembly file.
18726 @item -mbuild-constants
18727 @opindex mbuild-constants
18728 Normally GCC examines a 32- or 64-bit integer constant to
18729 see if it can construct it from smaller constants in two or three
18730 instructions. If it cannot, it outputs the constant as a literal and
18731 generates code to load it from the data segment at run time.
18733 Use this option to require GCC to construct @emph{all} integer constants
18734 using code, even if it takes more instructions (the maximum is six).
18736 You typically use this option to build a shared library dynamic
18737 loader. Itself a shared library, it must relocate itself in memory
18738 before it can find the variables and constants in its own data segment.
18756 Indicate whether GCC should generate code to use the optional BWX,
18757 CIX, FIX and MAX instruction sets. The default is to use the instruction
18758 sets supported by the CPU type specified via @option{-mcpu=} option or that
18759 of the CPU on which GCC was built if none is specified.
18762 @itemx -mfloat-ieee
18763 @opindex mfloat-vax
18764 @opindex mfloat-ieee
18765 Generate code that uses (does not use) VAX F and G floating-point
18766 arithmetic instead of IEEE single and double precision.
18768 @item -mexplicit-relocs
18769 @itemx -mno-explicit-relocs
18770 @opindex mexplicit-relocs
18771 @opindex mno-explicit-relocs
18772 Older Alpha assemblers provided no way to generate symbol relocations
18773 except via assembler macros. Use of these macros does not allow
18774 optimal instruction scheduling. GNU binutils as of version 2.12
18775 supports a new syntax that allows the compiler to explicitly mark
18776 which relocations should apply to which instructions. This option
18777 is mostly useful for debugging, as GCC detects the capabilities of
18778 the assembler when it is built and sets the default accordingly.
18781 @itemx -mlarge-data
18782 @opindex msmall-data
18783 @opindex mlarge-data
18784 When @option{-mexplicit-relocs} is in effect, static data is
18785 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
18786 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
18787 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
18788 16-bit relocations off of the @code{$gp} register. This limits the
18789 size of the small data area to 64KB, but allows the variables to be
18790 directly accessed via a single instruction.
18792 The default is @option{-mlarge-data}. With this option the data area
18793 is limited to just below 2GB@. Programs that require more than 2GB of
18794 data must use @code{malloc} or @code{mmap} to allocate the data in the
18795 heap instead of in the program's data segment.
18797 When generating code for shared libraries, @option{-fpic} implies
18798 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
18801 @itemx -mlarge-text
18802 @opindex msmall-text
18803 @opindex mlarge-text
18804 When @option{-msmall-text} is used, the compiler assumes that the
18805 code of the entire program (or shared library) fits in 4MB, and is
18806 thus reachable with a branch instruction. When @option{-msmall-data}
18807 is used, the compiler can assume that all local symbols share the
18808 same @code{$gp} value, and thus reduce the number of instructions
18809 required for a function call from 4 to 1.
18811 The default is @option{-mlarge-text}.
18813 @item -mcpu=@var{cpu_type}
18815 Set the instruction set and instruction scheduling parameters for
18816 machine type @var{cpu_type}. You can specify either the @samp{EV}
18817 style name or the corresponding chip number. GCC supports scheduling
18818 parameters for the EV4, EV5 and EV6 family of processors and
18819 chooses the default values for the instruction set from the processor
18820 you specify. If you do not specify a processor type, GCC defaults
18821 to the processor on which the compiler was built.
18823 Supported values for @var{cpu_type} are
18829 Schedules as an EV4 and has no instruction set extensions.
18833 Schedules as an EV5 and has no instruction set extensions.
18837 Schedules as an EV5 and supports the BWX extension.
18842 Schedules as an EV5 and supports the BWX and MAX extensions.
18846 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
18850 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
18853 Native toolchains also support the value @samp{native},
18854 which selects the best architecture option for the host processor.
18855 @option{-mcpu=native} has no effect if GCC does not recognize
18858 @item -mtune=@var{cpu_type}
18860 Set only the instruction scheduling parameters for machine type
18861 @var{cpu_type}. The instruction set is not changed.
18863 Native toolchains also support the value @samp{native},
18864 which selects the best architecture option for the host processor.
18865 @option{-mtune=native} has no effect if GCC does not recognize
18868 @item -mmemory-latency=@var{time}
18869 @opindex mmemory-latency
18870 Sets the latency the scheduler should assume for typical memory
18871 references as seen by the application. This number is highly
18872 dependent on the memory access patterns used by the application
18873 and the size of the external cache on the machine.
18875 Valid options for @var{time} are
18879 A decimal number representing clock cycles.
18885 The compiler contains estimates of the number of clock cycles for
18886 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
18887 (also called Dcache, Scache, and Bcache), as well as to main memory.
18888 Note that L3 is only valid for EV5.
18894 @subsection FR30 Options
18895 @cindex FR30 Options
18897 These options are defined specifically for the FR30 port.
18901 @item -msmall-model
18902 @opindex msmall-model
18903 Use the small address space model. This can produce smaller code, but
18904 it does assume that all symbolic values and addresses fit into a
18909 Assume that runtime support has been provided and so there is no need
18910 to include the simulator library (@file{libsim.a}) on the linker
18916 @subsection FT32 Options
18917 @cindex FT32 Options
18919 These options are defined specifically for the FT32 port.
18925 Specifies that the program will be run on the simulator. This causes
18926 an alternate runtime startup and library to be linked.
18927 You must not use this option when generating programs that will run on
18928 real hardware; you must provide your own runtime library for whatever
18929 I/O functions are needed.
18933 Enable Local Register Allocation. This is still experimental for FT32,
18934 so by default the compiler uses standard reload.
18938 Do not use div and mod instructions.
18942 Enable use of the extended instructions of the FT32B processor.
18946 Compress all code using the Ft32B code compression scheme.
18950 Do not generate code that reads program memory.
18955 @subsection FRV Options
18956 @cindex FRV Options
18962 Only use the first 32 general-purpose registers.
18967 Use all 64 general-purpose registers.
18972 Use only the first 32 floating-point registers.
18977 Use all 64 floating-point registers.
18980 @opindex mhard-float
18982 Use hardware instructions for floating-point operations.
18985 @opindex msoft-float
18987 Use library routines for floating-point operations.
18992 Dynamically allocate condition code registers.
18997 Do not try to dynamically allocate condition code registers, only
18998 use @code{icc0} and @code{fcc0}.
19003 Change ABI to use double word insns.
19008 Do not use double word instructions.
19013 Use floating-point double instructions.
19016 @opindex mno-double
19018 Do not use floating-point double instructions.
19023 Use media instructions.
19028 Do not use media instructions.
19033 Use multiply and add/subtract instructions.
19036 @opindex mno-muladd
19038 Do not use multiply and add/subtract instructions.
19043 Select the FDPIC ABI, which uses function descriptors to represent
19044 pointers to functions. Without any PIC/PIE-related options, it
19045 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
19046 assumes GOT entries and small data are within a 12-bit range from the
19047 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
19048 are computed with 32 bits.
19049 With a @samp{bfin-elf} target, this option implies @option{-msim}.
19052 @opindex minline-plt
19054 Enable inlining of PLT entries in function calls to functions that are
19055 not known to bind locally. It has no effect without @option{-mfdpic}.
19056 It's enabled by default if optimizing for speed and compiling for
19057 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
19058 optimization option such as @option{-O3} or above is present in the
19064 Assume a large TLS segment when generating thread-local code.
19069 Do not assume a large TLS segment when generating thread-local code.
19074 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
19075 that is known to be in read-only sections. It's enabled by default,
19076 except for @option{-fpic} or @option{-fpie}: even though it may help
19077 make the global offset table smaller, it trades 1 instruction for 4.
19078 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
19079 one of which may be shared by multiple symbols, and it avoids the need
19080 for a GOT entry for the referenced symbol, so it's more likely to be a
19081 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
19083 @item -multilib-library-pic
19084 @opindex multilib-library-pic
19086 Link with the (library, not FD) pic libraries. It's implied by
19087 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
19088 @option{-fpic} without @option{-mfdpic}. You should never have to use
19092 @opindex mlinked-fp
19094 Follow the EABI requirement of always creating a frame pointer whenever
19095 a stack frame is allocated. This option is enabled by default and can
19096 be disabled with @option{-mno-linked-fp}.
19099 @opindex mlong-calls
19101 Use indirect addressing to call functions outside the current
19102 compilation unit. This allows the functions to be placed anywhere
19103 within the 32-bit address space.
19105 @item -malign-labels
19106 @opindex malign-labels
19108 Try to align labels to an 8-byte boundary by inserting NOPs into the
19109 previous packet. This option only has an effect when VLIW packing
19110 is enabled. It doesn't create new packets; it merely adds NOPs to
19113 @item -mlibrary-pic
19114 @opindex mlibrary-pic
19116 Generate position-independent EABI code.
19121 Use only the first four media accumulator registers.
19126 Use all eight media accumulator registers.
19131 Pack VLIW instructions.
19136 Do not pack VLIW instructions.
19139 @opindex mno-eflags
19141 Do not mark ABI switches in e_flags.
19144 @opindex mcond-move
19146 Enable the use of conditional-move instructions (default).
19148 This switch is mainly for debugging the compiler and will likely be removed
19149 in a future version.
19151 @item -mno-cond-move
19152 @opindex mno-cond-move
19154 Disable the use of conditional-move instructions.
19156 This switch is mainly for debugging the compiler and will likely be removed
19157 in a future version.
19162 Enable the use of conditional set instructions (default).
19164 This switch is mainly for debugging the compiler and will likely be removed
19165 in a future version.
19170 Disable the use of conditional set instructions.
19172 This switch is mainly for debugging the compiler and will likely be removed
19173 in a future version.
19176 @opindex mcond-exec
19178 Enable the use of conditional execution (default).
19180 This switch is mainly for debugging the compiler and will likely be removed
19181 in a future version.
19183 @item -mno-cond-exec
19184 @opindex mno-cond-exec
19186 Disable the use of conditional execution.
19188 This switch is mainly for debugging the compiler and will likely be removed
19189 in a future version.
19191 @item -mvliw-branch
19192 @opindex mvliw-branch
19194 Run a pass to pack branches into VLIW instructions (default).
19196 This switch is mainly for debugging the compiler and will likely be removed
19197 in a future version.
19199 @item -mno-vliw-branch
19200 @opindex mno-vliw-branch
19202 Do not run a pass to pack branches into VLIW instructions.
19204 This switch is mainly for debugging the compiler and will likely be removed
19205 in a future version.
19207 @item -mmulti-cond-exec
19208 @opindex mmulti-cond-exec
19210 Enable optimization of @code{&&} and @code{||} in conditional execution
19213 This switch is mainly for debugging the compiler and will likely be removed
19214 in a future version.
19216 @item -mno-multi-cond-exec
19217 @opindex mno-multi-cond-exec
19219 Disable optimization of @code{&&} and @code{||} in conditional execution.
19221 This switch is mainly for debugging the compiler and will likely be removed
19222 in a future version.
19224 @item -mnested-cond-exec
19225 @opindex mnested-cond-exec
19227 Enable nested conditional execution optimizations (default).
19229 This switch is mainly for debugging the compiler and will likely be removed
19230 in a future version.
19232 @item -mno-nested-cond-exec
19233 @opindex mno-nested-cond-exec
19235 Disable nested conditional execution optimizations.
19237 This switch is mainly for debugging the compiler and will likely be removed
19238 in a future version.
19240 @item -moptimize-membar
19241 @opindex moptimize-membar
19243 This switch removes redundant @code{membar} instructions from the
19244 compiler-generated code. It is enabled by default.
19246 @item -mno-optimize-membar
19247 @opindex mno-optimize-membar
19249 This switch disables the automatic removal of redundant @code{membar}
19250 instructions from the generated code.
19252 @item -mtomcat-stats
19253 @opindex mtomcat-stats
19255 Cause gas to print out tomcat statistics.
19257 @item -mcpu=@var{cpu}
19260 Select the processor type for which to generate code. Possible values are
19261 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
19262 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
19266 @node GNU/Linux Options
19267 @subsection GNU/Linux Options
19269 These @samp{-m} options are defined for GNU/Linux targets:
19274 Use the GNU C library. This is the default except
19275 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
19276 @samp{*-*-linux-*android*} targets.
19280 Use uClibc C library. This is the default on
19281 @samp{*-*-linux-*uclibc*} targets.
19285 Use the musl C library. This is the default on
19286 @samp{*-*-linux-*musl*} targets.
19290 Use Bionic C library. This is the default on
19291 @samp{*-*-linux-*android*} targets.
19295 Compile code compatible with Android platform. This is the default on
19296 @samp{*-*-linux-*android*} targets.
19298 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
19299 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
19300 this option makes the GCC driver pass Android-specific options to the linker.
19301 Finally, this option causes the preprocessor macro @code{__ANDROID__}
19304 @item -tno-android-cc
19305 @opindex tno-android-cc
19306 Disable compilation effects of @option{-mandroid}, i.e., do not enable
19307 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
19308 @option{-fno-rtti} by default.
19310 @item -tno-android-ld
19311 @opindex tno-android-ld
19312 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
19313 linking options to the linker.
19317 @node H8/300 Options
19318 @subsection H8/300 Options
19320 These @samp{-m} options are defined for the H8/300 implementations:
19325 Shorten some address references at link time, when possible; uses the
19326 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
19327 ld, Using ld}, for a fuller description.
19331 Generate code for the H8/300H@.
19335 Generate code for the H8S@.
19339 Generate code for the H8S and H8/300H in the normal mode. This switch
19340 must be used either with @option{-mh} or @option{-ms}.
19344 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
19348 Extended registers are stored on stack before execution of function
19349 with monitor attribute. Default option is @option{-mexr}.
19350 This option is valid only for H8S targets.
19354 Extended registers are not stored on stack before execution of function
19355 with monitor attribute. Default option is @option{-mno-exr}.
19356 This option is valid only for H8S targets.
19360 Make @code{int} data 32 bits by default.
19363 @opindex malign-300
19364 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
19365 The default for the H8/300H and H8S is to align longs and floats on
19367 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
19368 This option has no effect on the H8/300.
19372 @subsection HPPA Options
19373 @cindex HPPA Options
19375 These @samp{-m} options are defined for the HPPA family of computers:
19378 @item -march=@var{architecture-type}
19380 Generate code for the specified architecture. The choices for
19381 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
19382 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
19383 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
19384 architecture option for your machine. Code compiled for lower numbered
19385 architectures runs on higher numbered architectures, but not the
19388 @item -mpa-risc-1-0
19389 @itemx -mpa-risc-1-1
19390 @itemx -mpa-risc-2-0
19391 @opindex mpa-risc-1-0
19392 @opindex mpa-risc-1-1
19393 @opindex mpa-risc-2-0
19394 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
19396 @item -mcaller-copies
19397 @opindex mcaller-copies
19398 The caller copies function arguments passed by hidden reference. This
19399 option should be used with care as it is not compatible with the default
19400 32-bit runtime. However, only aggregates larger than eight bytes are
19401 passed by hidden reference and the option provides better compatibility
19404 @item -mjump-in-delay
19405 @opindex mjump-in-delay
19406 This option is ignored and provided for compatibility purposes only.
19408 @item -mdisable-fpregs
19409 @opindex mdisable-fpregs
19410 Prevent floating-point registers from being used in any manner. This is
19411 necessary for compiling kernels that perform lazy context switching of
19412 floating-point registers. If you use this option and attempt to perform
19413 floating-point operations, the compiler aborts.
19415 @item -mdisable-indexing
19416 @opindex mdisable-indexing
19417 Prevent the compiler from using indexing address modes. This avoids some
19418 rather obscure problems when compiling MIG generated code under MACH@.
19420 @item -mno-space-regs
19421 @opindex mno-space-regs
19422 Generate code that assumes the target has no space registers. This allows
19423 GCC to generate faster indirect calls and use unscaled index address modes.
19425 Such code is suitable for level 0 PA systems and kernels.
19427 @item -mfast-indirect-calls
19428 @opindex mfast-indirect-calls
19429 Generate code that assumes calls never cross space boundaries. This
19430 allows GCC to emit code that performs faster indirect calls.
19432 This option does not work in the presence of shared libraries or nested
19435 @item -mfixed-range=@var{register-range}
19436 @opindex mfixed-range
19437 Generate code treating the given register range as fixed registers.
19438 A fixed register is one that the register allocator cannot use. This is
19439 useful when compiling kernel code. A register range is specified as
19440 two registers separated by a dash. Multiple register ranges can be
19441 specified separated by a comma.
19443 @item -mlong-load-store
19444 @opindex mlong-load-store
19445 Generate 3-instruction load and store sequences as sometimes required by
19446 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
19449 @item -mportable-runtime
19450 @opindex mportable-runtime
19451 Use the portable calling conventions proposed by HP for ELF systems.
19455 Enable the use of assembler directives only GAS understands.
19457 @item -mschedule=@var{cpu-type}
19459 Schedule code according to the constraints for the machine type
19460 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
19461 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
19462 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
19463 proper scheduling option for your machine. The default scheduling is
19467 @opindex mlinker-opt
19468 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
19469 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
19470 linkers in which they give bogus error messages when linking some programs.
19473 @opindex msoft-float
19474 Generate output containing library calls for floating point.
19475 @strong{Warning:} the requisite libraries are not available for all HPPA
19476 targets. Normally the facilities of the machine's usual C compiler are
19477 used, but this cannot be done directly in cross-compilation. You must make
19478 your own arrangements to provide suitable library functions for
19481 @option{-msoft-float} changes the calling convention in the output file;
19482 therefore, it is only useful if you compile @emph{all} of a program with
19483 this option. In particular, you need to compile @file{libgcc.a}, the
19484 library that comes with GCC, with @option{-msoft-float} in order for
19489 Generate the predefine, @code{_SIO}, for server IO@. The default is
19490 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
19491 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
19492 options are available under HP-UX and HI-UX@.
19496 Use options specific to GNU @command{ld}.
19497 This passes @option{-shared} to @command{ld} when
19498 building a shared library. It is the default when GCC is configured,
19499 explicitly or implicitly, with the GNU linker. This option does not
19500 affect which @command{ld} is called; it only changes what parameters
19501 are passed to that @command{ld}.
19502 The @command{ld} that is called is determined by the
19503 @option{--with-ld} configure option, GCC's program search path, and
19504 finally by the user's @env{PATH}. The linker used by GCC can be printed
19505 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
19506 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
19510 Use options specific to HP @command{ld}.
19511 This passes @option{-b} to @command{ld} when building
19512 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
19513 links. It is the default when GCC is configured, explicitly or
19514 implicitly, with the HP linker. This option does not affect
19515 which @command{ld} is called; it only changes what parameters are passed to that
19517 The @command{ld} that is called is determined by the @option{--with-ld}
19518 configure option, GCC's program search path, and finally by the user's
19519 @env{PATH}. The linker used by GCC can be printed using @samp{which
19520 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
19521 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
19524 @opindex mno-long-calls
19525 Generate code that uses long call sequences. This ensures that a call
19526 is always able to reach linker generated stubs. The default is to generate
19527 long calls only when the distance from the call site to the beginning
19528 of the function or translation unit, as the case may be, exceeds a
19529 predefined limit set by the branch type being used. The limits for
19530 normal calls are 7,600,000 and 240,000 bytes, respectively for the
19531 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
19534 Distances are measured from the beginning of functions when using the
19535 @option{-ffunction-sections} option, or when using the @option{-mgas}
19536 and @option{-mno-portable-runtime} options together under HP-UX with
19539 It is normally not desirable to use this option as it degrades
19540 performance. However, it may be useful in large applications,
19541 particularly when partial linking is used to build the application.
19543 The types of long calls used depends on the capabilities of the
19544 assembler and linker, and the type of code being generated. The
19545 impact on systems that support long absolute calls, and long pic
19546 symbol-difference or pc-relative calls should be relatively small.
19547 However, an indirect call is used on 32-bit ELF systems in pic code
19548 and it is quite long.
19550 @item -munix=@var{unix-std}
19552 Generate compiler predefines and select a startfile for the specified
19553 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
19554 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
19555 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
19556 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
19557 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
19560 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
19561 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
19562 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
19563 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
19564 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
19565 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
19567 It is @emph{important} to note that this option changes the interfaces
19568 for various library routines. It also affects the operational behavior
19569 of the C library. Thus, @emph{extreme} care is needed in using this
19572 Library code that is intended to operate with more than one UNIX
19573 standard must test, set and restore the variable @code{__xpg4_extended_mask}
19574 as appropriate. Most GNU software doesn't provide this capability.
19578 Suppress the generation of link options to search libdld.sl when the
19579 @option{-static} option is specified on HP-UX 10 and later.
19583 The HP-UX implementation of setlocale in libc has a dependency on
19584 libdld.sl. There isn't an archive version of libdld.sl. Thus,
19585 when the @option{-static} option is specified, special link options
19586 are needed to resolve this dependency.
19588 On HP-UX 10 and later, the GCC driver adds the necessary options to
19589 link with libdld.sl when the @option{-static} option is specified.
19590 This causes the resulting binary to be dynamic. On the 64-bit port,
19591 the linkers generate dynamic binaries by default in any case. The
19592 @option{-nolibdld} option can be used to prevent the GCC driver from
19593 adding these link options.
19597 Add support for multithreading with the @dfn{dce thread} library
19598 under HP-UX@. This option sets flags for both the preprocessor and
19602 @node IA-64 Options
19603 @subsection IA-64 Options
19604 @cindex IA-64 Options
19606 These are the @samp{-m} options defined for the Intel IA-64 architecture.
19610 @opindex mbig-endian
19611 Generate code for a big-endian target. This is the default for HP-UX@.
19613 @item -mlittle-endian
19614 @opindex mlittle-endian
19615 Generate code for a little-endian target. This is the default for AIX5
19621 @opindex mno-gnu-as
19622 Generate (or don't) code for the GNU assembler. This is the default.
19623 @c Also, this is the default if the configure option @option{--with-gnu-as}
19629 @opindex mno-gnu-ld
19630 Generate (or don't) code for the GNU linker. This is the default.
19631 @c Also, this is the default if the configure option @option{--with-gnu-ld}
19636 Generate code that does not use a global pointer register. The result
19637 is not position independent code, and violates the IA-64 ABI@.
19639 @item -mvolatile-asm-stop
19640 @itemx -mno-volatile-asm-stop
19641 @opindex mvolatile-asm-stop
19642 @opindex mno-volatile-asm-stop
19643 Generate (or don't) a stop bit immediately before and after volatile asm
19646 @item -mregister-names
19647 @itemx -mno-register-names
19648 @opindex mregister-names
19649 @opindex mno-register-names
19650 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
19651 the stacked registers. This may make assembler output more readable.
19657 Disable (or enable) optimizations that use the small data section. This may
19658 be useful for working around optimizer bugs.
19660 @item -mconstant-gp
19661 @opindex mconstant-gp
19662 Generate code that uses a single constant global pointer value. This is
19663 useful when compiling kernel code.
19667 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
19668 This is useful when compiling firmware code.
19670 @item -minline-float-divide-min-latency
19671 @opindex minline-float-divide-min-latency
19672 Generate code for inline divides of floating-point values
19673 using the minimum latency algorithm.
19675 @item -minline-float-divide-max-throughput
19676 @opindex minline-float-divide-max-throughput
19677 Generate code for inline divides of floating-point values
19678 using the maximum throughput algorithm.
19680 @item -mno-inline-float-divide
19681 @opindex mno-inline-float-divide
19682 Do not generate inline code for divides of floating-point values.
19684 @item -minline-int-divide-min-latency
19685 @opindex minline-int-divide-min-latency
19686 Generate code for inline divides of integer values
19687 using the minimum latency algorithm.
19689 @item -minline-int-divide-max-throughput
19690 @opindex minline-int-divide-max-throughput
19691 Generate code for inline divides of integer values
19692 using the maximum throughput algorithm.
19694 @item -mno-inline-int-divide
19695 @opindex mno-inline-int-divide
19696 Do not generate inline code for divides of integer values.
19698 @item -minline-sqrt-min-latency
19699 @opindex minline-sqrt-min-latency
19700 Generate code for inline square roots
19701 using the minimum latency algorithm.
19703 @item -minline-sqrt-max-throughput
19704 @opindex minline-sqrt-max-throughput
19705 Generate code for inline square roots
19706 using the maximum throughput algorithm.
19708 @item -mno-inline-sqrt
19709 @opindex mno-inline-sqrt
19710 Do not generate inline code for @code{sqrt}.
19713 @itemx -mno-fused-madd
19714 @opindex mfused-madd
19715 @opindex mno-fused-madd
19716 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
19717 instructions. The default is to use these instructions.
19719 @item -mno-dwarf2-asm
19720 @itemx -mdwarf2-asm
19721 @opindex mno-dwarf2-asm
19722 @opindex mdwarf2-asm
19723 Don't (or do) generate assembler code for the DWARF line number debugging
19724 info. This may be useful when not using the GNU assembler.
19726 @item -mearly-stop-bits
19727 @itemx -mno-early-stop-bits
19728 @opindex mearly-stop-bits
19729 @opindex mno-early-stop-bits
19730 Allow stop bits to be placed earlier than immediately preceding the
19731 instruction that triggered the stop bit. This can improve instruction
19732 scheduling, but does not always do so.
19734 @item -mfixed-range=@var{register-range}
19735 @opindex mfixed-range
19736 Generate code treating the given register range as fixed registers.
19737 A fixed register is one that the register allocator cannot use. This is
19738 useful when compiling kernel code. A register range is specified as
19739 two registers separated by a dash. Multiple register ranges can be
19740 specified separated by a comma.
19742 @item -mtls-size=@var{tls-size}
19744 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
19747 @item -mtune=@var{cpu-type}
19749 Tune the instruction scheduling for a particular CPU, Valid values are
19750 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
19751 and @samp{mckinley}.
19757 Generate code for a 32-bit or 64-bit environment.
19758 The 32-bit environment sets int, long and pointer to 32 bits.
19759 The 64-bit environment sets int to 32 bits and long and pointer
19760 to 64 bits. These are HP-UX specific flags.
19762 @item -mno-sched-br-data-spec
19763 @itemx -msched-br-data-spec
19764 @opindex mno-sched-br-data-spec
19765 @opindex msched-br-data-spec
19766 (Dis/En)able data speculative scheduling before reload.
19767 This results in generation of @code{ld.a} instructions and
19768 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
19769 The default setting is disabled.
19771 @item -msched-ar-data-spec
19772 @itemx -mno-sched-ar-data-spec
19773 @opindex msched-ar-data-spec
19774 @opindex mno-sched-ar-data-spec
19775 (En/Dis)able data speculative scheduling after reload.
19776 This results in generation of @code{ld.a} instructions and
19777 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
19778 The default setting is enabled.
19780 @item -mno-sched-control-spec
19781 @itemx -msched-control-spec
19782 @opindex mno-sched-control-spec
19783 @opindex msched-control-spec
19784 (Dis/En)able control speculative scheduling. This feature is
19785 available only during region scheduling (i.e.@: before reload).
19786 This results in generation of the @code{ld.s} instructions and
19787 the corresponding check instructions @code{chk.s}.
19788 The default setting is disabled.
19790 @item -msched-br-in-data-spec
19791 @itemx -mno-sched-br-in-data-spec
19792 @opindex msched-br-in-data-spec
19793 @opindex mno-sched-br-in-data-spec
19794 (En/Dis)able speculative scheduling of the instructions that
19795 are dependent on the data speculative loads before reload.
19796 This is effective only with @option{-msched-br-data-spec} enabled.
19797 The default setting is enabled.
19799 @item -msched-ar-in-data-spec
19800 @itemx -mno-sched-ar-in-data-spec
19801 @opindex msched-ar-in-data-spec
19802 @opindex mno-sched-ar-in-data-spec
19803 (En/Dis)able speculative scheduling of the instructions that
19804 are dependent on the data speculative loads after reload.
19805 This is effective only with @option{-msched-ar-data-spec} enabled.
19806 The default setting is enabled.
19808 @item -msched-in-control-spec
19809 @itemx -mno-sched-in-control-spec
19810 @opindex msched-in-control-spec
19811 @opindex mno-sched-in-control-spec
19812 (En/Dis)able speculative scheduling of the instructions that
19813 are dependent on the control speculative loads.
19814 This is effective only with @option{-msched-control-spec} enabled.
19815 The default setting is enabled.
19817 @item -mno-sched-prefer-non-data-spec-insns
19818 @itemx -msched-prefer-non-data-spec-insns
19819 @opindex mno-sched-prefer-non-data-spec-insns
19820 @opindex msched-prefer-non-data-spec-insns
19821 If enabled, data-speculative instructions are chosen for schedule
19822 only if there are no other choices at the moment. This makes
19823 the use of the data speculation much more conservative.
19824 The default setting is disabled.
19826 @item -mno-sched-prefer-non-control-spec-insns
19827 @itemx -msched-prefer-non-control-spec-insns
19828 @opindex mno-sched-prefer-non-control-spec-insns
19829 @opindex msched-prefer-non-control-spec-insns
19830 If enabled, control-speculative instructions are chosen for schedule
19831 only if there are no other choices at the moment. This makes
19832 the use of the control speculation much more conservative.
19833 The default setting is disabled.
19835 @item -mno-sched-count-spec-in-critical-path
19836 @itemx -msched-count-spec-in-critical-path
19837 @opindex mno-sched-count-spec-in-critical-path
19838 @opindex msched-count-spec-in-critical-path
19839 If enabled, speculative dependencies are considered during
19840 computation of the instructions priorities. This makes the use of the
19841 speculation a bit more conservative.
19842 The default setting is disabled.
19844 @item -msched-spec-ldc
19845 @opindex msched-spec-ldc
19846 Use a simple data speculation check. This option is on by default.
19848 @item -msched-control-spec-ldc
19849 @opindex msched-spec-ldc
19850 Use a simple check for control speculation. This option is on by default.
19852 @item -msched-stop-bits-after-every-cycle
19853 @opindex msched-stop-bits-after-every-cycle
19854 Place a stop bit after every cycle when scheduling. This option is on
19857 @item -msched-fp-mem-deps-zero-cost
19858 @opindex msched-fp-mem-deps-zero-cost
19859 Assume that floating-point stores and loads are not likely to cause a conflict
19860 when placed into the same instruction group. This option is disabled by
19863 @item -msel-sched-dont-check-control-spec
19864 @opindex msel-sched-dont-check-control-spec
19865 Generate checks for control speculation in selective scheduling.
19866 This flag is disabled by default.
19868 @item -msched-max-memory-insns=@var{max-insns}
19869 @opindex msched-max-memory-insns
19870 Limit on the number of memory insns per instruction group, giving lower
19871 priority to subsequent memory insns attempting to schedule in the same
19872 instruction group. Frequently useful to prevent cache bank conflicts.
19873 The default value is 1.
19875 @item -msched-max-memory-insns-hard-limit
19876 @opindex msched-max-memory-insns-hard-limit
19877 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
19878 disallowing more than that number in an instruction group.
19879 Otherwise, the limit is ``soft'', meaning that non-memory operations
19880 are preferred when the limit is reached, but memory operations may still
19886 @subsection LM32 Options
19887 @cindex LM32 options
19889 These @option{-m} options are defined for the LatticeMico32 architecture:
19892 @item -mbarrel-shift-enabled
19893 @opindex mbarrel-shift-enabled
19894 Enable barrel-shift instructions.
19896 @item -mdivide-enabled
19897 @opindex mdivide-enabled
19898 Enable divide and modulus instructions.
19900 @item -mmultiply-enabled
19901 @opindex multiply-enabled
19902 Enable multiply instructions.
19904 @item -msign-extend-enabled
19905 @opindex msign-extend-enabled
19906 Enable sign extend instructions.
19908 @item -muser-enabled
19909 @opindex muser-enabled
19910 Enable user-defined instructions.
19915 @subsection M32C Options
19916 @cindex M32C options
19919 @item -mcpu=@var{name}
19921 Select the CPU for which code is generated. @var{name} may be one of
19922 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
19923 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
19924 the M32C/80 series.
19928 Specifies that the program will be run on the simulator. This causes
19929 an alternate runtime library to be linked in which supports, for
19930 example, file I/O@. You must not use this option when generating
19931 programs that will run on real hardware; you must provide your own
19932 runtime library for whatever I/O functions are needed.
19934 @item -memregs=@var{number}
19936 Specifies the number of memory-based pseudo-registers GCC uses
19937 during code generation. These pseudo-registers are used like real
19938 registers, so there is a tradeoff between GCC's ability to fit the
19939 code into available registers, and the performance penalty of using
19940 memory instead of registers. Note that all modules in a program must
19941 be compiled with the same value for this option. Because of that, you
19942 must not use this option with GCC's default runtime libraries.
19946 @node M32R/D Options
19947 @subsection M32R/D Options
19948 @cindex M32R/D options
19950 These @option{-m} options are defined for Renesas M32R/D architectures:
19955 Generate code for the M32R/2@.
19959 Generate code for the M32R/X@.
19963 Generate code for the M32R@. This is the default.
19965 @item -mmodel=small
19966 @opindex mmodel=small
19967 Assume all objects live in the lower 16MB of memory (so that their addresses
19968 can be loaded with the @code{ld24} instruction), and assume all subroutines
19969 are reachable with the @code{bl} instruction.
19970 This is the default.
19972 The addressability of a particular object can be set with the
19973 @code{model} attribute.
19975 @item -mmodel=medium
19976 @opindex mmodel=medium
19977 Assume objects may be anywhere in the 32-bit address space (the compiler
19978 generates @code{seth/add3} instructions to load their addresses), and
19979 assume all subroutines are reachable with the @code{bl} instruction.
19981 @item -mmodel=large
19982 @opindex mmodel=large
19983 Assume objects may be anywhere in the 32-bit address space (the compiler
19984 generates @code{seth/add3} instructions to load their addresses), and
19985 assume subroutines may not be reachable with the @code{bl} instruction
19986 (the compiler generates the much slower @code{seth/add3/jl}
19987 instruction sequence).
19990 @opindex msdata=none
19991 Disable use of the small data area. Variables are put into
19992 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
19993 @code{section} attribute has been specified).
19994 This is the default.
19996 The small data area consists of sections @code{.sdata} and @code{.sbss}.
19997 Objects may be explicitly put in the small data area with the
19998 @code{section} attribute using one of these sections.
20000 @item -msdata=sdata
20001 @opindex msdata=sdata
20002 Put small global and static data in the small data area, but do not
20003 generate special code to reference them.
20006 @opindex msdata=use
20007 Put small global and static data in the small data area, and generate
20008 special instructions to reference them.
20012 @cindex smaller data references
20013 Put global and static objects less than or equal to @var{num} bytes
20014 into the small data or BSS sections instead of the normal data or BSS
20015 sections. The default value of @var{num} is 8.
20016 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
20017 for this option to have any effect.
20019 All modules should be compiled with the same @option{-G @var{num}} value.
20020 Compiling with different values of @var{num} may or may not work; if it
20021 doesn't the linker gives an error message---incorrect code is not
20026 Makes the M32R-specific code in the compiler display some statistics
20027 that might help in debugging programs.
20029 @item -malign-loops
20030 @opindex malign-loops
20031 Align all loops to a 32-byte boundary.
20033 @item -mno-align-loops
20034 @opindex mno-align-loops
20035 Do not enforce a 32-byte alignment for loops. This is the default.
20037 @item -missue-rate=@var{number}
20038 @opindex missue-rate=@var{number}
20039 Issue @var{number} instructions per cycle. @var{number} can only be 1
20042 @item -mbranch-cost=@var{number}
20043 @opindex mbranch-cost=@var{number}
20044 @var{number} can only be 1 or 2. If it is 1 then branches are
20045 preferred over conditional code, if it is 2, then the opposite applies.
20047 @item -mflush-trap=@var{number}
20048 @opindex mflush-trap=@var{number}
20049 Specifies the trap number to use to flush the cache. The default is
20050 12. Valid numbers are between 0 and 15 inclusive.
20052 @item -mno-flush-trap
20053 @opindex mno-flush-trap
20054 Specifies that the cache cannot be flushed by using a trap.
20056 @item -mflush-func=@var{name}
20057 @opindex mflush-func=@var{name}
20058 Specifies the name of the operating system function to call to flush
20059 the cache. The default is @samp{_flush_cache}, but a function call
20060 is only used if a trap is not available.
20062 @item -mno-flush-func
20063 @opindex mno-flush-func
20064 Indicates that there is no OS function for flushing the cache.
20068 @node M680x0 Options
20069 @subsection M680x0 Options
20070 @cindex M680x0 options
20072 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
20073 The default settings depend on which architecture was selected when
20074 the compiler was configured; the defaults for the most common choices
20078 @item -march=@var{arch}
20080 Generate code for a specific M680x0 or ColdFire instruction set
20081 architecture. Permissible values of @var{arch} for M680x0
20082 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
20083 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
20084 architectures are selected according to Freescale's ISA classification
20085 and the permissible values are: @samp{isaa}, @samp{isaaplus},
20086 @samp{isab} and @samp{isac}.
20088 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
20089 code for a ColdFire target. The @var{arch} in this macro is one of the
20090 @option{-march} arguments given above.
20092 When used together, @option{-march} and @option{-mtune} select code
20093 that runs on a family of similar processors but that is optimized
20094 for a particular microarchitecture.
20096 @item -mcpu=@var{cpu}
20098 Generate code for a specific M680x0 or ColdFire processor.
20099 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
20100 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
20101 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
20102 below, which also classifies the CPUs into families:
20104 @multitable @columnfractions 0.20 0.80
20105 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
20106 @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}
20107 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
20108 @item @samp{5206e} @tab @samp{5206e}
20109 @item @samp{5208} @tab @samp{5207} @samp{5208}
20110 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
20111 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
20112 @item @samp{5216} @tab @samp{5214} @samp{5216}
20113 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
20114 @item @samp{5225} @tab @samp{5224} @samp{5225}
20115 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
20116 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
20117 @item @samp{5249} @tab @samp{5249}
20118 @item @samp{5250} @tab @samp{5250}
20119 @item @samp{5271} @tab @samp{5270} @samp{5271}
20120 @item @samp{5272} @tab @samp{5272}
20121 @item @samp{5275} @tab @samp{5274} @samp{5275}
20122 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
20123 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
20124 @item @samp{5307} @tab @samp{5307}
20125 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
20126 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
20127 @item @samp{5407} @tab @samp{5407}
20128 @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}
20131 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
20132 @var{arch} is compatible with @var{cpu}. Other combinations of
20133 @option{-mcpu} and @option{-march} are rejected.
20135 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
20136 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
20137 where the value of @var{family} is given by the table above.
20139 @item -mtune=@var{tune}
20141 Tune the code for a particular microarchitecture within the
20142 constraints set by @option{-march} and @option{-mcpu}.
20143 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
20144 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
20145 and @samp{cpu32}. The ColdFire microarchitectures
20146 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
20148 You can also use @option{-mtune=68020-40} for code that needs
20149 to run relatively well on 68020, 68030 and 68040 targets.
20150 @option{-mtune=68020-60} is similar but includes 68060 targets
20151 as well. These two options select the same tuning decisions as
20152 @option{-m68020-40} and @option{-m68020-60} respectively.
20154 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
20155 when tuning for 680x0 architecture @var{arch}. It also defines
20156 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
20157 option is used. If GCC is tuning for a range of architectures,
20158 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
20159 it defines the macros for every architecture in the range.
20161 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
20162 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
20163 of the arguments given above.
20169 Generate output for a 68000. This is the default
20170 when the compiler is configured for 68000-based systems.
20171 It is equivalent to @option{-march=68000}.
20173 Use this option for microcontrollers with a 68000 or EC000 core,
20174 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
20178 Generate output for a 68010. This is the default
20179 when the compiler is configured for 68010-based systems.
20180 It is equivalent to @option{-march=68010}.
20186 Generate output for a 68020. This is the default
20187 when the compiler is configured for 68020-based systems.
