1 @c Copyright (C) 1988-2019 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{} 1988-2019 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 @gol
201 -fopenacc -fopenacc-dim=@var{geom} @gol
202 -fopenmp -fopenmp-simd @gol
203 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness} @gol
204 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
205 -fsigned-bitfields -fsigned-char @gol
206 -funsigned-bitfields -funsigned-char}
208 @item C++ Language Options
209 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
210 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
211 -faligned-new=@var{n} -fargs-in-order=@var{n} -fchar8_t -fcheck-new @gol
212 -fconstexpr-depth=@var{n} -fconstexpr-cache-depth=@var{n} @gol
213 -fconstexpr-loop-limit=@var{n} -fconstexpr-ops-limit=@var{n} @gol
214 -fno-elide-constructors @gol
215 -fno-enforce-eh-specs @gol
216 -fno-gnu-keywords @gol
217 -fno-implicit-templates @gol
218 -fno-implicit-inline-templates @gol
219 -fno-implement-inlines -fms-extensions @gol
220 -fnew-inheriting-ctors @gol
221 -fnew-ttp-matching @gol
222 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
223 -fno-optional-diags -fpermissive @gol
224 -fno-pretty-templates @gol
225 -fno-rtti -fsized-deallocation @gol
226 -ftemplate-backtrace-limit=@var{n} @gol
227 -ftemplate-depth=@var{n} @gol
228 -fno-threadsafe-statics -fuse-cxa-atexit @gol
229 -fno-weak -nostdinc++ @gol
230 -fvisibility-inlines-hidden @gol
231 -fvisibility-ms-compat @gol
232 -fext-numeric-literals @gol
233 -Wabi=@var{n} -Wabi-tag -Wcomma-subscript -Wconversion-null @gol
234 -Wctor-dtor-privacy @gol
235 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor @gol
236 -Wliteral-suffix @gol
237 -Wmultiple-inheritance -Wno-init-list-lifetime @gol
238 -Wnamespaces -Wnarrowing @gol
239 -Wpessimizing-move -Wredundant-move @gol
240 -Wnoexcept -Wnoexcept-type -Wclass-memaccess @gol
241 -Wnon-virtual-dtor -Wreorder -Wregister @gol
242 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
243 -Wno-non-template-friend -Wold-style-cast @gol
244 -Woverloaded-virtual -Wno-pmf-conversions @gol
245 -Wno-class-conversion -Wno-terminate @gol
246 -Wsign-promo -Wvirtual-inheritance -Wvolatile}
248 @item Objective-C and Objective-C++ Language Options
249 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
250 Objective-C and Objective-C++ Dialects}.
251 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
252 -fgnu-runtime -fnext-runtime @gol
253 -fno-nil-receivers @gol
254 -fobjc-abi-version=@var{n} @gol
255 -fobjc-call-cxx-cdtors @gol
256 -fobjc-direct-dispatch @gol
257 -fobjc-exceptions @gol
260 -fobjc-std=objc1 @gol
261 -fno-local-ivars @gol
262 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
263 -freplace-objc-classes @gol
266 -Wassign-intercept @gol
267 -Wno-protocol -Wselector @gol
268 -Wstrict-selector-match @gol
269 -Wundeclared-selector}
271 @item Diagnostic Message Formatting Options
272 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
273 @gccoptlist{-fmessage-length=@var{n} @gol
274 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
275 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
276 -fdiagnostics-format=@r{[}text@r{|}json@r{]} @gol
277 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
278 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers @gol
279 -fdiagnostics-minimum-margin-width=@var{width} @gol
280 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch @gol
281 -fdiagnostics-show-template-tree -fno-elide-type @gol
284 @item Warning Options
285 @xref{Warning Options,,Options to Request or Suppress Warnings}.
286 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
287 -pedantic-errors @gol
288 -w -Wextra -Wall -Waddress -Waddress-of-packed-member @gol
289 -Waggregate-return -Waligned-new @gol
290 -Walloc-zero -Walloc-size-larger-than=@var{byte-size} @gol
291 -Walloca -Walloca-larger-than=@var{byte-size} @gol
292 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
293 -Wno-attributes -Wattribute-alias=@var{n} @gol
294 -Wbool-compare -Wbool-operation @gol
295 -Wno-builtin-declaration-mismatch @gol
296 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat @gol
297 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat @gol
299 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual @gol
300 -Wchar-subscripts -Wcatch-value -Wcatch-value=@var{n} @gol
301 -Wclobbered -Wcomment -Wconditionally-supported @gol
302 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
303 -Wdelete-incomplete @gol
304 -Wno-attribute-warning @gol
305 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
306 -Wdisabled-optimization @gol
307 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
308 -Wno-div-by-zero -Wdouble-promotion @gol
309 -Wduplicated-branches -Wduplicated-cond @gol
310 -Wempty-body -Wenum-compare -Wenum-conversion @gol
311 -Wno-endif-labels -Wexpansion-to-defined @gol
312 -Werror -Werror=* -Wextra-semi -Wfatal-errors @gol
313 -Wfloat-equal -Wformat -Wformat=2 @gol
314 -Wno-format-contains-nul -Wno-format-extra-args @gol
315 -Wformat-nonliteral -Wformat-overflow=@var{n} @gol
316 -Wformat-security -Wformat-signedness -Wformat-truncation=@var{n} @gol
317 -Wformat-y2k -Wframe-address @gol
318 -Wframe-larger-than=@var{byte-size} -Wno-free-nonheap-object @gol
319 -Wjump-misses-init @gol
320 -Whsa -Wif-not-aligned @gol
321 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
322 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=@var{n} @gol
323 -Wimplicit-function-declaration -Wimplicit-int @gol
324 -Winaccessible-base @gol
325 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context @gol
326 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
327 -Winvalid-pch -Wlarger-than=@var{byte-size} @gol
328 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
329 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
330 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces @gol
331 -Wmissing-field-initializers -Wmissing-format-attribute @gol
332 -Wmissing-include-dirs -Wmissing-noreturn -Wmissing-profile @gol
333 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare @gol
334 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
335 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
336 -Woverride-init-side-effects -Woverlength-strings @gol
337 -Wpacked -Wpacked-bitfield-compat -Wpacked-not-aligned -Wpadded @gol
338 -Wparentheses -Wno-pedantic-ms-format @gol
339 -Wplacement-new -Wplacement-new=@var{n} @gol
340 -Wpointer-arith -Wpointer-compare -Wno-pointer-to-int-cast @gol
341 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls @gol
342 -Wrestrict -Wno-return-local-addr @gol
343 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
344 -Wshadow=global, -Wshadow=local, -Wshadow=compatible-local @gol
345 -Wshift-overflow -Wshift-overflow=@var{n} @gol
346 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
347 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
348 -Wno-scalar-storage-order -Wsizeof-pointer-div @gol
349 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
350 -Wstack-protector -Wstack-usage=@var{byte-size} -Wstrict-aliasing @gol
351 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
352 -Wstringop-overflow=@var{n} -Wstringop-truncation -Wsubobject-linkage @gol
353 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}malloc@r{]} @gol
354 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
355 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
356 -Wswitch-unreachable -Wsync-nand @gol
357 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
358 -Wtype-limits -Wundef @gol
359 -Wuninitialized -Wunknown-pragmas @gol
360 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
361 -Wunused-label -Wunused-local-typedefs -Wunused-macros @gol
362 -Wunused-parameter -Wno-unused-result @gol
363 -Wunused-value -Wunused-variable @gol
364 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
365 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
366 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
367 -Wvla -Wvla-larger-than=@var{byte-size} -Wvolatile-register-var @gol
369 -Wzero-as-null-pointer-constant}
371 @item C and Objective-C-only Warning Options
372 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
373 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
374 -Wold-style-declaration -Wold-style-definition @gol
375 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
376 -Wdeclaration-after-statement -Wpointer-sign}
378 @item Debugging Options
379 @xref{Debugging Options,,Options for Debugging Your Program}.
380 @gccoptlist{-g -g@var{level} -gdwarf -gdwarf-@var{version} @gol
381 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
382 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
383 -gas-loc-support -gno-as-loc-support @gol
384 -gas-locview-support -gno-as-locview-support @gol
385 -gcolumn-info -gno-column-info @gol
386 -gstatement-frontiers -gno-statement-frontiers @gol
387 -gvariable-location-views -gno-variable-location-views @gol
388 -ginternal-reset-location-views -gno-internal-reset-location-views @gol
389 -ginline-points -gno-inline-points @gol
390 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
391 -gsplit-dwarf -gdescribe-dies -gno-describe-dies @gol
392 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
393 -fno-eliminate-unused-debug-types @gol
394 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
395 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
396 -fno-eliminate-unused-debug-symbols -femit-class-debug-always @gol
397 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
398 -fvar-tracking -fvar-tracking-assignments}
400 @item Optimization Options
401 @xref{Optimize Options,,Options that Control Optimization}.
402 @gccoptlist{-faggressive-loop-optimizations @gol
403 -falign-functions[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
404 -falign-jumps[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
405 -falign-labels[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
406 -falign-loops[=@var{n}[:@var{m}:[@var{n2}[:@var{m2}]]]] @gol
407 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
408 -fauto-inc-dec -fbranch-probabilities @gol
410 -fcombine-stack-adjustments -fconserve-stack @gol
411 -fcompare-elim -fcprop-registers -fcrossjumping @gol
412 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
413 -fcx-limited-range @gol
414 -fdata-sections -fdce -fdelayed-branch @gol
415 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
416 -fdevirtualize-at-ltrans -fdse @gol
417 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
418 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
420 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
421 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
422 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
423 -fif-conversion2 -findirect-inlining @gol
424 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
425 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
426 -fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const @gol
427 -fipa-reference -fipa-reference-addressable @gol
428 -fipa-stack-alignment -fipa-icf -fira-algorithm=@var{algorithm} @gol
429 -flive-patching=@var{level} @gol
430 -fira-region=@var{region} -fira-hoist-pressure @gol
431 -fira-loop-pressure -fno-ira-share-save-slots @gol
432 -fno-ira-share-spill-slots @gol
433 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
434 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
435 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage @gol
436 -floop-block -floop-interchange -floop-strip-mine @gol
437 -floop-unroll-and-jam -floop-nest-optimize @gol
438 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
439 -flto-partition=@var{alg} -fmerge-all-constants @gol
440 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
441 -fmove-loop-invariants -fno-branch-count-reg @gol
442 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
443 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
444 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock @gol
445 -fno-sched-spec -fno-signed-zeros @gol
446 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
447 -fomit-frame-pointer -foptimize-sibling-calls @gol
448 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
449 -fprefetch-loop-arrays @gol
450 -fprofile-correction @gol
451 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
452 -fprofile-reorder-functions @gol
453 -freciprocal-math -free -frename-registers -freorder-blocks @gol
454 -freorder-blocks-algorithm=@var{algorithm} @gol
455 -freorder-blocks-and-partition -freorder-functions @gol
456 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
457 -frounding-math -fsave-optimization-record @gol
458 -fsched2-use-superblocks -fsched-pressure @gol
459 -fsched-spec-load -fsched-spec-load-dangerous @gol
460 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
461 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
462 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
463 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
464 -fschedule-fusion @gol
465 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
466 -fselective-scheduling -fselective-scheduling2 @gol
467 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
468 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate @gol
469 -fsignaling-nans @gol
470 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops@gol
472 -fsplit-wide-types -fsplit-wide-types-early -fssa-backprop -fssa-phiopt @gol
473 -fstdarg-opt -fstore-merging -fstrict-aliasing @gol
474 -fthread-jumps -ftracer -ftree-bit-ccp @gol
475 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
476 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
477 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting @gol
478 -ftree-loop-if-convert -ftree-loop-im @gol
479 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
480 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
481 -ftree-loop-vectorize @gol
482 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
483 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra @gol
484 -ftree-switch-conversion -ftree-tail-merge @gol
485 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
486 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
487 -funsafe-math-optimizations -funswitch-loops @gol
488 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
489 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
490 --param @var{name}=@var{value}
491 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
493 @item Program Instrumentation Options
494 @xref{Instrumentation Options,,Program Instrumentation Options}.
495 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
496 -fprofile-abs-path @gol
497 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
498 -fprofile-note=@var{path} -fprofile-update=@var{method} @gol
499 -fprofile-filter-files=@var{regex} -fprofile-exclude-files=@var{regex} @gol
500 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
501 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
502 -fsanitize-undefined-trap-on-error -fbounds-check @gol
503 -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]} @gol
504 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
505 -fstack-protector-explicit -fstack-check @gol
506 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
507 -fno-stack-limit -fsplit-stack @gol
508 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
509 -fvtv-counts -fvtv-debug @gol
510 -finstrument-functions @gol
511 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
512 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
514 @item Preprocessor Options
515 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
516 @gccoptlist{-A@var{question}=@var{answer} @gol
517 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
518 -C -CC -D@var{macro}@r{[}=@var{defn}@r{]} @gol
519 -dD -dI -dM -dN -dU @gol
520 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers @gol
521 -fexec-charset=@var{charset} -fextended-identifiers @gol
522 -finput-charset=@var{charset} -fmacro-prefix-map=@var{old}=@var{new} @gol
523 -fmax-include-depth=@var{depth} @gol
524 -fno-canonical-system-headers -fpch-deps -fpch-preprocess @gol
525 -fpreprocessed -ftabstop=@var{width} -ftrack-macro-expansion @gol
526 -fwide-exec-charset=@var{charset} -fworking-directory @gol
527 -H -imacros @var{file} -include @var{file} @gol
528 -M -MD -MF -MG -MM -MMD -MP -MQ -MT @gol
529 -no-integrated-cpp -P -pthread -remap @gol
530 -traditional -traditional-cpp -trigraphs @gol
531 -U@var{macro} -undef @gol
532 -Wp,@var{option} -Xpreprocessor @var{option}}
534 @item Assembler Options
535 @xref{Assembler Options,,Passing Options to the Assembler}.
536 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
539 @xref{Link Options,,Options for Linking}.
540 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
541 -nostartfiles -nodefaultlibs -nolibc -nostdlib @gol
542 -e @var{entry} --entry=@var{entry} @gol
543 -pie -pthread -r -rdynamic @gol
544 -s -static -static-pie -static-libgcc -static-libstdc++ @gol
545 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
546 -shared -shared-libgcc -symbolic @gol
547 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
548 -u @var{symbol} -z @var{keyword}}
550 @item Directory Options
551 @xref{Directory Options,,Options for Directory Search}.