20188 It is equivalent to @option{-march=68020}.
20192 Generate output for a 68030. This is the default when the compiler is
20193 configured for 68030-based systems. It is equivalent to
20194 @option{-march=68030}.
20198 Generate output for a 68040. This is the default when the compiler is
20199 configured for 68040-based systems. It is equivalent to
20200 @option{-march=68040}.
20202 This option inhibits the use of 68881/68882 instructions that have to be
20203 emulated by software on the 68040. Use this option if your 68040 does not
20204 have code to emulate those instructions.
20208 Generate output for a 68060. This is the default when the compiler is
20209 configured for 68060-based systems. It is equivalent to
20210 @option{-march=68060}.
20212 This option inhibits the use of 68020 and 68881/68882 instructions that
20213 have to be emulated by software on the 68060. Use this option if your 68060
20214 does not have code to emulate those instructions.
20218 Generate output for a CPU32. This is the default
20219 when the compiler is configured for CPU32-based systems.
20220 It is equivalent to @option{-march=cpu32}.
20222 Use this option for microcontrollers with a
20223 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
20224 68336, 68340, 68341, 68349 and 68360.
20228 Generate output for a 520X ColdFire CPU@. This is the default
20229 when the compiler is configured for 520X-based systems.
20230 It is equivalent to @option{-mcpu=5206}, and is now deprecated
20231 in favor of that option.
20233 Use this option for microcontroller with a 5200 core, including
20234 the MCF5202, MCF5203, MCF5204 and MCF5206.
20238 Generate output for a 5206e ColdFire CPU@. The option is now
20239 deprecated in favor of the equivalent @option{-mcpu=5206e}.
20243 Generate output for a member of the ColdFire 528X family.
20244 The option is now deprecated in favor of the equivalent
20245 @option{-mcpu=528x}.
20249 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
20250 in favor of the equivalent @option{-mcpu=5307}.
20254 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
20255 in favor of the equivalent @option{-mcpu=5407}.
20259 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
20260 This includes use of hardware floating-point instructions.
20261 The option is equivalent to @option{-mcpu=547x}, and is now
20262 deprecated in favor of that option.
20266 Generate output for a 68040, without using any of the new instructions.
20267 This results in code that can run relatively efficiently on either a
20268 68020/68881 or a 68030 or a 68040. The generated code does use the
20269 68881 instructions that are emulated on the 68040.
20271 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
20275 Generate output for a 68060, without using any of the new instructions.
20276 This results in code that can run relatively efficiently on either a
20277 68020/68881 or a 68030 or a 68040. The generated code does use the
20278 68881 instructions that are emulated on the 68060.
20280 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
20284 @opindex mhard-float
20286 Generate floating-point instructions. This is the default for 68020
20287 and above, and for ColdFire devices that have an FPU@. It defines the
20288 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
20289 on ColdFire targets.
20292 @opindex msoft-float
20293 Do not generate floating-point instructions; use library calls instead.
20294 This is the default for 68000, 68010, and 68832 targets. It is also
20295 the default for ColdFire devices that have no FPU.
20301 Generate (do not generate) ColdFire hardware divide and remainder
20302 instructions. If @option{-march} is used without @option{-mcpu},
20303 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
20304 architectures. Otherwise, the default is taken from the target CPU
20305 (either the default CPU, or the one specified by @option{-mcpu}). For
20306 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
20307 @option{-mcpu=5206e}.
20309 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
20313 Consider type @code{int} to be 16 bits wide, like @code{short int}.
20314 Additionally, parameters passed on the stack are also aligned to a
20315 16-bit boundary even on targets whose API mandates promotion to 32-bit.
20319 Do not consider type @code{int} to be 16 bits wide. This is the default.
20322 @itemx -mno-bitfield
20323 @opindex mnobitfield
20324 @opindex mno-bitfield
20325 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
20326 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
20330 Do use the bit-field instructions. The @option{-m68020} option implies
20331 @option{-mbitfield}. This is the default if you use a configuration
20332 designed for a 68020.
20336 Use a different function-calling convention, in which functions
20337 that take a fixed number of arguments return with the @code{rtd}
20338 instruction, which pops their arguments while returning. This
20339 saves one instruction in the caller since there is no need to pop
20340 the arguments there.
20342 This calling convention is incompatible with the one normally
20343 used on Unix, so you cannot use it if you need to call libraries
20344 compiled with the Unix compiler.
20346 Also, you must provide function prototypes for all functions that
20347 take variable numbers of arguments (including @code{printf});
20348 otherwise incorrect code is generated for calls to those
20351 In addition, seriously incorrect code results if you call a
20352 function with too many arguments. (Normally, extra arguments are
20353 harmlessly ignored.)
20355 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
20356 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
20360 Do not use the calling conventions selected by @option{-mrtd}.
20361 This is the default.
20364 @itemx -mno-align-int
20365 @opindex malign-int
20366 @opindex mno-align-int
20367 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
20368 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
20369 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
20370 Aligning variables on 32-bit boundaries produces code that runs somewhat
20371 faster on processors with 32-bit busses at the expense of more memory.
20373 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
20374 aligns structures containing the above types differently than
20375 most published application binary interface specifications for the m68k.
20379 Use the pc-relative addressing mode of the 68000 directly, instead of
20380 using a global offset table. At present, this option implies @option{-fpic},
20381 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
20382 not presently supported with @option{-mpcrel}, though this could be supported for
20383 68020 and higher processors.
20385 @item -mno-strict-align
20386 @itemx -mstrict-align
20387 @opindex mno-strict-align
20388 @opindex mstrict-align
20389 Do not (do) assume that unaligned memory references are handled by
20393 Generate code that allows the data segment to be located in a different
20394 area of memory from the text segment. This allows for execute-in-place in
20395 an environment without virtual memory management. This option implies
20398 @item -mno-sep-data
20399 Generate code that assumes that the data segment follows the text segment.
20400 This is the default.
20402 @item -mid-shared-library
20403 Generate code that supports shared libraries via the library ID method.
20404 This allows for execute-in-place and shared libraries in an environment
20405 without virtual memory management. This option implies @option{-fPIC}.
20407 @item -mno-id-shared-library
20408 Generate code that doesn't assume ID-based shared libraries are being used.
20409 This is the default.
20411 @item -mshared-library-id=n
20412 Specifies the identification number of the ID-based shared library being
20413 compiled. Specifying a value of 0 generates more compact code; specifying
20414 other values forces the allocation of that number to the current
20415 library, but is no more space- or time-efficient than omitting this option.
20421 When generating position-independent code for ColdFire, generate code
20422 that works if the GOT has more than 8192 entries. This code is
20423 larger and slower than code generated without this option. On M680x0
20424 processors, this option is not needed; @option{-fPIC} suffices.
20426 GCC normally uses a single instruction to load values from the GOT@.
20427 While this is relatively efficient, it only works if the GOT
20428 is smaller than about 64k. Anything larger causes the linker
20429 to report an error such as:
20431 @cindex relocation truncated to fit (ColdFire)
20433 relocation truncated to fit: R_68K_GOT16O foobar
20436 If this happens, you should recompile your code with @option{-mxgot}.
20437 It should then work with very large GOTs. However, code generated with
20438 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
20439 the value of a global symbol.
20441 Note that some linkers, including newer versions of the GNU linker,
20442 can create multiple GOTs and sort GOT entries. If you have such a linker,
20443 you should only need to use @option{-mxgot} when compiling a single
20444 object file that accesses more than 8192 GOT entries. Very few do.
20446 These options have no effect unless GCC is generating
20447 position-independent code.
20449 @item -mlong-jump-table-offsets
20450 @opindex mlong-jump-table-offsets
20451 Use 32-bit offsets in @code{switch} tables. The default is to use
20456 @node MCore Options
20457 @subsection MCore Options
20458 @cindex MCore options
20460 These are the @samp{-m} options defined for the Motorola M*Core
20466 @itemx -mno-hardlit
20468 @opindex mno-hardlit
20469 Inline constants into the code stream if it can be done in two
20470 instructions or less.
20476 Use the divide instruction. (Enabled by default).
20478 @item -mrelax-immediate
20479 @itemx -mno-relax-immediate
20480 @opindex mrelax-immediate
20481 @opindex mno-relax-immediate
20482 Allow arbitrary-sized immediates in bit operations.
20484 @item -mwide-bitfields
20485 @itemx -mno-wide-bitfields
20486 @opindex mwide-bitfields
20487 @opindex mno-wide-bitfields
20488 Always treat bit-fields as @code{int}-sized.
20490 @item -m4byte-functions
20491 @itemx -mno-4byte-functions
20492 @opindex m4byte-functions
20493 @opindex mno-4byte-functions
20494 Force all functions to be aligned to a 4-byte boundary.
20496 @item -mcallgraph-data
20497 @itemx -mno-callgraph-data
20498 @opindex mcallgraph-data
20499 @opindex mno-callgraph-data
20500 Emit callgraph information.
20503 @itemx -mno-slow-bytes
20504 @opindex mslow-bytes
20505 @opindex mno-slow-bytes
20506 Prefer word access when reading byte quantities.
20508 @item -mlittle-endian
20509 @itemx -mbig-endian
20510 @opindex mlittle-endian
20511 @opindex mbig-endian
20512 Generate code for a little-endian target.
20518 Generate code for the 210 processor.
20522 Assume that runtime support has been provided and so omit the
20523 simulator library (@file{libsim.a)} from the linker command line.
20525 @item -mstack-increment=@var{size}
20526 @opindex mstack-increment
20527 Set the maximum amount for a single stack increment operation. Large
20528 values can increase the speed of programs that contain functions
20529 that need a large amount of stack space, but they can also trigger a
20530 segmentation fault if the stack is extended too much. The default
20536 @subsection MeP Options
20537 @cindex MeP options
20543 Enables the @code{abs} instruction, which is the absolute difference
20544 between two registers.
20548 Enables all the optional instructions---average, multiply, divide, bit
20549 operations, leading zero, absolute difference, min/max, clip, and
20555 Enables the @code{ave} instruction, which computes the average of two
20558 @item -mbased=@var{n}
20560 Variables of size @var{n} bytes or smaller are placed in the
20561 @code{.based} section by default. Based variables use the @code{$tp}
20562 register as a base register, and there is a 128-byte limit to the
20563 @code{.based} section.
20567 Enables the bit operation instructions---bit test (@code{btstm}), set
20568 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
20569 test-and-set (@code{tas}).
20571 @item -mc=@var{name}
20573 Selects which section constant data is placed in. @var{name} may
20574 be @samp{tiny}, @samp{near}, or @samp{far}.
20578 Enables the @code{clip} instruction. Note that @option{-mclip} is not
20579 useful unless you also provide @option{-mminmax}.
20581 @item -mconfig=@var{name}
20583 Selects one of the built-in core configurations. Each MeP chip has
20584 one or more modules in it; each module has a core CPU and a variety of
20585 coprocessors, optional instructions, and peripherals. The
20586 @code{MeP-Integrator} tool, not part of GCC, provides these
20587 configurations through this option; using this option is the same as
20588 using all the corresponding command-line options. The default
20589 configuration is @samp{default}.
20593 Enables the coprocessor instructions. By default, this is a 32-bit
20594 coprocessor. Note that the coprocessor is normally enabled via the
20595 @option{-mconfig=} option.
20599 Enables the 32-bit coprocessor's instructions.
20603 Enables the 64-bit coprocessor's instructions.
20607 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
20611 Causes constant variables to be placed in the @code{.near} section.
20615 Enables the @code{div} and @code{divu} instructions.
20619 Generate big-endian code.
20623 Generate little-endian code.
20625 @item -mio-volatile
20626 @opindex mio-volatile
20627 Tells the compiler that any variable marked with the @code{io}
20628 attribute is to be considered volatile.
20632 Causes variables to be assigned to the @code{.far} section by default.
20636 Enables the @code{leadz} (leading zero) instruction.
20640 Causes variables to be assigned to the @code{.near} section by default.
20644 Enables the @code{min} and @code{max} instructions.
20648 Enables the multiplication and multiply-accumulate instructions.
20652 Disables all the optional instructions enabled by @option{-mall-opts}.
20656 Enables the @code{repeat} and @code{erepeat} instructions, used for
20657 low-overhead looping.
20661 Causes all variables to default to the @code{.tiny} section. Note
20662 that there is a 65536-byte limit to this section. Accesses to these
20663 variables use the @code{%gp} base register.
20667 Enables the saturation instructions. Note that the compiler does not
20668 currently generate these itself, but this option is included for
20669 compatibility with other tools, like @code{as}.
20673 Link the SDRAM-based runtime instead of the default ROM-based runtime.
20677 Link the simulator run-time libraries.
20681 Link the simulator runtime libraries, excluding built-in support
20682 for reset and exception vectors and tables.
20686 Causes all functions to default to the @code{.far} section. Without
20687 this option, functions default to the @code{.near} section.
20689 @item -mtiny=@var{n}
20691 Variables that are @var{n} bytes or smaller are allocated to the
20692 @code{.tiny} section. These variables use the @code{$gp} base
20693 register. The default for this option is 4, but note that there's a
20694 65536-byte limit to the @code{.tiny} section.
20698 @node MicroBlaze Options
20699 @subsection MicroBlaze Options
20700 @cindex MicroBlaze Options
20705 @opindex msoft-float
20706 Use software emulation for floating point (default).
20709 @opindex mhard-float
20710 Use hardware floating-point instructions.
20714 Do not optimize block moves, use @code{memcpy}.
20716 @item -mno-clearbss
20717 @opindex mno-clearbss
20718 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
20720 @item -mcpu=@var{cpu-type}
20722 Use features of, and schedule code for, the given CPU.
20723 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
20724 where @var{X} is a major version, @var{YY} is the minor version, and
20725 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
20726 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}.
20728 @item -mxl-soft-mul
20729 @opindex mxl-soft-mul
20730 Use software multiply emulation (default).
20732 @item -mxl-soft-div
20733 @opindex mxl-soft-div
20734 Use software emulation for divides (default).
20736 @item -mxl-barrel-shift
20737 @opindex mxl-barrel-shift
20738 Use the hardware barrel shifter.
20740 @item -mxl-pattern-compare
20741 @opindex mxl-pattern-compare
20742 Use pattern compare instructions.
20744 @item -msmall-divides
20745 @opindex msmall-divides
20746 Use table lookup optimization for small signed integer divisions.
20748 @item -mxl-stack-check
20749 @opindex mxl-stack-check
20750 This option is deprecated. Use @option{-fstack-check} instead.
20753 @opindex mxl-gp-opt
20754 Use GP-relative @code{.sdata}/@code{.sbss} sections.
20756 @item -mxl-multiply-high
20757 @opindex mxl-multiply-high
20758 Use multiply high instructions for high part of 32x32 multiply.
20760 @item -mxl-float-convert
20761 @opindex mxl-float-convert
20762 Use hardware floating-point conversion instructions.
20764 @item -mxl-float-sqrt
20765 @opindex mxl-float-sqrt
20766 Use hardware floating-point square root instruction.
20769 @opindex mbig-endian
20770 Generate code for a big-endian target.
20772 @item -mlittle-endian
20773 @opindex mlittle-endian
20774 Generate code for a little-endian target.
20777 @opindex mxl-reorder
20778 Use reorder instructions (swap and byte reversed load/store).
20780 @item -mxl-mode-@var{app-model}
20781 Select application model @var{app-model}. Valid models are
20784 normal executable (default), uses startup code @file{crt0.o}.
20786 @item -mpic-data-is-text-relative
20787 @opindex mpic-data-is-text-relative
20788 Assume that the displacement between the text and data segments is fixed
20789 at static link time. This allows data to be referenced by offset from start of
20790 text address instead of GOT since PC-relative addressing is not supported.
20793 for use with Xilinx Microprocessor Debugger (XMD) based
20794 software intrusive debug agent called xmdstub. This uses startup file
20795 @file{crt1.o} and sets the start address of the program to 0x800.
20798 for applications that are loaded using a bootloader.
20799 This model uses startup file @file{crt2.o} which does not contain a processor
20800 reset vector handler. This is suitable for transferring control on a
20801 processor reset to the bootloader rather than the application.
20804 for applications that do not require any of the
20805 MicroBlaze vectors. This option may be useful for applications running
20806 within a monitoring application. This model uses @file{crt3.o} as a startup file.
20809 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
20810 @option{-mxl-mode-@var{app-model}}.
20815 @subsection MIPS Options
20816 @cindex MIPS options
20822 Generate big-endian code.
20826 Generate little-endian code. This is the default for @samp{mips*el-*-*}
20829 @item -march=@var{arch}
20831 Generate code that runs on @var{arch}, which can be the name of a
20832 generic MIPS ISA, or the name of a particular processor.
20834 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
20835 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
20836 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
20837 @samp{mips64r5} and @samp{mips64r6}.
20838 The processor names are:
20839 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
20840 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
20841 @samp{5kc}, @samp{5kf},
20843 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
20844 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
20845 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
20846 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
20847 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
20848 @samp{i6400}, @samp{i6500},
20850 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464},
20851 @samp{gs464e}, @samp{gs264e},
20853 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
20854 @samp{m5100}, @samp{m5101},
20855 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
20857 @samp{p5600}, @samp{p6600},
20858 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
20859 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
20860 @samp{rm7000}, @samp{rm9000},
20861 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
20864 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
20865 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
20866 @samp{xlr} and @samp{xlp}.
20867 The special value @samp{from-abi} selects the
20868 most compatible architecture for the selected ABI (that is,
20869 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
20871 The native Linux/GNU toolchain also supports the value @samp{native},
20872 which selects the best architecture option for the host processor.
20873 @option{-march=native} has no effect if GCC does not recognize
20876 In processor names, a final @samp{000} can be abbreviated as @samp{k}
20877 (for example, @option{-march=r2k}). Prefixes are optional, and
20878 @samp{vr} may be written @samp{r}.
20880 Names of the form @samp{@var{n}f2_1} refer to processors with
20881 FPUs clocked at half the rate of the core, names of the form
20882 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
20883 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
20884 processors with FPUs clocked a ratio of 3:2 with respect to the core.
20885 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
20886 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
20887 accepted as synonyms for @samp{@var{n}f1_1}.
20889 GCC defines two macros based on the value of this option. The first
20890 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
20891 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
20892 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
20893 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
20894 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
20896 Note that the @code{_MIPS_ARCH} macro uses the processor names given
20897 above. In other words, it has the full prefix and does not
20898 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
20899 the macro names the resolved architecture (either @code{"mips1"} or
20900 @code{"mips3"}). It names the default architecture when no
20901 @option{-march} option is given.
20903 @item -mtune=@var{arch}
20905 Optimize for @var{arch}. Among other things, this option controls
20906 the way instructions are scheduled, and the perceived cost of arithmetic
20907 operations. The list of @var{arch} values is the same as for
20910 When this option is not used, GCC optimizes for the processor
20911 specified by @option{-march}. By using @option{-march} and
20912 @option{-mtune} together, it is possible to generate code that
20913 runs on a family of processors, but optimize the code for one
20914 particular member of that family.
20916 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
20917 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
20918 @option{-march} ones described above.
20922 Equivalent to @option{-march=mips1}.
20926 Equivalent to @option{-march=mips2}.
20930 Equivalent to @option{-march=mips3}.
20934 Equivalent to @option{-march=mips4}.
20938 Equivalent to @option{-march=mips32}.
20942 Equivalent to @option{-march=mips32r3}.
20946 Equivalent to @option{-march=mips32r5}.
20950 Equivalent to @option{-march=mips32r6}.
20954 Equivalent to @option{-march=mips64}.
20958 Equivalent to @option{-march=mips64r2}.
20962 Equivalent to @option{-march=mips64r3}.
20966 Equivalent to @option{-march=mips64r5}.
20970 Equivalent to @option{-march=mips64r6}.
20975 @opindex mno-mips16
20976 Generate (do not generate) MIPS16 code. If GCC is targeting a
20977 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
20979 MIPS16 code generation can also be controlled on a per-function basis
20980 by means of @code{mips16} and @code{nomips16} attributes.
20981 @xref{Function Attributes}, for more information.
20983 @item -mflip-mips16
20984 @opindex mflip-mips16
20985 Generate MIPS16 code on alternating functions. This option is provided
20986 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
20987 not intended for ordinary use in compiling user code.
20989 @item -minterlink-compressed
20990 @itemx -mno-interlink-compressed
20991 @opindex minterlink-compressed
20992 @opindex mno-interlink-compressed
20993 Require (do not require) that code using the standard (uncompressed) MIPS ISA
20994 be link-compatible with MIPS16 and microMIPS code, and vice versa.
20996 For example, code using the standard ISA encoding cannot jump directly
20997 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
20998 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
20999 knows that the target of the jump is not compressed.
21001 @item -minterlink-mips16
21002 @itemx -mno-interlink-mips16
21003 @opindex minterlink-mips16
21004 @opindex mno-interlink-mips16
21005 Aliases of @option{-minterlink-compressed} and
21006 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
21007 and are retained for backwards compatibility.
21019 Generate code for the given ABI@.
21021 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
21022 generates 64-bit code when you select a 64-bit architecture, but you
21023 can use @option{-mgp32} to get 32-bit code instead.
21025 For information about the O64 ABI, see
21026 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
21028 GCC supports a variant of the o32 ABI in which floating-point registers
21029 are 64 rather than 32 bits wide. You can select this combination with
21030 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
21031 and @code{mfhc1} instructions and is therefore only supported for
21032 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
21034 The register assignments for arguments and return values remain the
21035 same, but each scalar value is passed in a single 64-bit register
21036 rather than a pair of 32-bit registers. For example, scalar
21037 floating-point values are returned in @samp{$f0} only, not a
21038 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
21039 remains the same in that the even-numbered double-precision registers
21042 Two additional variants of the o32 ABI are supported to enable
21043 a transition from 32-bit to 64-bit registers. These are FPXX
21044 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
21045 The FPXX extension mandates that all code must execute correctly
21046 when run using 32-bit or 64-bit registers. The code can be interlinked
21047 with either FP32 or FP64, but not both.
21048 The FP64A extension is similar to the FP64 extension but forbids the
21049 use of odd-numbered single-precision registers. This can be used
21050 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
21051 processors and allows both FP32 and FP64A code to interlink and
21052 run in the same process without changing FPU modes.
21055 @itemx -mno-abicalls
21057 @opindex mno-abicalls
21058 Generate (do not generate) code that is suitable for SVR4-style
21059 dynamic objects. @option{-mabicalls} is the default for SVR4-based
21064 Generate (do not generate) code that is fully position-independent,
21065 and that can therefore be linked into shared libraries. This option
21066 only affects @option{-mabicalls}.
21068 All @option{-mabicalls} code has traditionally been position-independent,
21069 regardless of options like @option{-fPIC} and @option{-fpic}. However,
21070 as an extension, the GNU toolchain allows executables to use absolute
21071 accesses for locally-binding symbols. It can also use shorter GP
21072 initialization sequences and generate direct calls to locally-defined
21073 functions. This mode is selected by @option{-mno-shared}.
21075 @option{-mno-shared} depends on binutils 2.16 or higher and generates
21076 objects that can only be linked by the GNU linker. However, the option
21077 does not affect the ABI of the final executable; it only affects the ABI
21078 of relocatable objects. Using @option{-mno-shared} generally makes
21079 executables both smaller and quicker.
21081 @option{-mshared} is the default.
21087 Assume (do not assume) that the static and dynamic linkers
21088 support PLTs and copy relocations. This option only affects
21089 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
21090 has no effect without @option{-msym32}.
21092 You can make @option{-mplt} the default by configuring
21093 GCC with @option{--with-mips-plt}. The default is
21094 @option{-mno-plt} otherwise.
21100 Lift (do not lift) the usual restrictions on the size of the global
21103 GCC normally uses a single instruction to load values from the GOT@.
21104 While this is relatively efficient, it only works if the GOT
21105 is smaller than about 64k. Anything larger causes the linker
21106 to report an error such as:
21108 @cindex relocation truncated to fit (MIPS)
21110 relocation truncated to fit: R_MIPS_GOT16 foobar
21113 If this happens, you should recompile your code with @option{-mxgot}.
21114 This works with very large GOTs, although the code is also
21115 less efficient, since it takes three instructions to fetch the
21116 value of a global symbol.
21118 Note that some linkers can create multiple GOTs. If you have such a
21119 linker, you should only need to use @option{-mxgot} when a single object
21120 file accesses more than 64k's worth of GOT entries. Very few do.
21122 These options have no effect unless GCC is generating position
21127 Assume that general-purpose registers are 32 bits wide.
21131 Assume that general-purpose registers are 64 bits wide.
21135 Assume that floating-point registers are 32 bits wide.
21139 Assume that floating-point registers are 64 bits wide.
21143 Do not assume the width of floating-point registers.
21146 @opindex mhard-float
21147 Use floating-point coprocessor instructions.
21150 @opindex msoft-float
21151 Do not use floating-point coprocessor instructions. Implement
21152 floating-point calculations using library calls instead.
21156 Equivalent to @option{-msoft-float}, but additionally asserts that the
21157 program being compiled does not perform any floating-point operations.
21158 This option is presently supported only by some bare-metal MIPS
21159 configurations, where it may select a special set of libraries
21160 that lack all floating-point support (including, for example, the
21161 floating-point @code{printf} formats).
21162 If code compiled with @option{-mno-float} accidentally contains
21163 floating-point operations, it is likely to suffer a link-time
21164 or run-time failure.
21166 @item -msingle-float
21167 @opindex msingle-float
21168 Assume that the floating-point coprocessor only supports single-precision
21171 @item -mdouble-float
21172 @opindex mdouble-float
21173 Assume that the floating-point coprocessor supports double-precision
21174 operations. This is the default.
21177 @itemx -mno-odd-spreg
21178 @opindex modd-spreg
21179 @opindex mno-odd-spreg
21180 Enable the use of odd-numbered single-precision floating-point registers
21181 for the o32 ABI. This is the default for processors that are known to
21182 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
21186 @itemx -mabs=legacy
21188 @opindex mabs=legacy
21189 These options control the treatment of the special not-a-number (NaN)
21190 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
21191 @code{neg.@i{fmt}} machine instructions.
21193 By default or when @option{-mabs=legacy} is used the legacy
21194 treatment is selected. In this case these instructions are considered
21195 arithmetic and avoided where correct operation is required and the
21196 input operand might be a NaN. A longer sequence of instructions that
21197 manipulate the sign bit of floating-point datum manually is used
21198 instead unless the @option{-ffinite-math-only} option has also been
21201 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
21202 this case these instructions are considered non-arithmetic and therefore
21203 operating correctly in all cases, including in particular where the
21204 input operand is a NaN. These instructions are therefore always used
21205 for the respective operations.
21208 @itemx -mnan=legacy
21210 @opindex mnan=legacy
21211 These options control the encoding of the special not-a-number (NaN)
21212 IEEE 754 floating-point data.
21214 The @option{-mnan=legacy} option selects the legacy encoding. In this
21215 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
21216 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
21217 by the first bit of their trailing significand field being 1.
21219 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
21220 this case qNaNs are denoted by the first bit of their trailing
21221 significand field being 1, whereas sNaNs are denoted by the first bit of
21222 their trailing significand field being 0.
21224 The default is @option{-mnan=legacy} unless GCC has been configured with
21225 @option{--with-nan=2008}.
21231 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
21232 implement atomic memory built-in functions. When neither option is
21233 specified, GCC uses the instructions if the target architecture
21236 @option{-mllsc} is useful if the runtime environment can emulate the
21237 instructions and @option{-mno-llsc} can be useful when compiling for
21238 nonstandard ISAs. You can make either option the default by
21239 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
21240 respectively. @option{--with-llsc} is the default for some
21241 configurations; see the installation documentation for details.
21247 Use (do not use) revision 1 of the MIPS DSP ASE@.
21248 @xref{MIPS DSP Built-in Functions}. This option defines the
21249 preprocessor macro @code{__mips_dsp}. It also defines
21250 @code{__mips_dsp_rev} to 1.
21256 Use (do not use) revision 2 of the MIPS DSP ASE@.
21257 @xref{MIPS DSP Built-in Functions}. This option defines the
21258 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
21259 It also defines @code{__mips_dsp_rev} to 2.
21262 @itemx -mno-smartmips
21263 @opindex msmartmips
21264 @opindex mno-smartmips
21265 Use (do not use) the MIPS SmartMIPS ASE.
21267 @item -mpaired-single
21268 @itemx -mno-paired-single
21269 @opindex mpaired-single
21270 @opindex mno-paired-single
21271 Use (do not use) paired-single floating-point instructions.
21272 @xref{MIPS Paired-Single Support}. This option requires
21273 hardware floating-point support to be enabled.
21279 Use (do not use) MIPS Digital Media Extension instructions.
21280 This option can only be used when generating 64-bit code and requires
21281 hardware floating-point support to be enabled.
21286 @opindex mno-mips3d
21287 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
21288 The option @option{-mips3d} implies @option{-mpaired-single}.
21291 @itemx -mno-micromips
21292 @opindex mmicromips
21293 @opindex mno-mmicromips
21294 Generate (do not generate) microMIPS code.
21296 MicroMIPS code generation can also be controlled on a per-function basis
21297 by means of @code{micromips} and @code{nomicromips} attributes.
21298 @xref{Function Attributes}, for more information.
21304 Use (do not use) MT Multithreading instructions.
21310 Use (do not use) the MIPS MCU ASE instructions.
21316 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
21322 Use (do not use) the MIPS Virtualization (VZ) instructions.
21328 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
21334 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions.
21340 Use (do not use) the MIPS Global INValidate (GINV) instructions.
21342 @item -mloongson-mmi
21343 @itemx -mno-loongson-mmi
21344 @opindex mloongson-mmi
21345 @opindex mno-loongson-mmi
21346 Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
21348 @item -mloongson-ext
21349 @itemx -mno-loongson-ext
21350 @opindex mloongson-ext
21351 @opindex mno-loongson-ext
21352 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions.
21354 @item -mloongson-ext2
21355 @itemx -mno-loongson-ext2
21356 @opindex mloongson-ext2
21357 @opindex mno-loongson-ext2
21358 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
21362 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
21363 an explanation of the default and the way that the pointer size is
21368 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
21370 The default size of @code{int}s, @code{long}s and pointers depends on
21371 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
21372 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
21373 32-bit @code{long}s. Pointers are the same size as @code{long}s,
21374 or the same size as integer registers, whichever is smaller.
21380 Assume (do not assume) that all symbols have 32-bit values, regardless
21381 of the selected ABI@. This option is useful in combination with
21382 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
21383 to generate shorter and faster references to symbolic addresses.
21387 Put definitions of externally-visible data in a small data section
21388 if that data is no bigger than @var{num} bytes. GCC can then generate
21389 more efficient accesses to the data; see @option{-mgpopt} for details.
21391 The default @option{-G} option depends on the configuration.
21393 @item -mlocal-sdata
21394 @itemx -mno-local-sdata
21395 @opindex mlocal-sdata
21396 @opindex mno-local-sdata
21397 Extend (do not extend) the @option{-G} behavior to local data too,
21398 such as to static variables in C@. @option{-mlocal-sdata} is the
21399 default for all configurations.
21401 If the linker complains that an application is using too much small data,
21402 you might want to try rebuilding the less performance-critical parts with
21403 @option{-mno-local-sdata}. You might also want to build large
21404 libraries with @option{-mno-local-sdata}, so that the libraries leave
21405 more room for the main program.
21407 @item -mextern-sdata
21408 @itemx -mno-extern-sdata
21409 @opindex mextern-sdata
21410 @opindex mno-extern-sdata
21411 Assume (do not assume) that externally-defined data is in
21412 a small data section if the size of that data is within the @option{-G} limit.
21413 @option{-mextern-sdata} is the default for all configurations.
21415 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
21416 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
21417 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
21418 is placed in a small data section. If @var{Var} is defined by another
21419 module, you must either compile that module with a high-enough
21420 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
21421 definition. If @var{Var} is common, you must link the application
21422 with a high-enough @option{-G} setting.
21424 The easiest way of satisfying these restrictions is to compile
21425 and link every module with the same @option{-G} option. However,
21426 you may wish to build a library that supports several different
21427 small data limits. You can do this by compiling the library with
21428 the highest supported @option{-G} setting and additionally using
21429 @option{-mno-extern-sdata} to stop the library from making assumptions
21430 about externally-defined data.
21436 Use (do not use) GP-relative accesses for symbols that are known to be
21437 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
21438 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
21441 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
21442 might not hold the value of @code{_gp}. For example, if the code is
21443 part of a library that might be used in a boot monitor, programs that
21444 call boot monitor routines pass an unknown value in @code{$gp}.
21445 (In such situations, the boot monitor itself is usually compiled
21446 with @option{-G0}.)
21448 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
21449 @option{-mno-extern-sdata}.
21451 @item -membedded-data
21452 @itemx -mno-embedded-data
21453 @opindex membedded-data
21454 @opindex mno-embedded-data
21455 Allocate variables to the read-only data section first if possible, then
21456 next in the small data section if possible, otherwise in data. This gives
21457 slightly slower code than the default, but reduces the amount of RAM required
21458 when executing, and thus may be preferred for some embedded systems.
21460 @item -muninit-const-in-rodata
21461 @itemx -mno-uninit-const-in-rodata
21462 @opindex muninit-const-in-rodata
21463 @opindex mno-uninit-const-in-rodata
21464 Put uninitialized @code{const} variables in the read-only data section.
21465 This option is only meaningful in conjunction with @option{-membedded-data}.
21467 @item -mcode-readable=@var{setting}
21468 @opindex mcode-readable
21469 Specify whether GCC may generate code that reads from executable sections.
21470 There are three possible settings:
21473 @item -mcode-readable=yes
21474 Instructions may freely access executable sections. This is the
21477 @item -mcode-readable=pcrel
21478 MIPS16 PC-relative load instructions can access executable sections,
21479 but other instructions must not do so. This option is useful on 4KSc
21480 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
21481 It is also useful on processors that can be configured to have a dual
21482 instruction/data SRAM interface and that, like the M4K, automatically
21483 redirect PC-relative loads to the instruction RAM.
21485 @item -mcode-readable=no
21486 Instructions must not access executable sections. This option can be
21487 useful on targets that are configured to have a dual instruction/data
21488 SRAM interface but that (unlike the M4K) do not automatically redirect
21489 PC-relative loads to the instruction RAM.
21492 @item -msplit-addresses
21493 @itemx -mno-split-addresses
21494 @opindex msplit-addresses
21495 @opindex mno-split-addresses
21496 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
21497 relocation operators. This option has been superseded by
21498 @option{-mexplicit-relocs} but is retained for backwards compatibility.
21500 @item -mexplicit-relocs
21501 @itemx -mno-explicit-relocs
21502 @opindex mexplicit-relocs
21503 @opindex mno-explicit-relocs
21504 Use (do not use) assembler relocation operators when dealing with symbolic
21505 addresses. The alternative, selected by @option{-mno-explicit-relocs},
21506 is to use assembler macros instead.
21508 @option{-mexplicit-relocs} is the default if GCC was configured
21509 to use an assembler that supports relocation operators.
21511 @item -mcheck-zero-division
21512 @itemx -mno-check-zero-division
21513 @opindex mcheck-zero-division
21514 @opindex mno-check-zero-division
21515 Trap (do not trap) on integer division by zero.
21517 The default is @option{-mcheck-zero-division}.
21519 @item -mdivide-traps
21520 @itemx -mdivide-breaks
21521 @opindex mdivide-traps
21522 @opindex mdivide-breaks
21523 MIPS systems check for division by zero by generating either a
21524 conditional trap or a break instruction. Using traps results in
21525 smaller code, but is only supported on MIPS II and later. Also, some
21526 versions of the Linux kernel have a bug that prevents trap from
21527 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
21528 allow conditional traps on architectures that support them and
21529 @option{-mdivide-breaks} to force the use of breaks.
21531 The default is usually @option{-mdivide-traps}, but this can be
21532 overridden at configure time using @option{--with-divide=breaks}.
21533 Divide-by-zero checks can be completely disabled using
21534 @option{-mno-check-zero-division}.
21536 @item -mload-store-pairs
21537 @itemx -mno-load-store-pairs
21538 @opindex mload-store-pairs
21539 @opindex mno-load-store-pairs
21540 Enable (disable) an optimization that pairs consecutive load or store
21541 instructions to enable load/store bonding. This option is enabled by
21542 default but only takes effect when the selected architecture is known
21543 to support bonding.
21548 @opindex mno-memcpy
21549 Force (do not force) the use of @code{memcpy} for non-trivial block
21550 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
21551 most constant-sized copies.
21554 @itemx -mno-long-calls
21555 @opindex mlong-calls
21556 @opindex mno-long-calls
21557 Disable (do not disable) use of the @code{jal} instruction. Calling
21558 functions using @code{jal} is more efficient but requires the caller
21559 and callee to be in the same 256 megabyte segment.
21561 This option has no effect on abicalls code. The default is
21562 @option{-mno-long-calls}.
21568 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
21569 instructions, as provided by the R4650 ISA@.
21575 Enable (disable) use of the @code{madd} and @code{msub} integer
21576 instructions. The default is @option{-mimadd} on architectures
21577 that support @code{madd} and @code{msub} except for the 74k
21578 architecture where it was found to generate slower code.
21581 @itemx -mno-fused-madd
21582 @opindex mfused-madd
21583 @opindex mno-fused-madd
21584 Enable (disable) use of the floating-point multiply-accumulate
21585 instructions, when they are available. The default is
21586 @option{-mfused-madd}.
21588 On the R8000 CPU when multiply-accumulate instructions are used,
21589 the intermediate product is calculated to infinite precision
21590 and is not subject to the FCSR Flush to Zero bit. This may be
21591 undesirable in some circumstances. On other processors the result
21592 is numerically identical to the equivalent computation using
21593 separate multiply, add, subtract and negate instructions.
21597 Tell the MIPS assembler to not run its preprocessor over user
21598 assembler files (with a @samp{.s} suffix) when assembling them.
21601 @itemx -mno-fix-24k
21603 @opindex mno-fix-24k
21604 Work around the 24K E48 (lost data on stores during refill) errata.
21605 The workarounds are implemented by the assembler rather than by GCC@.
21608 @itemx -mno-fix-r4000
21609 @opindex mfix-r4000
21610 @opindex mno-fix-r4000
21611 Work around certain R4000 CPU errata:
21614 A double-word or a variable shift may give an incorrect result if executed
21615 immediately after starting an integer division.
21617 A double-word or a variable shift may give an incorrect result if executed
21618 while an integer multiplication is in progress.
21620 An integer division may give an incorrect result if started in a delay slot
21621 of a taken branch or a jump.
21625 @itemx -mno-fix-r4400
21626 @opindex mfix-r4400
21627 @opindex mno-fix-r4400
21628 Work around certain R4400 CPU errata:
21631 A double-word or a variable shift may give an incorrect result if executed
21632 immediately after starting an integer division.
21636 @itemx -mno-fix-r10000
21637 @opindex mfix-r10000
21638 @opindex mno-fix-r10000
21639 Work around certain R10000 errata:
21642 @code{ll}/@code{sc} sequences may not behave atomically on revisions
21643 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
21646 This option can only be used if the target architecture supports
21647 branch-likely instructions. @option{-mfix-r10000} is the default when
21648 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
21652 @itemx -mno-fix-rm7000
21653 @opindex mfix-rm7000
21654 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
21655 workarounds are implemented by the assembler rather than by GCC@.
21658 @itemx -mno-fix-vr4120
21659 @opindex mfix-vr4120
21660 Work around certain VR4120 errata:
21663 @code{dmultu} does not always produce the correct result.
21665 @code{div} and @code{ddiv} do not always produce the correct result if one
21666 of the operands is negative.
21668 The workarounds for the division errata rely on special functions in
21669 @file{libgcc.a}. At present, these functions are only provided by
21670 the @code{mips64vr*-elf} configurations.
21672 Other VR4120 errata require a NOP to be inserted between certain pairs of
21673 instructions. These errata are handled by the assembler, not by GCC itself.
21676 @opindex mfix-vr4130
21677 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
21678 workarounds are implemented by the assembler rather than by GCC,
21679 although GCC avoids using @code{mflo} and @code{mfhi} if the
21680 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
21681 instructions are available instead.
21684 @itemx -mno-fix-sb1
21686 Work around certain SB-1 CPU core errata.
21687 (This flag currently works around the SB-1 revision 2
21688 ``F1'' and ``F2'' floating-point errata.)
21690 @item -mr10k-cache-barrier=@var{setting}
21691 @opindex mr10k-cache-barrier
21692 Specify whether GCC should insert cache barriers to avoid the
21693 side effects of speculation on R10K processors.
21695 In common with many processors, the R10K tries to predict the outcome
21696 of a conditional branch and speculatively executes instructions from
21697 the ``taken'' branch. It later aborts these instructions if the
21698 predicted outcome is wrong. However, on the R10K, even aborted
21699 instructions can have side effects.
21701 This problem only affects kernel stores and, depending on the system,
21702 kernel loads. As an example, a speculatively-executed store may load
21703 the target memory into cache and mark the cache line as dirty, even if
21704 the store itself is later aborted. If a DMA operation writes to the
21705 same area of memory before the ``dirty'' line is flushed, the cached
21706 data overwrites the DMA-ed data. See the R10K processor manual
21707 for a full description, including other potential problems.
21709 One workaround is to insert cache barrier instructions before every memory
21710 access that might be speculatively executed and that might have side
21711 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
21712 controls GCC's implementation of this workaround. It assumes that
21713 aborted accesses to any byte in the following regions does not have
21718 the memory occupied by the current function's stack frame;
21721 the memory occupied by an incoming stack argument;
21724 the memory occupied by an object with a link-time-constant address.
21727 It is the kernel's responsibility to ensure that speculative
21728 accesses to these regions are indeed safe.
21730 If the input program contains a function declaration such as:
21736 then the implementation of @code{foo} must allow @code{j foo} and
21737 @code{jal foo} to be executed speculatively. GCC honors this
21738 restriction for functions it compiles itself. It expects non-GCC
21739 functions (such as hand-written assembly code) to do the same.
21741 The option has three forms:
21744 @item -mr10k-cache-barrier=load-store
21745 Insert a cache barrier before a load or store that might be
21746 speculatively executed and that might have side effects even
21749 @item -mr10k-cache-barrier=store
21750 Insert a cache barrier before a store that might be speculatively
21751 executed and that might have side effects even if aborted.
21753 @item -mr10k-cache-barrier=none
21754 Disable the insertion of cache barriers. This is the default setting.
21757 @item -mflush-func=@var{func}
21758 @itemx -mno-flush-func
21759 @opindex mflush-func
21760 Specifies the function to call to flush the I and D caches, or to not
21761 call any such function. If called, the function must take the same
21762 arguments as the common @code{_flush_func}, that is, the address of the
21763 memory range for which the cache is being flushed, the size of the
21764 memory range, and the number 3 (to flush both caches). The default
21765 depends on the target GCC was configured for, but commonly is either
21766 @code{_flush_func} or @code{__cpu_flush}.
21768 @item mbranch-cost=@var{num}
21769 @opindex mbranch-cost
21770 Set the cost of branches to roughly @var{num} ``simple'' instructions.
21771 This cost is only a heuristic and is not guaranteed to produce
21772 consistent results across releases. A zero cost redundantly selects
21773 the default, which is based on the @option{-mtune} setting.
21775 @item -mbranch-likely
21776 @itemx -mno-branch-likely
21777 @opindex mbranch-likely
21778 @opindex mno-branch-likely
21779 Enable or disable use of Branch Likely instructions, regardless of the
21780 default for the selected architecture. By default, Branch Likely
21781 instructions may be generated if they are supported by the selected
21782 architecture. An exception is for the MIPS32 and MIPS64 architectures
21783 and processors that implement those architectures; for those, Branch
21784 Likely instructions are not be generated by default because the MIPS32
21785 and MIPS64 architectures specifically deprecate their use.
21787 @item -mcompact-branches=never
21788 @itemx -mcompact-branches=optimal
21789 @itemx -mcompact-branches=always
21790 @opindex mcompact-branches=never
21791 @opindex mcompact-branches=optimal
21792 @opindex mcompact-branches=always
21793 These options control which form of branches will be generated. The
21794 default is @option{-mcompact-branches=optimal}.
21796 The @option{-mcompact-branches=never} option ensures that compact branch
21797 instructions will never be generated.
21799 The @option{-mcompact-branches=always} option ensures that a compact
21800 branch instruction will be generated if available. If a compact branch
21801 instruction is not available, a delay slot form of the branch will be
21804 This option is supported from MIPS Release 6 onwards.
21806 The @option{-mcompact-branches=optimal} option will cause a delay slot
21807 branch to be used if one is available in the current ISA and the delay
21808 slot is successfully filled. If the delay slot is not filled, a compact
21809 branch will be chosen if one is available.
21811 @item -mfp-exceptions
21812 @itemx -mno-fp-exceptions
21813 @opindex mfp-exceptions
21814 Specifies whether FP exceptions are enabled. This affects how
21815 FP instructions are scheduled for some processors.
21816 The default is that FP exceptions are
21819 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
21820 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
21823 @item -mvr4130-align
21824 @itemx -mno-vr4130-align
21825 @opindex mvr4130-align
21826 The VR4130 pipeline is two-way superscalar, but can only issue two
21827 instructions together if the first one is 8-byte aligned. When this
21828 option is enabled, GCC aligns pairs of instructions that it
21829 thinks should execute in parallel.
21831 This option only has an effect when optimizing for the VR4130.
21832 It normally makes code faster, but at the expense of making it bigger.
21833 It is enabled by default at optimization level @option{-O3}.
21838 Enable (disable) generation of @code{synci} instructions on
21839 architectures that support it. The @code{synci} instructions (if
21840 enabled) are generated when @code{__builtin___clear_cache} is
21843 This option defaults to @option{-mno-synci}, but the default can be
21844 overridden by configuring GCC with @option{--with-synci}.
21846 When compiling code for single processor systems, it is generally safe
21847 to use @code{synci}. However, on many multi-core (SMP) systems, it
21848 does not invalidate the instruction caches on all cores and may lead
21849 to undefined behavior.
21851 @item -mrelax-pic-calls
21852 @itemx -mno-relax-pic-calls
21853 @opindex mrelax-pic-calls
21854 Try to turn PIC calls that are normally dispatched via register
21855 @code{$25} into direct calls. This is only possible if the linker can
21856 resolve the destination at link time and if the destination is within
21857 range for a direct call.
21859 @option{-mrelax-pic-calls} is the default if GCC was configured to use
21860 an assembler and a linker that support the @code{.reloc} assembly
21861 directive and @option{-mexplicit-relocs} is in effect. With
21862 @option{-mno-explicit-relocs}, this optimization can be performed by the
21863 assembler and the linker alone without help from the compiler.
21865 @item -mmcount-ra-address
21866 @itemx -mno-mcount-ra-address
21867 @opindex mmcount-ra-address
21868 @opindex mno-mcount-ra-address
21869 Emit (do not emit) code that allows @code{_mcount} to modify the
21870 calling function's return address. When enabled, this option extends
21871 the usual @code{_mcount} interface with a new @var{ra-address}
21872 parameter, which has type @code{intptr_t *} and is passed in register
21873 @code{$12}. @code{_mcount} can then modify the return address by
21874 doing both of the following:
21877 Returning the new address in register @code{$31}.
21879 Storing the new address in @code{*@var{ra-address}},
21880 if @var{ra-address} is nonnull.
21883 The default is @option{-mno-mcount-ra-address}.
21885 @item -mframe-header-opt
21886 @itemx -mno-frame-header-opt
21887 @opindex mframe-header-opt
21888 Enable (disable) frame header optimization in the o32 ABI. When using the
21889 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
21890 function to write out register arguments. When enabled, this optimization
21891 will suppress the allocation of the frame header if it can be determined that
21894 This optimization is off by default at all optimization levels.
21897 @itemx -mno-lxc1-sxc1
21898 @opindex mlxc1-sxc1
21899 When applicable, enable (disable) the generation of @code{lwxc1},
21900 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
21905 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
21906 @code{madd.d} and related instructions. Enabled by default.
21911 @subsection MMIX Options
21912 @cindex MMIX Options
21914 These options are defined for the MMIX:
21918 @itemx -mno-libfuncs
21920 @opindex mno-libfuncs
21921 Specify that intrinsic library functions are being compiled, passing all
21922 values in registers, no matter the size.
21925 @itemx -mno-epsilon
21927 @opindex mno-epsilon
21928 Generate floating-point comparison instructions that compare with respect
21929 to the @code{rE} epsilon register.
21931 @item -mabi=mmixware
21933 @opindex mabi=mmixware
21935 Generate code that passes function parameters and return values that (in
21936 the called function) are seen as registers @code{$0} and up, as opposed to
21937 the GNU ABI which uses global registers @code{$231} and up.
21939 @item -mzero-extend
21940 @itemx -mno-zero-extend
21941 @opindex mzero-extend
21942 @opindex mno-zero-extend
21943 When reading data from memory in sizes shorter than 64 bits, use (do not
21944 use) zero-extending load instructions by default, rather than
21945 sign-extending ones.
21948 @itemx -mno-knuthdiv
21950 @opindex mno-knuthdiv
21951 Make the result of a division yielding a remainder have the same sign as
21952 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
21953 remainder follows the sign of the dividend. Both methods are
21954 arithmetically valid, the latter being almost exclusively used.
21956 @item -mtoplevel-symbols
21957 @itemx -mno-toplevel-symbols
21958 @opindex mtoplevel-symbols
21959 @opindex mno-toplevel-symbols
21960 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
21961 code can be used with the @code{PREFIX} assembly directive.
21965 Generate an executable in the ELF format, rather than the default
21966 @samp{mmo} format used by the @command{mmix} simulator.
21968 @item -mbranch-predict
21969 @itemx -mno-branch-predict
21970 @opindex mbranch-predict
21971 @opindex mno-branch-predict
21972 Use (do not use) the probable-branch instructions, when static branch
21973 prediction indicates a probable branch.
21975 @item -mbase-addresses
21976 @itemx -mno-base-addresses
21977 @opindex mbase-addresses
21978 @opindex mno-base-addresses
21979 Generate (do not generate) code that uses @emph{base addresses}. Using a
21980 base address automatically generates a request (handled by the assembler
21981 and the linker) for a constant to be set up in a global register. The
21982 register is used for one or more base address requests within the range 0
21983 to 255 from the value held in the register. The generally leads to short
21984 and fast code, but the number of different data items that can be
21985 addressed is limited. This means that a program that uses lots of static
21986 data may require @option{-mno-base-addresses}.
21988 @item -msingle-exit
21989 @itemx -mno-single-exit
21990 @opindex msingle-exit
21991 @opindex mno-single-exit
21992 Force (do not force) generated code to have a single exit point in each
21996 @node MN10300 Options
21997 @subsection MN10300 Options
21998 @cindex MN10300 options
22000 These @option{-m} options are defined for Matsushita MN10300 architectures:
22005 Generate code to avoid bugs in the multiply instructions for the MN10300
22006 processors. This is the default.
22008 @item -mno-mult-bug
22009 @opindex mno-mult-bug
22010 Do not generate code to avoid bugs in the multiply instructions for the
22011 MN10300 processors.
22015 Generate code using features specific to the AM33 processor.
22019 Do not generate code using features specific to the AM33 processor. This
22024 Generate code using features specific to the AM33/2.0 processor.
22028 Generate code using features specific to the AM34 processor.
22030 @item -mtune=@var{cpu-type}
22032 Use the timing characteristics of the indicated CPU type when
22033 scheduling instructions. This does not change the targeted processor
22034 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
22035 @samp{am33-2} or @samp{am34}.
22037 @item -mreturn-pointer-on-d0
22038 @opindex mreturn-pointer-on-d0
22039 When generating a function that returns a pointer, return the pointer
22040 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
22041 only in @code{a0}, and attempts to call such functions without a prototype
22042 result in errors. Note that this option is on by default; use
22043 @option{-mno-return-pointer-on-d0} to disable it.
22047 Do not link in the C run-time initialization object file.
22051 Indicate to the linker that it should perform a relaxation optimization pass
22052 to shorten branches, calls and absolute memory addresses. This option only
22053 has an effect when used on the command line for the final link step.
22055 This option makes symbolic debugging impossible.
22059 Allow the compiler to generate @emph{Long Instruction Word}
22060 instructions if the target is the @samp{AM33} or later. This is the
22061 default. This option defines the preprocessor macro @code{__LIW__}.
22065 Do not allow the compiler to generate @emph{Long Instruction Word}
22066 instructions. This option defines the preprocessor macro
22071 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
22072 instructions if the target is the @samp{AM33} or later. This is the
22073 default. This option defines the preprocessor macro @code{__SETLB__}.
22077 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
22078 instructions. This option defines the preprocessor macro
22079 @code{__NO_SETLB__}.
22083 @node Moxie Options
22084 @subsection Moxie Options
22085 @cindex Moxie Options
22091 Generate big-endian code. This is the default for @samp{moxie-*-*}
22096 Generate little-endian code.
22100 Generate mul.x and umul.x instructions. This is the default for
22101 @samp{moxiebox-*-*} configurations.
22105 Do not link in the C run-time initialization object file.
22109 @node MSP430 Options
22110 @subsection MSP430 Options
22111 @cindex MSP430 Options
22113 These options are defined for the MSP430:
22119 Force assembly output to always use hex constants. Normally such
22120 constants are signed decimals, but this option is available for
22121 testsuite and/or aesthetic purposes.
22125 Select the MCU to target. This is used to create a C preprocessor
22126 symbol based upon the MCU name, converted to upper case and pre- and
22127 post-fixed with @samp{__}. This in turn is used by the
22128 @file{msp430.h} header file to select an MCU-specific supplementary
22131 The option also sets the ISA to use. If the MCU name is one that is
22132 known to only support the 430 ISA then that is selected, otherwise the
22133 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
22134 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
22135 name selects the 430X ISA.
22137 In addition an MCU-specific linker script is added to the linker
22138 command line. The script's name is the name of the MCU with
22139 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
22140 command line defines the C preprocessor symbol @code{__XXX__} and
22141 cause the linker to search for a script called @file{xxx.ld}.
22143 This option is also passed on to the assembler.
22146 @itemx -mno-warn-mcu
22148 @opindex mno-warn-mcu
22149 This option enables or disables warnings about conflicts between the
22150 MCU name specified by the @option{-mmcu} option and the ISA set by the
22151 @option{-mcpu} option and/or the hardware multiply support set by the
22152 @option{-mhwmult} option. It also toggles warnings about unrecognized
22153 MCU names. This option is on by default.
22157 Specifies the ISA to use. Accepted values are @samp{msp430},
22158 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
22159 @option{-mmcu=} option should be used to select the ISA.
22163 Link to the simulator runtime libraries and linker script. Overrides
22164 any scripts that would be selected by the @option{-mmcu=} option.
22168 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
22172 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
22176 This option is passed to the assembler and linker, and allows the
22177 linker to perform certain optimizations that cannot be done until
22182 Describes the type of hardware multiply supported by the target.
22183 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
22184 for the original 16-bit-only multiply supported by early MCUs.
22185 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
22186 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
22187 A value of @samp{auto} can also be given. This tells GCC to deduce
22188 the hardware multiply support based upon the MCU name provided by the
22189 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
22190 the MCU name is not recognized then no hardware multiply support is
22191 assumed. @code{auto} is the default setting.
22193 Hardware multiplies are normally performed by calling a library
22194 routine. This saves space in the generated code. When compiling at
22195 @option{-O3} or higher however the hardware multiplier is invoked
22196 inline. This makes for bigger, but faster code.
22198 The hardware multiply routines disable interrupts whilst running and
22199 restore the previous interrupt state when they finish. This makes
22200 them safe to use inside interrupt handlers as well as in normal code.
22204 Enable the use of a minimum runtime environment - no static
22205 initializers or constructors. This is intended for memory-constrained
22206 devices. The compiler includes special symbols in some objects
22207 that tell the linker and runtime which code fragments are required.
22209 @item -mcode-region=
22210 @itemx -mdata-region=
22211 @opindex mcode-region
22212 @opindex mdata-region
22213 These options tell the compiler where to place functions and data that
22214 do not have one of the @code{lower}, @code{upper}, @code{either} or
22215 @code{section} attributes. Possible values are @code{lower},
22216 @code{upper}, @code{either} or @code{any}. The first three behave
22217 like the corresponding attribute. The fourth possible value -
22218 @code{any} - is the default. It leaves placement entirely up to the
22219 linker script and how it assigns the standard sections
22220 (@code{.text}, @code{.data}, etc) to the memory regions.
22222 @item -msilicon-errata=
22223 @opindex msilicon-errata
22224 This option passes on a request to assembler to enable the fixes for
22225 the named silicon errata.
22227 @item -msilicon-errata-warn=
22228 @opindex msilicon-errata-warn
22229 This option passes on a request to the assembler to enable warning
22230 messages when a silicon errata might need to be applied.
22234 @node NDS32 Options
22235 @subsection NDS32 Options
22236 @cindex NDS32 Options
22238 These options are defined for NDS32 implementations:
22243 @opindex mbig-endian
22244 Generate code in big-endian mode.
22246 @item -mlittle-endian
22247 @opindex mlittle-endian
22248 Generate code in little-endian mode.
22250 @item -mreduced-regs
22251 @opindex mreduced-regs
22252 Use reduced-set registers for register allocation.
22255 @opindex mfull-regs
22256 Use full-set registers for register allocation.
22260 Generate conditional move instructions.
22264 Do not generate conditional move instructions.
22268 Generate performance extension instructions.
22270 @item -mno-ext-perf
22271 @opindex mno-perf-ext
22272 Do not generate performance extension instructions.
22276 Generate performance extension 2 instructions.
22278 @item -mno-ext-perf2
22279 @opindex mno-perf-ext
22280 Do not generate performance extension 2 instructions.
22284 Generate string extension instructions.
22286 @item -mno-ext-string
22287 @opindex mno-perf-ext
22288 Do not generate string extension instructions.
22292 Generate v3 push25/pop25 instructions.
22295 @opindex mno-v3push
22296 Do not generate v3 push25/pop25 instructions.
22300 Generate 16-bit instructions.
22303 @opindex mno-16-bit
22304 Do not generate 16-bit instructions.
22306 @item -misr-vector-size=@var{num}
22307 @opindex misr-vector-size
22308 Specify the size of each interrupt vector, which must be 4 or 16.
22310 @item -mcache-block-size=@var{num}
22311 @opindex mcache-block-size
22312 Specify the size of each cache block,
22313 which must be a power of 2 between 4 and 512.
22315 @item -march=@var{arch}
22317 Specify the name of the target architecture.
22319 @item -mcmodel=@var{code-model}
22321 Set the code model to one of
22324 All the data and read-only data segments must be within 512KB addressing space.
22325 The text segment must be within 16MB addressing space.
22326 @item @samp{medium}
22327 The data segment must be within 512KB while the read-only data segment can be
22328 within 4GB addressing space. The text segment should be still within 16MB
22331 All the text and data segments can be within 4GB addressing space.
22335 @opindex mctor-dtor
22336 Enable constructor/destructor feature.
22340 Guide linker to relax instructions.
22344 @node Nios II Options
22345 @subsection Nios II Options
22346 @cindex Nios II options
22347 @cindex Altera Nios II options
22349 These are the options defined for the Altera Nios II processor.
22355 @cindex smaller data references
22356 Put global and static objects less than or equal to @var{num} bytes
22357 into the small data or BSS sections instead of the normal data or BSS
22358 sections. The default value of @var{num} is 8.
22360 @item -mgpopt=@var{option}
22365 Generate (do not generate) GP-relative accesses. The following
22366 @var{option} names are recognized:
22371 Do not generate GP-relative accesses.
22374 Generate GP-relative accesses for small data objects that are not
22375 external, weak, or uninitialized common symbols.
22376 Also use GP-relative addressing for objects that
22377 have been explicitly placed in a small data section via a @code{section}
22381 As for @samp{local}, but also generate GP-relative accesses for
22382 small data objects that are external, weak, or common. If you use this option,
22383 you must ensure that all parts of your program (including libraries) are
22384 compiled with the same @option{-G} setting.
22387 Generate GP-relative accesses for all data objects in the program. If you
22388 use this option, the entire data and BSS segments
22389 of your program must fit in 64K of memory and you must use an appropriate
22390 linker script to allocate them within the addressable range of the
22394 Generate GP-relative addresses for function pointers as well as data
22395 pointers. If you use this option, the entire text, data, and BSS segments
22396 of your program must fit in 64K of memory and you must use an appropriate
22397 linker script to allocate them within the addressable range of the
22402 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
22403 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
22405 The default is @option{-mgpopt} except when @option{-fpic} or
22406 @option{-fPIC} is specified to generate position-independent code.
22407 Note that the Nios II ABI does not permit GP-relative accesses from
22410 You may need to specify @option{-mno-gpopt} explicitly when building
22411 programs that include large amounts of small data, including large
22412 GOT data sections. In this case, the 16-bit offset for GP-relative
22413 addressing may not be large enough to allow access to the entire
22414 small data section.
22416 @item -mgprel-sec=@var{regexp}
22417 @opindex mgprel-sec
22418 This option specifies additional section names that can be accessed via
22419 GP-relative addressing. It is most useful in conjunction with
22420 @code{section} attributes on variable declarations
22421 (@pxref{Common Variable Attributes}) and a custom linker script.
22422 The @var{regexp} is a POSIX Extended Regular Expression.
22424 This option does not affect the behavior of the @option{-G} option, and
22425 the specified sections are in addition to the standard @code{.sdata}
22426 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
22428 @item -mr0rel-sec=@var{regexp}
22429 @opindex mr0rel-sec
22430 This option specifies names of sections that can be accessed via a
22431 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
22432 of the 32-bit address space. It is most useful in conjunction with
22433 @code{section} attributes on variable declarations
22434 (@pxref{Common Variable Attributes}) and a custom linker script.
22435 The @var{regexp} is a POSIX Extended Regular Expression.
22437 In contrast to the use of GP-relative addressing for small data,
22438 zero-based addressing is never generated by default and there are no
22439 conventional section names used in standard linker scripts for sections
22440 in the low or high areas of memory.
22446 Generate little-endian (default) or big-endian (experimental) code,
22449 @item -march=@var{arch}
22451 This specifies the name of the target Nios II architecture. GCC uses this
22452 name to determine what kind of instructions it can emit when generating
22453 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
22455 The preprocessor macro @code{__nios2_arch__} is available to programs,
22456 with value 1 or 2, indicating the targeted ISA level.
22458 @item -mbypass-cache
22459 @itemx -mno-bypass-cache
22460 @opindex mno-bypass-cache
22461 @opindex mbypass-cache
22462 Force all load and store instructions to always bypass cache by
22463 using I/O variants of the instructions. The default is not to
22466 @item -mno-cache-volatile
22467 @itemx -mcache-volatile
22468 @opindex mcache-volatile
22469 @opindex mno-cache-volatile
22470 Volatile memory access bypass the cache using the I/O variants of
22471 the load and store instructions. The default is not to bypass the cache.
22473 @item -mno-fast-sw-div
22474 @itemx -mfast-sw-div
22475 @opindex mno-fast-sw-div
22476 @opindex mfast-sw-div
22477 Do not use table-based fast divide for small numbers. The default
22478 is to use the fast divide at @option{-O3} and above.
22482 @itemx -mno-hw-mulx
22486 @opindex mno-hw-mul
22488 @opindex mno-hw-mulx
22490 @opindex mno-hw-div
22492 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
22493 instructions by the compiler. The default is to emit @code{mul}
22494 and not emit @code{div} and @code{mulx}.
22500 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
22501 CDX (code density) instructions. Enabling these instructions also
22502 requires @option{-march=r2}. Since these instructions are optional
22503 extensions to the R2 architecture, the default is not to emit them.
22505 @item -mcustom-@var{insn}=@var{N}
22506 @itemx -mno-custom-@var{insn}
22507 @opindex mcustom-@var{insn}
22508 @opindex mno-custom-@var{insn}
22509 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
22510 custom instruction with encoding @var{N} when generating code that uses
22511 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
22512 instruction 253 for single-precision floating-point add operations instead
22513 of the default behavior of using a library call.
22515 The following values of @var{insn} are supported. Except as otherwise
22516 noted, floating-point operations are expected to be implemented with
22517 normal IEEE 754 semantics and correspond directly to the C operators or the
22518 equivalent GCC built-in functions (@pxref{Other Builtins}).
22520 Single-precision floating point:
22523 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
22524 Binary arithmetic operations.
22530 Unary absolute value.
22532 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
22533 Comparison operations.
22535 @item @samp{fmins}, @samp{fmaxs}
22536 Floating-point minimum and maximum. These instructions are only
22537 generated if @option{-ffinite-math-only} is specified.
22539 @item @samp{fsqrts}
22540 Unary square root operation.
22542 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
22543 Floating-point trigonometric and exponential functions. These instructions
22544 are only generated if @option{-funsafe-math-optimizations} is also specified.
22548 Double-precision floating point:
22551 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
22552 Binary arithmetic operations.
22558 Unary absolute value.
22560 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
22561 Comparison operations.
22563 @item @samp{fmind}, @samp{fmaxd}
22564 Double-precision minimum and maximum. These instructions are only
22565 generated if @option{-ffinite-math-only} is specified.
22567 @item @samp{fsqrtd}
22568 Unary square root operation.
22570 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
22571 Double-precision trigonometric and exponential functions. These instructions
22572 are only generated if @option{-funsafe-math-optimizations} is also specified.
22578 @item @samp{fextsd}
22579 Conversion from single precision to double precision.
22581 @item @samp{ftruncds}
22582 Conversion from double precision to single precision.
22584 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
22585 Conversion from floating point to signed or unsigned integer types, with
22586 truncation towards zero.
22589 Conversion from single-precision floating point to signed integer,
22590 rounding to the nearest integer and ties away from zero.
22591 This corresponds to the @code{__builtin_lroundf} function when
22592 @option{-fno-math-errno} is used.
22594 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
22595 Conversion from signed or unsigned integer types to floating-point types.
22599 In addition, all of the following transfer instructions for internal
22600 registers X and Y must be provided to use any of the double-precision
22601 floating-point instructions. Custom instructions taking two
22602 double-precision source operands expect the first operand in the
22603 64-bit register X. The other operand (or only operand of a unary
22604 operation) is given to the custom arithmetic instruction with the
22605 least significant half in source register @var{src1} and the most
22606 significant half in @var{src2}. A custom instruction that returns a
22607 double-precision result returns the most significant 32 bits in the
22608 destination register and the other half in 32-bit register Y.
22609 GCC automatically generates the necessary code sequences to write
22610 register X and/or read register Y when double-precision floating-point
22611 instructions are used.
22616 Write @var{src1} into the least significant half of X and @var{src2} into
22617 the most significant half of X.
22620 Write @var{src1} into Y.
22622 @item @samp{frdxhi}, @samp{frdxlo}
22623 Read the most or least (respectively) significant half of X and store it in
22627 Read the value of Y and store it into @var{dest}.
22630 Note that you can gain more local control over generation of Nios II custom
22631 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
22632 and @code{target("no-custom-@var{insn}")} function attributes
22633 (@pxref{Function Attributes})
22634 or pragmas (@pxref{Function Specific Option Pragmas}).
22636 @item -mcustom-fpu-cfg=@var{name}
22637 @opindex mcustom-fpu-cfg
22639 This option enables a predefined, named set of custom instruction encodings
22640 (see @option{-mcustom-@var{insn}} above).
22641 Currently, the following sets are defined:
22643 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
22644 @gccoptlist{-mcustom-fmuls=252 @gol
22645 -mcustom-fadds=253 @gol
22646 -mcustom-fsubs=254 @gol
22647 -fsingle-precision-constant}
22649 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
22650 @gccoptlist{-mcustom-fmuls=252 @gol
22651 -mcustom-fadds=253 @gol
22652 -mcustom-fsubs=254 @gol
22653 -mcustom-fdivs=255 @gol
22654 -fsingle-precision-constant}
22656 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
22657 @gccoptlist{-mcustom-floatus=243 @gol
22658 -mcustom-fixsi=244 @gol
22659 -mcustom-floatis=245 @gol
22660 -mcustom-fcmpgts=246 @gol
22661 -mcustom-fcmples=249 @gol
22662 -mcustom-fcmpeqs=250 @gol
22663 -mcustom-fcmpnes=251 @gol
22664 -mcustom-fmuls=252 @gol
22665 -mcustom-fadds=253 @gol
22666 -mcustom-fsubs=254 @gol
22667 -mcustom-fdivs=255 @gol
22668 -fsingle-precision-constant}
22670 Custom instruction assignments given by individual
22671 @option{-mcustom-@var{insn}=} options override those given by
22672 @option{-mcustom-fpu-cfg=}, regardless of the
22673 order of the options on the command line.
22675 Note that you can gain more local control over selection of a FPU
22676 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
22677 function attribute (@pxref{Function Attributes})
22678 or pragma (@pxref{Function Specific Option Pragmas}).
22682 These additional @samp{-m} options are available for the Altera Nios II
22683 ELF (bare-metal) target:
22689 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
22690 startup and termination code, and is typically used in conjunction with
22691 @option{-msys-crt0=} to specify the location of the alternate startup code
22692 provided by the HAL BSP.
22696 Link with a limited version of the C library, @option{-lsmallc}, rather than
22699 @item -msys-crt0=@var{startfile}
22701 @var{startfile} is the file name of the startfile (crt0) to use
22702 when linking. This option is only useful in conjunction with @option{-mhal}.
22704 @item -msys-lib=@var{systemlib}
22706 @var{systemlib} is the library name of the library that provides
22707 low-level system calls required by the C library,
22708 e.g.@: @code{read} and @code{write}.
22709 This option is typically used to link with a library provided by a HAL BSP.
22713 @node Nvidia PTX Options
22714 @subsection Nvidia PTX Options
22715 @cindex Nvidia PTX options
22716 @cindex nvptx options
22718 These options are defined for Nvidia PTX:
22726 Generate code for 32-bit or 64-bit ABI.
22728 @item -misa=@var{ISA-string}
22730 Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA
22731 strings must be lower-case. Valid ISA strings include @samp{sm_30} and
22732 @samp{sm_35}. The default ISA is sm_30.
22735 @opindex mmainkernel
22736 Link in code for a __main kernel. This is for stand-alone instead of
22737 offloading execution.
22741 Apply partitioned execution optimizations. This is the default when any
22742 level of optimization is selected.
22745 @opindex msoft-stack
22746 Generate code that does not use @code{.local} memory
22747 directly for stack storage. Instead, a per-warp stack pointer is
22748 maintained explicitly. This enables variable-length stack allocation (with
22749 variable-length arrays or @code{alloca}), and when global memory is used for
22750 underlying storage, makes it possible to access automatic variables from other
22751 threads, or with atomic instructions. This code generation variant is used
22752 for OpenMP offloading, but the option is exposed on its own for the purpose
22753 of testing the compiler; to generate code suitable for linking into programs
22754 using OpenMP offloading, use option @option{-mgomp}.
22756 @item -muniform-simt
22757 @opindex muniform-simt
22758 Switch to code generation variant that allows to execute all threads in each
22759 warp, while maintaining memory state and side effects as if only one thread
22760 in each warp was active outside of OpenMP SIMD regions. All atomic operations
22761 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
22762 current lane index equals the master lane index), and the register being
22763 assigned is copied via a shuffle instruction from the master lane. Outside of
22764 SIMD regions lane 0 is the master; inside, each thread sees itself as the
22765 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
22766 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
22767 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
22768 with current lane index to compute the master lane index.
22772 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
22773 @option{-muniform-simt} options, and selects corresponding multilib variant.
22777 @node OpenRISC Options
22778 @subsection OpenRISC Options
22779 @cindex OpenRISC Options
22781 These options are defined for OpenRISC:
22785 @item -mboard=@var{name}
22787 Configure a board specific runtime. This will be passed to the linker for
22788 newlib board library linking. The default is @code{or1ksim}.
22792 For compatibility, it's always newlib for elf now.
22796 Generate code for hardware which supports divide instructions. This is the
22801 Generate code for hardware which supports multiply instructions. This is the
22806 Generate code for hardware which supports the conditional move (@code{l.cmov})
22811 Generate code for hardware which supports rotate right instructions.
22815 Generate code for hardware which supports sign-extension instructions.
22819 Generate code for hardware which supports set flag immediate (@code{l.sf*i})
22824 Generate code for hardware which supports shift immediate related instructions
22825 (i.e. @code{l.srai}, @code{l.srli}, @code{l.slli}, @code{1.rori}). Note, to
22826 enable generation of the @code{l.rori} instruction the @option{-mror} flag must
22831 Generate code for hardware which requires divide instruction emulation.
22835 Generate code for hardware which requires multiply instruction emulation.
22839 @node PDP-11 Options
22840 @subsection PDP-11 Options
22841 @cindex PDP-11 Options
22843 These options are defined for the PDP-11:
22848 Use hardware FPP floating point. This is the default. (FIS floating
22849 point on the PDP-11/40 is not supported.) Implies -m45.
22852 @opindex msoft-float
22853 Do not use hardware floating point.
22857 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
22861 Return floating-point results in memory. This is the default.
22865 Generate code for a PDP-11/40. Implies -msoft-float -mno-split.
22869 Generate code for a PDP-11/45. This is the default.
22873 Generate code for a PDP-11/10. Implies -msoft-float -mno-split.
22879 Use 16-bit @code{int}. This is the default.
22885 Use 32-bit @code{int}.
22889 Target has split instruction and data space. Implies -m45.
22893 Use Unix assembler syntax.
22897 Use DEC assembler syntax.
22901 Use GNU assembler syntax. This is the default.
22905 Use the new LRA register allocator. By default, the old ``reload''
22909 @node picoChip Options
22910 @subsection picoChip Options
22911 @cindex picoChip options
22913 These @samp{-m} options are defined for picoChip implementations:
22917 @item -mae=@var{ae_type}
22919 Set the instruction set, register set, and instruction scheduling
22920 parameters for array element type @var{ae_type}. Supported values
22921 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
22923 @option{-mae=ANY} selects a completely generic AE type. Code
22924 generated with this option runs on any of the other AE types. The
22925 code is not as efficient as it would be if compiled for a specific
22926 AE type, and some types of operation (e.g., multiplication) do not
22927 work properly on all types of AE.
22929 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
22930 for compiled code, and is the default.
22932 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
22933 option may suffer from poor performance of byte (char) manipulation,
22934 since the DSP AE does not provide hardware support for byte load/stores.
22936 @item -msymbol-as-address
22937 Enable the compiler to directly use a symbol name as an address in a
22938 load/store instruction, without first loading it into a
22939 register. Typically, the use of this option generates larger
22940 programs, which run faster than when the option isn't used. However, the
22941 results vary from program to program, so it is left as a user option,
22942 rather than being permanently enabled.
22944 @item -mno-inefficient-warnings
22945 Disables warnings about the generation of inefficient code. These
22946 warnings can be generated, for example, when compiling code that
22947 performs byte-level memory operations on the MAC AE type. The MAC AE has
22948 no hardware support for byte-level memory operations, so all byte
22949 load/stores must be synthesized from word load/store operations. This is
22950 inefficient and a warning is generated to indicate
22951 that you should rewrite the code to avoid byte operations, or to target
22952 an AE type that has the necessary hardware support. This option disables
22957 @node PowerPC Options
22958 @subsection PowerPC Options
22959 @cindex PowerPC options
22961 These are listed under @xref{RS/6000 and PowerPC Options}.
22963 @node PowerPC SPE Options
22964 @subsection PowerPC SPE Options
22965 @cindex PowerPC SPE options
22967 These @samp{-m} options are defined for PowerPC SPE:
22972 @itemx -mno-popcntb
22976 @opindex mno-popcntb
22977 You use these options to specify which instructions are available on the
22978 processor you are using. The default value of these options is
22979 determined when configuring GCC@. Specifying the
22980 @option{-mcpu=@var{cpu_type}} overrides the specification of these
22981 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
22982 rather than the options listed above.
22984 The @option{-mmfcrf} option allows GCC to generate the move from
22985 condition register field instruction implemented on the POWER4
22986 processor and other processors that support the PowerPC V2.01
22988 The @option{-mpopcntb} option allows GCC to generate the popcount and
22989 double-precision FP reciprocal estimate instruction implemented on the
22990 POWER5 processor and other processors that support the PowerPC V2.02
22993 @item -mcpu=@var{cpu_type}
22995 Set architecture type, register usage, and
22996 instruction scheduling parameters for machine type @var{cpu_type}.
22997 Supported values for @var{cpu_type} are @samp{8540}, @samp{8548},
23000 @option{-mcpu=powerpc} specifies pure 32-bit PowerPC (either
23001 endian), with an appropriate, generic processor model assumed for
23002 scheduling purposes.
23004 Specifying @samp{native} as cpu type detects and selects the
23005 architecture option that corresponds to the host processor of the
23006 system performing the compilation.
23007 @option{-mcpu=native} has no effect if GCC does not recognize the
23010 The other options specify a specific processor. Code generated under
23011 those options runs best on that processor, and may not run at all on
23014 The @option{-mcpu} options automatically enable or disable the
23017 @gccoptlist{-mhard-float -mmfcrf -mmultiple @gol
23018 -mpopcntb -mpopcntd @gol
23019 -msingle-float -mdouble-float @gol
23022 The particular options set for any particular CPU varies between
23023 compiler versions, depending on what setting seems to produce optimal
23024 code for that CPU; it doesn't necessarily reflect the actual hardware's
23025 capabilities. If you wish to set an individual option to a particular
23026 value, you may specify it after the @option{-mcpu} option, like
23027 @option{-mcpu=8548}.
23029 @item -mtune=@var{cpu_type}
23031 Set the instruction scheduling parameters for machine type
23032 @var{cpu_type}, but do not set the architecture type or register usage,
23033 as @option{-mcpu=@var{cpu_type}} does. The same
23034 values for @var{cpu_type} are used for @option{-mtune} as for
23035 @option{-mcpu}. If both are specified, the code generated uses the
23036 architecture and registers set by @option{-mcpu}, but the
23037 scheduling parameters set by @option{-mtune}.
23040 @opindex msecure-plt
23041 Generate code that allows @command{ld} and @command{ld.so}
23042 to build executables and shared
23043 libraries with non-executable @code{.plt} and @code{.got} sections.
23045 32-bit SYSV ABI option.
23049 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
23051 requires @code{.plt} and @code{.got}
23052 sections that are both writable and executable.
23053 This is a PowerPC 32-bit SYSV ABI option.
23059 This switch enables or disables the generation of ISEL instructions.
23061 @item -misel=@var{yes/no}
23062 This switch has been deprecated. Use @option{-misel} and
23063 @option{-mno-isel} instead.
23069 This switch enables or disables the generation of SPE simd
23072 @item -mspe=@var{yes/no}
23073 This option has been deprecated. Use @option{-mspe} and
23074 @option{-mno-spe} instead.
23077 @itemx -mno-float128
23079 @opindex mno-float128
23080 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
23081 and use either software emulation for IEEE 128-bit floating point or
23082 hardware instructions.
23084 @item -mfloat-gprs=@var{yes/single/double/no}
23085 @itemx -mfloat-gprs
23086 @opindex mfloat-gprs
23087 This switch enables or disables the generation of floating-point
23088 operations on the general-purpose registers for architectures that
23091 The argument @samp{yes} or @samp{single} enables the use of
23092 single-precision floating-point operations.
23094 The argument @samp{double} enables the use of single and
23095 double-precision floating-point operations.
23097 The argument @samp{no} disables floating-point operations on the
23098 general-purpose registers.