552 @gccoptlist{-B@var{prefix} -I@var{dir} -I- @gol
553 -idirafter @var{dir} @gol
554 -imacros @var{file} -imultilib @var{dir} @gol
555 -iplugindir=@var{dir} -iprefix @var{file} @gol
556 -iquote @var{dir} -isysroot @var{dir} -isystem @var{dir} @gol
557 -iwithprefix @var{dir} -iwithprefixbefore @var{dir} @gol
558 -L@var{dir} -no-canonical-prefixes --no-sysroot-suffix @gol
559 -nostdinc -nostdinc++ --sysroot=@var{dir}}
561 @item Code Generation Options
562 @xref{Code Gen Options,,Options for Code Generation Conventions}.
563 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
564 -ffixed-@var{reg} -fexceptions @gol
565 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
566 -fasynchronous-unwind-tables @gol
568 -finhibit-size-directive -fno-common -fno-ident @gol
569 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
570 -fno-jump-tables @gol
571 -frecord-gcc-switches @gol
572 -freg-struct-return -fshort-enums -fshort-wchar @gol
573 -fverbose-asm -fpack-struct[=@var{n}] @gol
574 -fleading-underscore -ftls-model=@var{model} @gol
575 -fstack-reuse=@var{reuse_level} @gol
576 -ftrampolines -ftrapv -fwrapv @gol
577 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
578 -fstrict-volatile-bitfields -fsync-libcalls}
580 @item Developer Options
581 @xref{Developer Options,,GCC Developer Options}.
582 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
583 -dumpfullversion -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
584 -fdbg-cnt=@var{counter-value-list} @gol
585 -fdisable-ipa-@var{pass_name} @gol
586 -fdisable-rtl-@var{pass_name} @gol
587 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
588 -fdisable-tree-@var{pass_name} @gol
589 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
590 -fdump-debug -fdump-earlydebug @gol
591 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
592 -fdump-final-insns@r{[}=@var{file}@r{]} @gol
593 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
595 -fdump-lang-@var{switch} @gol
596 -fdump-lang-@var{switch}-@var{options} @gol
597 -fdump-lang-@var{switch}-@var{options}=@var{filename} @gol
599 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
600 -fdump-statistics @gol
602 -fdump-tree-@var{switch} @gol
603 -fdump-tree-@var{switch}-@var{options} @gol
604 -fdump-tree-@var{switch}-@var{options}=@var{filename} @gol
605 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
606 -fenable-@var{kind}-@var{pass} @gol
607 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
608 -fira-verbose=@var{n} @gol
609 -flto-report -flto-report-wpa -fmem-report-wpa @gol
610 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
611 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
612 -fprofile-report @gol
613 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
614 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
615 -fstats -fstack-usage -ftime-report -ftime-report-details @gol
616 -fvar-tracking-assignments-toggle -gtoggle @gol
617 -print-file-name=@var{library} -print-libgcc-file-name @gol
618 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
619 -print-prog-name=@var{program} -print-search-dirs -Q @gol
620 -print-sysroot -print-sysroot-headers-suffix @gol
621 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
623 @item Machine-Dependent Options
624 @xref{Submodel Options,,Machine-Dependent Options}.
625 @c This list is ordered alphanumerically by subsection name.
626 @c Try and put the significant identifier (CPU or system) first,
627 @c so users have a clue at guessing where the ones they want will be.
629 @emph{AArch64 Options}
630 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
631 -mgeneral-regs-only @gol
632 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
633 -mstrict-align -mno-strict-align @gol
634 -momit-leaf-frame-pointer @gol
635 -mtls-dialect=desc -mtls-dialect=traditional @gol
636 -mtls-size=@var{size} @gol
637 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 @gol
638 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div @gol
639 -mpc-relative-literal-loads @gol
640 -msign-return-address=@var{scope} @gol
641 -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}
642 +@var{b-key}]|@var{bti} @gol
643 -march=@var{name} -mcpu=@var{name} -mtune=@var{name} @gol
644 -moverride=@var{string} -mverbose-cost-dump @gol
645 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{sysreg} @gol
646 -mstack-protector-guard-offset=@var{offset} -mtrack-speculation }
648 @emph{Adapteva Epiphany Options}
649 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
650 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
651 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
652 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
653 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
654 -msplit-vecmove-early -m1reg-@var{reg}}
656 @emph{AMD GCN Options}
657 @gccoptlist{-march=@var{gpu} -mtune=@var{gpu} -mstack-size=@var{bytes}}
660 @gccoptlist{-mbarrel-shifter -mjli-always @gol
661 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
662 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
663 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
664 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
665 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
666 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
667 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved @gol
668 -mrgf-banked-regs -mlpc-width=@var{width} -G @var{num} @gol
669 -mvolatile-cache -mtp-regno=@var{regno} @gol
670 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
671 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
672 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
673 -mlra-priority-compact mlra-priority-noncompact -mmillicode @gol
674 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
675 -mtune=@var{cpu} -mmultcost=@var{num} -mcode-density-frame @gol
676 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
677 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu} -mrf16 -mbranch-index}