23100 This option is currently only available on the MPC854x.
23103 @itemx -mno-fp-in-toc
23104 @itemx -mno-sum-in-toc
23105 @itemx -mminimal-toc
23107 @opindex mno-fp-in-toc
23108 @opindex mno-sum-in-toc
23109 @opindex mminimal-toc
23110 Modify generation of the TOC (Table Of Contents), which is created for
23111 every executable file. The @option{-mfull-toc} option is selected by
23112 default. In that case, GCC allocates at least one TOC entry for
23113 each unique non-automatic variable reference in your program. GCC
23114 also places floating-point constants in the TOC@. However, only
23115 16,384 entries are available in the TOC@.
23117 If you receive a linker error message that saying you have overflowed
23118 the available TOC space, you can reduce the amount of TOC space used
23119 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
23120 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
23121 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
23122 generate code to calculate the sum of an address and a constant at
23123 run time instead of putting that sum into the TOC@. You may specify one
23124 or both of these options. Each causes GCC to produce very slightly
23125 slower and larger code at the expense of conserving TOC space.
23127 If you still run out of space in the TOC even when you specify both of
23128 these options, specify @option{-mminimal-toc} instead. This option causes
23129 GCC to make only one TOC entry for every file. When you specify this
23130 option, GCC produces code that is slower and larger but which
23131 uses extremely little TOC space. You may wish to use this option
23132 only on files that contain less frequently-executed code.
23136 Disables the 64-bit ABI. GCC defaults to @option{-maix32}.
23139 @itemx -mno-xl-compat
23140 @opindex mxl-compat
23141 @opindex mno-xl-compat
23142 Produce code that conforms more closely to IBM XL compiler semantics
23143 when using AIX-compatible ABI@. Pass floating-point arguments to
23144 prototyped functions beyond the register save area (RSA) on the stack
23145 in addition to argument FPRs. Do not assume that most significant
23146 double in 128-bit long double value is properly rounded when comparing
23147 values and converting to double. Use XL symbol names for long double
23150 The AIX calling convention was extended but not initially documented to
23151 handle an obscure K&R C case of calling a function that takes the
23152 address of its arguments with fewer arguments than declared. IBM XL
23153 compilers access floating-point arguments that do not fit in the
23154 RSA from the stack when a subroutine is compiled without
23155 optimization. Because always storing floating-point arguments on the
23156 stack is inefficient and rarely needed, this option is not enabled by
23157 default and only is necessary when calling subroutines compiled by IBM
23158 XL compilers without optimization.
23160 @item -malign-natural
23161 @itemx -malign-power
23162 @opindex malign-natural
23163 @opindex malign-power
23164 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
23165 @option{-malign-natural} overrides the ABI-defined alignment of larger
23166 types, such as floating-point doubles, on their natural size-based boundary.
23167 The option @option{-malign-power} instructs GCC to follow the ABI-specified
23168 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
23170 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
23174 @itemx -mhard-float
23175 @opindex msoft-float
23176 @opindex mhard-float
23177 Generate code that does not use (uses) the floating-point register set.
23178 Software floating-point emulation is provided if you use the
23179 @option{-msoft-float} option, and pass the option to GCC when linking.
23181 @item -msingle-float
23182 @itemx -mdouble-float
23183 @opindex msingle-float
23184 @opindex mdouble-float
23185 Generate code for single- or double-precision floating-point operations.
23186 @option{-mdouble-float} implies @option{-msingle-float}.
23189 @itemx -mno-multiple
23191 @opindex mno-multiple
23192 Generate code that uses (does not use) the load multiple word
23193 instructions and the store multiple word instructions. These
23194 instructions are generated by default on POWER systems, and not
23195 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
23196 PowerPC systems, since those instructions do not work when the
23197 processor is in little-endian mode. The exceptions are PPC740 and
23198 PPC750 which permit these instructions in little-endian mode.
23203 @opindex mno-update
23204 Generate code that uses (does not use) the load or store instructions
23205 that update the base register to the address of the calculated memory
23206 location. These instructions are generated by default. If you use
23207 @option{-mno-update}, there is a small window between the time that the
23208 stack pointer is updated and the address of the previous frame is
23209 stored, which means code that walks the stack frame across interrupts or
23210 signals may get corrupted data.
23212 @item -mavoid-indexed-addresses
23213 @itemx -mno-avoid-indexed-addresses
23214 @opindex mavoid-indexed-addresses
23215 @opindex mno-avoid-indexed-addresses
23216 Generate code that tries to avoid (not avoid) the use of indexed load
23217 or store instructions. These instructions can incur a performance
23218 penalty on Power6 processors in certain situations, such as when
23219 stepping through large arrays that cross a 16M boundary. This option
23220 is enabled by default when targeting Power6 and disabled otherwise.
23223 @itemx -mno-fused-madd
23224 @opindex mfused-madd
23225 @opindex mno-fused-madd
23226 Generate code that uses (does not use) the floating-point multiply and
23227 accumulate instructions. These instructions are generated by default
23228 if hardware floating point is used. The machine-dependent
23229 @option{-mfused-madd} option is now mapped to the machine-independent
23230 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
23231 mapped to @option{-ffp-contract=off}.
23233 @item -mno-strict-align
23234 @itemx -mstrict-align
23235 @opindex mno-strict-align
23236 @opindex mstrict-align
23237 On System V.4 and embedded PowerPC systems do not (do) assume that
23238 unaligned memory references are handled by the system.
23240 @item -mrelocatable
23241 @itemx -mno-relocatable
23242 @opindex mrelocatable
23243 @opindex mno-relocatable
23244 Generate code that allows (does not allow) a static executable to be
23245 relocated to a different address at run time. A simple embedded
23246 PowerPC system loader should relocate the entire contents of
23247 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
23248 a table of 32-bit addresses generated by this option. For this to
23249 work, all objects linked together must be compiled with
23250 @option{-mrelocatable} or @option{-mrelocatable-lib}.
23251 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
23253 @item -mrelocatable-lib
23254 @itemx -mno-relocatable-lib
23255 @opindex mrelocatable-lib
23256 @opindex mno-relocatable-lib
23257 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
23258 @code{.fixup} section to allow static executables to be relocated at
23259 run time, but @option{-mrelocatable-lib} does not use the smaller stack
23260 alignment of @option{-mrelocatable}. Objects compiled with
23261 @option{-mrelocatable-lib} may be linked with objects compiled with
23262 any combination of the @option{-mrelocatable} options.
23268 On System V.4 and embedded PowerPC systems do not (do) assume that
23269 register 2 contains a pointer to a global area pointing to the addresses
23270 used in the program.
23273 @itemx -mlittle-endian
23275 @opindex mlittle-endian
23276 On System V.4 and embedded PowerPC systems compile code for the
23277 processor in little-endian mode. The @option{-mlittle-endian} option is
23278 the same as @option{-mlittle}.
23281 @itemx -mbig-endian
23283 @opindex mbig-endian
23284 On System V.4 and embedded PowerPC systems compile code for the
23285 processor in big-endian mode. The @option{-mbig-endian} option is
23286 the same as @option{-mbig}.
23288 @item -mdynamic-no-pic
23289 @opindex mdynamic-no-pic
23290 On Darwin and Mac OS X systems, compile code so that it is not
23291 relocatable, but that its external references are relocatable. The
23292 resulting code is suitable for applications, but not shared
23295 @item -msingle-pic-base
23296 @opindex msingle-pic-base
23297 Treat the register used for PIC addressing as read-only, rather than
23298 loading it in the prologue for each function. The runtime system is
23299 responsible for initializing this register with an appropriate value
23300 before execution begins.
23302 @item -mprioritize-restricted-insns=@var{priority}
23303 @opindex mprioritize-restricted-insns
23304 This option controls the priority that is assigned to
23305 dispatch-slot restricted instructions during the second scheduling
23306 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
23307 or @samp{2} to assign no, highest, or second-highest (respectively)
23308 priority to dispatch-slot restricted
23311 @item -msched-costly-dep=@var{dependence_type}
23312 @opindex msched-costly-dep
23313 This option controls which dependences are considered costly
23314 by the target during instruction scheduling. The argument
23315 @var{dependence_type} takes one of the following values:
23319 No dependence is costly.
23322 All dependences are costly.
23324 @item @samp{true_store_to_load}
23325 A true dependence from store to load is costly.
23327 @item @samp{store_to_load}
23328 Any dependence from store to load is costly.
23331 Any dependence for which the latency is greater than or equal to
23332 @var{number} is costly.
23335 @item -minsert-sched-nops=@var{scheme}
23336 @opindex minsert-sched-nops
23337 This option controls which NOP insertion scheme is used during
23338 the second scheduling pass. The argument @var{scheme} takes one of the
23346 Pad with NOPs any dispatch group that has vacant issue slots,
23347 according to the scheduler's grouping.
23349 @item @samp{regroup_exact}
23350 Insert NOPs to force costly dependent insns into
23351 separate groups. Insert exactly as many NOPs as needed to force an insn
23352 to a new group, according to the estimated processor grouping.
23355 Insert NOPs to force costly dependent insns into
23356 separate groups. Insert @var{number} NOPs to force an insn to a new group.
23360 @opindex mcall-sysv
23361 On System V.4 and embedded PowerPC systems compile code using calling
23362 conventions that adhere to the March 1995 draft of the System V
23363 Application Binary Interface, PowerPC processor supplement. This is the
23364 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
23366 @item -mcall-sysv-eabi
23368 @opindex mcall-sysv-eabi
23369 @opindex mcall-eabi
23370 Specify both @option{-mcall-sysv} and @option{-meabi} options.
23372 @item -mcall-sysv-noeabi
23373 @opindex mcall-sysv-noeabi
23374 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
23376 @item -mcall-aixdesc
23378 On System V.4 and embedded PowerPC systems compile code for the AIX
23382 @opindex mcall-linux
23383 On System V.4 and embedded PowerPC systems compile code for the
23384 Linux-based GNU system.
23386 @item -mcall-freebsd
23387 @opindex mcall-freebsd
23388 On System V.4 and embedded PowerPC systems compile code for the
23389 FreeBSD operating system.
23391 @item -mcall-netbsd
23392 @opindex mcall-netbsd
23393 On System V.4 and embedded PowerPC systems compile code for the
23394 NetBSD operating system.
23396 @item -mcall-openbsd
23397 @opindex mcall-netbsd
23398 On System V.4 and embedded PowerPC systems compile code for the
23399 OpenBSD operating system.
23401 @item -maix-struct-return
23402 @opindex maix-struct-return
23403 Return all structures in memory (as specified by the AIX ABI)@.
23405 @item -msvr4-struct-return
23406 @opindex msvr4-struct-return
23407 Return structures smaller than 8 bytes in registers (as specified by the
23410 @item -mabi=@var{abi-type}
23412 Extend the current ABI with a particular extension, or remove such extension.
23413 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
23414 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
23415 @samp{elfv1}, @samp{elfv2}@.
23419 Extend the current ABI with SPE ABI extensions. This does not change
23420 the default ABI, instead it adds the SPE ABI extensions to the current
23424 @opindex mabi=no-spe
23425 Disable Book-E SPE ABI extensions for the current ABI@.
23427 @item -mabi=ibmlongdouble
23428 @opindex mabi=ibmlongdouble
23429 Change the current ABI to use IBM extended-precision long double.
23430 This is not likely to work if your system defaults to using IEEE
23431 extended-precision long double. If you change the long double type
23432 from IEEE extended-precision, the compiler will issue a warning unless
23433 you use the @option{-Wno-psabi} option.
23435 @item -mabi=ieeelongdouble
23436 @opindex mabi=ieeelongdouble
23437 Change the current ABI to use IEEE extended-precision long double.
23438 This is not likely to work if your system defaults to using IBM
23439 extended-precision long double. If you change the long double type
23440 from IBM extended-precision, the compiler will issue a warning unless
23441 you use the @option{-Wno-psabi} option.
23444 @opindex mabi=elfv1
23445 Change the current ABI to use the ELFv1 ABI.
23446 This is the default ABI for big-endian PowerPC 64-bit Linux.
23447 Overriding the default ABI requires special system support and is
23448 likely to fail in spectacular ways.
23451 @opindex mabi=elfv2
23452 Change the current ABI to use the ELFv2 ABI.
23453 This is the default ABI for little-endian PowerPC 64-bit Linux.
23454 Overriding the default ABI requires special system support and is
23455 likely to fail in spectacular ways.
23457 @item -mgnu-attribute
23458 @itemx -mno-gnu-attribute
23459 @opindex mgnu-attribute
23460 @opindex mno-gnu-attribute
23461 Emit .gnu_attribute assembly directives to set tag/value pairs in a
23462 .gnu.attributes section that specify ABI variations in function
23463 parameters or return values.
23466 @itemx -mno-prototype
23467 @opindex mprototype
23468 @opindex mno-prototype
23469 On System V.4 and embedded PowerPC systems assume that all calls to
23470 variable argument functions are properly prototyped. Otherwise, the
23471 compiler must insert an instruction before every non-prototyped call to
23472 set or clear bit 6 of the condition code register (@code{CR}) to
23473 indicate whether floating-point values are passed in the floating-point
23474 registers in case the function takes variable arguments. With
23475 @option{-mprototype}, only calls to prototyped variable argument functions
23476 set or clear the bit.
23480 On embedded PowerPC systems, assume that the startup module is called
23481 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
23482 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
23487 On embedded PowerPC systems, assume that the startup module is called
23488 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
23493 On embedded PowerPC systems, assume that the startup module is called
23494 @file{crt0.o} and the standard C libraries are @file{libads.a} and
23497 @item -myellowknife
23498 @opindex myellowknife
23499 On embedded PowerPC systems, assume that the startup module is called
23500 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
23505 On System V.4 and embedded PowerPC systems, specify that you are
23506 compiling for a VxWorks system.
23510 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
23511 header to indicate that @samp{eabi} extended relocations are used.
23517 On System V.4 and embedded PowerPC systems do (do not) adhere to the
23518 Embedded Applications Binary Interface (EABI), which is a set of
23519 modifications to the System V.4 specifications. Selecting @option{-meabi}
23520 means that the stack is aligned to an 8-byte boundary, a function
23521 @code{__eabi} is called from @code{main} to set up the EABI
23522 environment, and the @option{-msdata} option can use both @code{r2} and
23523 @code{r13} to point to two separate small data areas. Selecting
23524 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
23525 no EABI initialization function is called from @code{main}, and the
23526 @option{-msdata} option only uses @code{r13} to point to a single
23527 small data area. The @option{-meabi} option is on by default if you
23528 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
23531 @opindex msdata=eabi
23532 On System V.4 and embedded PowerPC systems, put small initialized
23533 @code{const} global and static data in the @code{.sdata2} section, which
23534 is pointed to by register @code{r2}. Put small initialized
23535 non-@code{const} global and static data in the @code{.sdata} section,
23536 which is pointed to by register @code{r13}. Put small uninitialized
23537 global and static data in the @code{.sbss} section, which is adjacent to
23538 the @code{.sdata} section. The @option{-msdata=eabi} option is
23539 incompatible with the @option{-mrelocatable} option. The
23540 @option{-msdata=eabi} option also sets the @option{-memb} option.
23543 @opindex msdata=sysv
23544 On System V.4 and embedded PowerPC systems, put small global and static
23545 data in the @code{.sdata} section, which is pointed to by register
23546 @code{r13}. Put small uninitialized global and static data in the
23547 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
23548 The @option{-msdata=sysv} option is incompatible with the
23549 @option{-mrelocatable} option.
23551 @item -msdata=default
23553 @opindex msdata=default
23555 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
23556 compile code the same as @option{-msdata=eabi}, otherwise compile code the
23557 same as @option{-msdata=sysv}.
23560 @opindex msdata=data
23561 On System V.4 and embedded PowerPC systems, put small global
23562 data in the @code{.sdata} section. Put small uninitialized global
23563 data in the @code{.sbss} section. Do not use register @code{r13}
23564 to address small data however. This is the default behavior unless
23565 other @option{-msdata} options are used.
23569 @opindex msdata=none
23571 On embedded PowerPC systems, put all initialized global and static data
23572 in the @code{.data} section, and all uninitialized data in the
23573 @code{.bss} section.
23575 @item -mblock-move-inline-limit=@var{num}
23576 @opindex mblock-move-inline-limit
23577 Inline all block moves (such as calls to @code{memcpy} or structure
23578 copies) less than or equal to @var{num} bytes. The minimum value for
23579 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
23580 targets. The default value is target-specific.
23584 @cindex smaller data references (PowerPC)
23585 @cindex .sdata/.sdata2 references (PowerPC)
23586 On embedded PowerPC systems, put global and static items less than or
23587 equal to @var{num} bytes into the small data or BSS sections instead of
23588 the normal data or BSS section. By default, @var{num} is 8. The
23589 @option{-G @var{num}} switch is also passed to the linker.
23590 All modules should be compiled with the same @option{-G @var{num}} value.
23593 @itemx -mno-regnames
23595 @opindex mno-regnames
23596 On System V.4 and embedded PowerPC systems do (do not) emit register
23597 names in the assembly language output using symbolic forms.
23600 @itemx -mno-longcall
23602 @opindex mno-longcall
23603 By default assume that all calls are far away so that a longer and more
23604 expensive calling sequence is required. This is required for calls
23605 farther than 32 megabytes (33,554,432 bytes) from the current location.
23606 A short call is generated if the compiler knows
23607 the call cannot be that far away. This setting can be overridden by
23608 the @code{shortcall} function attribute, or by @code{#pragma
23611 Some linkers are capable of detecting out-of-range calls and generating
23612 glue code on the fly. On these systems, long calls are unnecessary and
23613 generate slower code. As of this writing, the AIX linker can do this,
23614 as can the GNU linker for PowerPC/64. It is planned to add this feature
23615 to the GNU linker for 32-bit PowerPC systems as well.
23617 In the future, GCC may ignore all longcall specifications
23618 when the linker is known to generate glue.
23620 @item -mtls-markers
23621 @itemx -mno-tls-markers
23622 @opindex mtls-markers
23623 @opindex mno-tls-markers
23624 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
23625 specifying the function argument. The relocation allows the linker to
23626 reliably associate function call with argument setup instructions for
23627 TLS optimization, which in turn allows GCC to better schedule the
23633 This option enables use of the reciprocal estimate and
23634 reciprocal square root estimate instructions with additional
23635 Newton-Raphson steps to increase precision instead of doing a divide or
23636 square root and divide for floating-point arguments. You should use
23637 the @option{-ffast-math} option when using @option{-mrecip} (or at
23638 least @option{-funsafe-math-optimizations},
23639 @option{-ffinite-math-only}, @option{-freciprocal-math} and
23640 @option{-fno-trapping-math}). Note that while the throughput of the
23641 sequence is generally higher than the throughput of the non-reciprocal
23642 instruction, the precision of the sequence can be decreased by up to 2
23643 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
23646 @item -mrecip=@var{opt}
23647 @opindex mrecip=opt
23648 This option controls which reciprocal estimate instructions
23649 may be used. @var{opt} is a comma-separated list of options, which may
23650 be preceded by a @code{!} to invert the option:
23655 Enable all estimate instructions.
23658 Enable the default instructions, equivalent to @option{-mrecip}.
23661 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23664 Enable the reciprocal approximation instructions for both
23665 single and double precision.
23668 Enable the single-precision reciprocal approximation instructions.
23671 Enable the double-precision reciprocal approximation instructions.
23674 Enable the reciprocal square root approximation instructions for both
23675 single and double precision.
23678 Enable the single-precision reciprocal square root approximation instructions.
23681 Enable the double-precision reciprocal square root approximation instructions.
23685 So, for example, @option{-mrecip=all,!rsqrtd} enables
23686 all of the reciprocal estimate instructions, except for the
23687 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
23688 which handle the double-precision reciprocal square root calculations.
23690 @item -mrecip-precision
23691 @itemx -mno-recip-precision
23692 @opindex mrecip-precision
23693 Assume (do not assume) that the reciprocal estimate instructions
23694 provide higher-precision estimates than is mandated by the PowerPC
23695 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
23696 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
23697 The double-precision square root estimate instructions are not generated by
23698 default on low-precision machines, since they do not provide an
23699 estimate that converges after three steps.
23701 @item -mpointers-to-nested-functions
23702 @itemx -mno-pointers-to-nested-functions
23703 @opindex mpointers-to-nested-functions
23704 Generate (do not generate) code to load up the static chain register
23705 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
23706 systems where a function pointer points to a 3-word descriptor giving
23707 the function address, TOC value to be loaded in register @code{r2}, and
23708 static chain value to be loaded in register @code{r11}. The
23709 @option{-mpointers-to-nested-functions} is on by default. You cannot
23710 call through pointers to nested functions or pointers
23711 to functions compiled in other languages that use the static chain if
23712 you use @option{-mno-pointers-to-nested-functions}.
23714 @item -msave-toc-indirect
23715 @itemx -mno-save-toc-indirect
23716 @opindex msave-toc-indirect
23717 Generate (do not generate) code to save the TOC value in the reserved
23718 stack location in the function prologue if the function calls through
23719 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
23720 saved in the prologue, it is saved just before the call through the
23721 pointer. The @option{-mno-save-toc-indirect} option is the default.
23723 @item -mcompat-align-parm
23724 @itemx -mno-compat-align-parm
23725 @opindex mcompat-align-parm
23726 Generate (do not generate) code to pass structure parameters with a
23727 maximum alignment of 64 bits, for compatibility with older versions
23730 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
23731 structure parameter on a 128-bit boundary when that structure contained
23732 a member requiring 128-bit alignment. This is corrected in more
23733 recent versions of GCC. This option may be used to generate code
23734 that is compatible with functions compiled with older versions of
23737 The @option{-mno-compat-align-parm} option is the default.
23739 @item -mstack-protector-guard=@var{guard}
23740 @itemx -mstack-protector-guard-reg=@var{reg}
23741 @itemx -mstack-protector-guard-offset=@var{offset}
23742 @itemx -mstack-protector-guard-symbol=@var{symbol}
23743 @opindex mstack-protector-guard
23744 @opindex mstack-protector-guard-reg
23745 @opindex mstack-protector-guard-offset
23746 @opindex mstack-protector-guard-symbol
23747 Generate stack protection code using canary at @var{guard}. Supported
23748 locations are @samp{global} for global canary or @samp{tls} for per-thread
23749 canary in the TLS block (the default with GNU libc version 2.4 or later).
23751 With the latter choice the options
23752 @option{-mstack-protector-guard-reg=@var{reg}} and
23753 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
23754 which register to use as base register for reading the canary, and from what
23755 offset from that base register. The default for those is as specified in the
23756 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
23757 the offset with a symbol reference to a canary in the TLS block.
23761 @node RISC-V Options
23762 @subsection RISC-V Options
23763 @cindex RISC-V Options
23765 These command-line options are defined for RISC-V targets:
23768 @item -mbranch-cost=@var{n}
23769 @opindex mbranch-cost
23770 Set the cost of branches to roughly @var{n} instructions.
23775 When generating PIC code, do or don't allow the use of PLTs. Ignored for
23776 non-PIC. The default is @option{-mplt}.
23778 @item -mabi=@var{ABI-string}
23780 Specify integer and floating-point calling convention. @var{ABI-string}
23781 contains two parts: the size of integer types and the registers used for
23782 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
23783 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
23784 32-bit), and that floating-point values up to 64 bits wide are passed in F
23785 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
23786 allows the compiler to generate code that uses the F and D extensions but only
23787 allows floating-point values up to 32 bits long to be passed in registers; or
23788 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
23789 passed in registers.
23791 The default for this argument is system dependent, users who want a specific
23792 calling convention should specify one explicitly. The valid calling
23793 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
23794 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
23795 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
23796 invalid because the ABI requires 64-bit values be passed in F registers, but F
23797 registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can
23798 only be used with the @samp{rv32e} architecture. This ABI is not well
23799 specified at present, and is subject to change.
23804 Do or don't use hardware floating-point divide and square root instructions.
23805 This requires the F or D extensions for floating-point registers. The default
23806 is to use them if the specified architecture has these instructions.
23811 Do or don't use hardware instructions for integer division. This requires the
23812 M extension. The default is to use them if the specified architecture has
23813 these instructions.
23815 @item -march=@var{ISA-string}
23817 Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be
23818 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and
23821 @item -mtune=@var{processor-string}
23823 Optimize the output for the given processor, specified by microarchitecture
23826 @item -mpreferred-stack-boundary=@var{num}
23827 @opindex mpreferred-stack-boundary
23828 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23829 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23830 the default is 4 (16 bytes or 128-bits).
23832 @strong{Warning:} If you use this switch, then you must build all modules with
23833 the same value, including any libraries. This includes the system libraries
23834 and startup modules.
23836 @item -msmall-data-limit=@var{n}
23837 @opindex msmall-data-limit
23838 Put global and static data smaller than @var{n} bytes into a special section
23841 @item -msave-restore
23842 @itemx -mno-save-restore
23843 @opindex msave-restore
23844 Do or don't use smaller but slower prologue and epilogue code that uses
23845 library function calls. The default is to use fast inline prologues and
23848 @item -mstrict-align
23849 @itemx -mno-strict-align
23850 @opindex mstrict-align
23851 Do not or do generate unaligned memory accesses. The default is set depending
23852 on whether the processor we are optimizing for supports fast unaligned access
23855 @item -mcmodel=medlow
23856 @opindex mcmodel=medlow
23857 Generate code for the medium-low code model. The program and its statically
23858 defined symbols must lie within a single 2 GiB address range and must lie
23859 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
23860 statically or dynamically linked. This is the default code model.
23862 @item -mcmodel=medany
23863 @opindex mcmodel=medany
23864 Generate code for the medium-any code model. The program and its statically
23865 defined symbols must be within any single 2 GiB address range. Programs can be
23866 statically or dynamically linked.
23868 @item -mexplicit-relocs
23869 @itemx -mno-exlicit-relocs
23870 Use or do not use assembler relocation operators when dealing with symbolic
23871 addresses. The alternative is to use assembler macros instead, which may
23872 limit optimization.
23876 Take advantage of linker relaxations to reduce the number of instructions
23877 required to materialize symbol addresses. The default is to take advantage of
23878 linker relaxations.
23883 @subsection RL78 Options
23884 @cindex RL78 Options
23890 Links in additional target libraries to support operation within a
23899 Specifies the type of hardware multiplication and division support to
23900 be used. The simplest is @code{none}, which uses software for both
23901 multiplication and division. This is the default. The @code{g13}
23902 value is for the hardware multiply/divide peripheral found on the
23903 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
23904 the multiplication and division instructions supported by the RL78/G14
23905 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
23906 the value @code{mg10} is an alias for @code{none}.
23908 In addition a C preprocessor macro is defined, based upon the setting
23909 of this option. Possible values are: @code{__RL78_MUL_NONE__},
23910 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
23917 Specifies the RL78 core to target. The default is the G14 core, also
23918 known as an S3 core or just RL78. The G13 or S2 core does not have
23919 multiply or divide instructions, instead it uses a hardware peripheral
23920 for these operations. The G10 or S1 core does not have register
23921 banks, so it uses a different calling convention.
23923 If this option is set it also selects the type of hardware multiply
23924 support to use, unless this is overridden by an explicit
23925 @option{-mmul=none} option on the command line. Thus specifying
23926 @option{-mcpu=g13} enables the use of the G13 hardware multiply
23927 peripheral and specifying @option{-mcpu=g10} disables the use of
23928 hardware multiplications altogether.
23930 Note, although the RL78/G14 core is the default target, specifying
23931 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
23932 change the behavior of the toolchain since it also enables G14
23933 hardware multiply support. If these options are not specified on the
23934 command line then software multiplication routines will be used even
23935 though the code targets the RL78 core. This is for backwards
23936 compatibility with older toolchains which did not have hardware
23937 multiply and divide support.
23939 In addition a C preprocessor macro is defined, based upon the setting
23940 of this option. Possible values are: @code{__RL78_G10__},
23941 @code{__RL78_G13__} or @code{__RL78_G14__}.
23951 These are aliases for the corresponding @option{-mcpu=} option. They
23952 are provided for backwards compatibility.
23956 Allow the compiler to use all of the available registers. By default
23957 registers @code{r24..r31} are reserved for use in interrupt handlers.
23958 With this option enabled these registers can be used in ordinary
23961 @item -m64bit-doubles
23962 @itemx -m32bit-doubles
23963 @opindex m64bit-doubles
23964 @opindex m32bit-doubles
23965 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
23966 or 32 bits (@option{-m32bit-doubles}) in size. The default is
23967 @option{-m32bit-doubles}.
23969 @item -msave-mduc-in-interrupts
23970 @itemx -mno-save-mduc-in-interrupts
23971 @opindex msave-mduc-in-interrupts
23972 @opindex mno-save-mduc-in-interrupts
23973 Specifies that interrupt handler functions should preserve the
23974 MDUC registers. This is only necessary if normal code might use
23975 the MDUC registers, for example because it performs multiplication
23976 and division operations. The default is to ignore the MDUC registers
23977 as this makes the interrupt handlers faster. The target option -mg13
23978 needs to be passed for this to work as this feature is only available
23979 on the G13 target (S2 core). The MDUC registers will only be saved
23980 if the interrupt handler performs a multiplication or division
23981 operation or it calls another function.
23985 @node RS/6000 and PowerPC Options
23986 @subsection IBM RS/6000 and PowerPC Options
23987 @cindex RS/6000 and PowerPC Options
23988 @cindex IBM RS/6000 and PowerPC Options
23990 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
23992 @item -mpowerpc-gpopt
23993 @itemx -mno-powerpc-gpopt
23994 @itemx -mpowerpc-gfxopt
23995 @itemx -mno-powerpc-gfxopt
23998 @itemx -mno-powerpc64
24002 @itemx -mno-popcntb
24004 @itemx -mno-popcntd
24013 @itemx -mno-hard-dfp
24014 @opindex mpowerpc-gpopt
24015 @opindex mno-powerpc-gpopt
24016 @opindex mpowerpc-gfxopt
24017 @opindex mno-powerpc-gfxopt
24018 @opindex mpowerpc64
24019 @opindex mno-powerpc64
24023 @opindex mno-popcntb
24025 @opindex mno-popcntd
24031 @opindex mno-mfpgpr
24033 @opindex mno-hard-dfp
24034 You use these options to specify which instructions are available on the
24035 processor you are using. The default value of these options is
24036 determined when configuring GCC@. Specifying the
24037 @option{-mcpu=@var{cpu_type}} overrides the specification of these
24038 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
24039 rather than the options listed above.
24041 Specifying @option{-mpowerpc-gpopt} allows
24042 GCC to use the optional PowerPC architecture instructions in the
24043 General Purpose group, including floating-point square root. Specifying
24044 @option{-mpowerpc-gfxopt} allows GCC to
24045 use the optional PowerPC architecture instructions in the Graphics
24046 group, including floating-point select.
24048 The @option{-mmfcrf} option allows GCC to generate the move from
24049 condition register field instruction implemented on the POWER4
24050 processor and other processors that support the PowerPC V2.01
24052 The @option{-mpopcntb} option allows GCC to generate the popcount and
24053 double-precision FP reciprocal estimate instruction implemented on the
24054 POWER5 processor and other processors that support the PowerPC V2.02
24056 The @option{-mpopcntd} option allows GCC to generate the popcount
24057 instruction implemented on the POWER7 processor and other processors
24058 that support the PowerPC V2.06 architecture.
24059 The @option{-mfprnd} option allows GCC to generate the FP round to
24060 integer instructions implemented on the POWER5+ processor and other
24061 processors that support the PowerPC V2.03 architecture.
24062 The @option{-mcmpb} option allows GCC to generate the compare bytes
24063 instruction implemented on the POWER6 processor and other processors
24064 that support the PowerPC V2.05 architecture.
24065 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
24066 general-purpose register instructions implemented on the POWER6X
24067 processor and other processors that support the extended PowerPC V2.05
24069 The @option{-mhard-dfp} option allows GCC to generate the decimal
24070 floating-point instructions implemented on some POWER processors.
24072 The @option{-mpowerpc64} option allows GCC to generate the additional
24073 64-bit instructions that are found in the full PowerPC64 architecture
24074 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
24075 @option{-mno-powerpc64}.
24077 @item -mcpu=@var{cpu_type}
24079 Set architecture type, register usage, and
24080 instruction scheduling parameters for machine type @var{cpu_type}.
24081 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
24082 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
24083 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
24084 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
24085 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
24086 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
24087 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
24088 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
24089 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
24090 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
24091 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
24092 @samp{rs64}, and @samp{native}.
24094 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
24095 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
24096 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
24097 architecture machine types, with an appropriate, generic processor
24098 model assumed for scheduling purposes.
24100 Specifying @samp{native} as cpu type detects and selects the
24101 architecture option that corresponds to the host processor of the
24102 system performing the compilation.
24103 @option{-mcpu=native} has no effect if GCC does not recognize the
24106 The other options specify a specific processor. Code generated under
24107 those options runs best on that processor, and may not run at all on
24110 The @option{-mcpu} options automatically enable or disable the
24113 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
24114 -mpopcntb -mpopcntd -mpowerpc64 @gol
24115 -mpowerpc-gpopt -mpowerpc-gfxopt @gol
24116 -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
24117 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
24118 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
24120 The particular options set for any particular CPU varies between
24121 compiler versions, depending on what setting seems to produce optimal
24122 code for that CPU; it doesn't necessarily reflect the actual hardware's
24123 capabilities. If you wish to set an individual option to a particular
24124 value, you may specify it after the @option{-mcpu} option, like
24125 @option{-mcpu=970 -mno-altivec}.
24127 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
24128 not enabled or disabled by the @option{-mcpu} option at present because
24129 AIX does not have full support for these options. You may still
24130 enable or disable them individually if you're sure it'll work in your
24133 @item -mtune=@var{cpu_type}
24135 Set the instruction scheduling parameters for machine type
24136 @var{cpu_type}, but do not set the architecture type or register usage,
24137 as @option{-mcpu=@var{cpu_type}} does. The same
24138 values for @var{cpu_type} are used for @option{-mtune} as for
24139 @option{-mcpu}. If both are specified, the code generated uses the
24140 architecture and registers set by @option{-mcpu}, but the
24141 scheduling parameters set by @option{-mtune}.
24143 @item -mcmodel=small
24144 @opindex mcmodel=small
24145 Generate PowerPC64 code for the small model: The TOC is limited to
24148 @item -mcmodel=medium
24149 @opindex mcmodel=medium
24150 Generate PowerPC64 code for the medium model: The TOC and other static
24151 data may be up to a total of 4G in size. This is the default for 64-bit
24154 @item -mcmodel=large
24155 @opindex mcmodel=large
24156 Generate PowerPC64 code for the large model: The TOC may be up to 4G
24157 in size. Other data and code is only limited by the 64-bit address
24161 @itemx -mno-altivec
24163 @opindex mno-altivec
24164 Generate code that uses (does not use) AltiVec instructions, and also
24165 enable the use of built-in functions that allow more direct access to
24166 the AltiVec instruction set. You may also need to set
24167 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
24170 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
24171 @option{-maltivec=be}, the element order for AltiVec intrinsics such
24172 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
24173 match array element order corresponding to the endianness of the
24174 target. That is, element zero identifies the leftmost element in a
24175 vector register when targeting a big-endian platform, and identifies
24176 the rightmost element in a vector register when targeting a
24177 little-endian platform.
24180 @opindex maltivec=be
24181 Generate AltiVec instructions using big-endian element order,
24182 regardless of whether the target is big- or little-endian. This is
24183 the default when targeting a big-endian platform. Using this option
24184 is currently deprecated. Support for this feature will be removed in
24187 The element order is used to interpret element numbers in AltiVec
24188 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
24189 @code{vec_insert}. By default, these match array element order
24190 corresponding to the endianness for the target.
24193 @opindex maltivec=le
24194 Generate AltiVec instructions using little-endian element order,
24195 regardless of whether the target is big- or little-endian. This is
24196 the default when targeting a little-endian platform. This option is
24197 currently ignored when targeting a big-endian platform.
24199 The element order is used to interpret element numbers in AltiVec
24200 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
24201 @code{vec_insert}. By default, these match array element order
24202 corresponding to the endianness for the target.
24207 @opindex mno-vrsave
24208 Generate VRSAVE instructions when generating AltiVec code.
24211 @opindex msecure-plt
24212 Generate code that allows @command{ld} and @command{ld.so}
24213 to build executables and shared
24214 libraries with non-executable @code{.plt} and @code{.got} sections.
24216 32-bit SYSV ABI option.
24220 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
24222 requires @code{.plt} and @code{.got}
24223 sections that are both writable and executable.
24224 This is a PowerPC 32-bit SYSV ABI option.
24230 This switch enables or disables the generation of ISEL instructions.
24236 Generate code that uses (does not use) vector/scalar (VSX)
24237 instructions, and also enable the use of built-in functions that allow
24238 more direct access to the VSX instruction set.
24243 @opindex mno-crypto
24244 Enable the use (disable) of the built-in functions that allow direct
24245 access to the cryptographic instructions that were added in version
24246 2.07 of the PowerPC ISA.
24252 Enable (disable) the use of the built-in functions that allow direct
24253 access to the Hardware Transactional Memory (HTM) instructions that
24254 were added in version 2.07 of the PowerPC ISA.
24256 @item -mpower8-fusion
24257 @itemx -mno-power8-fusion
24258 @opindex mpower8-fusion
24259 @opindex mno-power8-fusion
24260 Generate code that keeps (does not keeps) some integer operations
24261 adjacent so that the instructions can be fused together on power8 and
24264 @item -mpower8-vector
24265 @itemx -mno-power8-vector
24266 @opindex mpower8-vector
24267 @opindex mno-power8-vector
24268 Generate code that uses (does not use) the vector and scalar
24269 instructions that were added in version 2.07 of the PowerPC ISA. Also
24270 enable the use of built-in functions that allow more direct access to
24271 the vector instructions.
24273 @item -mquad-memory
24274 @itemx -mno-quad-memory
24275 @opindex mquad-memory
24276 @opindex mno-quad-memory
24277 Generate code that uses (does not use) the non-atomic quad word memory
24278 instructions. The @option{-mquad-memory} option requires use of
24281 @item -mquad-memory-atomic
24282 @itemx -mno-quad-memory-atomic
24283 @opindex mquad-memory-atomic
24284 @opindex mno-quad-memory-atomic
24285 Generate code that uses (does not use) the atomic quad word memory
24286 instructions. The @option{-mquad-memory-atomic} option requires use of
24290 @itemx -mno-float128
24292 @opindex mno-float128
24293 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
24294 and use either software emulation for IEEE 128-bit floating point or
24295 hardware instructions.
24297 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
24298 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
24299 use the IEEE 128-bit floating point support. The IEEE 128-bit
24300 floating point support only works on PowerPC Linux systems.
24302 The default for @option{-mfloat128} is enabled on PowerPC Linux
24303 systems using the VSX instruction set, and disabled on other systems.
24305 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
24306 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
24307 point support will also enable the generation of ISA 3.0 IEEE 128-bit
24308 floating point instructions. Otherwise, if you do not specify to
24309 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
24310 system, IEEE 128-bit floating point will be done with software
24313 @item -mfloat128-hardware
24314 @itemx -mno-float128-hardware
24315 @opindex mfloat128-hardware
24316 @opindex mno-float128-hardware
24317 Enable/disable using ISA 3.0 hardware instructions to support the
24318 @var{__float128} data type.
24320 The default for @option{-mfloat128-hardware} is enabled on PowerPC
24321 Linux systems using the ISA 3.0 instruction set, and disabled on other
24328 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
24329 targets (including GNU/Linux). The 32-bit environment sets int, long
24330 and pointer to 32 bits and generates code that runs on any PowerPC
24331 variant. The 64-bit environment sets int to 32 bits and long and
24332 pointer to 64 bits, and generates code for PowerPC64, as for
24333 @option{-mpowerpc64}.
24336 @itemx -mno-fp-in-toc
24337 @itemx -mno-sum-in-toc
24338 @itemx -mminimal-toc
24340 @opindex mno-fp-in-toc
24341 @opindex mno-sum-in-toc
24342 @opindex mminimal-toc
24343 Modify generation of the TOC (Table Of Contents), which is created for
24344 every executable file. The @option{-mfull-toc} option is selected by
24345 default. In that case, GCC allocates at least one TOC entry for
24346 each unique non-automatic variable reference in your program. GCC
24347 also places floating-point constants in the TOC@. However, only
24348 16,384 entries are available in the TOC@.