680 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
681 -mabi=@var{name} @gol
682 -mapcs-stack-check -mno-apcs-stack-check @gol
683 -mapcs-reentrant -mno-apcs-reentrant @gol
684 -mgeneral-regs-only @gol
685 -msched-prolog -mno-sched-prolog @gol
686 -mlittle-endian -mbig-endian @gol
688 -mfloat-abi=@var{name} @gol
689 -mfp16-format=@var{name}
690 -mthumb-interwork -mno-thumb-interwork @gol
691 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
692 -mtune=@var{name} -mprint-tune-info @gol
693 -mstructure-size-boundary=@var{n} @gol
694 -mabort-on-noreturn @gol
695 -mlong-calls -mno-long-calls @gol
696 -msingle-pic-base -mno-single-pic-base @gol
697 -mpic-register=@var{reg} @gol
698 -mnop-fun-dllimport @gol
699 -mpoke-function-name @gol
700 -mthumb -marm -mflip-thumb @gol
701 -mtpcs-frame -mtpcs-leaf-frame @gol
702 -mcaller-super-interworking -mcallee-super-interworking @gol
703 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
704 -mword-relocations @gol
705 -mfix-cortex-m3-ldrd @gol
706 -munaligned-access @gol
707 -mneon-for-64bits @gol
708 -mslow-flash-data @gol
709 -masm-syntax-unified @gol
711 -mverbose-cost-dump @gol
717 @gccoptlist{-mmcu=@var{mcu} -mabsdata -maccumulate-args @gol
718 -mbranch-cost=@var{cost} @gol
719 -mcall-prologues -mgas-isr-prologues -mint8 @gol
720 -mn_flash=@var{size} -mno-interrupts @gol
721 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack @gol
722 -mfract-convert-truncate @gol
723 -mshort-calls -nodevicelib @gol
724 -Waddr-space-convert -Wmisspelled-isr}
726 @emph{Blackfin Options}
727 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
728 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
729 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
730 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
731 -mno-id-shared-library -mshared-library-id=@var{n} @gol
732 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
733 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
734 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
738 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
739 -msim -msdata=@var{sdata-type}}
742 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
743 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
744 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
745 -mstack-align -mdata-align -mconst-align @gol
746 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
747 -melf -maout -melinux -mlinux -sim -sim2 @gol
748 -mmul-bug-workaround -mno-mul-bug-workaround}
751 @gccoptlist{-mmac @gol
752 -mcr16cplus -mcr16c @gol
753 -msim -mint32 -mbit-ops
754 -mdata-model=@var{model}}
757 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
758 -mbig-endian -EB -mlittle-endian -EL @gol
759 -mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
760 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
761 -mdsp -medsp -mvdsp @gol
762 -mdiv -msmart -mhigh-registers -manchor @gol
763 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
764 -mbranch-cost=@var{n} -mcse-cc -msched-prolog}
766 @emph{Darwin Options}
767 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
768 -arch_only -bind_at_load -bundle -bundle_loader @gol
769 -client_name -compatibility_version -current_version @gol
771 -dependency-file -dylib_file -dylinker_install_name @gol
772 -dynamic -dynamiclib -exported_symbols_list @gol
773 -filelist -flat_namespace -force_cpusubtype_ALL @gol
774 -force_flat_namespace -headerpad_max_install_names @gol
776 -image_base -init -install_name -keep_private_externs @gol
777 -multi_module -multiply_defined -multiply_defined_unused @gol
778 -noall_load -no_dead_strip_inits_and_terms @gol
779 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
780 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
781 -private_bundle -read_only_relocs -sectalign @gol
782 -sectobjectsymbols -whyload -seg1addr @gol
783 -sectcreate -sectobjectsymbols -sectorder @gol
784 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
785 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
786 -segprot -segs_read_only_addr -segs_read_write_addr @gol
787 -single_module -static -sub_library -sub_umbrella @gol
788 -twolevel_namespace -umbrella -undefined @gol
789 -unexported_symbols_list -weak_reference_mismatches @gol
790 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
791 -mkernel -mone-byte-bool}
793 @emph{DEC Alpha Options}
794 @gccoptlist{-mno-fp-regs -msoft-float @gol
795 -mieee -mieee-with-inexact -mieee-conformant @gol
796 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
797 -mtrap-precision=@var{mode} -mbuild-constants @gol
798 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
799 -mbwx -mmax -mfix -mcix @gol
800 -mfloat-vax -mfloat-ieee @gol
801 -mexplicit-relocs -msmall-data -mlarge-data @gol
802 -msmall-text -mlarge-text @gol
803 -mmemory-latency=@var{time}}
806 @gccoptlist{-mbig-endian -mlittle-endian -mkernel=@var{version}
807 -mframe-limit=@var{bytes}}
810 @gccoptlist{-msmall-model -mno-lsim}
813 @gccoptlist{-msim -mlra -mnodiv -mft32b -mcompress -mnopm}
816 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
817 -mhard-float -msoft-float @gol
818 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
819 -mdouble -mno-double @gol
820 -mmedia -mno-media -mmuladd -mno-muladd @gol
821 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
822 -mlinked-fp -mlong-calls -malign-labels @gol
823 -mlibrary-pic -macc-4 -macc-8 @gol
824 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
825 -moptimize-membar -mno-optimize-membar @gol
826 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
827 -mvliw-branch -mno-vliw-branch @gol
828 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
829 -mno-nested-cond-exec -mtomcat-stats @gol
833 @emph{GNU/Linux Options}
834 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
835 -tno-android-cc -tno-android-ld}
837 @emph{H8/300 Options}
838 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
841 @gccoptlist{-march=@var{architecture-type} @gol
842 -mcaller-copies -mdisable-fpregs -mdisable-indexing @gol
843 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
844 -mfixed-range=@var{register-range} @gol
845 -mjump-in-delay -mlinker-opt -mlong-calls @gol
846 -mlong-load-store -mno-disable-fpregs @gol
847 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
848 -mno-jump-in-delay -mno-long-load-store @gol
849 -mno-portable-runtime -mno-soft-float @gol
850 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
851 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
852 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
853 -munix=@var{unix-std} -nolibdld -static -threads}
856 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
857 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
858 -mconstant-gp -mauto-pic -mfused-madd @gol
859 -minline-float-divide-min-latency @gol
860 -minline-float-divide-max-throughput @gol
861 -mno-inline-float-divide @gol
862 -minline-int-divide-min-latency @gol
863 -minline-int-divide-max-throughput @gol
864 -mno-inline-int-divide @gol
865 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
866 -mno-inline-sqrt @gol
867 -mdwarf2-asm -mearly-stop-bits @gol
868 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
869 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
870 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
871 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
872 -msched-spec-ldc -msched-spec-control-ldc @gol
873 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
874 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
875 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
876 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
879 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
880 -msign-extend-enabled -muser-enabled}
882 @emph{M32R/D Options}
883 @gccoptlist{-m32r2 -m32rx -m32r @gol
885 -malign-loops -mno-align-loops @gol
886 -missue-rate=@var{number} @gol
887 -mbranch-cost=@var{number} @gol
888 -mmodel=@var{code-size-model-type} @gol
889 -msdata=@var{sdata-type} @gol
890 -mno-flush-func -mflush-func=@var{name} @gol
891 -mno-flush-trap -mflush-trap=@var{number} @gol
895 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
897 @emph{M680x0 Options}
898 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
899 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
900 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
901 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
902 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
903 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
904 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
905 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
906 -mxgot -mno-xgot -mlong-jump-table-offsets}
909 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
910 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
911 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
912 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
913 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
916 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
917 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
918 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
919 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
922 @emph{MicroBlaze Options}
923 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
924 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
925 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
926 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
927 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model} @gol
928 -mpic-data-is-text-relative}
931 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
932 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
933 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
934 -mips16 -mno-mips16 -mflip-mips16 @gol
935 -minterlink-compressed -mno-interlink-compressed @gol
936 -minterlink-mips16 -mno-interlink-mips16 @gol
937 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
938 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
939 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
940 -mno-float -msingle-float -mdouble-float @gol
941 -modd-spreg -mno-odd-spreg @gol
942 -mabs=@var{mode} -mnan=@var{encoding} @gol
943 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
946 -mvirt -mno-virt @gol
949 -mginv -mno-ginv @gol
950 -mmicromips -mno-micromips @gol
952 -mloongson-mmi -mno-loongson-mmi @gol
953 -mloongson-ext -mno-loongson-ext @gol
954 -mloongson-ext2 -mno-loongson-ext2 @gol
955 -mfpu=@var{fpu-type} @gol
956 -msmartmips -mno-smartmips @gol
957 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
958 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
959 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
960 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
961 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
962 -membedded-data -mno-embedded-data @gol
963 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
964 -mcode-readable=@var{setting} @gol
965 -msplit-addresses -mno-split-addresses @gol
966 -mexplicit-relocs -mno-explicit-relocs @gol
967 -mcheck-zero-division -mno-check-zero-division @gol
968 -mdivide-traps -mdivide-breaks @gol
969 -mload-store-pairs -mno-load-store-pairs @gol
970 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
971 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
972 -mfix-24k -mno-fix-24k @gol
973 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
974 -mfix-r5900 -mno-fix-r5900 @gol
975 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
976 -mfix-vr4120 -mno-fix-vr4120 @gol
977 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
978 -mflush-func=@var{func} -mno-flush-func @gol
979 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
980 -mcompact-branches=@var{policy} @gol
981 -mfp-exceptions -mno-fp-exceptions @gol
982 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
983 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 @gol
984 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
985 -mframe-header-opt -mno-frame-header-opt}
988 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
989 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
990 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
991 -mno-base-addresses -msingle-exit -mno-single-exit}
993 @emph{MN10300 Options}
994 @gccoptlist{-mmult-bug -mno-mult-bug @gol
995 -mno-am33 -mam33 -mam33-2 -mam34 @gol
996 -mtune=@var{cpu-type} @gol
997 -mreturn-pointer-on-d0 @gol
998 -mno-crt0 -mrelax -mliw -msetlb}
1000 @emph{Moxie Options}
1001 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
1003 @emph{MSP430 Options}
1004 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
1006 -mcode-region= -mdata-region= @gol
1007 -msilicon-errata= -msilicon-errata-warn= @gol
1010 @emph{NDS32 Options}
1011 @gccoptlist{-mbig-endian -mlittle-endian @gol
1012 -mreduced-regs -mfull-regs @gol
1013 -mcmov -mno-cmov @gol
1014 -mext-perf -mno-ext-perf @gol
1015 -mext-perf2 -mno-ext-perf2 @gol
1016 -mext-string -mno-ext-string @gol
1017 -mv3push -mno-v3push @gol
1018 -m16bit -mno-16bit @gol
1019 -misr-vector-size=@var{num} @gol
1020 -mcache-block-size=@var{num} @gol
1021 -march=@var{arch} @gol
1022 -mcmodel=@var{code-model} @gol
1023 -mctor-dtor -mrelax}
1025 @emph{Nios II Options}
1026 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
1027 -mgprel-sec=@var{regexp} -mr0rel-sec=@var{regexp} @gol
1029 -mno-bypass-cache -mbypass-cache @gol
1030 -mno-cache-volatile -mcache-volatile @gol
1031 -mno-fast-sw-div -mfast-sw-div @gol
1032 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
1033 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
1034 -mcustom-fpu-cfg=@var{name} @gol
1035 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
1036 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
1038 @emph{Nvidia PTX Options}
1039 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
1041 @emph{OpenRISC Options}
1042 @gccoptlist{-mboard=@var{name} -mnewlib -mhard-mul -mhard-div @gol
1043 -msoft-mul -msoft-div @gol
1044 -msoft-float -mhard-float -mdouble-float -munordered-float @gol
1045 -mcmov -mror -mrori -msext -msfimm -mshftimm}
1047 @emph{PDP-11 Options}
1048 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
1049 -mint32 -mno-int16 -mint16 -mno-int32 @gol
1050 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra}
1052 @emph{picoChip Options}
1053 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
1054 -msymbol-as-address -mno-inefficient-warnings}
1056 @emph{PowerPC Options}
1057 See RS/6000 and PowerPC Options.
1060 @gccoptlist{-mmcu=@var{mcu} -minrt -mno-relax -mloop @gol
1061 -mabi=@var{variant} @gol}
1063 @emph{RISC-V Options}
1064 @gccoptlist{-mbranch-cost=@var{N-instruction} @gol
1066 -mabi=@var{ABI-string} @gol
1067 -mfdiv -mno-fdiv @gol
1069 -march=@var{ISA-string} @gol
1070 -mtune=@var{processor-string} @gol
1071 -mpreferred-stack-boundary=@var{num} @gol
1072 -msmall-data-limit=@var{N-bytes} @gol
1073 -msave-restore -mno-save-restore @gol
1074 -mstrict-align -mno-strict-align @gol
1075 -mcmodel=medlow -mcmodel=medany @gol
1076 -mexplicit-relocs -mno-explicit-relocs @gol
1077 -mrelax -mno-relax @gol
1078 -mriscv-attribute -mmo-riscv-attribute @gol
1079 -malign-data=@var{type}}
1082 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
1083 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
1084 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
1086 @emph{RS/6000 and PowerPC Options}
1087 @gccoptlist{-mcpu=@var{cpu-type} @gol
1088 -mtune=@var{cpu-type} @gol
1089 -mcmodel=@var{code-model} @gol
1091 -maltivec -mno-altivec @gol
1092 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
1093 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
1094 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
1095 -mfprnd -mno-fprnd @gol
1096 -mcmpb -mno-cmpb -mhard-dfp -mno-hard-dfp @gol
1097 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
1098 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
1099 -malign-power -malign-natural @gol
1100 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
1101 -mupdate -mno-update @gol
1102 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
1103 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
1104 -mstrict-align -mno-strict-align -mrelocatable @gol
1105 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
1106 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
1107 -mdynamic-no-pic -mswdiv -msingle-pic-base @gol
1108 -mprioritize-restricted-insns=@var{priority} @gol
1109 -msched-costly-dep=@var{dependence_type} @gol
1110 -minsert-sched-nops=@var{scheme} @gol
1111 -mcall-aixdesc -mcall-eabi -mcall-freebsd @gol
1112 -mcall-linux -mcall-netbsd -mcall-openbsd @gol
1113 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi @gol
1114 -mtraceback=@var{traceback_type} @gol
1115 -maix-struct-return -msvr4-struct-return @gol
1116 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
1117 -mlongcall -mno-longcall -mpltseq -mno-pltseq @gol
1118 -mblock-move-inline-limit=@var{num} @gol
1119 -mblock-compare-inline-limit=@var{num} @gol
1120 -mblock-compare-inline-loop-limit=@var{num} @gol
1121 -mstring-compare-inline-limit=@var{num} @gol
1122 -misel -mno-isel @gol
1123 -mvrsave -mno-vrsave @gol
1124 -mmulhw -mno-mulhw @gol
1125 -mdlmzb -mno-dlmzb @gol
1126 -mprototype -mno-prototype @gol
1127 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1128 -msdata=@var{opt} -mreadonly-in-sdata -mvxworks -G @var{num} @gol
1129 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1130 -mno-recip-precision @gol
1131 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1132 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1133 -msave-toc-indirect -mno-save-toc-indirect @gol
1134 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1135 -mcrypto -mno-crypto -mhtm -mno-htm @gol
1136 -mquad-memory -mno-quad-memory @gol
1137 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1138 -mcompat-align-parm -mno-compat-align-parm @gol
1139 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1140 -mgnu-attribute -mno-gnu-attribute @gol
1141 -mstack-protector-guard=@var{guard} -mstack-protector-guard-reg=@var{reg} @gol
1142 -mstack-protector-guard-offset=@var{offset} -mpcrel -mno-pcrel}
1145 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1147 -mbig-endian-data -mlittle-endian-data @gol
1150 -mas100-syntax -mno-as100-syntax@gol
1152 -mmax-constant-size=@gol
1155 -mallow-string-insns -mno-allow-string-insns@gol
1157 -mno-warn-multiple-fast-interrupts@gol
1158 -msave-acc-in-interrupts}
1160 @emph{S/390 and zSeries Options}
1161 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1162 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1163 -mlong-double-64 -mlong-double-128 @gol
1164 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1165 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1166 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1167 -mhtm -mvx -mzvector @gol
1168 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1169 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1170 -mhotpatch=@var{halfwords},@var{halfwords}}
1172 @emph{Score Options}
1173 @gccoptlist{-meb -mel @gol
1177 -mscore5 -mscore5u -mscore7 -mscore7d}
1180 @gccoptlist{-m1 -m2 -m2e @gol
1181 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1183 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1184 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1185 -mb -ml -mdalign -mrelax @gol
1186 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1187 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1188 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1189 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1190 -maccumulate-outgoing-args @gol
1191 -matomic-model=@var{atomic-model} @gol
1192 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1193 -mcbranch-force-delay-slot @gol
1194 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1195 -mpretend-cmove -mtas}
1197 @emph{Solaris 2 Options}
1198 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1201 @emph{SPARC Options}
1202 @gccoptlist{-mcpu=@var{cpu-type} @gol
1203 -mtune=@var{cpu-type} @gol
1204 -mcmodel=@var{code-model} @gol
1205 -mmemory-model=@var{mem-model} @gol
1206 -m32 -m64 -mapp-regs -mno-app-regs @gol
1207 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1208 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1209 -mhard-quad-float -msoft-quad-float @gol
1210 -mstack-bias -mno-stack-bias @gol
1211 -mstd-struct-return -mno-std-struct-return @gol
1212 -munaligned-doubles -mno-unaligned-doubles @gol
1213 -muser-mode -mno-user-mode @gol
1214 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1215 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1216 -mvis4 -mno-vis4 -mvis4b -mno-vis4b @gol
1217 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld @gol
1218 -mpopc -mno-popc -msubxc -mno-subxc @gol
1219 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc @gol
1222 @emph{System V Options}
1223 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1225 @emph{TILE-Gx Options}
1226 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1227 -mcmodel=@var{code-model}}
1229 @emph{TILEPro Options}
1230 @gccoptlist{-mcpu=@var{cpu} -m32}
1233 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1234 -mprolog-function -mno-prolog-function -mspace @gol
1235 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1236 -mapp-regs -mno-app-regs @gol
1237 -mdisable-callt -mno-disable-callt @gol
1238 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1239 -mv850e -mv850 -mv850e3v5 @gol
1250 @gccoptlist{-mg -mgnu -munix}
1252 @emph{Visium Options}
1253 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1254 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1257 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1258 -mpointer-size=@var{size}}
1260 @emph{VxWorks Options}
1261 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1262 -Xbind-lazy -Xbind-now}
1265 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1266 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1267 -mfpmath=@var{unit} @gol
1268 -masm=@var{dialect} -mno-fancy-math-387 @gol
1269 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1270 -mno-wide-multiply -mrtd -malign-double @gol
1271 -mpreferred-stack-boundary=@var{num} @gol
1272 -mincoming-stack-boundary=@var{num} @gol
1273 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1274 -mrecip -mrecip=@var{opt} @gol
1275 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=@var{opt} @gol
1276 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1277 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1278 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1279 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd @gol
1280 -mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves @gol
1281 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol
1282 -madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp @gol
1283 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg @gol
1284 -mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 -mavx512bf16 -menqcmd @gol
1285 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq @gol
1286 -mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid @gol
1287 -mrdseed -msgx -mavx512vp2intersect@gol
1288 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1289 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1290 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1291 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1292 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1293 -mregparm=@var{num} -msseregparm @gol
1294 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1295 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1296 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1297 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1298 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1299 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1300 -minstrument-return=@var{type} -mfentry-name=@var{name} -mfentry-section=@var{name} @gol
1301 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1302 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1303 -mstack-protector-guard-reg=@var{reg} @gol
1304 -mstack-protector-guard-offset=@var{offset} @gol
1305 -mstack-protector-guard-symbol=@var{symbol} @gol
1306 -mgeneral-regs-only -mcall-ms2sysv-xlogues @gol
1307 -mindirect-branch=@var{choice} -mfunction-return=@var{choice} @gol
1308 -mindirect-branch-register}
1310 @emph{x86 Windows Options}
1311 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1312 -mnop-fun-dllimport -mthread @gol
1313 -municode -mwin32 -mwindows -fno-set-stack-executable}
1315 @emph{Xstormy16 Options}
1318 @emph{Xtensa Options}
1319 @gccoptlist{-mconst16 -mno-const16 @gol
1320 -mfused-madd -mno-fused-madd @gol
1322 -mserialize-volatile -mno-serialize-volatile @gol
1323 -mtext-section-literals -mno-text-section-literals @gol
1324 -mauto-litpools -mno-auto-litpools @gol
1325 -mtarget-align -mno-target-align @gol
1326 -mlongcalls -mno-longcalls}
1328 @emph{zSeries Options}
1329 See S/390 and zSeries Options.
1333 @node Overall Options
1334 @section Options Controlling the Kind of Output
1336 Compilation can involve up to four stages: preprocessing, compilation
1337 proper, assembly and linking, always in that order. GCC is capable of
1338 preprocessing and compiling several files either into several
1339 assembler input files, or into one assembler input file; then each
1340 assembler input file produces an object file, and linking combines all
1341 the object files (those newly compiled, and those specified as input)
1342 into an executable file.
1344 @cindex file name suffix
1345 For any given input file, the file name suffix determines what kind of
1346 compilation is done:
1350 C source code that must be preprocessed.
1353 C source code that should not be preprocessed.
1356 C++ source code that should not be preprocessed.
1359 Objective-C source code. Note that you must link with the @file{libobjc}
1360 library to make an Objective-C program work.
1363 Objective-C source code that should not be preprocessed.
1367 Objective-C++ source code. Note that you must link with the @file{libobjc}
1368 library to make an Objective-C++ program work. Note that @samp{.M} refers
1369 to a literal capital M@.
1371 @item @var{file}.mii
1372 Objective-C++ source code that should not be preprocessed.
1375 C, C++, Objective-C or Objective-C++ header file to be turned into a
1376 precompiled header (default), or C, C++ header file to be turned into an
1377 Ada spec (via the @option{-fdump-ada-spec} switch).
1380 @itemx @var{file}.cp
1381 @itemx @var{file}.cxx
1382 @itemx @var{file}.cpp
1383 @itemx @var{file}.CPP
1384 @itemx @var{file}.c++
1386 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1387 the last two letters must both be literally @samp{x}. Likewise,
1388 @samp{.C} refers to a literal capital C@.
1392 Objective-C++ source code that must be preprocessed.
1394 @item @var{file}.mii
1395 Objective-C++ source code that should not be preprocessed.
1399 @itemx @var{file}.hp
1400 @itemx @var{file}.hxx
1401 @itemx @var{file}.hpp
1402 @itemx @var{file}.HPP
1403 @itemx @var{file}.h++
1404 @itemx @var{file}.tcc
1405 C++ header file to be turned into a precompiled header or Ada spec.
1408 @itemx @var{file}.for
1409 @itemx @var{file}.ftn
1410 Fixed form Fortran source code that should not be preprocessed.
1413 @itemx @var{file}.FOR
1414 @itemx @var{file}.fpp
1415 @itemx @var{file}.FPP
1416 @itemx @var{file}.FTN
1417 Fixed form Fortran source code that must be preprocessed (with the traditional
1420 @item @var{file}.f90
1421 @itemx @var{file}.f95
1422 @itemx @var{file}.f03
1423 @itemx @var{file}.f08
1424 Free form Fortran source code that should not be preprocessed.
1426 @item @var{file}.F90
1427 @itemx @var{file}.F95
1428 @itemx @var{file}.F03
1429 @itemx @var{file}.F08
1430 Free form Fortran source code that must be preprocessed (with the
1431 traditional preprocessor).
1436 @item @var{file}.brig
1437 BRIG files (binary representation of HSAIL).
1446 D documentation code (Ddoc).
1448 @item @var{file}.ads
1449 Ada source code file that contains a library unit declaration (a
1450 declaration of a package, subprogram, or generic, or a generic
1451 instantiation), or a library unit renaming declaration (a package,
1452 generic, or subprogram renaming declaration). Such files are also
1455 @item @var{file}.adb
1456 Ada source code file containing a library unit body (a subprogram or
1457 package body). Such files are also called @dfn{bodies}.
1459 @c GCC also knows about some suffixes for languages not yet included:
1467 @itemx @var{file}.sx
1468 Assembler code that must be preprocessed.
1471 An object file to be fed straight into linking.
1472 Any file name with no recognized suffix is treated this way.
1476 You can specify the input language explicitly with the @option{-x} option:
1479 @item -x @var{language}
1480 Specify explicitly the @var{language} for the following input files
1481 (rather than letting the compiler choose a default based on the file
1482 name suffix). This option applies to all following input files until
1483 the next @option{-x} option. Possible values for @var{language} are:
1485 c c-header cpp-output
1486 c++ c++-header c++-cpp-output
1487 objective-c objective-c-header objective-c-cpp-output
1488 objective-c++ objective-c++-header objective-c++-cpp-output
1489 assembler assembler-with-cpp
1492 f77 f77-cpp-input f95 f95-cpp-input
1498 Turn off any specification of a language, so that subsequent files are
1499 handled according to their file name suffixes (as they are if @option{-x}
1500 has not been used at all).
1503 If you only want some of the stages of compilation, you can use
1504 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1505 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1506 @command{gcc} is to stop. Note that some combinations (for example,
1507 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1512 Compile or assemble the source files, but do not link. The linking
1513 stage simply is not done. The ultimate output is in the form of an
1514 object file for each source file.
1516 By default, the object file name for a source file is made by replacing
1517 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1519 Unrecognized input files, not requiring compilation or assembly, are
1524 Stop after the stage of compilation proper; do not assemble. The output
1525 is in the form of an assembler code file for each non-assembler input
1528 By default, the assembler file name for a source file is made by
1529 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1531 Input files that don't require compilation are ignored.
1535 Stop after the preprocessing stage; do not run the compiler proper. The
1536 output is in the form of preprocessed source code, which is sent to the
1539 Input files that don't require preprocessing are ignored.
1541 @cindex output file option
1544 Place output in file @var{file}. This applies to whatever
1545 sort of output is being produced, whether it be an executable file,
1546 an object file, an assembler file or preprocessed C code.
1548 If @option{-o} is not specified, the default is to put an executable
1549 file in @file{a.out}, the object file for
1550 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1551 assembler file in @file{@var{source}.s}, a precompiled header file in
1552 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1557 Print (on standard error output) the commands executed to run the stages
1558 of compilation. Also print the version number of the compiler driver
1559 program and of the preprocessor and the compiler proper.
1563 Like @option{-v} except the commands are not executed and arguments
1564 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1565 This is useful for shell scripts to capture the driver-generated command lines.
1569 Print (on the standard output) a description of the command-line options
1570 understood by @command{gcc}. If the @option{-v} option is also specified
1571 then @option{--help} is also passed on to the various processes
1572 invoked by @command{gcc}, so that they can display the command-line options
1573 they accept. If the @option{-Wextra} option has also been specified
1574 (prior to the @option{--help} option), then command-line options that
1575 have no documentation associated with them are also displayed.
1578 @opindex target-help
1579 Print (on the standard output) a description of target-specific command-line
1580 options for each tool. For some targets extra target-specific
1581 information may also be printed.
1583 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1584 Print (on the standard output) a description of the command-line
1585 options understood by the compiler that fit into all specified classes
1586 and qualifiers. These are the supported classes:
1589 @item @samp{optimizers}
1590 Display all of the optimization options supported by the
1593 @item @samp{warnings}
1594 Display all of the options controlling warning messages
1595 produced by the compiler.
1598 Display target-specific options. Unlike the
1599 @option{--target-help} option however, target-specific options of the
1600 linker and assembler are not displayed. This is because those
1601 tools do not currently support the extended @option{--help=} syntax.
1604 Display the values recognized by the @option{--param}
1607 @item @var{language}
1608 Display the options supported for @var{language}, where
1609 @var{language} is the name of one of the languages supported in this
1613 Display the options that are common to all languages.
1616 These are the supported qualifiers:
1619 @item @samp{undocumented}
1620 Display only those options that are undocumented.
1623 Display options taking an argument that appears after an equal
1624 sign in the same continuous piece of text, such as:
1625 @samp{--help=target}.
1627 @item @samp{separate}
1628 Display options taking an argument that appears as a separate word
1629 following the original option, such as: @samp{-o output-file}.
1632 Thus for example to display all the undocumented target-specific
1633 switches supported by the compiler, use:
1636 --help=target,undocumented
1639 The sense of a qualifier can be inverted by prefixing it with the
1640 @samp{^} character, so for example to display all binary warning
1641 options (i.e., ones that are either on or off and that do not take an
1642 argument) that have a description, use:
1645 --help=warnings,^joined,^undocumented
1648 The argument to @option{--help=} should not consist solely of inverted
1651 Combining several classes is possible, although this usually
1652 restricts the output so much that there is nothing to display. One
1653 case where it does work, however, is when one of the classes is
1654 @var{target}. For example, to display all the target-specific
1655 optimization options, use:
1658 --help=target,optimizers
1661 The @option{--help=} option can be repeated on the command line. Each
1662 successive use displays its requested class of options, skipping
1663 those that have already been displayed. If @option{--help} is also
1664 specified anywhere on the command line then this takes precedence
1665 over any @option{--help=} option.
1667 If the @option{-Q} option appears on the command line before the
1668 @option{--help=} option, then the descriptive text displayed by
1669 @option{--help=} is changed. Instead of describing the displayed
1670 options, an indication is given as to whether the option is enabled,
1671 disabled or set to a specific value (assuming that the compiler
1672 knows this at the point where the @option{--help=} option is used).
1674 Here is a truncated example from the ARM port of @command{gcc}:
1677 % gcc -Q -mabi=2 --help=target -c
1678 The following options are target specific:
1680 -mabort-on-noreturn [disabled]
1684 The output is sensitive to the effects of previous command-line
1685 options, so for example it is possible to find out which optimizations
1686 are enabled at @option{-O2} by using:
1689 -Q -O2 --help=optimizers
1692 Alternatively you can discover which binary optimizations are enabled
1693 by @option{-O3} by using:
1696 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1697 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1698 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1703 Display the version number and copyrights of the invoked GCC@.
1705 @item -pass-exit-codes
1706 @opindex pass-exit-codes
1707 Normally the @command{gcc} program exits with the code of 1 if any
1708 phase of the compiler returns a non-success return code. If you specify
1709 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1710 the numerically highest error produced by any phase returning an error
1711 indication. The C, C++, and Fortran front ends return 4 if an internal
1712 compiler error is encountered.
1716 Use pipes rather than temporary files for communication between the
1717 various stages of compilation. This fails to work on some systems where
1718 the assembler is unable to read from a pipe; but the GNU assembler has
1721 @item -specs=@var{file}
1723 Process @var{file} after the compiler reads in the standard @file{specs}
1724 file, in order to override the defaults which the @command{gcc} driver
1725 program uses when determining what switches to pass to @command{cc1},
1726 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1727 @option{-specs=@var{file}} can be specified on the command line, and they
1728 are processed in order, from left to right. @xref{Spec Files}, for
1729 information about the format of the @var{file}.
1733 Invoke all subcommands under a wrapper program. The name of the
1734 wrapper program and its parameters are passed as a comma separated
1738 gcc -c t.c -wrapper gdb,--args
1742 This invokes all subprograms of @command{gcc} under
1743 @samp{gdb --args}, thus the invocation of @command{cc1} is
1744 @samp{gdb --args cc1 @dots{}}.
1746 @item -ffile-prefix-map=@var{old}=@var{new}
1747 @opindex ffile-prefix-map
1748 When compiling files residing in directory @file{@var{old}}, record
1749 any references to them in the result of the compilation as if the
1750 files resided in directory @file{@var{new}} instead. Specifying this
1751 option is equivalent to specifying all the individual
1752 @option{-f*-prefix-map} options. This can be used to make reproducible
1753 builds that are location independent. See also
1754 @option{-fmacro-prefix-map} and @option{-fdebug-prefix-map}.
1756 @item -fplugin=@var{name}.so
1758 Load the plugin code in file @var{name}.so, assumed to be a
1759 shared object to be dlopen'd by the compiler. The base name of
1760 the shared object file is used to identify the plugin for the
1761 purposes of argument parsing (See
1762 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1763 Each plugin should define the callback functions specified in the
1766 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1767 @opindex fplugin-arg
1768 Define an argument called @var{key} with a value of @var{value}
1769 for the plugin called @var{name}.
1771 @item -fdump-ada-spec@r{[}-slim@r{]}
1772 @opindex fdump-ada-spec
1773 For C and C++ source and include files, generate corresponding Ada specs.
1774 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1775 GNAT User's Guide}, which provides detailed documentation on this feature.
1777 @item -fada-spec-parent=@var{unit}
1778 @opindex fada-spec-parent
1779 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1780 Ada specs as child units of parent @var{unit}.
1782 @item -fdump-go-spec=@var{file}
1783 @opindex fdump-go-spec
1784 For input files in any language, generate corresponding Go
1785 declarations in @var{file}. This generates Go @code{const},
1786 @code{type}, @code{var}, and @code{func} declarations which may be a
1787 useful way to start writing a Go interface to code written in some
1790 @include @value{srcdir}/../libiberty/at-file.texi
1794 @section Compiling C++ Programs
1796 @cindex suffixes for C++ source
1797 @cindex C++ source file suffixes
1798 C++ source files conventionally use one of the suffixes @samp{.C},
1799 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1800 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1801 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1802 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1803 files with these names and compiles them as C++ programs even if you
1804 call the compiler the same way as for compiling C programs (usually
1805 with the name @command{gcc}).
1809 However, the use of @command{gcc} does not add the C++ library.
1810 @command{g++} is a program that calls GCC and automatically specifies linking
1811 against the C++ library. It treats @samp{.c},
1812 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1813 files unless @option{-x} is used. This program is also useful when
1814 precompiling a C header file with a @samp{.h} extension for use in C++
1815 compilations. On many systems, @command{g++} is also installed with
1816 the name @command{c++}.
1818 @cindex invoking @command{g++}
1819 When you compile C++ programs, you may specify many of the same
1820 command-line options that you use for compiling programs in any
1821 language; or command-line options meaningful for C and related
1822 languages; or options that are meaningful only for C++ programs.
1823 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1824 explanations of options for languages related to C@.
1825 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1826 explanations of options that are meaningful only for C++ programs.
1828 @node C Dialect Options
1829 @section Options Controlling C Dialect
1830 @cindex dialect options
1831 @cindex language dialect options
1832 @cindex options, dialect
1834 The following options control the dialect of C (or languages derived
1835 from C, such as C++, Objective-C and Objective-C++) that the compiler
1839 @cindex ANSI support
1843 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1844 equivalent to @option{-std=c++98}.
1846 This turns off certain features of GCC that are incompatible with ISO
1847 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1848 such as the @code{asm} and @code{typeof} keywords, and
1849 predefined macros such as @code{unix} and @code{vax} that identify the
1850 type of system you are using. It also enables the undesirable and
1851 rarely used ISO trigraph feature. For the C compiler,
1852 it disables recognition of C++ style @samp{//} comments as well as
1853 the @code{inline} keyword.
1855 The alternate keywords @code{__asm__}, @code{__extension__},
1856 @code{__inline__} and @code{__typeof__} continue to work despite
1857 @option{-ansi}. You would not want to use them in an ISO C program, of
1858 course, but it is useful to put them in header files that might be included
1859 in compilations done with @option{-ansi}. Alternate predefined macros
1860 such as @code{__unix__} and @code{__vax__} are also available, with or
1861 without @option{-ansi}.
1863 The @option{-ansi} option does not cause non-ISO programs to be
1864 rejected gratuitously. For that, @option{-Wpedantic} is required in
1865 addition to @option{-ansi}. @xref{Warning Options}.
1867 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1868 option is used. Some header files may notice this macro and refrain
1869 from declaring certain functions or defining certain macros that the
1870 ISO standard doesn't call for; this is to avoid interfering with any
1871 programs that might use these names for other things.
1873 Functions that are normally built in but do not have semantics
1874 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1875 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1876 built-in functions provided by GCC}, for details of the functions
1881 Determine the language standard. @xref{Standards,,Language Standards
1882 Supported by GCC}, for details of these standard versions. This option
1883 is currently only supported when compiling C or C++.
1885 The compiler can accept several base standards, such as @samp{c90} or
1886 @samp{c++98}, and GNU dialects of those standards, such as
1887 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1888 compiler accepts all programs following that standard plus those
1889 using GNU extensions that do not contradict it. For example,
1890 @option{-std=c90} turns off certain features of GCC that are
1891 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1892 keywords, but not other GNU extensions that do not have a meaning in
1893 ISO C90, such as omitting the middle term of a @code{?:}
1894 expression. On the other hand, when a GNU dialect of a standard is
1895 specified, all features supported by the compiler are enabled, even when
1896 those features change the meaning of the base standard. As a result, some
1897 strict-conforming programs may be rejected. The particular standard
1898 is used by @option{-Wpedantic} to identify which features are GNU
1899 extensions given that version of the standard. For example
1900 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1901 comments, while @option{-std=gnu99 -Wpedantic} does not.
1903 A value for this option must be provided; possible values are
1909 Support all ISO C90 programs (certain GNU extensions that conflict
1910 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1912 @item iso9899:199409
1913 ISO C90 as modified in amendment 1.
1919 ISO C99. This standard is substantially completely supported, modulo
1920 bugs and floating-point issues
1921 (mainly but not entirely relating to optional C99 features from
1922 Annexes F and G). See
1923 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1924 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1929 ISO C11, the 2011 revision of the ISO C standard. This standard is
1930 substantially completely supported, modulo bugs, floating-point issues
1931 (mainly but not entirely relating to optional C11 features from
1932 Annexes F and G) and the optional Annexes K (Bounds-checking
1933 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1939 ISO C17, the 2017 revision of the ISO C standard
1940 (published in 2018). This standard is
1941 same as C11 except for corrections of defects (all of which are also
1942 applied with @option{-std=c11}) and a new value of
1943 @code{__STDC_VERSION__}, and so is supported to the same extent as C11.
1946 The next version of the ISO C standard, still under development. The
1947 support for this version is experimental and incomplete.
1951 GNU dialect of ISO C90 (including some C99 features).
1955 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1959 GNU dialect of ISO C11.
1960 The name @samp{gnu1x} is deprecated.
1964 GNU dialect of ISO C17. This is the default for C code.
1967 The next version of the ISO C standard, still under development, plus
1968 GNU extensions. The support for this version is experimental and
1973 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1974 additional defect reports. Same as @option{-ansi} for C++ code.