24350 If you receive a linker error message that saying you have overflowed
24351 the available TOC space, you can reduce the amount of TOC space used
24352 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
24353 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
24354 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
24355 generate code to calculate the sum of an address and a constant at
24356 run time instead of putting that sum into the TOC@. You may specify one
24357 or both of these options. Each causes GCC to produce very slightly
24358 slower and larger code at the expense of conserving TOC space.
24360 If you still run out of space in the TOC even when you specify both of
24361 these options, specify @option{-mminimal-toc} instead. This option causes
24362 GCC to make only one TOC entry for every file. When you specify this
24363 option, GCC produces code that is slower and larger but which
24364 uses extremely little TOC space. You may wish to use this option
24365 only on files that contain less frequently-executed code.
24371 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
24372 @code{long} type, and the infrastructure needed to support them.
24373 Specifying @option{-maix64} implies @option{-mpowerpc64},
24374 while @option{-maix32} disables the 64-bit ABI and
24375 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
24378 @itemx -mno-xl-compat
24379 @opindex mxl-compat
24380 @opindex mno-xl-compat
24381 Produce code that conforms more closely to IBM XL compiler semantics
24382 when using AIX-compatible ABI@. Pass floating-point arguments to
24383 prototyped functions beyond the register save area (RSA) on the stack
24384 in addition to argument FPRs. Do not assume that most significant
24385 double in 128-bit long double value is properly rounded when comparing
24386 values and converting to double. Use XL symbol names for long double
24389 The AIX calling convention was extended but not initially documented to
24390 handle an obscure K&R C case of calling a function that takes the
24391 address of its arguments with fewer arguments than declared. IBM XL
24392 compilers access floating-point arguments that do not fit in the
24393 RSA from the stack when a subroutine is compiled without
24394 optimization. Because always storing floating-point arguments on the
24395 stack is inefficient and rarely needed, this option is not enabled by
24396 default and only is necessary when calling subroutines compiled by IBM
24397 XL compilers without optimization.
24401 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
24402 application written to use message passing with special startup code to
24403 enable the application to run. The system must have PE installed in the
24404 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
24405 must be overridden with the @option{-specs=} option to specify the
24406 appropriate directory location. The Parallel Environment does not
24407 support threads, so the @option{-mpe} option and the @option{-pthread}
24408 option are incompatible.
24410 @item -malign-natural
24411 @itemx -malign-power
24412 @opindex malign-natural
24413 @opindex malign-power
24414 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
24415 @option{-malign-natural} overrides the ABI-defined alignment of larger
24416 types, such as floating-point doubles, on their natural size-based boundary.
24417 The option @option{-malign-power} instructs GCC to follow the ABI-specified
24418 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
24420 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
24424 @itemx -mhard-float
24425 @opindex msoft-float
24426 @opindex mhard-float
24427 Generate code that does not use (uses) the floating-point register set.
24428 Software floating-point emulation is provided if you use the
24429 @option{-msoft-float} option, and pass the option to GCC when linking.
24432 @itemx -mno-multiple
24434 @opindex mno-multiple
24435 Generate code that uses (does not use) the load multiple word
24436 instructions and the store multiple word instructions. These
24437 instructions are generated by default on POWER systems, and not
24438 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
24439 PowerPC systems, since those instructions do not work when the
24440 processor is in little-endian mode. The exceptions are PPC740 and
24441 PPC750 which permit these instructions in little-endian mode.
24446 @opindex mno-update
24447 Generate code that uses (does not use) the load or store instructions
24448 that update the base register to the address of the calculated memory
24449 location. These instructions are generated by default. If you use
24450 @option{-mno-update}, there is a small window between the time that the
24451 stack pointer is updated and the address of the previous frame is
24452 stored, which means code that walks the stack frame across interrupts or
24453 signals may get corrupted data.
24455 @item -mavoid-indexed-addresses
24456 @itemx -mno-avoid-indexed-addresses
24457 @opindex mavoid-indexed-addresses
24458 @opindex mno-avoid-indexed-addresses
24459 Generate code that tries to avoid (not avoid) the use of indexed load
24460 or store instructions. These instructions can incur a performance
24461 penalty on Power6 processors in certain situations, such as when
24462 stepping through large arrays that cross a 16M boundary. This option
24463 is enabled by default when targeting Power6 and disabled otherwise.
24466 @itemx -mno-fused-madd
24467 @opindex mfused-madd
24468 @opindex mno-fused-madd
24469 Generate code that uses (does not use) the floating-point multiply and
24470 accumulate instructions. These instructions are generated by default
24471 if hardware floating point is used. The machine-dependent
24472 @option{-mfused-madd} option is now mapped to the machine-independent
24473 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24474 mapped to @option{-ffp-contract=off}.
24480 Generate code that uses (does not use) the half-word multiply and
24481 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
24482 These instructions are generated by default when targeting those
24489 Generate code that uses (does not use) the string-search @samp{dlmzb}
24490 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
24491 generated by default when targeting those processors.
24493 @item -mno-bit-align
24495 @opindex mno-bit-align
24496 @opindex mbit-align
24497 On System V.4 and embedded PowerPC systems do not (do) force structures
24498 and unions that contain bit-fields to be aligned to the base type of the
24501 For example, by default a structure containing nothing but 8
24502 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
24503 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
24504 the structure is aligned to a 1-byte boundary and is 1 byte in
24507 @item -mno-strict-align
24508 @itemx -mstrict-align
24509 @opindex mno-strict-align
24510 @opindex mstrict-align
24511 On System V.4 and embedded PowerPC systems do not (do) assume that
24512 unaligned memory references are handled by the system.
24514 @item -mrelocatable
24515 @itemx -mno-relocatable
24516 @opindex mrelocatable
24517 @opindex mno-relocatable
24518 Generate code that allows (does not allow) a static executable to be
24519 relocated to a different address at run time. A simple embedded
24520 PowerPC system loader should relocate the entire contents of
24521 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
24522 a table of 32-bit addresses generated by this option. For this to
24523 work, all objects linked together must be compiled with
24524 @option{-mrelocatable} or @option{-mrelocatable-lib}.
24525 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
24527 @item -mrelocatable-lib
24528 @itemx -mno-relocatable-lib
24529 @opindex mrelocatable-lib
24530 @opindex mno-relocatable-lib
24531 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
24532 @code{.fixup} section to allow static executables to be relocated at
24533 run time, but @option{-mrelocatable-lib} does not use the smaller stack
24534 alignment of @option{-mrelocatable}. Objects compiled with
24535 @option{-mrelocatable-lib} may be linked with objects compiled with
24536 any combination of the @option{-mrelocatable} options.
24542 On System V.4 and embedded PowerPC systems do not (do) assume that
24543 register 2 contains a pointer to a global area pointing to the addresses
24544 used in the program.
24547 @itemx -mlittle-endian
24549 @opindex mlittle-endian
24550 On System V.4 and embedded PowerPC systems compile code for the
24551 processor in little-endian mode. The @option{-mlittle-endian} option is
24552 the same as @option{-mlittle}.
24555 @itemx -mbig-endian
24557 @opindex mbig-endian
24558 On System V.4 and embedded PowerPC systems compile code for the
24559 processor in big-endian mode. The @option{-mbig-endian} option is
24560 the same as @option{-mbig}.
24562 @item -mdynamic-no-pic
24563 @opindex mdynamic-no-pic
24564 On Darwin and Mac OS X systems, compile code so that it is not
24565 relocatable, but that its external references are relocatable. The
24566 resulting code is suitable for applications, but not shared
24569 @item -msingle-pic-base
24570 @opindex msingle-pic-base
24571 Treat the register used for PIC addressing as read-only, rather than
24572 loading it in the prologue for each function. The runtime system is
24573 responsible for initializing this register with an appropriate value
24574 before execution begins.
24576 @item -mprioritize-restricted-insns=@var{priority}
24577 @opindex mprioritize-restricted-insns
24578 This option controls the priority that is assigned to
24579 dispatch-slot restricted instructions during the second scheduling
24580 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
24581 or @samp{2} to assign no, highest, or second-highest (respectively)
24582 priority to dispatch-slot restricted
24585 @item -msched-costly-dep=@var{dependence_type}
24586 @opindex msched-costly-dep
24587 This option controls which dependences are considered costly
24588 by the target during instruction scheduling. The argument
24589 @var{dependence_type} takes one of the following values:
24593 No dependence is costly.
24596 All dependences are costly.
24598 @item @samp{true_store_to_load}
24599 A true dependence from store to load is costly.
24601 @item @samp{store_to_load}
24602 Any dependence from store to load is costly.
24605 Any dependence for which the latency is greater than or equal to
24606 @var{number} is costly.
24609 @item -minsert-sched-nops=@var{scheme}
24610 @opindex minsert-sched-nops
24611 This option controls which NOP insertion scheme is used during
24612 the second scheduling pass. The argument @var{scheme} takes one of the
24620 Pad with NOPs any dispatch group that has vacant issue slots,
24621 according to the scheduler's grouping.
24623 @item @samp{regroup_exact}
24624 Insert NOPs to force costly dependent insns into
24625 separate groups. Insert exactly as many NOPs as needed to force an insn
24626 to a new group, according to the estimated processor grouping.
24629 Insert NOPs to force costly dependent insns into
24630 separate groups. Insert @var{number} NOPs to force an insn to a new group.
24634 @opindex mcall-sysv
24635 On System V.4 and embedded PowerPC systems compile code using calling
24636 conventions that adhere to the March 1995 draft of the System V
24637 Application Binary Interface, PowerPC processor supplement. This is the
24638 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
24640 @item -mcall-sysv-eabi
24642 @opindex mcall-sysv-eabi
24643 @opindex mcall-eabi
24644 Specify both @option{-mcall-sysv} and @option{-meabi} options.
24646 @item -mcall-sysv-noeabi
24647 @opindex mcall-sysv-noeabi
24648 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
24650 @item -mcall-aixdesc
24652 On System V.4 and embedded PowerPC systems compile code for the AIX
24656 @opindex mcall-linux
24657 On System V.4 and embedded PowerPC systems compile code for the
24658 Linux-based GNU system.
24660 @item -mcall-freebsd
24661 @opindex mcall-freebsd
24662 On System V.4 and embedded PowerPC systems compile code for the
24663 FreeBSD operating system.
24665 @item -mcall-netbsd
24666 @opindex mcall-netbsd
24667 On System V.4 and embedded PowerPC systems compile code for the
24668 NetBSD operating system.
24670 @item -mcall-openbsd
24671 @opindex mcall-netbsd
24672 On System V.4 and embedded PowerPC systems compile code for the
24673 OpenBSD operating system.
24675 @item -mtraceback=@var{traceback_type}
24676 @opindex mtraceback
24677 Select the type of traceback table. Valid values for @var{traceback_type}
24678 are @samp{full}, @samp{part}, and @samp{no}.
24680 @item -maix-struct-return
24681 @opindex maix-struct-return
24682 Return all structures in memory (as specified by the AIX ABI)@.
24684 @item -msvr4-struct-return
24685 @opindex msvr4-struct-return
24686 Return structures smaller than 8 bytes in registers (as specified by the
24689 @item -mabi=@var{abi-type}
24691 Extend the current ABI with a particular extension, or remove such extension.
24692 Valid values are @samp{altivec}, @samp{no-altivec},
24693 @samp{ibmlongdouble}, @samp{ieeelongdouble},
24694 @samp{elfv1}, @samp{elfv2}@.
24696 @item -mabi=ibmlongdouble
24697 @opindex mabi=ibmlongdouble
24698 Change the current ABI to use IBM extended-precision long double.
24699 This is not likely to work if your system defaults to using IEEE
24700 extended-precision long double. If you change the long double type
24701 from IEEE extended-precision, the compiler will issue a warning unless
24702 you use the @option{-Wno-psabi} option.
24704 @item -mabi=ieeelongdouble
24705 @opindex mabi=ieeelongdouble
24706 Change the current ABI to use IEEE extended-precision long double.
24707 This is not likely to work if your system defaults to using IBM
24708 extended-precision long double. If you change the long double type
24709 from IBM extended-precision, the compiler will issue a warning unless
24710 you use the @option{-Wno-psabi} option.
24713 @opindex mabi=elfv1
24714 Change the current ABI to use the ELFv1 ABI.
24715 This is the default ABI for big-endian PowerPC 64-bit Linux.
24716 Overriding the default ABI requires special system support and is
24717 likely to fail in spectacular ways.
24720 @opindex mabi=elfv2
24721 Change the current ABI to use the ELFv2 ABI.
24722 This is the default ABI for little-endian PowerPC 64-bit Linux.
24723 Overriding the default ABI requires special system support and is
24724 likely to fail in spectacular ways.
24726 @item -mgnu-attribute
24727 @itemx -mno-gnu-attribute
24728 @opindex mgnu-attribute
24729 @opindex mno-gnu-attribute
24730 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24731 .gnu.attributes section that specify ABI variations in function
24732 parameters or return values.
24735 @itemx -mno-prototype
24736 @opindex mprototype
24737 @opindex mno-prototype
24738 On System V.4 and embedded PowerPC systems assume that all calls to
24739 variable argument functions are properly prototyped. Otherwise, the
24740 compiler must insert an instruction before every non-prototyped call to
24741 set or clear bit 6 of the condition code register (@code{CR}) to
24742 indicate whether floating-point values are passed in the floating-point
24743 registers in case the function takes variable arguments. With
24744 @option{-mprototype}, only calls to prototyped variable argument functions
24745 set or clear the bit.
24749 On embedded PowerPC systems, assume that the startup module is called
24750 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24751 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24756 On embedded PowerPC systems, assume that the startup module is called
24757 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24762 On embedded PowerPC systems, assume that the startup module is called
24763 @file{crt0.o} and the standard C libraries are @file{libads.a} and
24766 @item -myellowknife
24767 @opindex myellowknife
24768 On embedded PowerPC systems, assume that the startup module is called
24769 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
24774 On System V.4 and embedded PowerPC systems, specify that you are
24775 compiling for a VxWorks system.
24779 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
24780 header to indicate that @samp{eabi} extended relocations are used.
24786 On System V.4 and embedded PowerPC systems do (do not) adhere to the
24787 Embedded Applications Binary Interface (EABI), which is a set of
24788 modifications to the System V.4 specifications. Selecting @option{-meabi}
24789 means that the stack is aligned to an 8-byte boundary, a function
24790 @code{__eabi} is called from @code{main} to set up the EABI
24791 environment, and the @option{-msdata} option can use both @code{r2} and
24792 @code{r13} to point to two separate small data areas. Selecting
24793 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
24794 no EABI initialization function is called from @code{main}, and the
24795 @option{-msdata} option only uses @code{r13} to point to a single
24796 small data area. The @option{-meabi} option is on by default if you
24797 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
24800 @opindex msdata=eabi
24801 On System V.4 and embedded PowerPC systems, put small initialized
24802 @code{const} global and static data in the @code{.sdata2} section, which
24803 is pointed to by register @code{r2}. Put small initialized
24804 non-@code{const} global and static data in the @code{.sdata} section,
24805 which is pointed to by register @code{r13}. Put small uninitialized
24806 global and static data in the @code{.sbss} section, which is adjacent to
24807 the @code{.sdata} section. The @option{-msdata=eabi} option is
24808 incompatible with the @option{-mrelocatable} option. The
24809 @option{-msdata=eabi} option also sets the @option{-memb} option.
24812 @opindex msdata=sysv
24813 On System V.4 and embedded PowerPC systems, put small global and static
24814 data in the @code{.sdata} section, which is pointed to by register
24815 @code{r13}. Put small uninitialized global and static data in the
24816 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
24817 The @option{-msdata=sysv} option is incompatible with the
24818 @option{-mrelocatable} option.
24820 @item -msdata=default
24822 @opindex msdata=default
24824 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
24825 compile code the same as @option{-msdata=eabi}, otherwise compile code the
24826 same as @option{-msdata=sysv}.
24829 @opindex msdata=data
24830 On System V.4 and embedded PowerPC systems, put small global
24831 data in the @code{.sdata} section. Put small uninitialized global
24832 data in the @code{.sbss} section. Do not use register @code{r13}
24833 to address small data however. This is the default behavior unless
24834 other @option{-msdata} options are used.
24838 @opindex msdata=none
24840 On embedded PowerPC systems, put all initialized global and static data
24841 in the @code{.data} section, and all uninitialized data in the
24842 @code{.bss} section.
24844 @item -mreadonly-in-sdata
24845 @opindex mreadonly-in-sdata
24846 @opindex mno-readonly-in-sdata
24847 Put read-only objects in the @code{.sdata} section as well. This is the
24850 @item -mblock-move-inline-limit=@var{num}
24851 @opindex mblock-move-inline-limit
24852 Inline all block moves (such as calls to @code{memcpy} or structure
24853 copies) less than or equal to @var{num} bytes. The minimum value for
24854 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
24855 targets. The default value is target-specific.
24857 @item -mblock-compare-inline-limit=@var{num}
24858 @opindex mblock-compare-inline-limit
24859 Generate non-looping inline code for all block compares (such as calls
24860 to @code{memcmp} or structure compares) less than or equal to @var{num}
24861 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
24862 block compare is disabled. The default value is target-specific.
24864 @item -mblock-compare-inline-loop-limit=@var{num}
24865 @opindex mblock-compare-inline-loop-limit
24866 Generate an inline expansion using loop code for all block compares that
24867 are less than or equal to @var{num} bytes, but greater than the limit
24868 for non-loop inline block compare expansion. If the block length is not
24869 constant, at most @var{num} bytes will be compared before @code{memcmp}
24870 is called to compare the remainder of the block. The default value is
24873 @item -mstring-compare-inline-limit=@var{num}
24874 @opindex mstring-compare-inline-limit
24875 Compare at most @var{num} string bytes with inline code.
24876 If the difference or end of string is not found at the
24877 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
24878 take care of the rest of the comparison. The default is 64 bytes.
24882 @cindex smaller data references (PowerPC)
24883 @cindex .sdata/.sdata2 references (PowerPC)
24884 On embedded PowerPC systems, put global and static items less than or
24885 equal to @var{num} bytes into the small data or BSS sections instead of
24886 the normal data or BSS section. By default, @var{num} is 8. The
24887 @option{-G @var{num}} switch is also passed to the linker.
24888 All modules should be compiled with the same @option{-G @var{num}} value.
24891 @itemx -mno-regnames
24893 @opindex mno-regnames
24894 On System V.4 and embedded PowerPC systems do (do not) emit register
24895 names in the assembly language output using symbolic forms.
24898 @itemx -mno-longcall
24900 @opindex mno-longcall
24901 By default assume that all calls are far away so that a longer and more
24902 expensive calling sequence is required. This is required for calls
24903 farther than 32 megabytes (33,554,432 bytes) from the current location.
24904 A short call is generated if the compiler knows
24905 the call cannot be that far away. This setting can be overridden by
24906 the @code{shortcall} function attribute, or by @code{#pragma
24909 Some linkers are capable of detecting out-of-range calls and generating
24910 glue code on the fly. On these systems, long calls are unnecessary and
24911 generate slower code. As of this writing, the AIX linker can do this,
24912 as can the GNU linker for PowerPC/64. It is planned to add this feature
24913 to the GNU linker for 32-bit PowerPC systems as well.
24915 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
24916 callee, L42}, plus a @dfn{branch island} (glue code). The two target
24917 addresses represent the callee and the branch island. The
24918 Darwin/PPC linker prefers the first address and generates a @code{bl
24919 callee} if the PPC @code{bl} instruction reaches the callee directly;
24920 otherwise, the linker generates @code{bl L42} to call the branch
24921 island. The branch island is appended to the body of the
24922 calling function; it computes the full 32-bit address of the callee
24925 On Mach-O (Darwin) systems, this option directs the compiler emit to
24926 the glue for every direct call, and the Darwin linker decides whether
24927 to use or discard it.
24929 In the future, GCC may ignore all longcall specifications
24930 when the linker is known to generate glue.
24932 @item -mtls-markers
24933 @itemx -mno-tls-markers
24934 @opindex mtls-markers
24935 @opindex mno-tls-markers
24936 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
24937 specifying the function argument. The relocation allows the linker to
24938 reliably associate function call with argument setup instructions for
24939 TLS optimization, which in turn allows GCC to better schedule the
24945 This option enables use of the reciprocal estimate and
24946 reciprocal square root estimate instructions with additional
24947 Newton-Raphson steps to increase precision instead of doing a divide or
24948 square root and divide for floating-point arguments. You should use
24949 the @option{-ffast-math} option when using @option{-mrecip} (or at
24950 least @option{-funsafe-math-optimizations},
24951 @option{-ffinite-math-only}, @option{-freciprocal-math} and
24952 @option{-fno-trapping-math}). Note that while the throughput of the
24953 sequence is generally higher than the throughput of the non-reciprocal
24954 instruction, the precision of the sequence can be decreased by up to 2
24955 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
24958 @item -mrecip=@var{opt}
24959 @opindex mrecip=opt
24960 This option controls which reciprocal estimate instructions
24961 may be used. @var{opt} is a comma-separated list of options, which may
24962 be preceded by a @code{!} to invert the option:
24967 Enable all estimate instructions.
24970 Enable the default instructions, equivalent to @option{-mrecip}.
24973 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24976 Enable the reciprocal approximation instructions for both
24977 single and double precision.
24980 Enable the single-precision reciprocal approximation instructions.
24983 Enable the double-precision reciprocal approximation instructions.
24986 Enable the reciprocal square root approximation instructions for both
24987 single and double precision.
24990 Enable the single-precision reciprocal square root approximation instructions.
24993 Enable the double-precision reciprocal square root approximation instructions.
24997 So, for example, @option{-mrecip=all,!rsqrtd} enables
24998 all of the reciprocal estimate instructions, except for the
24999 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
25000 which handle the double-precision reciprocal square root calculations.
25002 @item -mrecip-precision
25003 @itemx -mno-recip-precision
25004 @opindex mrecip-precision
25005 Assume (do not assume) that the reciprocal estimate instructions
25006 provide higher-precision estimates than is mandated by the PowerPC
25007 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
25008 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
25009 The double-precision square root estimate instructions are not generated by
25010 default on low-precision machines, since they do not provide an
25011 estimate that converges after three steps.
25013 @item -mveclibabi=@var{type}
25014 @opindex mveclibabi
25015 Specifies the ABI type to use for vectorizing intrinsics using an
25016 external library. The only type supported at present is @samp{mass},
25017 which specifies to use IBM's Mathematical Acceleration Subsystem
25018 (MASS) libraries for vectorizing intrinsics using external libraries.
25019 GCC currently emits calls to @code{acosd2}, @code{acosf4},
25020 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
25021 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
25022 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
25023 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
25024 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
25025 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
25026 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
25027 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
25028 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
25029 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
25030 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
25031 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
25032 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
25033 for power7. Both @option{-ftree-vectorize} and
25034 @option{-funsafe-math-optimizations} must also be enabled. The MASS
25035 libraries must be specified at link time.
25040 Generate (do not generate) the @code{friz} instruction when the
25041 @option{-funsafe-math-optimizations} option is used to optimize
25042 rounding of floating-point values to 64-bit integer and back to floating
25043 point. The @code{friz} instruction does not return the same value if
25044 the floating-point number is too large to fit in an integer.
25046 @item -mpointers-to-nested-functions
25047 @itemx -mno-pointers-to-nested-functions
25048 @opindex mpointers-to-nested-functions
25049 Generate (do not generate) code to load up the static chain register
25050 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
25051 systems where a function pointer points to a 3-word descriptor giving
25052 the function address, TOC value to be loaded in register @code{r2}, and
25053 static chain value to be loaded in register @code{r11}. The
25054 @option{-mpointers-to-nested-functions} is on by default. You cannot
25055 call through pointers to nested functions or pointers
25056 to functions compiled in other languages that use the static chain if
25057 you use @option{-mno-pointers-to-nested-functions}.
25059 @item -msave-toc-indirect
25060 @itemx -mno-save-toc-indirect
25061 @opindex msave-toc-indirect
25062 Generate (do not generate) code to save the TOC value in the reserved
25063 stack location in the function prologue if the function calls through
25064 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
25065 saved in the prologue, it is saved just before the call through the
25066 pointer. The @option{-mno-save-toc-indirect} option is the default.
25068 @item -mcompat-align-parm
25069 @itemx -mno-compat-align-parm
25070 @opindex mcompat-align-parm
25071 Generate (do not generate) code to pass structure parameters with a
25072 maximum alignment of 64 bits, for compatibility with older versions
25075 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
25076 structure parameter on a 128-bit boundary when that structure contained
25077 a member requiring 128-bit alignment. This is corrected in more
25078 recent versions of GCC. This option may be used to generate code
25079 that is compatible with functions compiled with older versions of
25082 The @option{-mno-compat-align-parm} option is the default.
25084 @item -mstack-protector-guard=@var{guard}
25085 @itemx -mstack-protector-guard-reg=@var{reg}
25086 @itemx -mstack-protector-guard-offset=@var{offset}
25087 @itemx -mstack-protector-guard-symbol=@var{symbol}
25088 @opindex mstack-protector-guard
25089 @opindex mstack-protector-guard-reg
25090 @opindex mstack-protector-guard-offset
25091 @opindex mstack-protector-guard-symbol
25092 Generate stack protection code using canary at @var{guard}. Supported
25093 locations are @samp{global} for global canary or @samp{tls} for per-thread
25094 canary in the TLS block (the default with GNU libc version 2.4 or later).
25096 With the latter choice the options
25097 @option{-mstack-protector-guard-reg=@var{reg}} and
25098 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
25099 which register to use as base register for reading the canary, and from what
25100 offset from that base register. The default for those is as specified in the
25101 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
25102 the offset with a symbol reference to a canary in the TLS block.
25106 @subsection RX Options
25109 These command-line options are defined for RX targets:
25112 @item -m64bit-doubles
25113 @itemx -m32bit-doubles
25114 @opindex m64bit-doubles
25115 @opindex m32bit-doubles
25116 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
25117 or 32 bits (@option{-m32bit-doubles}) in size. The default is
25118 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
25119 works on 32-bit values, which is why the default is
25120 @option{-m32bit-doubles}.
25126 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
25127 floating-point hardware. The default is enabled for the RX600
25128 series and disabled for the RX200 series.
25130 Floating-point instructions are only generated for 32-bit floating-point
25131 values, however, so the FPU hardware is not used for doubles if the
25132 @option{-m64bit-doubles} option is used.
25134 @emph{Note} If the @option{-fpu} option is enabled then
25135 @option{-funsafe-math-optimizations} is also enabled automatically.
25136 This is because the RX FPU instructions are themselves unsafe.
25138 @item -mcpu=@var{name}
25140 Selects the type of RX CPU to be targeted. Currently three types are
25141 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
25142 the specific @samp{RX610} CPU. The default is @samp{RX600}.
25144 The only difference between @samp{RX600} and @samp{RX610} is that the
25145 @samp{RX610} does not support the @code{MVTIPL} instruction.
25147 The @samp{RX200} series does not have a hardware floating-point unit
25148 and so @option{-nofpu} is enabled by default when this type is
25151 @item -mbig-endian-data
25152 @itemx -mlittle-endian-data
25153 @opindex mbig-endian-data
25154 @opindex mlittle-endian-data
25155 Store data (but not code) in the big-endian format. The default is
25156 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
25159 @item -msmall-data-limit=@var{N}
25160 @opindex msmall-data-limit
25161 Specifies the maximum size in bytes of global and static variables
25162 which can be placed into the small data area. Using the small data
25163 area can lead to smaller and faster code, but the size of area is
25164 limited and it is up to the programmer to ensure that the area does
25165 not overflow. Also when the small data area is used one of the RX's
25166 registers (usually @code{r13}) is reserved for use pointing to this
25167 area, so it is no longer available for use by the compiler. This
25168 could result in slower and/or larger code if variables are pushed onto
25169 the stack instead of being held in this register.
25171 Note, common variables (variables that have not been initialized) and
25172 constants are not placed into the small data area as they are assigned
25173 to other sections in the output executable.
25175 The default value is zero, which disables this feature. Note, this
25176 feature is not enabled by default with higher optimization levels
25177 (@option{-O2} etc) because of the potentially detrimental effects of
25178 reserving a register. It is up to the programmer to experiment and
25179 discover whether this feature is of benefit to their program. See the
25180 description of the @option{-mpid} option for a description of how the
25181 actual register to hold the small data area pointer is chosen.
25187 Use the simulator runtime. The default is to use the libgloss
25188 board-specific runtime.
25190 @item -mas100-syntax
25191 @itemx -mno-as100-syntax
25192 @opindex mas100-syntax
25193 @opindex mno-as100-syntax
25194 When generating assembler output use a syntax that is compatible with
25195 Renesas's AS100 assembler. This syntax can also be handled by the GAS
25196 assembler, but it has some restrictions so it is not generated by default.
25198 @item -mmax-constant-size=@var{N}
25199 @opindex mmax-constant-size
25200 Specifies the maximum size, in bytes, of a constant that can be used as
25201 an operand in a RX instruction. Although the RX instruction set does
25202 allow constants of up to 4 bytes in length to be used in instructions,
25203 a longer value equates to a longer instruction. Thus in some
25204 circumstances it can be beneficial to restrict the size of constants
25205 that are used in instructions. Constants that are too big are instead
25206 placed into a constant pool and referenced via register indirection.
25208 The value @var{N} can be between 0 and 4. A value of 0 (the default)
25209 or 4 means that constants of any size are allowed.
25213 Enable linker relaxation. Linker relaxation is a process whereby the
25214 linker attempts to reduce the size of a program by finding shorter
25215 versions of various instructions. Disabled by default.
25217 @item -mint-register=@var{N}
25218 @opindex mint-register
25219 Specify the number of registers to reserve for fast interrupt handler
25220 functions. The value @var{N} can be between 0 and 4. A value of 1
25221 means that register @code{r13} is reserved for the exclusive use
25222 of fast interrupt handlers. A value of 2 reserves @code{r13} and
25223 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
25224 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
25225 A value of 0, the default, does not reserve any registers.
25227 @item -msave-acc-in-interrupts
25228 @opindex msave-acc-in-interrupts
25229 Specifies that interrupt handler functions should preserve the
25230 accumulator register. This is only necessary if normal code might use
25231 the accumulator register, for example because it performs 64-bit
25232 multiplications. The default is to ignore the accumulator as this
25233 makes the interrupt handlers faster.
25239 Enables the generation of position independent data. When enabled any
25240 access to constant data is done via an offset from a base address
25241 held in a register. This allows the location of constant data to be
25242 determined at run time without requiring the executable to be
25243 relocated, which is a benefit to embedded applications with tight
25244 memory constraints. Data that can be modified is not affected by this
25247 Note, using this feature reserves a register, usually @code{r13}, for
25248 the constant data base address. This can result in slower and/or
25249 larger code, especially in complicated functions.
25251 The actual register chosen to hold the constant data base address
25252 depends upon whether the @option{-msmall-data-limit} and/or the
25253 @option{-mint-register} command-line options are enabled. Starting
25254 with register @code{r13} and proceeding downwards, registers are
25255 allocated first to satisfy the requirements of @option{-mint-register},
25256 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
25257 is possible for the small data area register to be @code{r8} if both
25258 @option{-mint-register=4} and @option{-mpid} are specified on the
25261 By default this feature is not enabled. The default can be restored
25262 via the @option{-mno-pid} command-line option.
25264 @item -mno-warn-multiple-fast-interrupts
25265 @itemx -mwarn-multiple-fast-interrupts
25266 @opindex mno-warn-multiple-fast-interrupts
25267 @opindex mwarn-multiple-fast-interrupts
25268 Prevents GCC from issuing a warning message if it finds more than one
25269 fast interrupt handler when it is compiling a file. The default is to
25270 issue a warning for each extra fast interrupt handler found, as the RX
25271 only supports one such interrupt.
25273 @item -mallow-string-insns
25274 @itemx -mno-allow-string-insns
25275 @opindex mallow-string-insns
25276 @opindex mno-allow-string-insns
25277 Enables or disables the use of the string manipulation instructions
25278 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
25279 @code{SWHILE} and also the @code{RMPA} instruction. These
25280 instructions may prefetch data, which is not safe to do if accessing
25281 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
25282 for more information).
25284 The default is to allow these instructions, but it is not possible for
25285 GCC to reliably detect all circumstances where a string instruction
25286 might be used to access an I/O register, so their use cannot be
25287 disabled automatically. Instead it is reliant upon the programmer to
25288 use the @option{-mno-allow-string-insns} option if their program
25289 accesses I/O space.
25291 When the instructions are enabled GCC defines the C preprocessor
25292 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
25293 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
25299 Use only (or not only) @code{JSR} instructions to access functions.
25300 This option can be used when code size exceeds the range of @code{BSR}
25301 instructions. Note that @option{-mno-jsr} does not mean to not use
25302 @code{JSR} but instead means that any type of branch may be used.
25305 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
25306 has special significance to the RX port when used with the
25307 @code{interrupt} function attribute. This attribute indicates a
25308 function intended to process fast interrupts. GCC ensures
25309 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
25310 and/or @code{r13} and only provided that the normal use of the
25311 corresponding registers have been restricted via the
25312 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
25315 @node S/390 and zSeries Options
25316 @subsection S/390 and zSeries Options
25317 @cindex S/390 and zSeries Options
25319 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
25323 @itemx -msoft-float
25324 @opindex mhard-float
25325 @opindex msoft-float
25326 Use (do not use) the hardware floating-point instructions and registers
25327 for floating-point operations. When @option{-msoft-float} is specified,
25328 functions in @file{libgcc.a} are used to perform floating-point
25329 operations. When @option{-mhard-float} is specified, the compiler
25330 generates IEEE floating-point instructions. This is the default.
25333 @itemx -mno-hard-dfp
25335 @opindex mno-hard-dfp
25336 Use (do not use) the hardware decimal-floating-point instructions for
25337 decimal-floating-point operations. When @option{-mno-hard-dfp} is
25338 specified, functions in @file{libgcc.a} are used to perform
25339 decimal-floating-point operations. When @option{-mhard-dfp} is
25340 specified, the compiler generates decimal-floating-point hardware
25341 instructions. This is the default for @option{-march=z9-ec} or higher.
25343 @item -mlong-double-64
25344 @itemx -mlong-double-128
25345 @opindex mlong-double-64
25346 @opindex mlong-double-128
25347 These switches control the size of @code{long double} type. A size
25348 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25349 type. This is the default.
25352 @itemx -mno-backchain
25353 @opindex mbackchain
25354 @opindex mno-backchain
25355 Store (do not store) the address of the caller's frame as backchain pointer
25356 into the callee's stack frame.
25357 A backchain may be needed to allow debugging using tools that do not understand
25358 DWARF call frame information.
25359 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
25360 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
25361 the backchain is placed into the topmost word of the 96/160 byte register
25364 In general, code compiled with @option{-mbackchain} is call-compatible with
25365 code compiled with @option{-mmo-backchain}; however, use of the backchain
25366 for debugging purposes usually requires that the whole binary is built with
25367 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
25368 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25369 to build a linux kernel use @option{-msoft-float}.
25371 The default is to not maintain the backchain.
25373 @item -mpacked-stack
25374 @itemx -mno-packed-stack
25375 @opindex mpacked-stack
25376 @opindex mno-packed-stack
25377 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
25378 specified, the compiler uses the all fields of the 96/160 byte register save
25379 area only for their default purpose; unused fields still take up stack space.
25380 When @option{-mpacked-stack} is specified, register save slots are densely
25381 packed at the top of the register save area; unused space is reused for other
25382 purposes, allowing for more efficient use of the available stack space.
25383 However, when @option{-mbackchain} is also in effect, the topmost word of
25384 the save area is always used to store the backchain, and the return address
25385 register is always saved two words below the backchain.
25387 As long as the stack frame backchain is not used, code generated with
25388 @option{-mpacked-stack} is call-compatible with code generated with
25389 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
25390 S/390 or zSeries generated code that uses the stack frame backchain at run
25391 time, not just for debugging purposes. Such code is not call-compatible
25392 with code compiled with @option{-mpacked-stack}. Also, note that the
25393 combination of @option{-mbackchain},
25394 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25395 to build a linux kernel use @option{-msoft-float}.
25397 The default is to not use the packed stack layout.
25400 @itemx -mno-small-exec
25401 @opindex msmall-exec
25402 @opindex mno-small-exec
25403 Generate (or do not generate) code using the @code{bras} instruction
25404 to do subroutine calls.
25405 This only works reliably if the total executable size does not
25406 exceed 64k. The default is to use the @code{basr} instruction instead,
25407 which does not have this limitation.
25413 When @option{-m31} is specified, generate code compliant to the
25414 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
25415 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
25416 particular to generate 64-bit instructions. For the @samp{s390}
25417 targets, the default is @option{-m31}, while the @samp{s390x}
25418 targets default to @option{-m64}.
25424 When @option{-mzarch} is specified, generate code using the
25425 instructions available on z/Architecture.
25426 When @option{-mesa} is specified, generate code using the
25427 instructions available on ESA/390. Note that @option{-mesa} is
25428 not possible with @option{-m64}.
25429 When generating code compliant to the GNU/Linux for S/390 ABI,
25430 the default is @option{-mesa}. When generating code compliant
25431 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
25437 The @option{-mhtm} option enables a set of builtins making use of
25438 instructions available with the transactional execution facility
25439 introduced with the IBM zEnterprise EC12 machine generation
25440 @ref{S/390 System z Built-in Functions}.
25441 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
25447 When @option{-mvx} is specified, generate code using the instructions
25448 available with the vector extension facility introduced with the IBM
25449 z13 machine generation.
25450 This option changes the ABI for some vector type values with regard to
25451 alignment and calling conventions. In case vector type values are
25452 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
25453 command will be added to mark the resulting binary with the ABI used.
25454 @option{-mvx} is enabled by default when using @option{-march=z13}.
25457 @itemx -mno-zvector
25459 @opindex mno-zvector
25460 The @option{-mzvector} option enables vector language extensions and
25461 builtins using instructions available with the vector extension
25462 facility introduced with the IBM z13 machine generation.
25463 This option adds support for @samp{vector} to be used as a keyword to
25464 define vector type variables and arguments. @samp{vector} is only
25465 available when GNU extensions are enabled. It will not be expanded
25466 when requesting strict standard compliance e.g.@: with @option{-std=c99}.
25467 In addition to the GCC low-level builtins @option{-mzvector} enables
25468 a set of builtins added for compatibility with AltiVec-style
25469 implementations like Power and Cell. In order to make use of these
25470 builtins the header file @file{vecintrin.h} needs to be included.
25471 @option{-mzvector} is disabled by default.
25477 Generate (or do not generate) code using the @code{mvcle} instruction
25478 to perform block moves. When @option{-mno-mvcle} is specified,
25479 use a @code{mvc} loop instead. This is the default unless optimizing for
25486 Print (or do not print) additional debug information when compiling.
25487 The default is to not print debug information.
25489 @item -march=@var{cpu-type}
25491 Generate code that runs on @var{cpu-type}, which is the name of a
25492 system representing a certain processor type. Possible values for
25493 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
25494 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
25495 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11}, and
25498 The default is @option{-march=z900}.
25500 Specifying @samp{native} as cpu type can be used to select the best
25501 architecture option for the host processor.
25502 @option{-march=native} has no effect if GCC does not recognize the
25505 @item -mtune=@var{cpu-type}
25507 Tune to @var{cpu-type} everything applicable about the generated code,
25508 except for the ABI and the set of available instructions.
25509 The list of @var{cpu-type} values is the same as for @option{-march}.
25510 The default is the value used for @option{-march}.
25513 @itemx -mno-tpf-trace
25514 @opindex mtpf-trace
25515 @opindex mno-tpf-trace
25516 Generate code that adds (does not add) in TPF OS specific branches to trace
25517 routines in the operating system. This option is off by default, even
25518 when compiling for the TPF OS@.