1978 GNU dialect of @option{-std=c++98}.
1982 The 2011 ISO C++ standard plus amendments.
1983 The name @samp{c++0x} is deprecated.
1987 GNU dialect of @option{-std=c++11}.
1988 The name @samp{gnu++0x} is deprecated.
1992 The 2014 ISO C++ standard plus amendments.
1993 The name @samp{c++1y} is deprecated.
1997 GNU dialect of @option{-std=c++14}.
1998 This is the default for C++ code.
1999 The name @samp{gnu++1y} is deprecated.
2003 The 2017 ISO C++ standard plus amendments.
2004 The name @samp{c++1z} is deprecated.
2008 GNU dialect of @option{-std=c++17}.
2009 The name @samp{gnu++1z} is deprecated.
2012 The next revision of the ISO C++ standard, tentatively planned for
2013 2020. Support is highly experimental, and will almost certainly
2014 change in incompatible ways in future releases.
2017 GNU dialect of @option{-std=c++2a}. Support is highly experimental,
2018 and will almost certainly change in incompatible ways in future
2022 @item -fgnu89-inline
2023 @opindex fgnu89-inline
2024 The option @option{-fgnu89-inline} tells GCC to use the traditional
2025 GNU semantics for @code{inline} functions when in C99 mode.
2026 @xref{Inline,,An Inline Function is As Fast As a Macro}.
2027 Using this option is roughly equivalent to adding the
2028 @code{gnu_inline} function attribute to all inline functions
2029 (@pxref{Function Attributes}).
2031 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
2032 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
2033 specifies the default behavior).
2034 This option is not supported in @option{-std=c90} or
2035 @option{-std=gnu90} mode.
2037 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
2038 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
2039 in effect for @code{inline} functions. @xref{Common Predefined
2040 Macros,,,cpp,The C Preprocessor}.
2042 @item -fpermitted-flt-eval-methods=@var{style}
2043 @opindex fpermitted-flt-eval-methods
2044 @opindex fpermitted-flt-eval-methods=c11
2045 @opindex fpermitted-flt-eval-methods=ts-18661-3
2046 ISO/IEC TS 18661-3 defines new permissible values for
2047 @code{FLT_EVAL_METHOD} that indicate that operations and constants with
2048 a semantic type that is an interchange or extended format should be
2049 evaluated to the precision and range of that type. These new values are
2050 a superset of those permitted under C99/C11, which does not specify the
2051 meaning of other positive values of @code{FLT_EVAL_METHOD}. As such, code
2052 conforming to C11 may not have been written expecting the possibility of
2055 @option{-fpermitted-flt-eval-methods} specifies whether the compiler
2056 should allow only the values of @code{FLT_EVAL_METHOD} specified in C99/C11,
2057 or the extended set of values specified in ISO/IEC TS 18661-3.
2059 @var{style} is either @code{c11} or @code{ts-18661-3} as appropriate.
2061 The default when in a standards compliant mode (@option{-std=c11} or similar)
2062 is @option{-fpermitted-flt-eval-methods=c11}. The default when in a GNU
2063 dialect (@option{-std=gnu11} or similar) is
2064 @option{-fpermitted-flt-eval-methods=ts-18661-3}.
2066 @item -aux-info @var{filename}
2068 Output to the given filename prototyped declarations for all functions
2069 declared and/or defined in a translation unit, including those in header
2070 files. This option is silently ignored in any language other than C@.
2072 Besides declarations, the file indicates, in comments, the origin of
2073 each declaration (source file and line), whether the declaration was
2074 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
2075 @samp{O} for old, respectively, in the first character after the line
2076 number and the colon), and whether it came from a declaration or a
2077 definition (@samp{C} or @samp{F}, respectively, in the following
2078 character). In the case of function definitions, a K&R-style list of
2079 arguments followed by their declarations is also provided, inside
2080 comments, after the declaration.
2082 @item -fallow-parameterless-variadic-functions
2083 @opindex fallow-parameterless-variadic-functions
2084 Accept variadic functions without named parameters.
2086 Although it is possible to define such a function, this is not very
2087 useful as it is not possible to read the arguments. This is only
2088 supported for C as this construct is allowed by C++.
2093 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
2094 keyword, so that code can use these words as identifiers. You can use
2095 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
2096 instead. @option{-ansi} implies @option{-fno-asm}.
2098 In C++, this switch only affects the @code{typeof} keyword, since
2099 @code{asm} and @code{inline} are standard keywords. You may want to
2100 use the @option{-fno-gnu-keywords} flag instead, which has the same
2101 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
2102 switch only affects the @code{asm} and @code{typeof} keywords, since
2103 @code{inline} is a standard keyword in ISO C99.
2106 @itemx -fno-builtin-@var{function}
2107 @opindex fno-builtin
2109 @cindex built-in functions
2110 Don't recognize built-in functions that do not begin with
2111 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
2112 functions provided by GCC}, for details of the functions affected,
2113 including those which are not built-in functions when @option{-ansi} or
2114 @option{-std} options for strict ISO C conformance are used because they
2115 do not have an ISO standard meaning.
2117 GCC normally generates special code to handle certain built-in functions
2118 more efficiently; for instance, calls to @code{alloca} may become single
2119 instructions which adjust the stack directly, and calls to @code{memcpy}
2120 may become inline copy loops. The resulting code is often both smaller
2121 and faster, but since the function calls no longer appear as such, you
2122 cannot set a breakpoint on those calls, nor can you change the behavior
2123 of the functions by linking with a different library. In addition,
2124 when a function is recognized as a built-in function, GCC may use
2125 information about that function to warn about problems with calls to
2126 that function, or to generate more efficient code, even if the
2127 resulting code still contains calls to that function. For example,
2128 warnings are given with @option{-Wformat} for bad calls to
2129 @code{printf} when @code{printf} is built in and @code{strlen} is
2130 known not to modify global memory.
2132 With the @option{-fno-builtin-@var{function}} option
2133 only the built-in function @var{function} is
2134 disabled. @var{function} must not begin with @samp{__builtin_}. If a
2135 function is named that is not built-in in this version of GCC, this
2136 option is ignored. There is no corresponding
2137 @option{-fbuiltin-@var{function}} option; if you wish to enable
2138 built-in functions selectively when using @option{-fno-builtin} or
2139 @option{-ffreestanding}, you may define macros such as:
2142 #define abs(n) __builtin_abs ((n))
2143 #define strcpy(d, s) __builtin_strcpy ((d), (s))
2149 Enable parsing of function definitions marked with @code{__GIMPLE}.
2150 This is an experimental feature that allows unit testing of GIMPLE
2155 @cindex hosted environment
2157 Assert that compilation targets a hosted environment. This implies
2158 @option{-fbuiltin}. A hosted environment is one in which the
2159 entire standard library is available, and in which @code{main} has a return
2160 type of @code{int}. Examples are nearly everything except a kernel.
2161 This is equivalent to @option{-fno-freestanding}.
2163 @item -ffreestanding
2164 @opindex ffreestanding
2165 @cindex hosted environment
2167 Assert that compilation targets a freestanding environment. This
2168 implies @option{-fno-builtin}. A freestanding environment
2169 is one in which the standard library may not exist, and program startup may
2170 not necessarily be at @code{main}. The most obvious example is an OS kernel.
2171 This is equivalent to @option{-fno-hosted}.
2173 @xref{Standards,,Language Standards Supported by GCC}, for details of
2174 freestanding and hosted environments.
2178 @cindex OpenACC accelerator programming
2179 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
2180 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
2181 compiler generates accelerated code according to the OpenACC Application
2182 Programming Interface v2.0 @w{@uref{https://www.openacc.org}}. This option
2183 implies @option{-pthread}, and thus is only supported on targets that
2184 have support for @option{-pthread}.
2186 @item -fopenacc-dim=@var{geom}
2187 @opindex fopenacc-dim
2188 @cindex OpenACC accelerator programming
2189 Specify default compute dimensions for parallel offload regions that do
2190 not explicitly specify. The @var{geom} value is a triple of
2191 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
2192 can be omitted, to use a target-specific default value.
2196 @cindex OpenMP parallel
2197 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
2198 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
2199 compiler generates parallel code according to the OpenMP Application
2200 Program Interface v4.5 @w{@uref{https://www.openmp.org}}. This option
2201 implies @option{-pthread}, and thus is only supported on targets that
2202 have support for @option{-pthread}. @option{-fopenmp} implies
2203 @option{-fopenmp-simd}.
2206 @opindex fopenmp-simd
2209 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2210 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2215 When the option @option{-fgnu-tm} is specified, the compiler
2216 generates code for the Linux variant of Intel's current Transactional
2217 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2218 an experimental feature whose interface may change in future versions
2219 of GCC, as the official specification changes. Please note that not
2220 all architectures are supported for this feature.
2222 For more information on GCC's support for transactional memory,
2223 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2224 Transactional Memory Library}.
2226 Note that the transactional memory feature is not supported with
2227 non-call exceptions (@option{-fnon-call-exceptions}).
2229 @item -fms-extensions
2230 @opindex fms-extensions
2231 Accept some non-standard constructs used in Microsoft header files.
2233 In C++ code, this allows member names in structures to be similar
2234 to previous types declarations.
2243 Some cases of unnamed fields in structures and unions are only
2244 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2245 fields within structs/unions}, for details.
2247 Note that this option is off for all targets except for x86
2248 targets using ms-abi.
2250 @item -fplan9-extensions
2251 @opindex fplan9-extensions
2252 Accept some non-standard constructs used in Plan 9 code.
2254 This enables @option{-fms-extensions}, permits passing pointers to
2255 structures with anonymous fields to functions that expect pointers to
2256 elements of the type of the field, and permits referring to anonymous
2257 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2258 struct/union fields within structs/unions}, for details. This is only
2259 supported for C, not C++.
2261 @item -fcond-mismatch
2262 @opindex fcond-mismatch
2263 Allow conditional expressions with mismatched types in the second and
2264 third arguments. The value of such an expression is void. This option
2265 is not supported for C++.
2267 @item -flax-vector-conversions
2268 @opindex flax-vector-conversions
2269 Allow implicit conversions between vectors with differing numbers of
2270 elements and/or incompatible element types. This option should not be
2273 @item -funsigned-char
2274 @opindex funsigned-char
2275 Let the type @code{char} be unsigned, like @code{unsigned char}.
2277 Each kind of machine has a default for what @code{char} should
2278 be. It is either like @code{unsigned char} by default or like
2279 @code{signed char} by default.
2281 Ideally, a portable program should always use @code{signed char} or
2282 @code{unsigned char} when it depends on the signedness of an object.
2283 But many programs have been written to use plain @code{char} and
2284 expect it to be signed, or expect it to be unsigned, depending on the
2285 machines they were written for. This option, and its inverse, let you
2286 make such a program work with the opposite default.
2288 The type @code{char} is always a distinct type from each of
2289 @code{signed char} or @code{unsigned char}, even though its behavior
2290 is always just like one of those two.
2293 @opindex fsigned-char
2294 Let the type @code{char} be signed, like @code{signed char}.
2296 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2297 the negative form of @option{-funsigned-char}. Likewise, the option
2298 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2300 @item -fsigned-bitfields
2301 @itemx -funsigned-bitfields
2302 @itemx -fno-signed-bitfields
2303 @itemx -fno-unsigned-bitfields
2304 @opindex fsigned-bitfields
2305 @opindex funsigned-bitfields
2306 @opindex fno-signed-bitfields
2307 @opindex fno-unsigned-bitfields
2308 These options control whether a bit-field is signed or unsigned, when the
2309 declaration does not use either @code{signed} or @code{unsigned}. By
2310 default, such a bit-field is signed, because this is consistent: the
2311 basic integer types such as @code{int} are signed types.
2313 @item -fsso-struct=@var{endianness}
2314 @opindex fsso-struct
2315 Set the default scalar storage order of structures and unions to the
2316 specified endianness. The accepted values are @samp{big-endian},
2317 @samp{little-endian} and @samp{native} for the native endianness of
2318 the target (the default). This option is not supported for C++.
2320 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2321 code that is not binary compatible with code generated without it if the
2322 specified endianness is not the native endianness of the target.
2325 @node C++ Dialect Options
2326 @section Options Controlling C++ Dialect
2328 @cindex compiler options, C++
2329 @cindex C++ options, command-line
2330 @cindex options, C++
2331 This section describes the command-line options that are only meaningful
2332 for C++ programs. You can also use most of the GNU compiler options
2333 regardless of what language your program is in. For example, you
2334 might compile a file @file{firstClass.C} like this:
2337 g++ -g -fstrict-enums -O -c firstClass.C
2341 In this example, only @option{-fstrict-enums} is an option meant
2342 only for C++ programs; you can use the other options with any
2343 language supported by GCC@.
2345 Some options for compiling C programs, such as @option{-std}, are also
2346 relevant for C++ programs.
2347 @xref{C Dialect Options,,Options Controlling C Dialect}.
2349 Here is a list of options that are @emph{only} for compiling C++ programs:
2353 @item -fabi-version=@var{n}
2354 @opindex fabi-version
2355 Use version @var{n} of the C++ ABI@. The default is version 0.
2357 Version 0 refers to the version conforming most closely to
2358 the C++ ABI specification. Therefore, the ABI obtained using version 0
2359 will change in different versions of G++ as ABI bugs are fixed.
2361 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2363 Version 2 is the version of the C++ ABI that first appeared in G++
2364 3.4, and was the default through G++ 4.9.
2366 Version 3 corrects an error in mangling a constant address as a
2369 Version 4, which first appeared in G++ 4.5, implements a standard
2370 mangling for vector types.
2372 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2373 attribute const/volatile on function pointer types, decltype of a
2374 plain decl, and use of a function parameter in the declaration of
2377 Version 6, which first appeared in G++ 4.7, corrects the promotion
2378 behavior of C++11 scoped enums and the mangling of template argument
2379 packs, const/static_cast, prefix ++ and --, and a class scope function
2380 used as a template argument.
2382 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2383 builtin type and corrects the mangling of lambdas in default argument
2386 Version 8, which first appeared in G++ 4.9, corrects the substitution
2387 behavior of function types with function-cv-qualifiers.
2389 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2392 Version 10, which first appeared in G++ 6.1, adds mangling of
2393 attributes that affect type identity, such as ia32 calling convention
2394 attributes (e.g.@: @samp{stdcall}).
2396 Version 11, which first appeared in G++ 7, corrects the mangling of
2397 sizeof... expressions and operator names. For multiple entities with
2398 the same name within a function, that are declared in different scopes,
2399 the mangling now changes starting with the twelfth occurrence. It also
2400 implies @option{-fnew-inheriting-ctors}.
2402 Version 12, which first appeared in G++ 8, corrects the calling
2403 conventions for empty classes on the x86_64 target and for classes
2404 with only deleted copy/move constructors. It accidentally changes the
2405 calling convention for classes with a deleted copy constructor and a
2406 trivial move constructor.
2408 Version 13, which first appeared in G++ 8.2, fixes the accidental
2409 change in version 12.
2411 See also @option{-Wabi}.
2413 @item -fabi-compat-version=@var{n}
2414 @opindex fabi-compat-version
2415 On targets that support strong aliases, G++
2416 works around mangling changes by creating an alias with the correct
2417 mangled name when defining a symbol with an incorrect mangled name.
2418 This switch specifies which ABI version to use for the alias.
2420 With @option{-fabi-version=0} (the default), this defaults to 11 (GCC 7
2421 compatibility). If another ABI version is explicitly selected, this
2422 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2423 use @option{-fabi-compat-version=2}.
2425 If this option is not provided but @option{-Wabi=@var{n}} is, that
2426 version is used for compatibility aliases. If this option is provided
2427 along with @option{-Wabi} (without the version), the version from this
2428 option is used for the warning.
2430 @item -fno-access-control
2431 @opindex fno-access-control
2432 @opindex faccess-control
2433 Turn off all access checking. This switch is mainly useful for working
2434 around bugs in the access control code.
2437 @opindex faligned-new
2438 Enable support for C++17 @code{new} of types that require more
2439 alignment than @code{void* ::operator new(std::size_t)} provides. A
2440 numeric argument such as @code{-faligned-new=32} can be used to
2441 specify how much alignment (in bytes) is provided by that function,
2442 but few users will need to override the default of
2443 @code{alignof(std::max_align_t)}.
2445 This flag is enabled by default for @option{-std=c++17}.
2450 @opindex fno-char8_t
2451 Enable support for @code{char8_t} as adopted for C++2a. This includes
2452 the addition of a new @code{char8_t} fundamental type, changes to the
2453 types of UTF-8 string and character literals, new signatures for
2454 user-defined literals, associated standard library updates, and new
2455 @code{__cpp_char8_t} and @code{__cpp_lib_char8_t} feature test macros.
2457 This option enables functions to be overloaded for ordinary and UTF-8
2461 int f(const char *); // #1
2462 int f(const char8_t *); // #2
2463 int v1 = f("text"); // Calls #1
2464 int v2 = f(u8"text"); // Calls #2
2468 and introduces new signatures for user-defined literals:
2471 int operator""_udl1(char8_t);
2472 int v3 = u8'x'_udl1;
2473 int operator""_udl2(const char8_t*, std::size_t);
2474 int v4 = u8"text"_udl2;
2475 template<typename T, T...> int operator""_udl3();
2476 int v5 = u8"text"_udl3;
2480 The change to the types of UTF-8 string and character literals introduces
2481 incompatibilities with ISO C++11 and later standards. For example, the
2482 following code is well-formed under ISO C++11, but is ill-formed when
2483 @option{-fchar8_t} is specified.
2486 char ca[] = u8"xx"; // error: char-array initialized from wide
2488 const char *cp = u8"xx";// error: invalid conversion from
2489 // `const char8_t*' to `const char*'
2491 auto v = f(u8"xx"); // error: invalid conversion from
2492 // `const char8_t*' to `const char*'
2493 std::string s@{u8"xx"@}; // error: no matching function for call to
2494 // `std::basic_string<char>::basic_string()'
2495 using namespace std::literals;
2496 s = u8"xx"s; // error: conversion from
2497 // `basic_string<char8_t>' to non-scalar
2498 // type `basic_string<char>' requested
2503 Check that the pointer returned by @code{operator new} is non-null
2504 before attempting to modify the storage allocated. This check is
2505 normally unnecessary because the C++ standard specifies that
2506 @code{operator new} only returns @code{0} if it is declared
2507 @code{throw()}, in which case the compiler always checks the
2508 return value even without this option. In all other cases, when
2509 @code{operator new} has a non-empty exception specification, memory
2510 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2511 @samp{new (nothrow)}.
2515 Enable support for the C++ Extensions for Concepts Technical
2516 Specification, ISO 19217 (2015), which allows code like
2519 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2520 template <Addable T> T add (T a, T b) @{ return a + b; @}
2523 @item -fconstexpr-depth=@var{n}
2524 @opindex fconstexpr-depth
2525 Set the maximum nested evaluation depth for C++11 constexpr functions
2526 to @var{n}. A limit is needed to detect endless recursion during
2527 constant expression evaluation. The minimum specified by the standard
2530 @item -fconstexpr-cache-depth=@var{n}
2531 @opindex fconstexpr-cache-depth
2532 Set the maximum level of nested evaluation depth for C++11 constexpr
2533 functions that will be cached to @var{n}. This is a heuristic that
2534 trades off compilation speed (when the cache avoids repeated
2535 calculations) against memory consumption (when the cache grows very
2536 large from highly recursive evaluations). The default is 8. Very few
2537 users are likely to want to adjust it, but if your code does heavy
2538 constexpr calculations you might want to experiment to find which
2539 value works best for you.
2541 @item -fconstexpr-loop-limit=@var{n}
2542 @opindex fconstexpr-loop-limit
2543 Set the maximum number of iterations for a loop in C++14 constexpr functions
2544 to @var{n}. A limit is needed to detect infinite loops during
2545 constant expression evaluation. The default is 262144 (1<<18).
2547 @item -fconstexpr-ops-limit=@var{n}
2548 @opindex fconstexpr-ops-limit
2549 Set the maximum number of operations during a single constexpr evaluation.
2550 Even when number of iterations of a single loop is limited with the above limit,
2551 if there are several nested loops and each of them has many iterations but still
2552 smaller than the above limit, or if in a body of some loop or even outside
2553 of a loop too many expressions need to be evaluated, the resulting constexpr
2554 evaluation might take too long.
2555 The default is 33554432 (1<<25).
2557 @item -fno-elide-constructors
2558 @opindex fno-elide-constructors
2559 @opindex felide-constructors
2560 The C++ standard allows an implementation to omit creating a temporary
2561 that is only used to initialize another object of the same type.
2562 Specifying this option disables that optimization, and forces G++ to
2563 call the copy constructor in all cases. This option also causes G++
2564 to call trivial member functions which otherwise would be expanded inline.
2566 In C++17, the compiler is required to omit these temporaries, but this
2567 option still affects trivial member functions.
2569 @item -fno-enforce-eh-specs
2570 @opindex fno-enforce-eh-specs
2571 @opindex fenforce-eh-specs
2572 Don't generate code to check for violation of exception specifications
2573 at run time. This option violates the C++ standard, but may be useful
2574 for reducing code size in production builds, much like defining
2575 @code{NDEBUG}. This does not give user code permission to throw
2576 exceptions in violation of the exception specifications; the compiler
2577 still optimizes based on the specifications, so throwing an
2578 unexpected exception results in undefined behavior at run time.
2580 @item -fextern-tls-init
2581 @itemx -fno-extern-tls-init
2582 @opindex fextern-tls-init
2583 @opindex fno-extern-tls-init
2584 The C++11 and OpenMP standards allow @code{thread_local} and
2585 @code{threadprivate} variables to have dynamic (runtime)
2586 initialization. To support this, any use of such a variable goes
2587 through a wrapper function that performs any necessary initialization.
2588 When the use and definition of the variable are in the same
2589 translation unit, this overhead can be optimized away, but when the
2590 use is in a different translation unit there is significant overhead
2591 even if the variable doesn't actually need dynamic initialization. If
2592 the programmer can be sure that no use of the variable in a
2593 non-defining TU needs to trigger dynamic initialization (either
2594 because the variable is statically initialized, or a use of the
2595 variable in the defining TU will be executed before any uses in
2596 another TU), they can avoid this overhead with the
2597 @option{-fno-extern-tls-init} option.
2599 On targets that support symbol aliases, the default is
2600 @option{-fextern-tls-init}. On targets that do not support symbol
2601 aliases, the default is @option{-fno-extern-tls-init}.
2603 @item -fno-gnu-keywords
2604 @opindex fno-gnu-keywords
2605 @opindex fgnu-keywords
2606 Do not recognize @code{typeof} as a keyword, so that code can use this
2607 word as an identifier. You can use the keyword @code{__typeof__} instead.
2608 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2609 @option{-std=c++98}, @option{-std=c++11}, etc.
2611 @item -fno-implicit-templates
2612 @opindex fno-implicit-templates
2613 @opindex fimplicit-templates
2614 Never emit code for non-inline templates that are instantiated
2615 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2616 If you use this option, you must take care to structure your code to
2617 include all the necessary explicit instantiations to avoid getting
2618 undefined symbols at link time.
2619 @xref{Template Instantiation}, for more information.
2621 @item -fno-implicit-inline-templates
2622 @opindex fno-implicit-inline-templates
2623 @opindex fimplicit-inline-templates
2624 Don't emit code for implicit instantiations of inline templates, either.
2625 The default is to handle inlines differently so that compiles with and
2626 without optimization need the same set of explicit instantiations.
2628 @item -fno-implement-inlines
2629 @opindex fno-implement-inlines
2630 @opindex fimplement-inlines
2631 To save space, do not emit out-of-line copies of inline functions
2632 controlled by @code{#pragma implementation}. This causes linker
2633 errors if these functions are not inlined everywhere they are called.
2635 @item -fms-extensions
2636 @opindex fms-extensions
2637 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2638 int and getting a pointer to member function via non-standard syntax.
2640 @item -fnew-inheriting-ctors
2641 @opindex fnew-inheriting-ctors
2642 Enable the P0136 adjustment to the semantics of C++11 constructor
2643 inheritance. This is part of C++17 but also considered to be a Defect
2644 Report against C++11 and C++14. This flag is enabled by default
2645 unless @option{-fabi-version=10} or lower is specified.
2647 @item -fnew-ttp-matching
2648 @opindex fnew-ttp-matching
2649 Enable the P0522 resolution to Core issue 150, template template
2650 parameters and default arguments: this allows a template with default
2651 template arguments as an argument for a template template parameter
2652 with fewer template parameters. This flag is enabled by default for
2653 @option{-std=c++17}.
2655 @item -fno-nonansi-builtins
2656 @opindex fno-nonansi-builtins
2657 @opindex fnonansi-builtins
2658 Disable built-in declarations of functions that are not mandated by
2659 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2660 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2663 @opindex fnothrow-opt
2664 Treat a @code{throw()} exception specification as if it were a
2665 @code{noexcept} specification to reduce or eliminate the text size
2666 overhead relative to a function with no exception specification. If
2667 the function has local variables of types with non-trivial
2668 destructors, the exception specification actually makes the
2669 function smaller because the EH cleanups for those variables can be
2670 optimized away. The semantic effect is that an exception thrown out of
2671 a function with such an exception specification results in a call
2672 to @code{terminate} rather than @code{unexpected}.
2674 @item -fno-operator-names
2675 @opindex fno-operator-names
2676 @opindex foperator-names
2677 Do not treat the operator name keywords @code{and}, @code{bitand},
2678 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2679 synonyms as keywords.
2681 @item -fno-optional-diags
2682 @opindex fno-optional-diags
2683 @opindex foptional-diags
2684 Disable diagnostics that the standard says a compiler does not need to
2685 issue. Currently, the only such diagnostic issued by G++ is the one for
2686 a name having multiple meanings within a class.
2689 @opindex fpermissive
2690 Downgrade some diagnostics about nonconformant code from errors to
2691 warnings. Thus, using @option{-fpermissive} allows some
2692 nonconforming code to compile.
2694 @item -fno-pretty-templates
2695 @opindex fno-pretty-templates
2696 @opindex fpretty-templates
2697 When an error message refers to a specialization of a function
2698 template, the compiler normally prints the signature of the
2699 template followed by the template arguments and any typedefs or
2700 typenames in the signature (e.g.@: @code{void f(T) [with T = int]}
2701 rather than @code{void f(int)}) so that it's clear which template is
2702 involved. When an error message refers to a specialization of a class
2703 template, the compiler omits any template arguments that match
2704 the default template arguments for that template. If either of these
2705 behaviors make it harder to understand the error message rather than
2706 easier, you can use @option{-fno-pretty-templates} to disable them.
2711 Disable generation of information about every class with virtual
2712 functions for use by the C++ run-time type identification features
2713 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2714 of the language, you can save some space by using this flag. Note that
2715 exception handling uses the same information, but G++ generates it as
2716 needed. The @code{dynamic_cast} operator can still be used for casts that
2717 do not require run-time type information, i.e.@: casts to @code{void *} or to
2718 unambiguous base classes.
2720 Mixing code compiled with @option{-frtti} with that compiled with
2721 @option{-fno-rtti} may not work. For example, programs may
2722 fail to link if a class compiled with @option{-fno-rtti} is used as a base
2723 for a class compiled with @option{-frtti}.
2725 @item -fsized-deallocation
2726 @opindex fsized-deallocation
2727 Enable the built-in global declarations
2729 void operator delete (void *, std::size_t) noexcept;
2730 void operator delete[] (void *, std::size_t) noexcept;
2732 as introduced in C++14. This is useful for user-defined replacement
2733 deallocation functions that, for example, use the size of the object
2734 to make deallocation faster. Enabled by default under
2735 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2736 warns about places that might want to add a definition.
2738 @item -fstrict-enums
2739 @opindex fstrict-enums
2740 Allow the compiler to optimize using the assumption that a value of
2741 enumerated type can only be one of the values of the enumeration (as
2742 defined in the C++ standard; basically, a value that can be
2743 represented in the minimum number of bits needed to represent all the
2744 enumerators). This assumption may not be valid if the program uses a
2745 cast to convert an arbitrary integer value to the enumerated type.
2747 @item -fstrong-eval-order
2748 @opindex fstrong-eval-order
2749 Evaluate member access, array subscripting, and shift expressions in
2750 left-to-right order, and evaluate assignment in right-to-left order,
2751 as adopted for C++17. Enabled by default with @option{-std=c++17}.
2752 @option{-fstrong-eval-order=some} enables just the ordering of member
2753 access and shift expressions, and is the default without
2754 @option{-std=c++17}.
2756 @item -ftemplate-backtrace-limit=@var{n}
2757 @opindex ftemplate-backtrace-limit
2758 Set the maximum number of template instantiation notes for a single
2759 warning or error to @var{n}. The default value is 10.
2761 @item -ftemplate-depth=@var{n}
2762 @opindex ftemplate-depth
2763 Set the maximum instantiation depth for template classes to @var{n}.
2764 A limit on the template instantiation depth is needed to detect
2765 endless recursions during template class instantiation. ANSI/ISO C++
2766 conforming programs must not rely on a maximum depth greater than 17
2767 (changed to 1024 in C++11). The default value is 900, as the compiler
2768 can run out of stack space before hitting 1024 in some situations.
2770 @item -fno-threadsafe-statics
2771 @opindex fno-threadsafe-statics
2772 @opindex fthreadsafe-statics
2773 Do not emit the extra code to use the routines specified in the C++
2774 ABI for thread-safe initialization of local statics. You can use this
2775 option to reduce code size slightly in code that doesn't need to be
2778 @item -fuse-cxa-atexit
2779 @opindex fuse-cxa-atexit
2780 Register destructors for objects with static storage duration with the
2781 @code{__cxa_atexit} function rather than the @code{atexit} function.