25521 @itemx -mno-fused-madd
25522 @opindex mfused-madd
25523 @opindex mno-fused-madd
25524 Generate code that uses (does not use) the floating-point multiply and
25525 accumulate instructions. These instructions are generated by default if
25526 hardware floating point is used.
25528 @item -mwarn-framesize=@var{framesize}
25529 @opindex mwarn-framesize
25530 Emit a warning if the current function exceeds the given frame size. Because
25531 this is a compile-time check it doesn't need to be a real problem when the program
25532 runs. It is intended to identify functions that most probably cause
25533 a stack overflow. It is useful to be used in an environment with limited stack
25534 size e.g.@: the linux kernel.
25536 @item -mwarn-dynamicstack
25537 @opindex mwarn-dynamicstack
25538 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
25539 arrays. This is generally a bad idea with a limited stack size.
25541 @item -mstack-guard=@var{stack-guard}
25542 @itemx -mstack-size=@var{stack-size}
25543 @opindex mstack-guard
25544 @opindex mstack-size
25545 If these options are provided the S/390 back end emits additional instructions in
25546 the function prologue that trigger a trap if the stack size is @var{stack-guard}
25547 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
25548 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
25549 the frame size of the compiled function is chosen.
25550 These options are intended to be used to help debugging stack overflow problems.
25551 The additionally emitted code causes only little overhead and hence can also be
25552 used in production-like systems without greater performance degradation. The given
25553 values have to be exact powers of 2 and @var{stack-size} has to be greater than
25554 @var{stack-guard} without exceeding 64k.
25555 In order to be efficient the extra code makes the assumption that the stack starts
25556 at an address aligned to the value given by @var{stack-size}.
25557 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
25559 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
25561 If the hotpatch option is enabled, a ``hot-patching'' function
25562 prologue is generated for all functions in the compilation unit.
25563 The funtion label is prepended with the given number of two-byte
25564 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
25565 the label, 2 * @var{post-halfwords} bytes are appended, using the
25566 largest NOP like instructions the architecture allows (maximum
25569 If both arguments are zero, hotpatching is disabled.
25571 This option can be overridden for individual functions with the
25572 @code{hotpatch} attribute.
25575 @node Score Options
25576 @subsection Score Options
25577 @cindex Score Options
25579 These options are defined for Score implementations:
25584 Compile code for big-endian mode. This is the default.
25588 Compile code for little-endian mode.
25592 Disable generation of @code{bcnz} instructions.
25596 Enable generation of unaligned load and store instructions.
25600 Enable the use of multiply-accumulate instructions. Disabled by default.
25604 Specify the SCORE5 as the target architecture.
25608 Specify the SCORE5U of the target architecture.
25612 Specify the SCORE7 as the target architecture. This is the default.
25616 Specify the SCORE7D as the target architecture.
25620 @subsection SH Options
25622 These @samp{-m} options are defined for the SH implementations:
25627 Generate code for the SH1.
25631 Generate code for the SH2.
25634 Generate code for the SH2e.
25638 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
25639 that the floating-point unit is not used.
25641 @item -m2a-single-only
25642 @opindex m2a-single-only
25643 Generate code for the SH2a-FPU, in such a way that no double-precision
25644 floating-point operations are used.
25647 @opindex m2a-single
25648 Generate code for the SH2a-FPU assuming the floating-point unit is in
25649 single-precision mode by default.
25653 Generate code for the SH2a-FPU assuming the floating-point unit is in
25654 double-precision mode by default.
25658 Generate code for the SH3.
25662 Generate code for the SH3e.
25666 Generate code for the SH4 without a floating-point unit.
25668 @item -m4-single-only
25669 @opindex m4-single-only
25670 Generate code for the SH4 with a floating-point unit that only
25671 supports single-precision arithmetic.
25675 Generate code for the SH4 assuming the floating-point unit is in
25676 single-precision mode by default.
25680 Generate code for the SH4.
25684 Generate code for SH4-100.
25686 @item -m4-100-nofpu
25687 @opindex m4-100-nofpu
25688 Generate code for SH4-100 in such a way that the
25689 floating-point unit is not used.
25691 @item -m4-100-single
25692 @opindex m4-100-single
25693 Generate code for SH4-100 assuming the floating-point unit is in
25694 single-precision mode by default.
25696 @item -m4-100-single-only
25697 @opindex m4-100-single-only
25698 Generate code for SH4-100 in such a way that no double-precision
25699 floating-point operations are used.
25703 Generate code for SH4-200.
25705 @item -m4-200-nofpu
25706 @opindex m4-200-nofpu
25707 Generate code for SH4-200 without in such a way that the
25708 floating-point unit is not used.
25710 @item -m4-200-single
25711 @opindex m4-200-single
25712 Generate code for SH4-200 assuming the floating-point unit is in
25713 single-precision mode by default.
25715 @item -m4-200-single-only
25716 @opindex m4-200-single-only
25717 Generate code for SH4-200 in such a way that no double-precision
25718 floating-point operations are used.
25722 Generate code for SH4-300.
25724 @item -m4-300-nofpu
25725 @opindex m4-300-nofpu
25726 Generate code for SH4-300 without in such a way that the
25727 floating-point unit is not used.
25729 @item -m4-300-single
25730 @opindex m4-300-single
25731 Generate code for SH4-300 in such a way that no double-precision
25732 floating-point operations are used.
25734 @item -m4-300-single-only
25735 @opindex m4-300-single-only
25736 Generate code for SH4-300 in such a way that no double-precision
25737 floating-point operations are used.
25741 Generate code for SH4-340 (no MMU, no FPU).
25745 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
25750 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
25751 floating-point unit is not used.
25753 @item -m4a-single-only
25754 @opindex m4a-single-only
25755 Generate code for the SH4a, in such a way that no double-precision
25756 floating-point operations are used.
25759 @opindex m4a-single
25760 Generate code for the SH4a assuming the floating-point unit is in
25761 single-precision mode by default.
25765 Generate code for the SH4a.
25769 Same as @option{-m4a-nofpu}, except that it implicitly passes
25770 @option{-dsp} to the assembler. GCC doesn't generate any DSP
25771 instructions at the moment.
25775 Compile code for the processor in big-endian mode.
25779 Compile code for the processor in little-endian mode.
25783 Align doubles at 64-bit boundaries. Note that this changes the calling
25784 conventions, and thus some functions from the standard C library do
25785 not work unless you recompile it first with @option{-mdalign}.
25789 Shorten some address references at link time, when possible; uses the
25790 linker option @option{-relax}.
25794 Use 32-bit offsets in @code{switch} tables. The default is to use
25799 Enable the use of bit manipulation instructions on SH2A.
25803 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
25804 alignment constraints.
25808 Comply with the calling conventions defined by Renesas.
25811 @opindex mno-renesas
25812 Comply with the calling conventions defined for GCC before the Renesas
25813 conventions were available. This option is the default for all
25814 targets of the SH toolchain.
25817 @opindex mnomacsave
25818 Mark the @code{MAC} register as call-clobbered, even if
25819 @option{-mrenesas} is given.
25825 Control the IEEE compliance of floating-point comparisons, which affects the
25826 handling of cases where the result of a comparison is unordered. By default
25827 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
25828 enabled @option{-mno-ieee} is implicitly set, which results in faster
25829 floating-point greater-equal and less-equal comparisons. The implicit settings
25830 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
25832 @item -minline-ic_invalidate
25833 @opindex minline-ic_invalidate
25834 Inline code to invalidate instruction cache entries after setting up
25835 nested function trampolines.
25836 This option has no effect if @option{-musermode} is in effect and the selected
25837 code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi}
25839 If the selected code generation option does not allow the use of the @code{icbi}
25840 instruction, and @option{-musermode} is not in effect, the inlined code
25841 manipulates the instruction cache address array directly with an associative
25842 write. This not only requires privileged mode at run time, but it also
25843 fails if the cache line had been mapped via the TLB and has become unmapped.
25847 Dump instruction size and location in the assembly code.
25850 @opindex mpadstruct
25851 This option is deprecated. It pads structures to multiple of 4 bytes,
25852 which is incompatible with the SH ABI@.
25854 @item -matomic-model=@var{model}
25855 @opindex matomic-model=@var{model}
25856 Sets the model of atomic operations and additional parameters as a comma
25857 separated list. For details on the atomic built-in functions see
25858 @ref{__atomic Builtins}. The following models and parameters are supported:
25863 Disable compiler generated atomic sequences and emit library calls for atomic
25864 operations. This is the default if the target is not @code{sh*-*-linux*}.
25867 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
25868 built-in functions. The generated atomic sequences require additional support
25869 from the interrupt/exception handling code of the system and are only suitable
25870 for SH3* and SH4* single-core systems. This option is enabled by default when
25871 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
25872 this option also partially utilizes the hardware atomic instructions
25873 @code{movli.l} and @code{movco.l} to create more efficient code, unless
25874 @samp{strict} is specified.
25877 Generate software atomic sequences that use a variable in the thread control
25878 block. This is a variation of the gUSA sequences which can also be used on
25879 SH1* and SH2* targets. The generated atomic sequences require additional
25880 support from the interrupt/exception handling code of the system and are only
25881 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
25882 parameter has to be specified as well.
25885 Generate software atomic sequences that temporarily disable interrupts by
25886 setting @code{SR.IMASK = 1111}. This model works only when the program runs
25887 in privileged mode and is only suitable for single-core systems. Additional
25888 support from the interrupt/exception handling code of the system is not
25889 required. This model is enabled by default when the target is
25890 @code{sh*-*-linux*} and SH1* or SH2*.
25893 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
25894 instructions only. This is only available on SH4A and is suitable for
25895 multi-core systems. Since the hardware instructions support only 32 bit atomic
25896 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
25897 Code compiled with this option is also compatible with other software
25898 atomic model interrupt/exception handling systems if executed on an SH4A
25899 system. Additional support from the interrupt/exception handling code of the
25900 system is not required for this model.
25903 This parameter specifies the offset in bytes of the variable in the thread
25904 control block structure that should be used by the generated atomic sequences
25905 when the @samp{soft-tcb} model has been selected. For other models this
25906 parameter is ignored. The specified value must be an integer multiple of four
25907 and in the range 0-1020.
25910 This parameter prevents mixed usage of multiple atomic models, even if they
25911 are compatible, and makes the compiler generate atomic sequences of the
25912 specified model only.
25918 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
25919 Notice that depending on the particular hardware and software configuration
25920 this can degrade overall performance due to the operand cache line flushes
25921 that are implied by the @code{tas.b} instruction. On multi-core SH4A
25922 processors the @code{tas.b} instruction must be used with caution since it
25923 can result in data corruption for certain cache configurations.
25926 @opindex mprefergot
25927 When generating position-independent code, emit function calls using
25928 the Global Offset Table instead of the Procedure Linkage Table.
25931 @itemx -mno-usermode
25933 @opindex mno-usermode
25934 Don't allow (allow) the compiler generating privileged mode code. Specifying
25935 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
25936 inlined code would not work in user mode. @option{-musermode} is the default
25937 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
25938 @option{-musermode} has no effect, since there is no user mode.
25940 @item -multcost=@var{number}
25941 @opindex multcost=@var{number}
25942 Set the cost to assume for a multiply insn.
25944 @item -mdiv=@var{strategy}
25945 @opindex mdiv=@var{strategy}
25946 Set the division strategy to be used for integer division operations.
25947 @var{strategy} can be one of:
25952 Calls a library function that uses the single-step division instruction
25953 @code{div1} to perform the operation. Division by zero calculates an
25954 unspecified result and does not trap. This is the default except for SH4,
25955 SH2A and SHcompact.
25958 Calls a library function that performs the operation in double precision
25959 floating point. Division by zero causes a floating-point exception. This is
25960 the default for SHcompact with FPU. Specifying this for targets that do not
25961 have a double precision FPU defaults to @code{call-div1}.
25964 Calls a library function that uses a lookup table for small divisors and
25965 the @code{div1} instruction with case distinction for larger divisors. Division
25966 by zero calculates an unspecified result and does not trap. This is the default
25967 for SH4. Specifying this for targets that do not have dynamic shift
25968 instructions defaults to @code{call-div1}.
25972 When a division strategy has not been specified the default strategy is
25973 selected based on the current target. For SH2A the default strategy is to
25974 use the @code{divs} and @code{divu} instructions instead of library function
25977 @item -maccumulate-outgoing-args
25978 @opindex maccumulate-outgoing-args
25979 Reserve space once for outgoing arguments in the function prologue rather
25980 than around each call. Generally beneficial for performance and size. Also
25981 needed for unwinding to avoid changing the stack frame around conditional code.
25983 @item -mdivsi3_libfunc=@var{name}
25984 @opindex mdivsi3_libfunc=@var{name}
25985 Set the name of the library function used for 32-bit signed division to
25987 This only affects the name used in the @samp{call} division strategies, and
25988 the compiler still expects the same sets of input/output/clobbered registers as
25989 if this option were not present.
25991 @item -mfixed-range=@var{register-range}
25992 @opindex mfixed-range
25993 Generate code treating the given register range as fixed registers.
25994 A fixed register is one that the register allocator can not use. This is
25995 useful when compiling kernel code. A register range is specified as
25996 two registers separated by a dash. Multiple register ranges can be
25997 specified separated by a comma.
25999 @item -mbranch-cost=@var{num}
26000 @opindex mbranch-cost=@var{num}
26001 Assume @var{num} to be the cost for a branch instruction. Higher numbers
26002 make the compiler try to generate more branch-free code if possible.
26003 If not specified the value is selected depending on the processor type that
26004 is being compiled for.
26007 @itemx -mno-zdcbranch
26008 @opindex mzdcbranch
26009 @opindex mno-zdcbranch
26010 Assume (do not assume) that zero displacement conditional branch instructions
26011 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
26012 compiler prefers zero displacement branch code sequences. This is
26013 enabled by default when generating code for SH4 and SH4A. It can be explicitly
26014 disabled by specifying @option{-mno-zdcbranch}.
26016 @item -mcbranch-force-delay-slot
26017 @opindex mcbranch-force-delay-slot
26018 Force the usage of delay slots for conditional branches, which stuffs the delay
26019 slot with a @code{nop} if a suitable instruction cannot be found. By default
26020 this option is disabled. It can be enabled to work around hardware bugs as
26021 found in the original SH7055.
26024 @itemx -mno-fused-madd
26025 @opindex mfused-madd
26026 @opindex mno-fused-madd
26027 Generate code that uses (does not use) the floating-point multiply and
26028 accumulate instructions. These instructions are generated by default
26029 if hardware floating point is used. The machine-dependent
26030 @option{-mfused-madd} option is now mapped to the machine-independent
26031 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
26032 mapped to @option{-ffp-contract=off}.
26038 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
26039 and cosine approximations. The option @option{-mfsca} must be used in
26040 combination with @option{-funsafe-math-optimizations}. It is enabled by default
26041 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
26042 approximations even if @option{-funsafe-math-optimizations} is in effect.
26048 Allow or disallow the compiler to emit the @code{fsrra} instruction for
26049 reciprocal square root approximations. The option @option{-mfsrra} must be used
26050 in combination with @option{-funsafe-math-optimizations} and
26051 @option{-ffinite-math-only}. It is enabled by default when generating code for
26052 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
26053 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
26056 @item -mpretend-cmove
26057 @opindex mpretend-cmove
26058 Prefer zero-displacement conditional branches for conditional move instruction
26059 patterns. This can result in faster code on the SH4 processor.
26063 Generate code using the FDPIC ABI.
26067 @node Solaris 2 Options
26068 @subsection Solaris 2 Options
26069 @cindex Solaris 2 options
26071 These @samp{-m} options are supported on Solaris 2:
26074 @item -mclear-hwcap
26075 @opindex mclear-hwcap
26076 @option{-mclear-hwcap} tells the compiler to remove the hardware
26077 capabilities generated by the Solaris assembler. This is only necessary
26078 when object files use ISA extensions not supported by the current
26079 machine, but check at runtime whether or not to use them.
26081 @item -mimpure-text
26082 @opindex mimpure-text
26083 @option{-mimpure-text}, used in addition to @option{-shared}, tells
26084 the compiler to not pass @option{-z text} to the linker when linking a
26085 shared object. Using this option, you can link position-dependent
26086 code into a shared object.
26088 @option{-mimpure-text} suppresses the ``relocations remain against
26089 allocatable but non-writable sections'' linker error message.
26090 However, the necessary relocations trigger copy-on-write, and the
26091 shared object is not actually shared across processes. Instead of
26092 using @option{-mimpure-text}, you should compile all source code with
26093 @option{-fpic} or @option{-fPIC}.
26097 These switches are supported in addition to the above on Solaris 2:
26102 This is a synonym for @option{-pthread}.
26105 @node SPARC Options
26106 @subsection SPARC Options
26107 @cindex SPARC options
26109 These @samp{-m} options are supported on the SPARC:
26112 @item -mno-app-regs
26114 @opindex mno-app-regs
26116 Specify @option{-mapp-regs} to generate output using the global registers
26117 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
26118 global register 1, each global register 2 through 4 is then treated as an
26119 allocable register that is clobbered by function calls. This is the default.
26121 To be fully SVR4 ABI-compliant at the cost of some performance loss,
26122 specify @option{-mno-app-regs}. You should compile libraries and system
26123 software with this option.
26129 With @option{-mflat}, the compiler does not generate save/restore instructions
26130 and uses a ``flat'' or single register window model. This model is compatible
26131 with the regular register window model. The local registers and the input
26132 registers (0--5) are still treated as ``call-saved'' registers and are
26133 saved on the stack as needed.
26135 With @option{-mno-flat} (the default), the compiler generates save/restore
26136 instructions (except for leaf functions). This is the normal operating mode.
26139 @itemx -mhard-float
26141 @opindex mhard-float
26142 Generate output containing floating-point instructions. This is the
26146 @itemx -msoft-float
26148 @opindex msoft-float
26149 Generate output containing library calls for floating point.
26150 @strong{Warning:} the requisite libraries are not available for all SPARC
26151 targets. Normally the facilities of the machine's usual C compiler are
26152 used, but this cannot be done directly in cross-compilation. You must make
26153 your own arrangements to provide suitable library functions for
26154 cross-compilation. The embedded targets @samp{sparc-*-aout} and
26155 @samp{sparclite-*-*} do provide software floating-point support.
26157 @option{-msoft-float} changes the calling convention in the output file;
26158 therefore, it is only useful if you compile @emph{all} of a program with
26159 this option. In particular, you need to compile @file{libgcc.a}, the
26160 library that comes with GCC, with @option{-msoft-float} in order for
26163 @item -mhard-quad-float
26164 @opindex mhard-quad-float
26165 Generate output containing quad-word (long double) floating-point
26168 @item -msoft-quad-float
26169 @opindex msoft-quad-float
26170 Generate output containing library calls for quad-word (long double)
26171 floating-point instructions. The functions called are those specified
26172 in the SPARC ABI@. This is the default.
26174 As of this writing, there are no SPARC implementations that have hardware
26175 support for the quad-word floating-point instructions. They all invoke
26176 a trap handler for one of these instructions, and then the trap handler
26177 emulates the effect of the instruction. Because of the trap handler overhead,
26178 this is much slower than calling the ABI library routines. Thus the
26179 @option{-msoft-quad-float} option is the default.
26181 @item -mno-unaligned-doubles
26182 @itemx -munaligned-doubles
26183 @opindex mno-unaligned-doubles
26184 @opindex munaligned-doubles
26185 Assume that doubles have 8-byte alignment. This is the default.
26187 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
26188 alignment only if they are contained in another type, or if they have an
26189 absolute address. Otherwise, it assumes they have 4-byte alignment.
26190 Specifying this option avoids some rare compatibility problems with code
26191 generated by other compilers. It is not the default because it results
26192 in a performance loss, especially for floating-point code.
26195 @itemx -mno-user-mode
26196 @opindex muser-mode
26197 @opindex mno-user-mode
26198 Do not generate code that can only run in supervisor mode. This is relevant
26199 only for the @code{casa} instruction emitted for the LEON3 processor. This
26202 @item -mfaster-structs
26203 @itemx -mno-faster-structs
26204 @opindex mfaster-structs
26205 @opindex mno-faster-structs
26206 With @option{-mfaster-structs}, the compiler assumes that structures
26207 should have 8-byte alignment. This enables the use of pairs of
26208 @code{ldd} and @code{std} instructions for copies in structure
26209 assignment, in place of twice as many @code{ld} and @code{st} pairs.
26210 However, the use of this changed alignment directly violates the SPARC
26211 ABI@. Thus, it's intended only for use on targets where the developer
26212 acknowledges that their resulting code is not directly in line with
26213 the rules of the ABI@.
26215 @item -mstd-struct-return
26216 @itemx -mno-std-struct-return
26217 @opindex mstd-struct-return
26218 @opindex mno-std-struct-return
26219 With @option{-mstd-struct-return}, the compiler generates checking code
26220 in functions returning structures or unions to detect size mismatches
26221 between the two sides of function calls, as per the 32-bit ABI@.
26223 The default is @option{-mno-std-struct-return}. This option has no effect
26230 Enable Local Register Allocation. This is the default for SPARC since GCC 7
26231 so @option{-mno-lra} needs to be passed to get old Reload.
26233 @item -mcpu=@var{cpu_type}
26235 Set the instruction set, register set, and instruction scheduling parameters
26236 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26237 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
26238 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
26239 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
26240 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
26241 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
26243 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
26244 which selects the best architecture option for the host processor.
26245 @option{-mcpu=native} has no effect if GCC does not recognize
26248 Default instruction scheduling parameters are used for values that select
26249 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
26250 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
26252 Here is a list of each supported architecture and their supported
26260 supersparc, hypersparc, leon, leon3
26263 f930, f934, sparclite86x
26269 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
26273 By default (unless configured otherwise), GCC generates code for the V7
26274 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
26275 additionally optimizes it for the Cypress CY7C602 chip, as used in the
26276 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
26277 SPARCStation 1, 2, IPX etc.
26279 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
26280 architecture. The only difference from V7 code is that the compiler emits
26281 the integer multiply and integer divide instructions which exist in SPARC-V8
26282 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
26283 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
26286 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
26287 the SPARC architecture. This adds the integer multiply, integer divide step
26288 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
26289 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
26290 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
26291 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
26292 MB86934 chip, which is the more recent SPARClite with FPU@.
26294 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
26295 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
26296 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
26297 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
26298 optimizes it for the TEMIC SPARClet chip.
26300 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
26301 architecture. This adds 64-bit integer and floating-point move instructions,
26302 3 additional floating-point condition code registers and conditional move
26303 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
26304 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
26305 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
26306 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
26307 @option{-mcpu=niagara}, the compiler additionally optimizes it for
26308 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
26309 additionally optimizes it for Sun UltraSPARC T2 chips. With
26310 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
26311 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
26312 additionally optimizes it for Sun UltraSPARC T4 chips. With
26313 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
26314 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
26315 additionally optimizes it for Oracle M8 chips.
26317 @item -mtune=@var{cpu_type}
26319 Set the instruction scheduling parameters for machine type
26320 @var{cpu_type}, but do not set the instruction set or register set that the
26321 option @option{-mcpu=@var{cpu_type}} does.
26323 The same values for @option{-mcpu=@var{cpu_type}} can be used for
26324 @option{-mtune=@var{cpu_type}}, but the only useful values are those
26325 that select a particular CPU implementation. Those are
26326 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
26327 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
26328 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
26329 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
26330 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
26331 and GNU/Linux toolchains, @samp{native} can also be used.
26336 @opindex mno-v8plus
26337 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
26338 difference from the V8 ABI is that the global and out registers are
26339 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
26340 mode for all SPARC-V9 processors.
26346 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
26347 Visual Instruction Set extensions. The default is @option{-mno-vis}.
26353 With @option{-mvis2}, GCC generates code that takes advantage of
26354 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
26355 default is @option{-mvis2} when targeting a cpu that supports such
26356 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
26357 also sets @option{-mvis}.
26363 With @option{-mvis3}, GCC generates code that takes advantage of
26364 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
26365 default is @option{-mvis3} when targeting a cpu that supports such
26366 instructions, such as niagara-3 and later. Setting @option{-mvis3}
26367 also sets @option{-mvis2} and @option{-mvis}.
26373 With @option{-mvis4}, GCC generates code that takes advantage of
26374 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
26375 default is @option{-mvis4} when targeting a cpu that supports such
26376 instructions, such as niagara-7 and later. Setting @option{-mvis4}
26377 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
26383 With @option{-mvis4b}, GCC generates code that takes advantage of
26384 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
26385 the additional VIS instructions introduced in the Oracle SPARC
26386 Architecture 2017. The default is @option{-mvis4b} when targeting a
26387 cpu that supports such instructions, such as m8 and later. Setting
26388 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
26389 @option{-mvis2} and @option{-mvis}.
26394 @opindex mno-cbcond
26395 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
26396 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
26397 when targeting a CPU that supports such instructions, such as Niagara-4 and
26404 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
26405 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
26406 when targeting a CPU that supports such instructions, such as Niagara-3 and
26412 @opindex mno-fsmuld
26413 With @option{-mfsmuld}, GCC generates code that takes advantage of the
26414 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
26415 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
26416 or V9 with FPU except @option{-mcpu=leon}.
26422 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
26423 Population Count instruction. The default is @option{-mpopc}
26424 when targeting a CPU that supports such an instruction, such as Niagara-2 and
26431 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
26432 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
26433 when targeting a CPU that supports such an instruction, such as Niagara-7 and
26437 @opindex mfix-at697f
26438 Enable the documented workaround for the single erratum of the Atmel AT697F
26439 processor (which corresponds to erratum #13 of the AT697E processor).
26442 @opindex mfix-ut699
26443 Enable the documented workarounds for the floating-point errata and the data
26444 cache nullify errata of the UT699 processor.
26447 @opindex mfix-ut700
26448 Enable the documented workaround for the back-to-back store errata of
26449 the UT699E/UT700 processor.
26451 @item -mfix-gr712rc
26452 @opindex mfix-gr712rc
26453 Enable the documented workaround for the back-to-back store errata of
26454 the GR712RC processor.
26457 These @samp{-m} options are supported in addition to the above
26458 on SPARC-V9 processors in 64-bit environments:
26465 Generate code for a 32-bit or 64-bit environment.
26466 The 32-bit environment sets int, long and pointer to 32 bits.
26467 The 64-bit environment sets int to 32 bits and long and pointer
26470 @item -mcmodel=@var{which}
26472 Set the code model to one of
26476 The Medium/Low code model: 64-bit addresses, programs
26477 must be linked in the low 32 bits of memory. Programs can be statically
26478 or dynamically linked.
26481 The Medium/Middle code model: 64-bit addresses, programs
26482 must be linked in the low 44 bits of memory, the text and data segments must
26483 be less than 2GB in size and the data segment must be located within 2GB of
26487 The Medium/Anywhere code model: 64-bit addresses, programs
26488 may be linked anywhere in memory, the text and data segments must be less
26489 than 2GB in size and the data segment must be located within 2GB of the
26493 The Medium/Anywhere code model for embedded systems:
26494 64-bit addresses, the text and data segments must be less than 2GB in
26495 size, both starting anywhere in memory (determined at link time). The
26496 global register %g4 points to the base of the data segment. Programs
26497 are statically linked and PIC is not supported.
26500 @item -mmemory-model=@var{mem-model}
26501 @opindex mmemory-model
26502 Set the memory model in force on the processor to one of
26506 The default memory model for the processor and operating system.
26509 Relaxed Memory Order
26512 Partial Store Order
26518 Sequential Consistency
26521 These memory models are formally defined in Appendix D of the SPARC-V9
26522 architecture manual, as set in the processor's @code{PSTATE.MM} field.
26525 @itemx -mno-stack-bias
26526 @opindex mstack-bias
26527 @opindex mno-stack-bias
26528 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
26529 frame pointer if present, are offset by @minus{}2047 which must be added back
26530 when making stack frame references. This is the default in 64-bit mode.
26531 Otherwise, assume no such offset is present.
26535 @subsection SPU Options
26536 @cindex SPU options
26538 These @samp{-m} options are supported on the SPU:
26542 @itemx -merror-reloc
26543 @opindex mwarn-reloc
26544 @opindex merror-reloc
26546 The loader for SPU does not handle dynamic relocations. By default, GCC
26547 gives an error when it generates code that requires a dynamic
26548 relocation. @option{-mno-error-reloc} disables the error,
26549 @option{-mwarn-reloc} generates a warning instead.
26552 @itemx -munsafe-dma
26554 @opindex munsafe-dma
26556 Instructions that initiate or test completion of DMA must not be
26557 reordered with respect to loads and stores of the memory that is being
26559 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
26560 memory accesses, but that can lead to inefficient code in places where the
26561 memory is known to not change. Rather than mark the memory as volatile,
26562 you can use @option{-msafe-dma} to tell the compiler to treat
26563 the DMA instructions as potentially affecting all memory.
26565 @item -mbranch-hints
26566 @opindex mbranch-hints
26568 By default, GCC generates a branch hint instruction to avoid
26569 pipeline stalls for always-taken or probably-taken branches. A hint
26570 is not generated closer than 8 instructions away from its branch.
26571 There is little reason to disable them, except for debugging purposes,
26572 or to make an object a little bit smaller.
26576 @opindex msmall-mem
26577 @opindex mlarge-mem
26579 By default, GCC generates code assuming that addresses are never larger
26580 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
26581 a full 32-bit address.
26586 By default, GCC links against startup code that assumes the SPU-style
26587 main function interface (which has an unconventional parameter list).
26588 With @option{-mstdmain}, GCC links your program against startup
26589 code that assumes a C99-style interface to @code{main}, including a
26590 local copy of @code{argv} strings.
26592 @item -mfixed-range=@var{register-range}
26593 @opindex mfixed-range
26594 Generate code treating the given register range as fixed registers.
26595 A fixed register is one that the register allocator cannot use. This is
26596 useful when compiling kernel code. A register range is specified as
26597 two registers separated by a dash. Multiple register ranges can be
26598 specified separated by a comma.
26604 Compile code assuming that pointers to the PPU address space accessed
26605 via the @code{__ea} named address space qualifier are either 32 or 64
26606 bits wide. The default is 32 bits. As this is an ABI-changing option,
26607 all object code in an executable must be compiled with the same setting.
26609 @item -maddress-space-conversion
26610 @itemx -mno-address-space-conversion
26611 @opindex maddress-space-conversion
26612 @opindex mno-address-space-conversion
26613 Allow/disallow treating the @code{__ea} address space as superset
26614 of the generic address space. This enables explicit type casts
26615 between @code{__ea} and generic pointer as well as implicit
26616 conversions of generic pointers to @code{__ea} pointers. The
26617 default is to allow address space pointer conversions.
26619 @item -mcache-size=@var{cache-size}
26620 @opindex mcache-size
26621 This option controls the version of libgcc that the compiler links to an
26622 executable and selects a software-managed cache for accessing variables
26623 in the @code{__ea} address space with a particular cache size. Possible
26624 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
26625 and @samp{128}. The default cache size is 64KB.
26627 @item -matomic-updates
26628 @itemx -mno-atomic-updates
26629 @opindex matomic-updates
26630 @opindex mno-atomic-updates
26631 This option controls the version of libgcc that the compiler links to an
26632 executable and selects whether atomic updates to the software-managed
26633 cache of PPU-side variables are used. If you use atomic updates, changes
26634 to a PPU variable from SPU code using the @code{__ea} named address space
26635 qualifier do not interfere with changes to other PPU variables residing
26636 in the same cache line from PPU code. If you do not use atomic updates,
26637 such interference may occur; however, writing back cache lines is
26638 more efficient. The default behavior is to use atomic updates.
26641 @itemx -mdual-nops=@var{n}
26642 @opindex mdual-nops
26643 By default, GCC inserts NOPs to increase dual issue when it expects
26644 it to increase performance. @var{n} can be a value from 0 to 10. A
26645 smaller @var{n} inserts fewer NOPs. 10 is the default, 0 is the
26646 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
26648 @item -mhint-max-nops=@var{n}
26649 @opindex mhint-max-nops
26650 Maximum number of NOPs to insert for a branch hint. A branch hint must
26651 be at least 8 instructions away from the branch it is affecting. GCC
26652 inserts up to @var{n} NOPs to enforce this, otherwise it does not
26653 generate the branch hint.
26655 @item -mhint-max-distance=@var{n}
26656 @opindex mhint-max-distance
26657 The encoding of the branch hint instruction limits the hint to be within
26658 256 instructions of the branch it is affecting. By default, GCC makes
26659 sure it is within 125.
26662 @opindex msafe-hints
26663 Work around a hardware bug that causes the SPU to stall indefinitely.
26664 By default, GCC inserts the @code{hbrp} instruction to make sure
26665 this stall won't happen.
26669 @node System V Options
26670 @subsection Options for System V
26672 These additional options are available on System V Release 4 for
26673 compatibility with other compilers on those systems:
26678 Create a shared object.
26679 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
26683 Identify the versions of each tool used by the compiler, in a
26684 @code{.ident} assembler directive in the output.
26688 Refrain from adding @code{.ident} directives to the output file (this is
26691 @item -YP,@var{dirs}
26693 Search the directories @var{dirs}, and no others, for libraries
26694 specified with @option{-l}.
26696 @item -Ym,@var{dir}
26698 Look in the directory @var{dir} to find the M4 preprocessor.
26699 The assembler uses this option.
26700 @c This is supposed to go with a -Yd for predefined M4 macro files, but
26701 @c the generic assembler that comes with Solaris takes just -Ym.
26704 @node TILE-Gx Options
26705 @subsection TILE-Gx Options
26706 @cindex TILE-Gx options
26708 These @samp{-m} options are supported on the TILE-Gx:
26711 @item -mcmodel=small
26712 @opindex mcmodel=small
26713 Generate code for the small model. The distance for direct calls is
26714 limited to 500M in either direction. PC-relative addresses are 32
26715 bits. Absolute addresses support the full address range.
26717 @item -mcmodel=large
26718 @opindex mcmodel=large
26719 Generate code for the large model. There is no limitation on call
26720 distance, pc-relative addresses, or absolute addresses.
26722 @item -mcpu=@var{name}
26724 Selects the type of CPU to be targeted. Currently the only supported
26725 type is @samp{tilegx}.
26731 Generate code for a 32-bit or 64-bit environment. The 32-bit
26732 environment sets int, long, and pointer to 32 bits. The 64-bit
26733 environment sets int to 32 bits and long and pointer to 64 bits.
26736 @itemx -mlittle-endian
26737 @opindex mbig-endian
26738 @opindex mlittle-endian
26739 Generate code in big/little endian mode, respectively.
26742 @node TILEPro Options
26743 @subsection TILEPro Options
26744 @cindex TILEPro options
26746 These @samp{-m} options are supported on the TILEPro:
26749 @item -mcpu=@var{name}
26751 Selects the type of CPU to be targeted. Currently the only supported
26752 type is @samp{tilepro}.
26756 Generate code for a 32-bit environment, which sets int, long, and
26757 pointer to 32 bits. This is the only supported behavior so the flag
26758 is essentially ignored.
26762 @subsection V850 Options
26763 @cindex V850 Options
26765 These @samp{-m} options are defined for V850 implementations:
26769 @itemx -mno-long-calls
26770 @opindex mlong-calls
26771 @opindex mno-long-calls
26772 Treat all calls as being far away (near). If calls are assumed to be
26773 far away, the compiler always loads the function's address into a
26774 register, and calls indirect through the pointer.
26780 Do not optimize (do optimize) basic blocks that use the same index
26781 pointer 4 or more times to copy pointer into the @code{ep} register, and
26782 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26783 option is on by default if you optimize.
26785 @item -mno-prolog-function
26786 @itemx -mprolog-function
26787 @opindex mno-prolog-function
26788 @opindex mprolog-function
26789 Do not use (do use) external functions to save and restore registers
26790 at the prologue and epilogue of a function. The external functions
26791 are slower, but use less code space if more than one function saves
26792 the same number of registers. The @option{-mprolog-function} option
26793 is on by default if you optimize.
26797 Try to make the code as small as possible. At present, this just turns
26798 on the @option{-mep} and @option{-mprolog-function} options.
26800 @item -mtda=@var{n}
26802 Put static or global variables whose size is @var{n} bytes or less into
26803 the tiny data area that register @code{ep} points to. The tiny data
26804 area can hold up to 256 bytes in total (128 bytes for byte references).
26806 @item -msda=@var{n}
26808 Put static or global variables whose size is @var{n} bytes or less into
26809 the small data area that register @code{gp} points to. The small data
26810 area can hold up to 64 kilobytes.
26812 @item -mzda=@var{n}
26814 Put static or global variables whose size is @var{n} bytes or less into
26815 the first 32 kilobytes of memory.
26819 Specify that the target processor is the V850.
26823 Specify that the target processor is the V850E3V5. The preprocessor
26824 constant @code{__v850e3v5__} is defined if this option is used.
26828 Specify that the target processor is the V850E3V5. This is an alias for
26829 the @option{-mv850e3v5} option.
26833 Specify that the target processor is the V850E2V3. The preprocessor
26834 constant @code{__v850e2v3__} is defined if this option is used.
26838 Specify that the target processor is the V850E2. The preprocessor
26839 constant @code{__v850e2__} is defined if this option is used.
26843 Specify that the target processor is the V850E1. The preprocessor
26844 constants @code{__v850e1__} and @code{__v850e__} are defined if
26845 this option is used.
26849 Specify that the target processor is the V850ES. This is an alias for
26850 the @option{-mv850e1} option.
26854 Specify that the target processor is the V850E@. The preprocessor
26855 constant @code{__v850e__} is defined if this option is used.
26857 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26858 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26859 are defined then a default target processor is chosen and the
26860 relevant @samp{__v850*__} preprocessor constant is defined.
26862 The preprocessor constants @code{__v850} and @code{__v851__} are always
26863 defined, regardless of which processor variant is the target.
26865 @item -mdisable-callt
26866 @itemx -mno-disable-callt
26867 @opindex mdisable-callt
26868 @opindex mno-disable-callt
26869 This option suppresses generation of the @code{CALLT} instruction for the
26870 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
26873 This option is enabled by default when the RH850 ABI is
26874 in use (see @option{-mrh850-abi}), and disabled by default when the
26875 GCC ABI is in use. If @code{CALLT} instructions are being generated
26876 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
26882 Pass on (or do not pass on) the @option{-mrelax} command-line option
26886 @itemx -mno-long-jumps
26887 @opindex mlong-jumps
26888 @opindex mno-long-jumps
26889 Disable (or re-enable) the generation of PC-relative jump instructions.
26892 @itemx -mhard-float
26893 @opindex msoft-float
26894 @opindex mhard-float
26895 Disable (or re-enable) the generation of hardware floating point
26896 instructions. This option is only significant when the target
26897 architecture is @samp{V850E2V3} or higher. If hardware floating point
26898 instructions are being generated then the C preprocessor symbol
26899 @code{__FPU_OK__} is defined, otherwise the symbol
26900 @code{__NO_FPU__} is defined.
26904 Enables the use of the e3v5 LOOP instruction. The use of this
26905 instruction is not enabled by default when the e3v5 architecture is
26906 selected because its use is still experimental.
26910 @opindex mrh850-abi
26912 Enables support for the RH850 version of the V850 ABI. This is the
26913 default. With this version of the ABI the following rules apply:
26917 Integer sized structures and unions are returned via a memory pointer
26918 rather than a register.
26921 Large structures and unions (more than 8 bytes in size) are passed by
26925 Functions are aligned to 16-bit boundaries.
26928 The @option{-m8byte-align} command-line option is supported.
26931 The @option{-mdisable-callt} command-line option is enabled by
26932 default. The @option{-mno-disable-callt} command-line option is not
26936 When this version of the ABI is enabled the C preprocessor symbol
26937 @code{__V850_RH850_ABI__} is defined.
26941 Enables support for the old GCC version of the V850 ABI. With this
26942 version of the ABI the following rules apply:
26946 Integer sized structures and unions are returned in register @code{r10}.
26949 Large structures and unions (more than 8 bytes in size) are passed by
26953 Functions are aligned to 32-bit boundaries, unless optimizing for
26957 The @option{-m8byte-align} command-line option is not supported.
26960 The @option{-mdisable-callt} command-line option is supported but not
26961 enabled by default.
26964 When this version of the ABI is enabled the C preprocessor symbol
26965 @code{__V850_GCC_ABI__} is defined.
26967 @item -m8byte-align
26968 @itemx -mno-8byte-align
26969 @opindex m8byte-align
26970 @opindex mno-8byte-align
26971 Enables support for @code{double} and @code{long long} types to be
26972 aligned on 8-byte boundaries. The default is to restrict the
26973 alignment of all objects to at most 4-bytes. When
26974 @option{-m8byte-align} is in effect the C preprocessor symbol
26975 @code{__V850_8BYTE_ALIGN__} is defined.
26978 @opindex mbig-switch
26979 Generate code suitable for big switch tables. Use this option only if
26980 the assembler/linker complain about out of range branches within a switch
26985 This option causes r2 and r5 to be used in the code generated by
26986 the compiler. This setting is the default.
26988 @item -mno-app-regs
26989 @opindex mno-app-regs
26990 This option causes r2 and r5 to be treated as fixed registers.
26995 @subsection VAX Options
26996 @cindex VAX options
26998 These @samp{-m} options are defined for the VAX:
27003 Do not output certain jump instructions (@code{aobleq} and so on)
27004 that the Unix assembler for the VAX cannot handle across long
27009 Do output those jump instructions, on the assumption that the
27010 GNU assembler is being used.
27014 Output code for G-format floating-point numbers instead of D-format.
27017 @node Visium Options
27018 @subsection Visium Options
27019 @cindex Visium options
27025 A program which performs file I/O and is destined to run on an MCM target
27026 should be linked with this option. It causes the libraries libc.a and
27027 libdebug.a to be linked. The program should be run on the target under
27028 the control of the GDB remote debugging stub.
27032 A program which performs file I/O and is destined to run on the simulator
27033 should be linked with option. This causes libraries libc.a and libsim.a to
27037 @itemx -mhard-float
27039 @opindex mhard-float
27040 Generate code containing floating-point instructions. This is the
27044 @itemx -msoft-float
27046 @opindex msoft-float
27047 Generate code containing library calls for floating-point.
27049 @option{-msoft-float} changes the calling convention in the output file;
27050 therefore, it is only useful if you compile @emph{all} of a program with
27051 this option. In particular, you need to compile @file{libgcc.a}, the
27052 library that comes with GCC, with @option{-msoft-float} in order for
27055 @item -mcpu=@var{cpu_type}
27057 Set the instruction set, register set, and instruction scheduling parameters
27058 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
27059 @samp{mcm}, @samp{gr5} and @samp{gr6}.
27061 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
27063 By default (unless configured otherwise), GCC generates code for the GR5
27064 variant of the Visium architecture.
27066 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
27067 architecture. The only difference from GR5 code is that the compiler will
27068 generate block move instructions.
27070 @item -mtune=@var{cpu_type}
27072 Set the instruction scheduling parameters for machine type @var{cpu_type},
27073 but do not set the instruction set or register set that the option
27074 @option{-mcpu=@var{cpu_type}} would.
27078 Generate code for the supervisor mode, where there are no restrictions on
27079 the access to general registers. This is the default.
27082 @opindex muser-mode
27083 Generate code for the user mode, where the access to some general registers
27084 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
27085 mode; on the GR6, only registers r29 to r31 are affected.
27089 @subsection VMS Options
27091 These @samp{-m} options are defined for the VMS implementations:
27094 @item -mvms-return-codes
27095 @opindex mvms-return-codes
27096 Return VMS condition codes from @code{main}. The default is to return POSIX-style
27097 condition (e.g.@: error) codes.
27099 @item -mdebug-main=@var{prefix}
27100 @opindex mdebug-main=@var{prefix}
27101 Flag the first routine whose name starts with @var{prefix} as the main
27102 routine for the debugger.
27106 Default to 64-bit memory allocation routines.
27108 @item -mpointer-size=@var{size}
27109 @opindex mpointer-size=@var{size}
27110 Set the default size of pointers. Possible options for @var{size} are
27111 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
27112 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
27113 The later option disables @code{pragma pointer_size}.
27116 @node VxWorks Options
27117 @subsection VxWorks Options
27118 @cindex VxWorks Options
27120 The options in this section are defined for all VxWorks targets.
27121 Options specific to the target hardware are listed with the other
27122 options for that target.
27127 GCC can generate code for both VxWorks kernels and real time processes
27128 (RTPs). This option switches from the former to the latter. It also
27129 defines the preprocessor macro @code{__RTP__}.
27132 @opindex non-static
27133 Link an RTP executable against shared libraries rather than static
27134 libraries. The options @option{-static} and @option{-shared} can
27135 also be used for RTPs (@pxref{Link Options}); @option{-static}
27142 These options are passed down to the linker. They are defined for
27143 compatibility with Diab.
27146 @opindex Xbind-lazy
27147 Enable lazy binding of function calls. This option is equivalent to
27148 @option{-Wl,-z,now} and is defined for compatibility with Diab.
27152 Disable lazy binding of function calls. This option is the default and
27153 is defined for compatibility with Diab.
27157 @subsection x86 Options
27158 @cindex x86 Options
27160 These @samp{-m} options are defined for the x86 family of computers.
27164 @item -march=@var{cpu-type}
27166 Generate instructions for the machine type @var{cpu-type}. In contrast to
27167 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
27168 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
27169 to generate code that may not run at all on processors other than the one
27170 indicated. Specifying @option{-march=@var{cpu-type}} implies
27171 @option{-mtune=@var{cpu-type}}.
27173 The choices for @var{cpu-type} are:
27177 This selects the CPU to generate code for at compilation time by determining
27178 the processor type of the compiling machine. Using @option{-march=native}
27179 enables all instruction subsets supported by the local machine (hence
27180 the result might not run on different machines). Using @option{-mtune=native}
27181 produces code optimized for the local machine under the constraints
27182 of the selected instruction set.
27185 A generic CPU with 64-bit extensions.
27188 Original Intel i386 CPU@.
27191 Intel i486 CPU@. (No scheduling is implemented for this chip.)
27195 Intel Pentium CPU with no MMX support.
27198 Intel Lakemont MCU, based on Intel Pentium CPU.
27201 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
27204 Intel Pentium Pro CPU@.
27207 When used with @option{-march}, the Pentium Pro
27208 instruction set is used, so the code runs on all i686 family chips.
27209 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
27212 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
27217 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
27221 Intel Pentium M; low-power version of Intel Pentium III CPU
27222 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
27226 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
27229 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
27233 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
27234 SSE2 and SSE3 instruction set support.
27237 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
27238 instruction set support.
27241 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27242 SSE4.1, SSE4.2 and POPCNT instruction set support.
27245 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27246 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
27249 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27250 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
27253 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27254 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
27255 instruction set support.
27258 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27259 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27260 BMI, BMI2 and F16C instruction set support.
27263 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27264 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27265 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
27268 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27269 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27270 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
27271 XSAVES instruction set support.
27274 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
27275 instruction set support.
27278 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27279 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
27282 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27283 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT and FSGSBASE
27284 instruction set support.
27286 @item goldmont-plus
27287 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27288 SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE,
27289 PTWRITE, RDPID, SGX and UMIP instruction set support.
27292 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27293 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE,
27294 RDPID, SGX, UMIP, GFNI-SSE, CLWB and ENCLV instruction set support.
27297 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27298 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27299 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
27300 AVX512CD instruction set support.
27303 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27304 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27305 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27306 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
27308 @item skylake-avx512
27309 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27310 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27311 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27312 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
27315 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27316 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27317 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27318 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27319 AVX512IFMA, SHA and UMIP instruction set support.
27321 @item icelake-client
27322 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27323 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27324 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27325 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27326 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27327 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
27329 @item icelake-server
27330 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27331 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27332 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27333 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27334 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27335 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
27339 AMD K6 CPU with MMX instruction set support.
27343 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
27346 @itemx athlon-tbird
27347 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
27353 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
27354 instruction set support.
27360 Processors based on the AMD K8 core with x86-64 instruction set support,
27361 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
27362 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
27363 instruction set extensions.)
27366 @itemx opteron-sse3
27367 @itemx athlon64-sse3
27368 Improved versions of AMD K8 cores with SSE3 instruction set support.
27372 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
27373 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
27374 instruction set extensions.)
27377 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
27378 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
27379 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
27381 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27382 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
27383 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
27386 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27387 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
27388 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
27389 64-bit instruction set extensions.
27391 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27392 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
27393 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
27394 SSE4.2, ABM and 64-bit instruction set extensions.
27397 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27398 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
27399 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
27400 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27401 instruction set extensions.
27403 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27404 supersets BMI, BMI2, ,CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED,
27405 MWAITX, SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A,
27406 SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27407 instruction set extensions.)
27411 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
27412 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
27413 instruction set extensions.)
27416 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
27417 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
27418 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
27421 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
27425 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
27426 instruction set support.
27429 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
27430 (No scheduling is implemented for this chip.)
27433 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
27434 (No scheduling is implemented for this chip.)
27437 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27438 (No scheduling is implemented for this chip.)
27441 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
27442 (No scheduling is implemented for this chip.)
27445 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
27446 (No scheduling is implemented for this chip.)
27449 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27450 (No scheduling is implemented for this chip.)
27453 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
27454 (No scheduling is implemented for this chip.)
27457 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
27458 AVX and AVX2 instruction set support.
27459 (No scheduling is implemented for this chip.)
27462 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27463 instruction set support.
27464 (No scheduling is implemented for this chip.)
27467 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27468 instruction set support.
27469 (No scheduling is implemented for this chip.)
27472 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27473 instruction set support.
27474 (No scheduling is implemented for this chip.)
27477 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27478 instruction set support.
27479 (No scheduling is implemented for this chip.)
27482 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27483 instruction set support.
27484 (No scheduling is implemented for this chip.)
27487 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27488 instruction set support.
27489 (No scheduling is implemented for this chip.)
27492 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
27495 @item -mtune=@var{cpu-type}
27497 Tune to @var{cpu-type} everything applicable about the generated code, except
27498 for the ABI and the set of available instructions.
27499 While picking a specific @var{cpu-type} schedules things appropriately
27500 for that particular chip, the compiler does not generate any code that
27501 cannot run on the default machine type unless you use a
27502 @option{-march=@var{cpu-type}} option.
27503 For example, if GCC is configured for i686-pc-linux-gnu
27504 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
27505 but still runs on i686 machines.
27507 The choices for @var{cpu-type} are the same as for @option{-march}.
27508 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
27512 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
27513 If you know the CPU on which your code will run, then you should use
27514 the corresponding @option{-mtune} or @option{-march} option instead of
27515 @option{-mtune=generic}. But, if you do not know exactly what CPU users
27516 of your application will have, then you should use this option.
27518 As new processors are deployed in the marketplace, the behavior of this
27519 option will change. Therefore, if you upgrade to a newer version of
27520 GCC, code generation controlled by this option will change to reflect
27522 that are most common at the time that version of GCC is released.
27524 There is no @option{-march=generic} option because @option{-march}
27525 indicates the instruction set the compiler can use, and there is no
27526 generic instruction set applicable to all processors. In contrast,
27527 @option{-mtune} indicates the processor (or, in this case, collection of
27528 processors) for which the code is optimized.
27531 Produce code optimized for the most current Intel processors, which are
27532 Haswell and Silvermont for this version of GCC. If you know the CPU
27533 on which your code will run, then you should use the corresponding
27534 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
27535 But, if you want your application performs better on both Haswell and
27536 Silvermont, then you should use this option.
27538 As new Intel processors are deployed in the marketplace, the behavior of
27539 this option will change. Therefore, if you upgrade to a newer version of
27540 GCC, code generation controlled by this option will change to reflect
27541 the most current Intel processors at the time that version of GCC is
27544 There is no @option{-march=intel} option because @option{-march} indicates
27545 the instruction set the compiler can use, and there is no common
27546 instruction set applicable to all processors. In contrast,
27547 @option{-mtune} indicates the processor (or, in this case, collection of
27548 processors) for which the code is optimized.
27551 @item -mcpu=@var{cpu-type}
27553 A deprecated synonym for @option{-mtune}.
27555 @item -mfpmath=@var{unit}
27557 Generate floating-point arithmetic for selected unit @var{unit}. The choices
27558 for @var{unit} are:
27562 Use the standard 387 floating-point coprocessor present on the majority of chips and
27563 emulated otherwise. Code compiled with this option runs almost everywhere.
27564 The temporary results are computed in 80-bit precision instead of the precision
27565 specified by the type, resulting in slightly different results compared to most
27566 of other chips. See @option{-ffloat-store} for more detailed description.
27568 This is the default choice for non-Darwin x86-32 targets.
27571 Use scalar floating-point instructions present in the SSE instruction set.
27572 This instruction set is supported by Pentium III and newer chips,
27573 and in the AMD line
27574 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
27575 instruction set supports only single-precision arithmetic, thus the double and
27576 extended-precision arithmetic are still done using 387. A later version, present
27577 only in Pentium 4 and AMD x86-64 chips, supports double-precision
27580 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
27581 or @option{-msse2} switches to enable SSE extensions and make this option
27582 effective. For the x86-64 compiler, these extensions are enabled by default.
27584 The resulting code should be considerably faster in the majority of cases and avoid
27585 the numerical instability problems of 387 code, but may break some existing
27586 code that expects temporaries to be 80 bits.
27588 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
27589 and the default choice for x86-32 targets with the SSE2 instruction set
27590 when @option{-ffast-math} is enabled.
27595 Attempt to utilize both instruction sets at once. This effectively doubles the
27596 amount of available registers, and on chips with separate execution units for
27597 387 and SSE the execution resources too. Use this option with care, as it is
27598 still experimental, because the GCC register allocator does not model separate
27599 functional units well, resulting in unstable performance.
27602 @item -masm=@var{dialect}
27603 @opindex masm=@var{dialect}
27604 Output assembly instructions using selected @var{dialect}. Also affects
27605 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
27606 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
27607 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
27608 not support @samp{intel}.
27611 @itemx -mno-ieee-fp
27613 @opindex mno-ieee-fp
27614 Control whether or not the compiler uses IEEE floating-point
27615 comparisons. These correctly handle the case where the result of a
27616 comparison is unordered.
27619 @itemx -mhard-float
27621 @opindex mhard-float
27622 Generate output containing 80387 instructions for floating point.
27625 @itemx -msoft-float
27627 @opindex msoft-float
27628 Generate output containing library calls for floating point.
27630 @strong{Warning:} the requisite libraries are not part of GCC@.
27631 Normally the facilities of the machine's usual C compiler are used, but
27632 this cannot be done directly in cross-compilation. You must make your
27633 own arrangements to provide suitable library functions for
27636 On machines where a function returns floating-point results in the 80387
27637 register stack, some floating-point opcodes may be emitted even if
27638 @option{-msoft-float} is used.
27640 @item -mno-fp-ret-in-387
27641 @opindex mno-fp-ret-in-387
27642 Do not use the FPU registers for return values of functions.
27644 The usual calling convention has functions return values of types
27645 @code{float} and @code{double} in an FPU register, even if there
27646 is no FPU@. The idea is that the operating system should emulate
27649 The option @option{-mno-fp-ret-in-387} causes such values to be returned
27650 in ordinary CPU registers instead.
27652 @item -mno-fancy-math-387
27653 @opindex mno-fancy-math-387
27654 Some 387 emulators do not support the @code{sin}, @code{cos} and
27655 @code{sqrt} instructions for the 387. Specify this option to avoid
27656 generating those instructions.
27657 This option is overridden when @option{-march}
27658 indicates that the target CPU always has an FPU and so the
27659 instruction does not need emulation. These
27660 instructions are not generated unless you also use the
27661 @option{-funsafe-math-optimizations} switch.
27663 @item -malign-double
27664 @itemx -mno-align-double
27665 @opindex malign-double
27666 @opindex mno-align-double
27667 Control whether GCC aligns @code{double}, @code{long double}, and
27668 @code{long long} variables on a two-word boundary or a one-word
27669 boundary. Aligning @code{double} variables on a two-word boundary
27670 produces code that runs somewhat faster on a Pentium at the
27671 expense of more memory.
27673 On x86-64, @option{-malign-double} is enabled by default.
27675 @strong{Warning:} if you use the @option{-malign-double} switch,
27676 structures containing the above types are aligned differently than
27677 the published application binary interface specifications for the x86-32
27678 and are not binary compatible with structures in code compiled
27679 without that switch.
27681 @item -m96bit-long-double
27682 @itemx -m128bit-long-double
27683 @opindex m96bit-long-double
27684 @opindex m128bit-long-double
27685 These switches control the size of @code{long double} type. The x86-32
27686 application binary interface specifies the size to be 96 bits,
27687 so @option{-m96bit-long-double} is the default in 32-bit mode.
27689 Modern architectures (Pentium and newer) prefer @code{long double}
27690 to be aligned to an 8- or 16-byte boundary. In arrays or structures
27691 conforming to the ABI, this is not possible. So specifying
27692 @option{-m128bit-long-double} aligns @code{long double}
27693 to a 16-byte boundary by padding the @code{long double} with an additional
27696 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
27697 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
27699 Notice that neither of these options enable any extra precision over the x87
27700 standard of 80 bits for a @code{long double}.
27702 @strong{Warning:} if you override the default value for your target ABI, this
27703 changes the size of
27704 structures and arrays containing @code{long double} variables,
27705 as well as modifying the function calling convention for functions taking
27706 @code{long double}. Hence they are not binary-compatible
27707 with code compiled without that switch.
27709 @item -mlong-double-64
27710 @itemx -mlong-double-80
27711 @itemx -mlong-double-128
27712 @opindex mlong-double-64
27713 @opindex mlong-double-80
27714 @opindex mlong-double-128
27715 These switches control the size of @code{long double} type. A size
27716 of 64 bits makes the @code{long double} type equivalent to the @code{double}
27717 type. This is the default for 32-bit Bionic C library. A size
27718 of 128 bits makes the @code{long double} type equivalent to the
27719 @code{__float128} type. This is the default for 64-bit Bionic C library.
27721 @strong{Warning:} if you override the default value for your target ABI, this
27722 changes the size of
27723 structures and arrays containing @code{long double} variables,
27724 as well as modifying the function calling convention for functions taking
27725 @code{long double}. Hence they are not binary-compatible
27726 with code compiled without that switch.
27728 @item -malign-data=@var{type}
27729 @opindex malign-data
27730 Control how GCC aligns variables. Supported values for @var{type} are
27731 @samp{compat} uses increased alignment value compatible uses GCC 4.8
27732 and earlier, @samp{abi} uses alignment value as specified by the
27733 psABI, and @samp{cacheline} uses increased alignment value to match
27734 the cache line size. @samp{compat} is the default.
27736 @item -mlarge-data-threshold=@var{threshold}
27737 @opindex mlarge-data-threshold
27738 When @option{-mcmodel=medium} is specified, data objects larger than
27739 @var{threshold} are placed in the large data section. This value must be the
27740 same across all objects linked into the binary, and defaults to 65535.
27744 Use a different function-calling convention, in which functions that
27745 take a fixed number of arguments return with the @code{ret @var{num}}
27746 instruction, which pops their arguments while returning. This saves one
27747 instruction in the caller since there is no need to pop the arguments
27750 You can specify that an individual function is called with this calling
27751 sequence with the function attribute @code{stdcall}. You can also
27752 override the @option{-mrtd} option by using the function attribute
27753 @code{cdecl}. @xref{Function Attributes}.
27755 @strong{Warning:} this calling convention is incompatible with the one
27756 normally used on Unix, so you cannot use it if you need to call
27757 libraries compiled with the Unix compiler.
27759 Also, you must provide function prototypes for all functions that
27760 take variable numbers of arguments (including @code{printf});
27761 otherwise incorrect code is generated for calls to those
27764 In addition, seriously incorrect code results if you call a
27765 function with too many arguments. (Normally, extra arguments are
27766 harmlessly ignored.)
27768 @item -mregparm=@var{num}
27770 Control how many registers are used to pass integer arguments. By
27771 default, no registers are used to pass arguments, and at most 3
27772 registers can be used. You can control this behavior for a specific
27773 function by using the function attribute @code{regparm}.
27774 @xref{Function Attributes}.
27776 @strong{Warning:} if you use this switch, and
27777 @var{num} is nonzero, then you must build all modules with the same
27778 value, including any libraries. This includes the system libraries and
27782 @opindex msseregparm
27783 Use SSE register passing conventions for float and double arguments
27784 and return values. You can control this behavior for a specific
27785 function by using the function attribute @code{sseregparm}.
27786 @xref{Function Attributes}.
27788 @strong{Warning:} if you use this switch then you must build all
27789 modules with the same value, including any libraries. This includes
27790 the system libraries and startup modules.
27792 @item -mvect8-ret-in-mem
27793 @opindex mvect8-ret-in-mem
27794 Return 8-byte vectors in memory instead of MMX registers. This is the
27795 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
27796 Studio compilers until version 12. Later compiler versions (starting
27797 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
27798 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
27799 you need to remain compatible with existing code produced by those
27800 previous compiler versions or older versions of GCC@.
27809 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27810 is specified, the significands of results of floating-point operations are
27811 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27812 significands of results of floating-point operations to 53 bits (double
27813 precision) and @option{-mpc80} rounds the significands of results of
27814 floating-point operations to 64 bits (extended double precision), which is
27815 the default. When this option is used, floating-point operations in higher
27816 precisions are not available to the programmer without setting the FPU
27817 control word explicitly.
27819 Setting the rounding of floating-point operations to less than the default
27820 80 bits can speed some programs by 2% or more. Note that some mathematical
27821 libraries assume that extended-precision (80-bit) floating-point operations
27822 are enabled by default; routines in such libraries could suffer significant
27823 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27824 when this option is used to set the precision to less than extended precision.
27826 @item -mstackrealign
27827 @opindex mstackrealign
27828 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27829 option generates an alternate prologue and epilogue that realigns the
27830 run-time stack if necessary. This supports mixing legacy codes that keep
27831 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27832 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27833 applicable to individual functions.
27835 @item -mpreferred-stack-boundary=@var{num}
27836 @opindex mpreferred-stack-boundary
27837 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27838 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27839 the default is 4 (16 bytes or 128 bits).
27841 @strong{Warning:} When generating code for the x86-64 architecture with
27842 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27843 used to keep the stack boundary aligned to 8 byte boundary. Since
27844 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27845 intended to be used in controlled environment where stack space is
27846 important limitation. This option leads to wrong code when functions
27847 compiled with 16 byte stack alignment (such as functions from a standard
27848 library) are called with misaligned stack. In this case, SSE
27849 instructions may lead to misaligned memory access traps. In addition,
27850 variable arguments are handled incorrectly for 16 byte aligned
27851 objects (including x87 long double and __int128), leading to wrong
27852 results. You must build all modules with
27853 @option{-mpreferred-stack-boundary=3}, including any libraries. This
27854 includes the system libraries and startup modules.
27856 @item -mincoming-stack-boundary=@var{num}
27857 @opindex mincoming-stack-boundary
27858 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
27859 boundary. If @option{-mincoming-stack-boundary} is not specified,
27860 the one specified by @option{-mpreferred-stack-boundary} is used.
27862 On Pentium and Pentium Pro, @code{double} and @code{long double} values
27863 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
27864 suffer significant run time performance penalties. On Pentium III, the
27865 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
27866 properly if it is not 16-byte aligned.
27868 To ensure proper alignment of this values on the stack, the stack boundary
27869 must be as aligned as that required by any value stored on the stack.
27870 Further, every function must be generated such that it keeps the stack
27871 aligned. Thus calling a function compiled with a higher preferred
27872 stack boundary from a function compiled with a lower preferred stack
27873 boundary most likely misaligns the stack. It is recommended that
27874 libraries that use callbacks always use the default setting.
27876 This extra alignment does consume extra stack space, and generally
27877 increases code size. Code that is sensitive to stack space usage, such
27878 as embedded systems and operating system kernels, may want to reduce the
27879 preferred alignment to @option{-mpreferred-stack-boundary=2}.
27936 @itemx -mavx512ifma
27937 @opindex mavx512ifma
27939 @itemx -mavx512vbmi
27940 @opindex mavx512vbmi
27951 @itemx -mclflushopt
27952 @opindex mclflushopt
27978 @itemx -mprefetchwt1
27979 @opindex mprefetchwt1
28037 @itemx -mavx512vbmi2
28038 @opindex mavx512vbmi2
28049 @itemx -mvpclmulqdq
28050 @opindex mvpclmulqdq
28052 @itemx -mavx512bitalg
28053 @opindex mavx512bitalg
28059 @opindex mmovdir64b
28061 @itemx -mavx512vpopcntdq
28062 @opindex mavx512vpopcntdq
28066 These switches enable the use of instructions in the MMX, SSE,
28067 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
28068 SHA, AES, PCLMUL, FSGSBASE, PTWRITE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
28069 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, BMI, BMI2, VAES, WAITPKG,
28070 FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MWAITX, PKU, IBT, SHSTK, AVX512VBMI2,
28071 GFNI, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B,
28072 AVX512VPOPCNTDQ, CLDEMOTE, 3DNow!@: or enhanced 3DNow!@: extended instruction
28073 sets. Each has a corresponding @option{-mno-} option to disable use of these
28076 These extensions are also available as built-in functions: see
28077 @ref{x86 Built-in Functions}, for details of the functions enabled and
28078 disabled by these switches.
28080 To generate SSE/SSE2 instructions automatically from floating-point
28081 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
28083 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
28084 generates new AVX instructions or AVX equivalence for all SSEx instructions
28087 These options enable GCC to use these extended instructions in
28088 generated code, even without @option{-mfpmath=sse}. Applications that
28089 perform run-time CPU detection must compile separate files for each
28090 supported architecture, using the appropriate flags. In particular,
28091 the file containing the CPU detection code should be compiled without
28094 @item -mdump-tune-features
28095 @opindex mdump-tune-features
28096 This option instructs GCC to dump the names of the x86 performance
28097 tuning features and default settings. The names can be used in
28098 @option{-mtune-ctrl=@var{feature-list}}.
28100 @item -mtune-ctrl=@var{feature-list}
28101 @opindex mtune-ctrl=@var{feature-list}
28102 This option is used to do fine grain control of x86 code generation features.
28103 @var{feature-list} is a comma separated list of @var{feature} names. See also
28104 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
28105 on if it is not preceded with @samp{^}, otherwise, it is turned off.
28106 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
28107 developers. Using it may lead to code paths not covered by testing and can
28108 potentially result in compiler ICEs or runtime errors.
28111 @opindex mno-default
28112 This option instructs GCC to turn off all tunable features. See also
28113 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
28117 This option instructs GCC to emit a @code{cld} instruction in the prologue
28118 of functions that use string instructions. String instructions depend on
28119 the DF flag to select between autoincrement or autodecrement mode. While the
28120 ABI specifies the DF flag to be cleared on function entry, some operating
28121 systems violate this specification by not clearing the DF flag in their
28122 exception dispatchers. The exception handler can be invoked with the DF flag
28123 set, which leads to wrong direction mode when string instructions are used.
28124 This option can be enabled by default on 32-bit x86 targets by configuring
28125 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
28126 instructions can be suppressed with the @option{-mno-cld} compiler option
28130 @opindex mvzeroupper
28131 This option instructs GCC to emit a @code{vzeroupper} instruction
28132 before a transfer of control flow out of the function to minimize
28133 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
28136 @item -mprefer-avx128
28137 @opindex mprefer-avx128
28138 This option instructs GCC to use 128-bit AVX instructions instead of
28139 256-bit AVX instructions in the auto-vectorizer.
28141 @item -mprefer-vector-width=@var{opt}
28142 @opindex mprefer-vector-width
28143 This option instructs GCC to use @var{opt}-bit vector width in instructions
28144 instead of default on the selected platform.
28148 No extra limitations applied to GCC other than defined by the selected platform.
28151 Prefer 128-bit vector width for instructions.
28154 Prefer 256-bit vector width for instructions.
28157 Prefer 512-bit vector width for instructions.
28162 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
28163 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
28164 objects. This is useful for atomic updates of data structures exceeding one
28165 machine word in size. The compiler uses this instruction to implement
28166 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
28167 128-bit integers, a library call is always used.
28171 This option enables generation of @code{SAHF} instructions in 64-bit code.
28172 Early Intel Pentium 4 CPUs with Intel 64 support,
28173 prior to the introduction of Pentium 4 G1 step in December 2005,
28174 lacked the @code{LAHF} and @code{SAHF} instructions
28175 which are supported by AMD64.
28176 These are load and store instructions, respectively, for certain status flags.
28177 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
28178 @code{drem}, and @code{remainder} built-in functions;
28179 see @ref{Other Builtins} for details.
28183 This option enables use of the @code{movbe} instruction to implement
28184 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
28188 The @option{-mshstk} option enables shadow stack built-in functions
28189 from x86 Control-flow Enforcement Technology (CET).
28193 This option enables built-in functions @code{__builtin_ia32_crc32qi},
28194 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
28195 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
28199 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
28200 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
28201 with an additional Newton-Raphson step
28202 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
28203 (and their vectorized
28204 variants) for single-precision floating-point arguments. These instructions
28205 are generated only when @option{-funsafe-math-optimizations} is enabled
28206 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
28207 Note that while the throughput of the sequence is higher than the throughput
28208 of the non-reciprocal instruction, the precision of the sequence can be
28209 decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994).
28211 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
28212 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
28213 combination), and doesn't need @option{-mrecip}.
28215 Also note that GCC emits the above sequence with additional Newton-Raphson step
28216 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
28217 already with @option{-ffast-math} (or the above option combination), and
28218 doesn't need @option{-mrecip}.
28220 @item -mrecip=@var{opt}
28221 @opindex mrecip=opt
28222 This option controls which reciprocal estimate instructions
28223 may be used. @var{opt} is a comma-separated list of options, which may
28224 be preceded by a @samp{!} to invert the option:
28228 Enable all estimate instructions.
28231 Enable the default instructions, equivalent to @option{-mrecip}.
28234 Disable all estimate instructions, equivalent to @option{-mno-recip}.
28237 Enable the approximation for scalar division.
28240 Enable the approximation for vectorized division.
28243 Enable the approximation for scalar square root.
28246 Enable the approximation for vectorized square root.
28249 So, for example, @option{-mrecip=all,!sqrt} enables
28250 all of the reciprocal approximations, except for square root.
28252 @item -mveclibabi=@var{type}
28253 @opindex mveclibabi
28254 Specifies the ABI type to use for vectorizing intrinsics using an
28255 external library. Supported values for @var{type} are @samp{svml}
28256 for the Intel short
28257 vector math library and @samp{acml} for the AMD math core library.
28258 To use this option, both @option{-ftree-vectorize} and
28259 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
28260 ABI-compatible library must be specified at link time.
28262 GCC currently emits calls to @code{vmldExp2},
28263 @code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2},
28264 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
28265 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
28266 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
28267 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4},
28268 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
28269 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
28270 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
28271 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
28272 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
28273 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
28274 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
28275 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
28276 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
28277 when @option{-mveclibabi=acml} is used.
28279 @item -mabi=@var{name}
28281 Generate code for the specified calling convention. Permissible values
28282 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
28283 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
28284 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
28285 You can control this behavior for specific functions by
28286 using the function attributes @code{ms_abi} and @code{sysv_abi}.
28287 @xref{Function Attributes}.
28289 @item -mforce-indirect-call
28290 @opindex mforce-indirect-call
28291 Force all calls to functions to be indirect. This is useful
28292 when using Intel Processor Trace where it generates more precise timing
28293 information for function calls.
28295 @item -mcall-ms2sysv-xlogues
28296 @opindex mcall-ms2sysv-xlogues
28297 @opindex mno-call-ms2sysv-xlogues
28298 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
28299 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
28300 default, the code for saving and restoring these registers is emitted inline,
28301 resulting in fairly lengthy prologues and epilogues. Using
28302 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
28303 use stubs in the static portion of libgcc to perform these saves and restores,
28304 thus reducing function size at the cost of a few extra instructions.
28306 @item -mtls-dialect=@var{type}
28307 @opindex mtls-dialect
28308 Generate code to access thread-local storage using the @samp{gnu} or
28309 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
28310 @samp{gnu2} is more efficient, but it may add compile- and run-time
28311 requirements that cannot be satisfied on all systems.
28314 @itemx -mno-push-args
28315 @opindex mpush-args
28316 @opindex mno-push-args
28317 Use PUSH operations to store outgoing parameters. This method is shorter
28318 and usually equally fast as method using SUB/MOV operations and is enabled
28319 by default. In some cases disabling it may improve performance because of
28320 improved scheduling and reduced dependencies.
28322 @item -maccumulate-outgoing-args
28323 @opindex maccumulate-outgoing-args
28324 If enabled, the maximum amount of space required for outgoing arguments is
28325 computed in the function prologue. This is faster on most modern CPUs
28326 because of reduced dependencies, improved scheduling and reduced stack usage
28327 when the preferred stack boundary is not equal to 2. The drawback is a notable
28328 increase in code size. This switch implies @option{-mno-push-args}.
28332 Support thread-safe exception handling on MinGW. Programs that rely
28333 on thread-safe exception handling must compile and link all code with the
28334 @option{-mthreads} option. When compiling, @option{-mthreads} defines
28335 @option{-D_MT}; when linking, it links in a special thread helper library
28336 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
28338 @item -mms-bitfields
28339 @itemx -mno-ms-bitfields
28340 @opindex mms-bitfields
28341 @opindex mno-ms-bitfields
28343 Enable/disable bit-field layout compatible with the native Microsoft
28346 If @code{packed} is used on a structure, or if bit-fields are used,
28347 it may be that the Microsoft ABI lays out the structure differently
28348 than the way GCC normally does. Particularly when moving packed
28349 data between functions compiled with GCC and the native Microsoft compiler
28350 (either via function call or as data in a file), it may be necessary to access
28353 This option is enabled by default for Microsoft Windows
28354 targets. This behavior can also be controlled locally by use of variable
28355 or type attributes. For more information, see @ref{x86 Variable Attributes}
28356 and @ref{x86 Type Attributes}.
28358 The Microsoft structure layout algorithm is fairly simple with the exception
28359 of the bit-field packing.
28360 The padding and alignment of members of structures and whether a bit-field
28361 can straddle a storage-unit boundary are determine by these rules:
28364 @item Structure members are stored sequentially in the order in which they are
28365 declared: the first member has the lowest memory address and the last member
28368 @item Every data object has an alignment requirement. The alignment requirement
28369 for all data except structures, unions, and arrays is either the size of the
28370 object or the current packing size (specified with either the
28371 @code{aligned} attribute or the @code{pack} pragma),
28372 whichever is less. For structures, unions, and arrays,
28373 the alignment requirement is the largest alignment requirement of its members.
28374 Every object is allocated an offset so that:
28377 offset % alignment_requirement == 0
28380 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
28381 unit if the integral types are the same size and if the next bit-field fits
28382 into the current allocation unit without crossing the boundary imposed by the
28383 common alignment requirements of the bit-fields.
28386 MSVC interprets zero-length bit-fields in the following ways:
28389 @item If a zero-length bit-field is inserted between two bit-fields that
28390 are normally coalesced, the bit-fields are not coalesced.
28397 unsigned long bf_1 : 12;
28399 unsigned long bf_2 : 12;
28404 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
28405 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
28407 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
28408 alignment of the zero-length bit-field is greater than the member that follows it,
28409 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
28430 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
28431 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
28432 bit-field does not affect the alignment of @code{bar} or, as a result, the size
28435 Taking this into account, it is important to note the following:
28438 @item If a zero-length bit-field follows a normal bit-field, the type of the
28439 zero-length bit-field may affect the alignment of the structure as whole. For
28440 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
28441 normal bit-field, and is of type short.
28443 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
28444 still affect the alignment of the structure:
28455 Here, @code{t4} takes up 4 bytes.
28458 @item Zero-length bit-fields following non-bit-field members are ignored:
28470 Here, @code{t5} takes up 2 bytes.
28474 @item -mno-align-stringops
28475 @opindex mno-align-stringops
28476 Do not align the destination of inlined string operations. This switch reduces
28477 code size and improves performance in case the destination is already aligned,
28478 but GCC doesn't know about it.
28480 @item -minline-all-stringops
28481 @opindex minline-all-stringops
28482 By default GCC inlines string operations only when the destination is
28483 known to be aligned to least a 4-byte boundary.
28484 This enables more inlining and increases code
28485 size, but may improve performance of code that depends on fast
28486 @code{memcpy}, @code{strlen},
28487 and @code{memset} for short lengths.
28489 @item -minline-stringops-dynamically
28490 @opindex minline-stringops-dynamically
28491 For string operations of unknown size, use run-time checks with
28492 inline code for small blocks and a library call for large blocks.
28494 @item -mstringop-strategy=@var{alg}
28495 @opindex mstringop-strategy=@var{alg}
28496 Override the internal decision heuristic for the particular algorithm to use
28497 for inlining string operations. The allowed values for @var{alg} are:
28503 Expand using i386 @code{rep} prefix of the specified size.
28507 @itemx unrolled_loop
28508 Expand into an inline loop.
28511 Always use a library call.
28514 @item -mmemcpy-strategy=@var{strategy}
28515 @opindex mmemcpy-strategy=@var{strategy}
28516 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
28517 should be inlined and what inline algorithm to use when the expected size
28518 of the copy operation is known. @var{strategy}
28519 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
28520 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
28521 the max byte size with which inline algorithm @var{alg} is allowed. For the last
28522 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
28523 in the list must be specified in increasing order. The minimal byte size for
28524 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
28527 @item -mmemset-strategy=@var{strategy}
28528 @opindex mmemset-strategy=@var{strategy}
28529 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
28530 @code{__builtin_memset} expansion.
28532 @item -momit-leaf-frame-pointer
28533 @opindex momit-leaf-frame-pointer
28534 Don't keep the frame pointer in a register for leaf functions. This
28535 avoids the instructions to save, set up, and restore frame pointers and
28536 makes an extra register available in leaf functions. The option
28537 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
28538 which might make debugging harder.
28540 @item -mtls-direct-seg-refs
28541 @itemx -mno-tls-direct-seg-refs
28542 @opindex mtls-direct-seg-refs
28543 Controls whether TLS variables may be accessed with offsets from the
28544 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
28545 or whether the thread base pointer must be added. Whether or not this
28546 is valid depends on the operating system, and whether it maps the
28547 segment to cover the entire TLS area.
28549 For systems that use the GNU C Library, the default is on.
28552 @itemx -mno-sse2avx
28554 Specify that the assembler should encode SSE instructions with VEX
28555 prefix. The option @option{-mavx} turns this on by default.
28560 If profiling is active (@option{-pg}), put the profiling
28561 counter call before the prologue.
28562 Note: On x86 architectures the attribute @code{ms_hook_prologue}
28563 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
28565 @item -mrecord-mcount
28566 @itemx -mno-record-mcount
28567 @opindex mrecord-mcount
28568 If profiling is active (@option{-pg}), generate a __mcount_loc section
28569 that contains pointers to each profiling call. This is useful for
28570 automatically patching and out calls.
28573 @itemx -mno-nop-mcount
28574 @opindex mnop-mcount
28575 If profiling is active (@option{-pg}), generate the calls to
28576 the profiling functions as NOPs. This is useful when they
28577 should be patched in later dynamically. This is likely only
28578 useful together with @option{-mrecord-mcount}.
28580 @item -mskip-rax-setup
28581 @itemx -mno-skip-rax-setup
28582 @opindex mskip-rax-setup
28583 When generating code for the x86-64 architecture with SSE extensions
28584 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
28585 register when there are no variable arguments passed in vector registers.