2782 This option is required for fully standards-compliant handling of static
2783 destructors, but only works if your C library supports
2784 @code{__cxa_atexit}.
2786 @item -fno-use-cxa-get-exception-ptr
2787 @opindex fno-use-cxa-get-exception-ptr
2788 @opindex fuse-cxa-get-exception-ptr
2789 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2790 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2791 if the runtime routine is not available.
2793 @item -fvisibility-inlines-hidden
2794 @opindex fvisibility-inlines-hidden
2795 This switch declares that the user does not attempt to compare
2796 pointers to inline functions or methods where the addresses of the two functions
2797 are taken in different shared objects.
2799 The effect of this is that GCC may, effectively, mark inline methods with
2800 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2801 appear in the export table of a DSO and do not require a PLT indirection
2802 when used within the DSO@. Enabling this option can have a dramatic effect
2803 on load and link times of a DSO as it massively reduces the size of the
2804 dynamic export table when the library makes heavy use of templates.
2806 The behavior of this switch is not quite the same as marking the
2807 methods as hidden directly, because it does not affect static variables
2808 local to the function or cause the compiler to deduce that
2809 the function is defined in only one shared object.
2811 You may mark a method as having a visibility explicitly to negate the
2812 effect of the switch for that method. For example, if you do want to
2813 compare pointers to a particular inline method, you might mark it as
2814 having default visibility. Marking the enclosing class with explicit
2815 visibility has no effect.
2817 Explicitly instantiated inline methods are unaffected by this option
2818 as their linkage might otherwise cross a shared library boundary.
2819 @xref{Template Instantiation}.
2821 @item -fvisibility-ms-compat
2822 @opindex fvisibility-ms-compat
2823 This flag attempts to use visibility settings to make GCC's C++
2824 linkage model compatible with that of Microsoft Visual Studio.
2826 The flag makes these changes to GCC's linkage model:
2830 It sets the default visibility to @code{hidden}, like
2831 @option{-fvisibility=hidden}.
2834 Types, but not their members, are not hidden by default.
2837 The One Definition Rule is relaxed for types without explicit
2838 visibility specifications that are defined in more than one
2839 shared object: those declarations are permitted if they are
2840 permitted when this option is not used.
2843 In new code it is better to use @option{-fvisibility=hidden} and
2844 export those classes that are intended to be externally visible.
2845 Unfortunately it is possible for code to rely, perhaps accidentally,
2846 on the Visual Studio behavior.
2848 Among the consequences of these changes are that static data members
2849 of the same type with the same name but defined in different shared
2850 objects are different, so changing one does not change the other;
2851 and that pointers to function members defined in different shared
2852 objects may not compare equal. When this flag is given, it is a
2853 violation of the ODR to define types with the same name differently.
2858 Do not use weak symbol support, even if it is provided by the linker.
2859 By default, G++ uses weak symbols if they are available. This
2860 option exists only for testing, and should not be used by end-users;
2861 it results in inferior code and has no benefits. This option may
2862 be removed in a future release of G++.
2866 Do not search for header files in the standard directories specific to
2867 C++, but do still search the other standard directories. (This option
2868 is used when building the C++ library.)
2871 In addition, these optimization, warning, and code generation options
2872 have meanings only for C++ programs:
2875 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2878 Warn when G++ it generates code that is probably not compatible with
2879 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2880 ABI with each major release, normally @option{-Wabi} will warn only if
2881 there is a check added later in a release series for an ABI issue
2882 discovered since the initial release. @option{-Wabi} will warn about
2883 more things if an older ABI version is selected (with
2884 @option{-fabi-version=@var{n}}).
2886 @option{-Wabi} can also be used with an explicit version number to
2887 warn about compatibility with a particular @option{-fabi-version}
2888 level, e.g.@: @option{-Wabi=2} to warn about changes relative to
2889 @option{-fabi-version=2}.
2891 If an explicit version number is provided and
2892 @option{-fabi-compat-version} is not specified, the version number
2893 from this option is used for compatibility aliases. If no explicit
2894 version number is provided with this option, but
2895 @option{-fabi-compat-version} is specified, that version number is
2896 used for ABI warnings.
2898 Although an effort has been made to warn about
2899 all such cases, there are probably some cases that are not warned about,
2900 even though G++ is generating incompatible code. There may also be
2901 cases where warnings are emitted even though the code that is generated
2904 You should rewrite your code to avoid these warnings if you are
2905 concerned about the fact that code generated by G++ may not be binary
2906 compatible with code generated by other compilers.
2908 Known incompatibilities in @option{-fabi-version=2} (which was the
2909 default from GCC 3.4 to 4.9) include:
2914 A template with a non-type template parameter of reference type was
2915 mangled incorrectly:
2918 template <int &> struct S @{@};
2922 This was fixed in @option{-fabi-version=3}.
2925 SIMD vector types declared using @code{__attribute ((vector_size))} were
2926 mangled in a non-standard way that does not allow for overloading of
2927 functions taking vectors of different sizes.
2929 The mangling was changed in @option{-fabi-version=4}.
2932 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2933 qualifiers, and @code{decltype} of a plain declaration was folded away.
2935 These mangling issues were fixed in @option{-fabi-version=5}.
2938 Scoped enumerators passed as arguments to a variadic function are
2939 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2940 On most targets this does not actually affect the parameter passing
2941 ABI, as there is no way to pass an argument smaller than @code{int}.
2943 Also, the ABI changed the mangling of template argument packs,
2944 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2945 a class scope function used as a template argument.
2947 These issues were corrected in @option{-fabi-version=6}.
2950 Lambdas in default argument scope were mangled incorrectly, and the
2951 ABI changed the mangling of @code{nullptr_t}.
2953 These issues were corrected in @option{-fabi-version=7}.
2956 When mangling a function type with function-cv-qualifiers, the
2957 un-qualified function type was incorrectly treated as a substitution
2960 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2963 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2964 unaligned accesses. Note that this did not affect the ABI of a
2965 function with a @code{nullptr_t} parameter, as parameters have a
2968 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2971 Target-specific attributes that affect the identity of a type, such as
2972 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2973 did not affect the mangled name, leading to name collisions when
2974 function pointers were used as template arguments.
2976 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2980 It also warns about psABI-related changes. The known psABI changes at this
2986 For SysV/x86-64, unions with @code{long double} members are
2987 passed in memory as specified in psABI. For example:
2997 @code{union U} is always passed in memory.
3001 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
3004 Warn when a type with an ABI tag is used in a context that does not
3005 have that ABI tag. See @ref{C++ Attributes} for more information
3008 @item -Wcomma-subscript @r{(C++ and Objective-C++ only)}
3009 @opindex Wcomma-subscript
3010 @opindex Wno-comma-subscript
3011 Warn about uses of a comma expression within a subscripting expression.
3012 This usage was deprecated in C++2a. However, a comma expression wrapped
3013 in @code{( )} is not deprecated. Example:
3017 void f(int *a, int b, int c) @{
3018 a[b,c]; // deprecated
3024 Enabled by default with @option{-std=c++2a}.
3026 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
3027 @opindex Wctor-dtor-privacy
3028 @opindex Wno-ctor-dtor-privacy
3029 Warn when a class seems unusable because all the constructors or
3030 destructors in that class are private, and it has neither friends nor
3031 public static member functions. Also warn if there are no non-private
3032 methods, and there's at least one private member function that isn't
3033 a constructor or destructor.
3035 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
3036 @opindex Wdelete-non-virtual-dtor
3037 @opindex Wno-delete-non-virtual-dtor
3038 Warn when @code{delete} is used to destroy an instance of a class that
3039 has virtual functions and non-virtual destructor. It is unsafe to delete
3040 an instance of a derived class through a pointer to a base class if the
3041 base class does not have a virtual destructor. This warning is enabled
3044 @item -Wdeprecated-copy @r{(C++ and Objective-C++ only)}
3045 @opindex Wdeprecated-copy
3046 @opindex Wno-deprecated-copy
3047 Warn that the implicit declaration of a copy constructor or copy
3048 assignment operator is deprecated if the class has a user-provided
3049 copy constructor or copy assignment operator, in C++11 and up. This
3050 warning is enabled by @option{-Wextra}. With
3051 @option{-Wdeprecated-copy-dtor}, also deprecate if the class has a
3052 user-provided destructor.
3054 @item -Wno-init-list-lifetime @r{(C++ and Objective-C++ only)}
3055 @opindex Winit-list-lifetime
3056 @opindex Wno-init-list-lifetime
3057 Do not warn about uses of @code{std::initializer_list} that are likely
3058 to result in dangling pointers. Since the underlying array for an
3059 @code{initializer_list} is handled like a normal C++ temporary object,
3060 it is easy to inadvertently keep a pointer to the array past the end
3061 of the array's lifetime. For example:
3065 If a function returns a temporary @code{initializer_list}, or a local
3066 @code{initializer_list} variable, the array's lifetime ends at the end
3067 of the return statement, so the value returned has a dangling pointer.
3070 If a new-expression creates an @code{initializer_list}, the array only
3071 lives until the end of the enclosing full-expression, so the
3072 @code{initializer_list} in the heap has a dangling pointer.
3075 When an @code{initializer_list} variable is assigned from a
3076 brace-enclosed initializer list, the temporary array created for the
3077 right side of the assignment only lives until the end of the
3078 full-expression, so at the next statement the @code{initializer_list}
3079 variable has a dangling pointer.
3082 // li's initial underlying array lives as long as li
3083 std::initializer_list<int> li = @{ 1,2,3 @};
3084 // assignment changes li to point to a temporary array
3086 // now the temporary is gone and li has a dangling pointer
3087 int i = li.begin()[0] // undefined behavior
3091 When a list constructor stores the @code{begin} pointer from the
3092 @code{initializer_list} argument, this doesn't extend the lifetime of
3093 the array, so if a class variable is constructed from a temporary
3094 @code{initializer_list}, the pointer is left dangling by the end of
3095 the variable declaration statement.
3099 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
3100 @opindex Wliteral-suffix
3101 @opindex Wno-literal-suffix
3102 Warn when a string or character literal is followed by a ud-suffix which does
3103 not begin with an underscore. As a conforming extension, GCC treats such
3104 suffixes as separate preprocessing tokens in order to maintain backwards
3105 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
3109 #define __STDC_FORMAT_MACROS
3110 #include <inttypes.h>
3115 printf("My int64: %" PRId64"\n", i64);
3119 In this case, @code{PRId64} is treated as a separate preprocessing token.
3121 Additionally, warn when a user-defined literal operator is declared with
3122 a literal suffix identifier that doesn't begin with an underscore. Literal
3123 suffix identifiers that don't begin with an underscore are reserved for
3124 future standardization.
3126 This warning is enabled by default.
3128 @item -Wlto-type-mismatch
3129 @opindex Wlto-type-mismatch
3130 @opindex Wno-lto-type-mismatch
3132 During the link-time optimization warn about type mismatches in
3133 global declarations from different compilation units.
3134 Requires @option{-flto} to be enabled. Enabled by default.
3136 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
3138 @opindex Wno-narrowing
3139 For C++11 and later standards, narrowing conversions are diagnosed by default,
3140 as required by the standard. A narrowing conversion from a constant produces
3141 an error, and a narrowing conversion from a non-constant produces a warning,
3142 but @option{-Wno-narrowing} suppresses the diagnostic.
3143 Note that this does not affect the meaning of well-formed code;
3144 narrowing conversions are still considered ill-formed in SFINAE contexts.
3146 With @option{-Wnarrowing} in C++98, warn when a narrowing
3147 conversion prohibited by C++11 occurs within
3151 int i = @{ 2.2 @}; // error: narrowing from double to int
3154 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
3156 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
3158 @opindex Wno-noexcept
3159 Warn when a noexcept-expression evaluates to false because of a call
3160 to a function that does not have a non-throwing exception
3161 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
3162 the compiler to never throw an exception.
3164 @item -Wnoexcept-type @r{(C++ and Objective-C++ only)}
3165 @opindex Wnoexcept-type
3166 @opindex Wno-noexcept-type
3167 Warn if the C++17 feature making @code{noexcept} part of a function
3168 type changes the mangled name of a symbol relative to C++14. Enabled
3169 by @option{-Wabi} and @option{-Wc++17-compat}.
3174 template <class T> void f(T t) @{ t(); @};
3176 void h() @{ f(g); @}
3180 In C++14, @code{f} calls @code{f<void(*)()>}, but in
3181 C++17 it calls @code{f<void(*)()noexcept>}.
3183 @item -Wclass-memaccess @r{(C++ and Objective-C++ only)}
3184 @opindex Wclass-memaccess
3185 @opindex Wno-class-memaccess
3186 Warn when the destination of a call to a raw memory function such as
3187 @code{memset} or @code{memcpy} is an object of class type, and when writing
3188 into such an object might bypass the class non-trivial or deleted constructor
3189 or copy assignment, violate const-correctness or encapsulation, or corrupt
3190 virtual table pointers. Modifying the representation of such objects may
3191 violate invariants maintained by member functions of the class. For example,
3192 the call to @code{memset} below is undefined because it modifies a non-trivial
3193 class object and is, therefore, diagnosed. The safe way to either initialize
3194 or clear the storage of objects of such types is by using the appropriate
3195 constructor or assignment operator, if one is available.
3197 std::string str = "abc";
3198 memset (&str, 0, sizeof str);
3200 The @option{-Wclass-memaccess} option is enabled by @option{-Wall}.
3201 Explicitly casting the pointer to the class object to @code{void *} or
3202 to a type that can be safely accessed by the raw memory function suppresses
3205 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
3206 @opindex Wnon-virtual-dtor
3207 @opindex Wno-non-virtual-dtor
3208 Warn when a class has virtual functions and an accessible non-virtual
3209 destructor itself or in an accessible polymorphic base class, in which
3210 case it is possible but unsafe to delete an instance of a derived
3211 class through a pointer to the class itself or base class. This
3212 warning is automatically enabled if @option{-Weffc++} is specified.
3214 @item -Wregister @r{(C++ and Objective-C++ only)}
3216 @opindex Wno-register
3217 Warn on uses of the @code{register} storage class specifier, except
3218 when it is part of the GNU @ref{Explicit Register Variables} extension.
3219 The use of the @code{register} keyword as storage class specifier has
3220 been deprecated in C++11 and removed in C++17.