28587 @strong{Warning:} Since RAX register is used to avoid unnecessarily
28588 saving vector registers on stack when passing variable arguments, the
28589 impacts of this option are callees may waste some stack space,
28590 misbehave or jump to a random location. GCC 4.4 or newer don't have
28591 those issues, regardless the RAX register value.
28594 @itemx -mno-8bit-idiv
28595 @opindex m8bit-idiv
28596 On some processors, like Intel Atom, 8-bit unsigned integer divide is
28597 much faster than 32-bit/64-bit integer divide. This option generates a
28598 run-time check. If both dividend and divisor are within range of 0
28599 to 255, 8-bit unsigned integer divide is used instead of
28600 32-bit/64-bit integer divide.
28602 @item -mavx256-split-unaligned-load
28603 @itemx -mavx256-split-unaligned-store
28604 @opindex mavx256-split-unaligned-load
28605 @opindex mavx256-split-unaligned-store
28606 Split 32-byte AVX unaligned load and store.
28608 @item -mstack-protector-guard=@var{guard}
28609 @itemx -mstack-protector-guard-reg=@var{reg}
28610 @itemx -mstack-protector-guard-offset=@var{offset}
28611 @opindex mstack-protector-guard
28612 @opindex mstack-protector-guard-reg
28613 @opindex mstack-protector-guard-offset
28614 Generate stack protection code using canary at @var{guard}. Supported
28615 locations are @samp{global} for global canary or @samp{tls} for per-thread
28616 canary in the TLS block (the default). This option has effect only when
28617 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
28619 With the latter choice the options
28620 @option{-mstack-protector-guard-reg=@var{reg}} and
28621 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
28622 which segment register (@code{%fs} or @code{%gs}) to use as base register
28623 for reading the canary, and from what offset from that base register.
28624 The default for those is as specified in the relevant ABI.
28626 @item -mgeneral-regs-only
28627 @opindex mgeneral-regs-only
28628 Generate code that uses only the general-purpose registers. This
28629 prevents the compiler from using floating-point, vector, mask and bound
28632 @item -mindirect-branch=@var{choice}
28633 @opindex mindirect-branch
28634 Convert indirect call and jump with @var{choice}. The default is
28635 @samp{keep}, which keeps indirect call and jump unmodified.
28636 @samp{thunk} converts indirect call and jump to call and return thunk.
28637 @samp{thunk-inline} converts indirect call and jump to inlined call
28638 and return thunk. @samp{thunk-extern} converts indirect call and jump
28639 to external call and return thunk provided in a separate object file.
28640 You can control this behavior for a specific function by using the
28641 function attribute @code{indirect_branch}. @xref{Function Attributes}.
28643 Note that @option{-mcmodel=large} is incompatible with
28644 @option{-mindirect-branch=thunk} and
28645 @option{-mindirect-branch=thunk-extern} since the thunk function may
28646 not be reachable in the large code model.
28648 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
28649 @option{-fcf-protection=branch} since the external thunk can not be modified
28650 to disable control-flow check.
28652 @item -mfunction-return=@var{choice}
28653 @opindex mfunction-return
28654 Convert function return with @var{choice}. The default is @samp{keep},
28655 which keeps function return unmodified. @samp{thunk} converts function
28656 return to call and return thunk. @samp{thunk-inline} converts function
28657 return to inlined call and return thunk. @samp{thunk-extern} converts
28658 function return to external call and return thunk provided in a separate
28659 object file. You can control this behavior for a specific function by
28660 using the function attribute @code{function_return}.
28661 @xref{Function Attributes}.
28663 Note that @option{-mcmodel=large} is incompatible with
28664 @option{-mfunction-return=thunk} and
28665 @option{-mfunction-return=thunk-extern} since the thunk function may
28666 not be reachable in the large code model.
28669 @item -mindirect-branch-register
28670 @opindex mindirect-branch-register
28671 Force indirect call and jump via register.
28675 These @samp{-m} switches are supported in addition to the above
28676 on x86-64 processors in 64-bit environments.
28689 Generate code for a 16-bit, 32-bit or 64-bit environment.
28690 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
28692 generates code that runs on any i386 system.
28694 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
28695 types to 64 bits, and generates code for the x86-64 architecture.
28696 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
28697 and @option{-mdynamic-no-pic} options.
28699 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
28701 generates code for the x86-64 architecture.
28703 The @option{-m16} option is the same as @option{-m32}, except for that
28704 it outputs the @code{.code16gcc} assembly directive at the beginning of
28705 the assembly output so that the binary can run in 16-bit mode.
28707 The @option{-miamcu} option generates code which conforms to Intel MCU
28708 psABI. It requires the @option{-m32} option to be turned on.
28710 @item -mno-red-zone
28711 @opindex mno-red-zone
28712 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
28713 by the x86-64 ABI; it is a 128-byte area beyond the location of the
28714 stack pointer that is not modified by signal or interrupt handlers
28715 and therefore can be used for temporary data without adjusting the stack
28716 pointer. The flag @option{-mno-red-zone} disables this red zone.
28718 @item -mcmodel=small
28719 @opindex mcmodel=small
28720 Generate code for the small code model: the program and its symbols must
28721 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28722 Programs can be statically or dynamically linked. This is the default
28725 @item -mcmodel=kernel
28726 @opindex mcmodel=kernel
28727 Generate code for the kernel code model. The kernel runs in the
28728 negative 2 GB of the address space.
28729 This model has to be used for Linux kernel code.
28731 @item -mcmodel=medium
28732 @opindex mcmodel=medium
28733 Generate code for the medium model: the program is linked in the lower 2
28734 GB of the address space. Small symbols are also placed there. Symbols
28735 with sizes larger than @option{-mlarge-data-threshold} are put into
28736 large data or BSS sections and can be located above 2GB. Programs can
28737 be statically or dynamically linked.
28739 @item -mcmodel=large
28740 @opindex mcmodel=large
28741 Generate code for the large model. This model makes no assumptions
28742 about addresses and sizes of sections.
28744 @item -maddress-mode=long
28745 @opindex maddress-mode=long
28746 Generate code for long address mode. This is only supported for 64-bit
28747 and x32 environments. It is the default address mode for 64-bit
28750 @item -maddress-mode=short
28751 @opindex maddress-mode=short
28752 Generate code for short address mode. This is only supported for 32-bit
28753 and x32 environments. It is the default address mode for 32-bit and
28757 @node x86 Windows Options
28758 @subsection x86 Windows Options
28759 @cindex x86 Windows Options
28760 @cindex Windows Options for x86
28762 These additional options are available for Microsoft Windows targets:
28768 specifies that a console application is to be generated, by
28769 instructing the linker to set the PE header subsystem type
28770 required for console applications.
28771 This option is available for Cygwin and MinGW targets and is
28772 enabled by default on those targets.
28776 This option is available for Cygwin and MinGW targets. It
28777 specifies that a DLL---a dynamic link library---is to be
28778 generated, enabling the selection of the required runtime
28779 startup object and entry point.
28781 @item -mnop-fun-dllimport
28782 @opindex mnop-fun-dllimport
28783 This option is available for Cygwin and MinGW targets. It
28784 specifies that the @code{dllimport} attribute should be ignored.
28788 This option is available for MinGW targets. It specifies
28789 that MinGW-specific thread support is to be used.
28793 This option is available for MinGW-w64 targets. It causes
28794 the @code{UNICODE} preprocessor macro to be predefined, and
28795 chooses Unicode-capable runtime startup code.
28799 This option is available for Cygwin and MinGW targets. It
28800 specifies that the typical Microsoft Windows predefined macros are to
28801 be set in the pre-processor, but does not influence the choice
28802 of runtime library/startup code.
28806 This option is available for Cygwin and MinGW targets. It
28807 specifies that a GUI application is to be generated by
28808 instructing the linker to set the PE header subsystem type
28811 @item -fno-set-stack-executable
28812 @opindex fno-set-stack-executable
28813 This option is available for MinGW targets. It specifies that
28814 the executable flag for the stack used by nested functions isn't
28815 set. This is necessary for binaries running in kernel mode of
28816 Microsoft Windows, as there the User32 API, which is used to set executable
28817 privileges, isn't available.
28819 @item -fwritable-relocated-rdata
28820 @opindex fno-writable-relocated-rdata
28821 This option is available for MinGW and Cygwin targets. It specifies
28822 that relocated-data in read-only section is put into the @code{.data}
28823 section. This is a necessary for older runtimes not supporting
28824 modification of @code{.rdata} sections for pseudo-relocation.
28826 @item -mpe-aligned-commons
28827 @opindex mpe-aligned-commons
28828 This option is available for Cygwin and MinGW targets. It
28829 specifies that the GNU extension to the PE file format that
28830 permits the correct alignment of COMMON variables should be
28831 used when generating code. It is enabled by default if
28832 GCC detects that the target assembler found during configuration
28833 supports the feature.
28836 See also under @ref{x86 Options} for standard options.
28838 @node Xstormy16 Options
28839 @subsection Xstormy16 Options
28840 @cindex Xstormy16 Options
28842 These options are defined for Xstormy16:
28847 Choose startup files and linker script suitable for the simulator.
28850 @node Xtensa Options
28851 @subsection Xtensa Options
28852 @cindex Xtensa Options
28854 These options are supported for Xtensa targets:
28858 @itemx -mno-const16
28860 @opindex mno-const16
28861 Enable or disable use of @code{CONST16} instructions for loading
28862 constant values. The @code{CONST16} instruction is currently not a
28863 standard option from Tensilica. When enabled, @code{CONST16}
28864 instructions are always used in place of the standard @code{L32R}
28865 instructions. The use of @code{CONST16} is enabled by default only if
28866 the @code{L32R} instruction is not available.
28869 @itemx -mno-fused-madd
28870 @opindex mfused-madd
28871 @opindex mno-fused-madd
28872 Enable or disable use of fused multiply/add and multiply/subtract
28873 instructions in the floating-point option. This has no effect if the
28874 floating-point option is not also enabled. Disabling fused multiply/add
28875 and multiply/subtract instructions forces the compiler to use separate
28876 instructions for the multiply and add/subtract operations. This may be
28877 desirable in some cases where strict IEEE 754-compliant results are
28878 required: the fused multiply add/subtract instructions do not round the
28879 intermediate result, thereby producing results with @emph{more} bits of
28880 precision than specified by the IEEE standard. Disabling fused multiply
28881 add/subtract instructions also ensures that the program output is not
28882 sensitive to the compiler's ability to combine multiply and add/subtract
28885 @item -mserialize-volatile
28886 @itemx -mno-serialize-volatile
28887 @opindex mserialize-volatile
28888 @opindex mno-serialize-volatile
28889 When this option is enabled, GCC inserts @code{MEMW} instructions before
28890 @code{volatile} memory references to guarantee sequential consistency.
28891 The default is @option{-mserialize-volatile}. Use
28892 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
28894 @item -mforce-no-pic
28895 @opindex mforce-no-pic
28896 For targets, like GNU/Linux, where all user-mode Xtensa code must be
28897 position-independent code (PIC), this option disables PIC for compiling
28900 @item -mtext-section-literals
28901 @itemx -mno-text-section-literals
28902 @opindex mtext-section-literals
28903 @opindex mno-text-section-literals
28904 These options control the treatment of literal pools. The default is
28905 @option{-mno-text-section-literals}, which places literals in a separate
28906 section in the output file. This allows the literal pool to be placed
28907 in a data RAM/ROM, and it also allows the linker to combine literal
28908 pools from separate object files to remove redundant literals and
28909 improve code size. With @option{-mtext-section-literals}, the literals
28910 are interspersed in the text section in order to keep them as close as
28911 possible to their references. This may be necessary for large assembly
28912 files. Literals for each function are placed right before that function.
28914 @item -mauto-litpools
28915 @itemx -mno-auto-litpools
28916 @opindex mauto-litpools
28917 @opindex mno-auto-litpools
28918 These options control the treatment of literal pools. The default is
28919 @option{-mno-auto-litpools}, which places literals in a separate
28920 section in the output file unless @option{-mtext-section-literals} is
28921 used. With @option{-mauto-litpools} the literals are interspersed in
28922 the text section by the assembler. Compiler does not produce explicit
28923 @code{.literal} directives and loads literals into registers with
28924 @code{MOVI} instructions instead of @code{L32R} to let the assembler
28925 do relaxation and place literals as necessary. This option allows
28926 assembler to create several literal pools per function and assemble
28927 very big functions, which may not be possible with
28928 @option{-mtext-section-literals}.
28930 @item -mtarget-align
28931 @itemx -mno-target-align
28932 @opindex mtarget-align
28933 @opindex mno-target-align
28934 When this option is enabled, GCC instructs the assembler to
28935 automatically align instructions to reduce branch penalties at the
28936 expense of some code density. The assembler attempts to widen density
28937 instructions to align branch targets and the instructions following call
28938 instructions. If there are not enough preceding safe density
28939 instructions to align a target, no widening is performed. The
28940 default is @option{-mtarget-align}. These options do not affect the
28941 treatment of auto-aligned instructions like @code{LOOP}, which the
28942 assembler always aligns, either by widening density instructions or
28943 by inserting NOP instructions.
28946 @itemx -mno-longcalls
28947 @opindex mlongcalls
28948 @opindex mno-longcalls
28949 When this option is enabled, GCC instructs the assembler to translate
28950 direct calls to indirect calls unless it can determine that the target
28951 of a direct call is in the range allowed by the call instruction. This
28952 translation typically occurs for calls to functions in other source
28953 files. Specifically, the assembler translates a direct @code{CALL}
28954 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
28955 The default is @option{-mno-longcalls}. This option should be used in
28956 programs where the call target can potentially be out of range. This
28957 option is implemented in the assembler, not the compiler, so the
28958 assembly code generated by GCC still shows direct call
28959 instructions---look at the disassembled object code to see the actual
28960 instructions. Note that the assembler uses an indirect call for
28961 every cross-file call, not just those that really are out of range.
28964 @node zSeries Options
28965 @subsection zSeries Options
28966 @cindex zSeries options
28968 These are listed under @xref{S/390 and zSeries Options}.
28974 @section Specifying Subprocesses and the Switches to Pass to Them
28977 @command{gcc} is a driver program. It performs its job by invoking a
28978 sequence of other programs to do the work of compiling, assembling and
28979 linking. GCC interprets its command-line parameters and uses these to
28980 deduce which programs it should invoke, and which command-line options
28981 it ought to place on their command lines. This behavior is controlled
28982 by @dfn{spec strings}. In most cases there is one spec string for each
28983 program that GCC can invoke, but a few programs have multiple spec
28984 strings to control their behavior. The spec strings built into GCC can
28985 be overridden by using the @option{-specs=} command-line switch to specify
28988 @dfn{Spec files} are plain-text files that are used to construct spec
28989 strings. They consist of a sequence of directives separated by blank
28990 lines. The type of directive is determined by the first non-whitespace
28991 character on the line, which can be one of the following:
28994 @item %@var{command}
28995 Issues a @var{command} to the spec file processor. The commands that can
28999 @item %include <@var{file}>
29000 @cindex @code{%include}
29001 Search for @var{file} and insert its text at the current point in the
29004 @item %include_noerr <@var{file}>
29005 @cindex @code{%include_noerr}
29006 Just like @samp{%include}, but do not generate an error message if the include
29007 file cannot be found.
29009 @item %rename @var{old_name} @var{new_name}
29010 @cindex @code{%rename}
29011 Rename the spec string @var{old_name} to @var{new_name}.
29015 @item *[@var{spec_name}]:
29016 This tells the compiler to create, override or delete the named spec
29017 string. All lines after this directive up to the next directive or
29018 blank line are considered to be the text for the spec string. If this
29019 results in an empty string then the spec is deleted. (Or, if the
29020 spec did not exist, then nothing happens.) Otherwise, if the spec
29021 does not currently exist a new spec is created. If the spec does
29022 exist then its contents are overridden by the text of this
29023 directive, unless the first character of that text is the @samp{+}
29024 character, in which case the text is appended to the spec.
29026 @item [@var{suffix}]:
29027 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
29028 and up to the next directive or blank line are considered to make up the
29029 spec string for the indicated suffix. When the compiler encounters an
29030 input file with the named suffix, it processes the spec string in
29031 order to work out how to compile that file. For example:
29035 z-compile -input %i
29038 This says that any input file whose name ends in @samp{.ZZ} should be
29039 passed to the program @samp{z-compile}, which should be invoked with the
29040 command-line switch @option{-input} and with the result of performing the
29041 @samp{%i} substitution. (See below.)
29043 As an alternative to providing a spec string, the text following a
29044 suffix directive can be one of the following:
29047 @item @@@var{language}
29048 This says that the suffix is an alias for a known @var{language}. This is
29049 similar to using the @option{-x} command-line switch to GCC to specify a
29050 language explicitly. For example:
29057 Says that .ZZ files are, in fact, C++ source files.
29060 This causes an error messages saying:
29063 @var{name} compiler not installed on this system.
29067 GCC already has an extensive list of suffixes built into it.
29068 This directive adds an entry to the end of the list of suffixes, but
29069 since the list is searched from the end backwards, it is effectively
29070 possible to override earlier entries using this technique.
29074 GCC has the following spec strings built into it. Spec files can
29075 override these strings or create their own. Note that individual
29076 targets can also add their own spec strings to this list.
29079 asm Options to pass to the assembler
29080 asm_final Options to pass to the assembler post-processor
29081 cpp Options to pass to the C preprocessor
29082 cc1 Options to pass to the C compiler
29083 cc1plus Options to pass to the C++ compiler
29084 endfile Object files to include at the end of the link
29085 link Options to pass to the linker
29086 lib Libraries to include on the command line to the linker
29087 libgcc Decides which GCC support library to pass to the linker
29088 linker Sets the name of the linker
29089 predefines Defines to be passed to the C preprocessor
29090 signed_char Defines to pass to CPP to say whether @code{char} is signed
29092 startfile Object files to include at the start of the link
29095 Here is a small example of a spec file:
29098 %rename lib old_lib
29101 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
29104 This example renames the spec called @samp{lib} to @samp{old_lib} and
29105 then overrides the previous definition of @samp{lib} with a new one.
29106 The new definition adds in some extra command-line options before
29107 including the text of the old definition.
29109 @dfn{Spec strings} are a list of command-line options to be passed to their
29110 corresponding program. In addition, the spec strings can contain
29111 @samp{%}-prefixed sequences to substitute variable text or to
29112 conditionally insert text into the command line. Using these constructs
29113 it is possible to generate quite complex command lines.
29115 Here is a table of all defined @samp{%}-sequences for spec
29116 strings. Note that spaces are not generated automatically around the
29117 results of expanding these sequences. Therefore you can concatenate them
29118 together or combine them with constant text in a single argument.
29122 Substitute one @samp{%} into the program name or argument.
29125 Substitute the name of the input file being processed.
29128 Substitute the basename of the input file being processed.
29129 This is the substring up to (and not including) the last period
29130 and not including the directory.
29133 This is the same as @samp{%b}, but include the file suffix (text after
29137 Marks the argument containing or following the @samp{%d} as a
29138 temporary file name, so that that file is deleted if GCC exits
29139 successfully. Unlike @samp{%g}, this contributes no text to the
29142 @item %g@var{suffix}
29143 Substitute a file name that has suffix @var{suffix} and is chosen
29144 once per compilation, and mark the argument in the same way as
29145 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
29146 name is now chosen in a way that is hard to predict even when previously
29147 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
29148 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
29149 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
29150 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
29151 was simply substituted with a file name chosen once per compilation,
29152 without regard to any appended suffix (which was therefore treated
29153 just like ordinary text), making such attacks more likely to succeed.
29155 @item %u@var{suffix}
29156 Like @samp{%g}, but generates a new temporary file name
29157 each time it appears instead of once per compilation.
29159 @item %U@var{suffix}
29160 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
29161 new one if there is no such last file name. In the absence of any
29162 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
29163 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
29164 involves the generation of two distinct file names, one
29165 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
29166 simply substituted with a file name chosen for the previous @samp{%u},
29167 without regard to any appended suffix.
29169 @item %j@var{suffix}
29170 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
29171 writable, and if @option{-save-temps} is not used;
29172 otherwise, substitute the name
29173 of a temporary file, just like @samp{%u}. This temporary file is not
29174 meant for communication between processes, but rather as a junk
29175 disposal mechanism.
29177 @item %|@var{suffix}
29178 @itemx %m@var{suffix}
29179 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
29180 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
29181 all. These are the two most common ways to instruct a program that it
29182 should read from standard input or write to standard output. If you
29183 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
29184 construct: see for example @file{f/lang-specs.h}.
29186 @item %.@var{SUFFIX}
29187 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
29188 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
29189 terminated by the next space or %.
29192 Marks the argument containing or following the @samp{%w} as the
29193 designated output file of this compilation. This puts the argument
29194 into the sequence of arguments that @samp{%o} substitutes.
29197 Substitutes the names of all the output files, with spaces
29198 automatically placed around them. You should write spaces
29199 around the @samp{%o} as well or the results are undefined.
29200 @samp{%o} is for use in the specs for running the linker.
29201 Input files whose names have no recognized suffix are not compiled
29202 at all, but they are included among the output files, so they are
29206 Substitutes the suffix for object files. Note that this is
29207 handled specially when it immediately follows @samp{%g, %u, or %U},
29208 because of the need for those to form complete file names. The
29209 handling is such that @samp{%O} is treated exactly as if it had already
29210 been substituted, except that @samp{%g, %u, and %U} do not currently
29211 support additional @var{suffix} characters following @samp{%O} as they do
29212 following, for example, @samp{.o}.
29215 Substitutes the standard macro predefinitions for the
29216 current target machine. Use this when running @command{cpp}.
29219 Like @samp{%p}, but puts @samp{__} before and after the name of each
29220 predefined macro, except for macros that start with @samp{__} or with
29221 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
29225 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
29226 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
29227 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
29228 and @option{-imultilib} as necessary.
29231 Current argument is the name of a library or startup file of some sort.
29232 Search for that file in a standard list of directories and substitute
29233 the full name found. The current working directory is included in the
29234 list of directories scanned.
29237 Current argument is the name of a linker script. Search for that file
29238 in the current list of directories to scan for libraries. If the file
29239 is located insert a @option{--script} option into the command line
29240 followed by the full path name found. If the file is not found then
29241 generate an error message. Note: the current working directory is not
29245 Print @var{str} as an error message. @var{str} is terminated by a newline.
29246 Use this when inconsistent options are detected.
29248 @item %(@var{name})
29249 Substitute the contents of spec string @var{name} at this point.
29251 @item %x@{@var{option}@}
29252 Accumulate an option for @samp{%X}.
29255 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
29259 Output the accumulated assembler options specified by @option{-Wa}.
29262 Output the accumulated preprocessor options specified by @option{-Wp}.
29265 Process the @code{asm} spec. This is used to compute the
29266 switches to be passed to the assembler.
29269 Process the @code{asm_final} spec. This is a spec string for
29270 passing switches to an assembler post-processor, if such a program is
29274 Process the @code{link} spec. This is the spec for computing the
29275 command line passed to the linker. Typically it makes use of the
29276 @samp{%L %G %S %D and %E} sequences.
29279 Dump out a @option{-L} option for each directory that GCC believes might
29280 contain startup files. If the target supports multilibs then the
29281 current multilib directory is prepended to each of these paths.
29284 Process the @code{lib} spec. This is a spec string for deciding which
29285 libraries are included on the command line to the linker.
29288 Process the @code{libgcc} spec. This is a spec string for deciding
29289 which GCC support library is included on the command line to the linker.
29292 Process the @code{startfile} spec. This is a spec for deciding which
29293 object files are the first ones passed to the linker. Typically
29294 this might be a file named @file{crt0.o}.
29297 Process the @code{endfile} spec. This is a spec string that specifies
29298 the last object files that are passed to the linker.
29301 Process the @code{cpp} spec. This is used to construct the arguments
29302 to be passed to the C preprocessor.
29305 Process the @code{cc1} spec. This is used to construct the options to be
29306 passed to the actual C compiler (@command{cc1}).
29309 Process the @code{cc1plus} spec. This is used to construct the options to be
29310 passed to the actual C++ compiler (@command{cc1plus}).
29313 Substitute the variable part of a matched option. See below.
29314 Note that each comma in the substituted string is replaced by
29318 Remove all occurrences of @code{-S} from the command line. Note---this
29319 command is position dependent. @samp{%} commands in the spec string
29320 before this one see @code{-S}, @samp{%} commands in the spec string
29321 after this one do not.
29323 @item %:@var{function}(@var{args})
29324 Call the named function @var{function}, passing it @var{args}.
29325 @var{args} is first processed as a nested spec string, then split
29326 into an argument vector in the usual fashion. The function returns
29327 a string which is processed as if it had appeared literally as part
29328 of the current spec.
29330 The following built-in spec functions are provided:
29333 @item @code{getenv}
29334 The @code{getenv} spec function takes two arguments: an environment
29335 variable name and a string. If the environment variable is not
29336 defined, a fatal error is issued. Otherwise, the return value is the
29337 value of the environment variable concatenated with the string. For
29338 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
29341 %:getenv(TOPDIR /include)
29344 expands to @file{/path/to/top/include}.
29346 @item @code{if-exists}
29347 The @code{if-exists} spec function takes one argument, an absolute
29348 pathname to a file. If the file exists, @code{if-exists} returns the
29349 pathname. Here is a small example of its usage:
29353 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
29356 @item @code{if-exists-else}
29357 The @code{if-exists-else} spec function is similar to the @code{if-exists}
29358 spec function, except that it takes two arguments. The first argument is
29359 an absolute pathname to a file. If the file exists, @code{if-exists-else}
29360 returns the pathname. If it does not exist, it returns the second argument.
29361 This way, @code{if-exists-else} can be used to select one file or another,
29362 based on the existence of the first. Here is a small example of its usage:
29366 crt0%O%s %:if-exists(crti%O%s) \
29367 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
29370 @item @code{replace-outfile}
29371 The @code{replace-outfile} spec function takes two arguments. It looks for the
29372 first argument in the outfiles array and replaces it with the second argument. Here
29373 is a small example of its usage:
29376 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
29379 @item @code{remove-outfile}
29380 The @code{remove-outfile} spec function takes one argument. It looks for the
29381 first argument in the outfiles array and removes it. Here is a small example
29385 %:remove-outfile(-lm)
29388 @item @code{pass-through-libs}
29389 The @code{pass-through-libs} spec function takes any number of arguments. It
29390 finds any @option{-l} options and any non-options ending in @file{.a} (which it
29391 assumes are the names of linker input library archive files) and returns a
29392 result containing all the found arguments each prepended by
29393 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
29394 intended to be passed to the LTO linker plugin.
29397 %:pass-through-libs(%G %L %G)
29400 @item @code{print-asm-header}
29401 The @code{print-asm-header} function takes no arguments and simply
29402 prints a banner like:
29408 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
29411 It is used to separate compiler options from assembler options
29412 in the @option{--target-help} output.
29416 Substitutes the @code{-S} switch, if that switch is given to GCC@.
29417 If that switch is not specified, this substitutes nothing. Note that
29418 the leading dash is omitted when specifying this option, and it is
29419 automatically inserted if the substitution is performed. Thus the spec
29420 string @samp{%@{foo@}} matches the command-line option @option{-foo}
29421 and outputs the command-line option @option{-foo}.
29424 Like %@{@code{S}@} but mark last argument supplied within as a file to be
29425 deleted on failure.
29428 Substitutes all the switches specified to GCC whose names start
29429 with @code{-S}, but which also take an argument. This is used for
29430 switches like @option{-o}, @option{-D}, @option{-I}, etc.
29431 GCC considers @option{-o foo} as being
29432 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
29433 text, including the space. Thus two arguments are generated.
29436 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
29437 (the order of @code{S} and @code{T} in the spec is not significant).
29438 There can be any number of ampersand-separated variables; for each the
29439 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
29442 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
29445 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
29448 Substitutes @code{X} if one or more switches whose names start with
29449 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
29450 once, no matter how many such switches appeared. However, if @code{%*}
29451 appears somewhere in @code{X}, then @code{X} is substituted once
29452 for each matching switch, with the @code{%*} replaced by the part of
29453 that switch matching the @code{*}.
29455 If @code{%*} appears as the last part of a spec sequence then a space
29456 is added after the end of the last substitution. If there is more
29457 text in the sequence, however, then a space is not generated. This
29458 allows the @code{%*} substitution to be used as part of a larger
29459 string. For example, a spec string like this:
29462 %@{mcu=*:--script=%*/memory.ld@}
29466 when matching an option like @option{-mcu=newchip} produces:
29469 --script=newchip/memory.ld
29473 Substitutes @code{X}, if processing a file with suffix @code{S}.
29476 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
29479 Substitutes @code{X}, if processing a file for language @code{S}.
29482 Substitutes @code{X}, if not processing a file for language @code{S}.
29485 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
29486 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
29487 @code{*} sequences as well, although they have a stronger binding than
29488 the @samp{|}. If @code{%*} appears in @code{X}, all of the
29489 alternatives must be starred, and only the first matching alternative
29492 For example, a spec string like this:
29495 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
29499 outputs the following command-line options from the following input
29500 command-line options:
29505 -d fred.c -foo -baz -boggle
29506 -d jim.d -bar -baz -boggle
29509 @item %@{S:X; T:Y; :D@}
29511 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
29512 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
29513 be as many clauses as you need. This may be combined with @code{.},
29514 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
29519 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
29520 or similar construct can use a backslash to ignore the special meaning
29521 of the character following it, thus allowing literal matching of a
29522 character that is otherwise specially treated. For example,
29523 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
29524 @option{-std=iso9899:1999} option is given.
29526 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
29527 construct may contain other nested @samp{%} constructs or spaces, or
29528 even newlines. They are processed as usual, as described above.
29529 Trailing white space in @code{X} is ignored. White space may also
29530 appear anywhere on the left side of the colon in these constructs,
29531 except between @code{.} or @code{*} and the corresponding word.
29533 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
29534 handled specifically in these constructs. If another value of
29535 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
29536 @option{-W} switch is found later in the command line, the earlier
29537 switch value is ignored, except with @{@code{S}*@} where @code{S} is
29538 just one letter, which passes all matching options.
29540 The character @samp{|} at the beginning of the predicate text is used to
29541 indicate that a command should be piped to the following command, but
29542 only if @option{-pipe} is specified.
29544 It is built into GCC which switches take arguments and which do not.
29545 (You might think it would be useful to generalize this to allow each
29546 compiler's spec to say which switches take arguments. But this cannot
29547 be done in a consistent fashion. GCC cannot even decide which input
29548 files have been specified without knowing which switches take arguments,
29549 and it must know which input files to compile in order to tell which
29552 GCC also knows implicitly that arguments starting in @option{-l} are to be
29553 treated as compiler output files, and passed to the linker in their
29554 proper position among the other output files.
29556 @node Environment Variables
29557 @section Environment Variables Affecting GCC
29558 @cindex environment variables
29560 @c man begin ENVIRONMENT
29561 This section describes several environment variables that affect how GCC
29562 operates. Some of them work by specifying directories or prefixes to use
29563 when searching for various kinds of files. Some are used to specify other
29564 aspects of the compilation environment.
29566 Note that you can also specify places to search using options such as
29567 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
29568 take precedence over places specified using environment variables, which
29569 in turn take precedence over those specified by the configuration of GCC@.
29570 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
29571 GNU Compiler Collection (GCC) Internals}.
29576 @c @itemx LC_COLLATE
29578 @c @itemx LC_MONETARY
29579 @c @itemx LC_NUMERIC
29584 @c @findex LC_COLLATE
29585 @findex LC_MESSAGES
29586 @c @findex LC_MONETARY
29587 @c @findex LC_NUMERIC
29591 These environment variables control the way that GCC uses
29592 localization information which allows GCC to work with different
29593 national conventions. GCC inspects the locale categories
29594 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
29595 so. These locale categories can be set to any value supported by your
29596 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
29597 Kingdom encoded in UTF-8.
29599 The @env{LC_CTYPE} environment variable specifies character
29600 classification. GCC uses it to determine the character boundaries in
29601 a string; this is needed for some multibyte encodings that contain quote
29602 and escape characters that are otherwise interpreted as a string
29605 The @env{LC_MESSAGES} environment variable specifies the language to
29606 use in diagnostic messages.
29608 If the @env{LC_ALL} environment variable is set, it overrides the value
29609 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
29610 and @env{LC_MESSAGES} default to the value of the @env{LANG}
29611 environment variable. If none of these variables are set, GCC
29612 defaults to traditional C English behavior.
29616 If @env{TMPDIR} is set, it specifies the directory to use for temporary
29617 files. GCC uses temporary files to hold the output of one stage of
29618 compilation which is to be used as input to the next stage: for example,
29619 the output of the preprocessor, which is the input to the compiler
29622 @item GCC_COMPARE_DEBUG
29623 @findex GCC_COMPARE_DEBUG
29624 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
29625 @option{-fcompare-debug} to the compiler driver. See the documentation
29626 of this option for more details.
29628 @item GCC_EXEC_PREFIX
29629 @findex GCC_EXEC_PREFIX
29630 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
29631 names of the subprograms executed by the compiler. No slash is added
29632 when this prefix is combined with the name of a subprogram, but you can
29633 specify a prefix that ends with a slash if you wish.
29635 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
29636 an appropriate prefix to use based on the pathname it is invoked with.
29638 If GCC cannot find the subprogram using the specified prefix, it
29639 tries looking in the usual places for the subprogram.
29641 The default value of @env{GCC_EXEC_PREFIX} is
29642 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
29643 the installed compiler. In many cases @var{prefix} is the value
29644 of @code{prefix} when you ran the @file{configure} script.
29646 Other prefixes specified with @option{-B} take precedence over this prefix.
29648 This prefix is also used for finding files such as @file{crt0.o} that are
29651 In addition, the prefix is used in an unusual way in finding the
29652 directories to search for header files. For each of the standard
29653 directories whose name normally begins with @samp{/usr/local/lib/gcc}
29654 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
29655 replacing that beginning with the specified prefix to produce an
29656 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
29657 @file{foo/bar} just before it searches the standard directory
29658 @file{/usr/local/lib/bar}.
29659 If a standard directory begins with the configured
29660 @var{prefix} then the value of @var{prefix} is replaced by
29661 @env{GCC_EXEC_PREFIX} when looking for header files.
29663 @item COMPILER_PATH
29664 @findex COMPILER_PATH
29665 The value of @env{COMPILER_PATH} is a colon-separated list of
29666 directories, much like @env{PATH}. GCC tries the directories thus
29667 specified when searching for subprograms, if it cannot find the
29668 subprograms using @env{GCC_EXEC_PREFIX}.
29671 @findex LIBRARY_PATH
29672 The value of @env{LIBRARY_PATH} is a colon-separated list of
29673 directories, much like @env{PATH}. When configured as a native compiler,
29674 GCC tries the directories thus specified when searching for special
29675 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
29676 using GCC also uses these directories when searching for ordinary
29677 libraries for the @option{-l} option (but directories specified with
29678 @option{-L} come first).
29682 @cindex locale definition
29683 This variable is used to pass locale information to the compiler. One way in
29684 which this information is used is to determine the character set to be used
29685 when character literals, string literals and comments are parsed in C and C++.
29686 When the compiler is configured to allow multibyte characters,
29687 the following values for @env{LANG} are recognized:
29691 Recognize JIS characters.
29693 Recognize SJIS characters.
29695 Recognize EUCJP characters.
29698 If @env{LANG} is not defined, or if it has some other value, then the
29699 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
29700 recognize and translate multibyte characters.
29704 Some additional environment variables affect the behavior of the
29707 @include cppenv.texi
29711 @node Precompiled Headers
29712 @section Using Precompiled Headers
29713 @cindex precompiled headers
29714 @cindex speed of compilation
29716 Often large projects have many header files that are included in every
29717 source file. The time the compiler takes to process these header files
29718 over and over again can account for nearly all of the time required to
29719 build the project. To make builds faster, GCC allows you to
29720 @dfn{precompile} a header file.
29722 To create a precompiled header file, simply compile it as you would any
29723 other file, if necessary using the @option{-x} option to make the driver
29724 treat it as a C or C++ header file. You may want to use a
29725 tool like @command{make} to keep the precompiled header up-to-date when
29726 the headers it contains change.
29728 A precompiled header file is searched for when @code{#include} is
29729 seen in the compilation. As it searches for the included file
29730 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29731 compiler looks for a precompiled header in each directory just before it
29732 looks for the include file in that directory. The name searched for is
29733 the name specified in the @code{#include} with @samp{.gch} appended. If
29734 the precompiled header file cannot be used, it is ignored.
29736 For instance, if you have @code{#include "all.h"}, and you have
29737 @file{all.h.gch} in the same directory as @file{all.h}, then the
29738 precompiled header file is used if possible, and the original
29739 header is used otherwise.
29741 Alternatively, you might decide to put the precompiled header file in a
29742 directory and use @option{-I} to ensure that directory is searched
29743 before (or instead of) the directory containing the original header.
29744 Then, if you want to check that the precompiled header file is always
29745 used, you can put a file of the same name as the original header in this
29746 directory containing an @code{#error} command.
29748 This also works with @option{-include}. So yet another way to use
29749 precompiled headers, good for projects not designed with precompiled
29750 header files in mind, is to simply take most of the header files used by
29751 a project, include them from another header file, precompile that header
29752 file, and @option{-include} the precompiled header. If the header files
29753 have guards against multiple inclusion, they are skipped because
29754 they've already been included (in the precompiled header).
29756 If you need to precompile the same header file for different
29757 languages, targets, or compiler options, you can instead make a
29758 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29759 header in the directory, perhaps using @option{-o}. It doesn't matter
29760 what you call the files in the directory; every precompiled header in
29761 the directory is considered. The first precompiled header
29762 encountered in the directory that is valid for this compilation is
29763 used; they're searched in no particular order.
29765 There are many other possibilities, limited only by your imagination,
29766 good sense, and the constraints of your build system.
29768 A precompiled header file can be used only when these conditions apply:
29772 Only one precompiled header can be used in a particular compilation.
29775 A precompiled header cannot be used once the first C token is seen. You
29776 can have preprocessor directives before a precompiled header; you cannot
29777 include a precompiled header from inside another header.
29780 The precompiled header file must be produced for the same language as
29781 the current compilation. You cannot use a C precompiled header for a C++
29785 The precompiled header file must have been produced by the same compiler
29786 binary as the current compilation is using.
29789 Any macros defined before the precompiled header is included must
29790 either be defined in the same way as when the precompiled header was
29791 generated, or must not affect the precompiled header, which usually
29792 means that they don't appear in the precompiled header at all.
29794 The @option{-D} option is one way to define a macro before a
29795 precompiled header is included; using a @code{#define} can also do it.
29796 There are also some options that define macros implicitly, like
29797 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
29800 @item If debugging information is output when using the precompiled
29801 header, using @option{-g} or similar, the same kind of debugging information
29802 must have been output when building the precompiled header. However,
29803 a precompiled header built using @option{-g} can be used in a compilation
29804 when no debugging information is being output.
29806 @item The same @option{-m} options must generally be used when building
29807 and using the precompiled header. @xref{Submodel Options},
29808 for any cases where this rule is relaxed.
29810 @item Each of the following options must be the same when building and using
29811 the precompiled header:
29813 @gccoptlist{-fexceptions}
29816 Some other command-line options starting with @option{-f},
29817 @option{-p}, or @option{-O} must be defined in the same way as when
29818 the precompiled header was generated. At present, it's not clear
29819 which options are safe to change and which are not; the safest choice
29820 is to use exactly the same options when generating and using the
29821 precompiled header. The following are known to be safe:
29823 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
29824 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
29825 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
29830 For all of these except the last, the compiler automatically
29831 ignores the precompiled header if the conditions aren't met. If you
29832 find an option combination that doesn't work and doesn't cause the
29833 precompiled header to be ignored, please consider filing a bug report,
29836 If you do use differing options when generating and using the
29837 precompiled header, the actual behavior is a mixture of the
29838 behavior for the options. For instance, if you use @option{-g} to
29839 generate the precompiled header but not when using it, you may or may
29840 not get debugging information for routines in the precompiled header.