3221 Enabled by default with @option{-std=c++17}.
3223 @item -Wreorder @r{(C++ and Objective-C++ only)}
3225 @opindex Wno-reorder
3226 @cindex reordering, warning
3227 @cindex warning for reordering of member initializers
3228 Warn when the order of member initializers given in the code does not
3229 match the order in which they must be executed. For instance:
3235 A(): j (0), i (1) @{ @}
3240 The compiler rearranges the member initializers for @code{i}
3241 and @code{j} to match the declaration order of the members, emitting
3242 a warning to that effect. This warning is enabled by @option{-Wall}.
3244 @item -Wno-pessimizing-move @r{(C++ and Objective-C++ only)}
3245 @opindex Wpessimizing-move
3246 @opindex Wno-pessimizing-move
3247 This warning warns when a call to @code{std::move} prevents copy
3248 elision. A typical scenario when copy elision can occur is when returning in
3249 a function with a class return type, when the expression being returned is the
3250 name of a non-volatile automatic object, and is not a function parameter, and
3251 has the same type as the function return type.
3261 return std::move (t);
3265 But in this example, the @code{std::move} call prevents copy elision.
3267 This warning is enabled by @option{-Wall}.
3269 @item -Wno-redundant-move @r{(C++ and Objective-C++ only)}
3270 @opindex Wredundant-move
3271 @opindex Wno-redundant-move
3272 This warning warns about redundant calls to @code{std::move}; that is, when
3273 a move operation would have been performed even without the @code{std::move}
3274 call. This happens because the compiler is forced to treat the object as if
3275 it were an rvalue in certain situations such as returning a local variable,
3276 where copy elision isn't applicable. Consider:
3285 return std::move (t);
3289 Here, the @code{std::move} call is redundant. Because G++ implements Core
3290 Issue 1579, another example is:
3293 struct T @{ // convertible to U
3303 return std::move (t);
3306 In this example, copy elision isn't applicable because the type of the
3307 expression being returned and the function return type differ, yet G++
3308 treats the return value as if it were designated by an rvalue.
3310 This warning is enabled by @option{-Wextra}.
3312 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
3313 @opindex fext-numeric-literals
3314 @opindex fno-ext-numeric-literals
3315 Accept imaginary, fixed-point, or machine-defined
3316 literal number suffixes as GNU extensions.
3317 When this option is turned off these suffixes are treated
3318 as C++11 user-defined literal numeric suffixes.
3319 This is on by default for all pre-C++11 dialects and all GNU dialects:
3320 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
3321 @option{-std=gnu++14}.
3322 This option is off by default
3323 for ISO C++11 onwards (@option{-std=c++11}, ...).
3326 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
3329 @item -Weffc++ @r{(C++ and Objective-C++ only)}
3332 Warn about violations of the following style guidelines from Scott Meyers'
3333 @cite{Effective C++} series of books:
3337 Define a copy constructor and an assignment operator for classes
3338 with dynamically-allocated memory.
3341 Prefer initialization to assignment in constructors.
3344 Have @code{operator=} return a reference to @code{*this}.
3347 Don't try to return a reference when you must return an object.
3350 Distinguish between prefix and postfix forms of increment and
3351 decrement operators.
3354 Never overload @code{&&}, @code{||}, or @code{,}.
3358 This option also enables @option{-Wnon-virtual-dtor}, which is also
3359 one of the effective C++ recommendations. However, the check is
3360 extended to warn about the lack of virtual destructor in accessible
3361 non-polymorphic bases classes too.
3363 When selecting this option, be aware that the standard library
3364 headers do not obey all of these guidelines; use @samp{grep -v}
3365 to filter out those warnings.
3367 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
3368 @opindex Wstrict-null-sentinel
3369 @opindex Wno-strict-null-sentinel
3370 Warn about the use of an uncasted @code{NULL} as sentinel. When
3371 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
3372 to @code{__null}. Although it is a null pointer constant rather than a
3373 null pointer, it is guaranteed to be of the same size as a pointer.
3374 But this use is not portable across different compilers.
3376 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
3377 @opindex Wno-non-template-friend
3378 @opindex Wnon-template-friend
3379 Disable warnings when non-template friend functions are declared
3380 within a template. In very old versions of GCC that predate implementation
3381 of the ISO standard, declarations such as
3382 @samp{friend int foo(int)}, where the name of the friend is an unqualified-id,
3383 could be interpreted as a particular specialization of a template
3384 function; the warning exists to diagnose compatibility problems,
3385 and is enabled by default.
3387 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
3388 @opindex Wold-style-cast
3389 @opindex Wno-old-style-cast
3390 Warn if an old-style (C-style) cast to a non-void type is used within
3391 a C++ program. The new-style casts (@code{dynamic_cast},
3392 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
3393 less vulnerable to unintended effects and much easier to search for.
3395 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
3396 @opindex Woverloaded-virtual
3397 @opindex Wno-overloaded-virtual
3398 @cindex overloaded virtual function, warning
3399 @cindex warning for overloaded virtual function
3400 Warn when a function declaration hides virtual functions from a
3401 base class. For example, in:
3408 struct B: public A @{
3413 the @code{A} class version of @code{f} is hidden in @code{B}, and code
3424 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
3425 @opindex Wno-pmf-conversions
3426 @opindex Wpmf-conversions
3427 Disable the diagnostic for converting a bound pointer to member function
3430 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
3431 @opindex Wsign-promo
3432 @opindex Wno-sign-promo
3433 Warn when overload resolution chooses a promotion from unsigned or
3434 enumerated type to a signed type, over a conversion to an unsigned type of
3435 the same size. Previous versions of G++ tried to preserve
3436 unsignedness, but the standard mandates the current behavior.
3438 @item -Wtemplates @r{(C++ and Objective-C++ only)}
3440 @opindex Wno-templates
3441 Warn when a primary template declaration is encountered. Some coding
3442 rules disallow templates, and this may be used to enforce that rule.
3443 The warning is inactive inside a system header file, such as the STL, so
3444 one can still use the STL. One may also instantiate or specialize
3447 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
3448 @opindex Wmultiple-inheritance
3449 @opindex Wno-multiple-inheritance
3450 Warn when a class is defined with multiple direct base classes. Some
3451 coding rules disallow multiple inheritance, and this may be used to
3452 enforce that rule. The warning is inactive inside a system header file,
3453 such as the STL, so one can still use the STL. One may also define
3454 classes that indirectly use multiple inheritance.
3456 @item -Wvirtual-inheritance
3457 @opindex Wvirtual-inheritance
3458 @opindex Wno-virtual-inheritance
3459 Warn when a class is defined with a virtual direct base class. Some
3460 coding rules disallow multiple inheritance, and this may be used to
3461 enforce that rule. The warning is inactive inside a system header file,
3462 such as the STL, so one can still use the STL. One may also define
3463 classes that indirectly use virtual inheritance.
3466 @opindex Wnamespaces
3467 @opindex Wno-namespaces
3468 Warn when a namespace definition is opened. Some coding rules disallow
3469 namespaces, and this may be used to enforce that rule. The warning is
3470 inactive inside a system header file, such as the STL, so one can still
3471 use the STL. One may also use using directives and qualified names.
3473 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3475 @opindex Wno-terminate
3476 Disable the warning about a throw-expression that will immediately
3477 result in a call to @code{terminate}.
3479 @item -Wno-class-conversion @r{(C++ and Objective-C++ only)}
3480 @opindex Wno-class-conversion
3481 @opindex Wclass-conversion
3482 Disable the warning about the case when a conversion function converts an
3483 object to the same type, to a base class of that type, or to void; such
3484 a conversion function will never be called.
3486 @item -Wvolatile @r{(C++ and Objective-C++ only)}
3488 @opindex Wno-volatile
3489 Warn about deprecated uses of the @code{volatile} qualifier. This includes
3490 postfix and prefix @code{++} and @code{--} expressions of
3491 @code{volatile}-qualified types, using simple assignments where the left
3492 operand is a @code{volatile}-qualified non-class type for their value,
3493 compound assignments where the left operand is a @code{volatile}-qualified
3494 non-class type, @code{volatile}-qualified function return type,
3495 @code{volatile}-qualified parameter type, and structured bindings of a
3496 @code{volatile}-qualified type. This usage was deprecated in C++20.
3498 Enabled by default with @option{-std=c++2a}.
3501 @node Objective-C and Objective-C++ Dialect Options
3502 @section Options Controlling Objective-C and Objective-C++ Dialects
3504 @cindex compiler options, Objective-C and Objective-C++
3505 @cindex Objective-C and Objective-C++ options, command-line
3506 @cindex options, Objective-C and Objective-C++
3507 (NOTE: This manual does not describe the Objective-C and Objective-C++
3508 languages themselves. @xref{Standards,,Language Standards
3509 Supported by GCC}, for references.)
3511 This section describes the command-line options that are only meaningful
3512 for Objective-C and Objective-C++ programs. You can also use most of
3513 the language-independent GNU compiler options.
3514 For example, you might compile a file @file{some_class.m} like this:
3517 gcc -g -fgnu-runtime -O -c some_class.m
3521 In this example, @option{-fgnu-runtime} is an option meant only for
3522 Objective-C and Objective-C++ programs; you can use the other options with
3523 any language supported by GCC@.
3525 Note that since Objective-C is an extension of the C language, Objective-C
3526 compilations may also use options specific to the C front-end (e.g.,
3527 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3528 C++-specific options (e.g., @option{-Wabi}).
3530 Here is a list of options that are @emph{only} for compiling Objective-C
3531 and Objective-C++ programs:
3534 @item -fconstant-string-class=@var{class-name}
3535 @opindex fconstant-string-class
3536 Use @var{class-name} as the name of the class to instantiate for each
3537 literal string specified with the syntax @code{@@"@dots{}"}. The default
3538 class name is @code{NXConstantString} if the GNU runtime is being used, and
3539 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3540 @option{-fconstant-cfstrings} option, if also present, overrides the
3541 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3542 to be laid out as constant CoreFoundation strings.
3545 @opindex fgnu-runtime
3546 Generate object code compatible with the standard GNU Objective-C
3547 runtime. This is the default for most types of systems.
3549 @item -fnext-runtime
3550 @opindex fnext-runtime
3551 Generate output compatible with the NeXT runtime. This is the default
3552 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3553 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3556 @item -fno-nil-receivers
3557 @opindex fno-nil-receivers
3558 @opindex fnil-receivers
3559 Assume that all Objective-C message dispatches (@code{[receiver
3560 message:arg]}) in this translation unit ensure that the receiver is
3561 not @code{nil}. This allows for more efficient entry points in the
3562 runtime to be used. This option is only available in conjunction with
3563 the NeXT runtime and ABI version 0 or 1.
3565 @item -fobjc-abi-version=@var{n}
3566 @opindex fobjc-abi-version
3567 Use version @var{n} of the Objective-C ABI for the selected runtime.
3568 This option is currently supported only for the NeXT runtime. In that
3569 case, Version 0 is the traditional (32-bit) ABI without support for
3570 properties and other Objective-C 2.0 additions. Version 1 is the
3571 traditional (32-bit) ABI with support for properties and other
3572 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3573 nothing is specified, the default is Version 0 on 32-bit target
3574 machines, and Version 2 on 64-bit target machines.
3576 @item -fobjc-call-cxx-cdtors
3577 @opindex fobjc-call-cxx-cdtors
3578 For each Objective-C class, check if any of its instance variables is a
3579 C++ object with a non-trivial default constructor. If so, synthesize a
3580 special @code{- (id) .cxx_construct} instance method which runs
3581 non-trivial default constructors on any such instance variables, in order,
3582 and then return @code{self}. Similarly, check if any instance variable
3583 is a C++ object with a non-trivial destructor, and if so, synthesize a
3584 special @code{- (void) .cxx_destruct} method which runs
3585 all such default destructors, in reverse order.
3587 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3588 methods thusly generated only operate on instance variables
3589 declared in the current Objective-C class, and not those inherited
3590 from superclasses. It is the responsibility of the Objective-C
3591 runtime to invoke all such methods in an object's inheritance
3592 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3593 by the runtime immediately after a new object instance is allocated;
3594 the @code{- (void) .cxx_destruct} methods are invoked immediately
3595 before the runtime deallocates an object instance.
3597 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3598 support for invoking the @code{- (id) .cxx_construct} and
3599 @code{- (void) .cxx_destruct} methods.
3601 @item -fobjc-direct-dispatch
3602 @opindex fobjc-direct-dispatch
3603 Allow fast jumps to the message dispatcher. On Darwin this is
3604 accomplished via the comm page.
3606 @item -fobjc-exceptions
3607 @opindex fobjc-exceptions
3608 Enable syntactic support for structured exception handling in
3609 Objective-C, similar to what is offered by C++. This option
3610 is required to use the Objective-C keywords @code{@@try},
3611 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3612 @code{@@synchronized}. This option is available with both the GNU
3613 runtime and the NeXT runtime (but not available in conjunction with
3614 the NeXT runtime on Mac OS X 10.2 and earlier).
3618 Enable garbage collection (GC) in Objective-C and Objective-C++
3619 programs. This option is only available with the NeXT runtime; the
3620 GNU runtime has a different garbage collection implementation that
3621 does not require special compiler flags.
3623 @item -fobjc-nilcheck
3624 @opindex fobjc-nilcheck
3625 For the NeXT runtime with version 2 of the ABI, check for a nil
3626 receiver in method invocations before doing the actual method call.
3627 This is the default and can be disabled using
3628 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3629 checked for nil in this way no matter what this flag is set to.
3630 Currently this flag does nothing when the GNU runtime, or an older
3631 version of the NeXT runtime ABI, is used.
3633 @item -fobjc-std=objc1
3635 Conform to the language syntax of Objective-C 1.0, the language
3636 recognized by GCC 4.0. This only affects the Objective-C additions to
3637 the C/C++ language; it does not affect conformance to C/C++ standards,
3638 which is controlled by the separate C/C++ dialect option flags. When
3639 this option is used with the Objective-C or Objective-C++ compiler,
3640 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3641 This is useful if you need to make sure that your Objective-C code can
3642 be compiled with older versions of GCC@.
3644 @item -freplace-objc-classes
3645 @opindex freplace-objc-classes
3646 Emit a special marker instructing @command{ld(1)} not to statically link in
3647 the resulting object file, and allow @command{dyld(1)} to load it in at
3648 run time instead. This is used in conjunction with the Fix-and-Continue
3649 debugging mode, where the object file in question may be recompiled and
3650 dynamically reloaded in the course of program execution, without the need
3651 to restart the program itself. Currently, Fix-and-Continue functionality
3652 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3657 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3658 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3659 compile time) with static class references that get initialized at load time,
3660 which improves run-time performance. Specifying the @option{-fzero-link} flag
3661 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3662 to be retained. This is useful in Zero-Link debugging mode, since it allows
3663 for individual class implementations to be modified during program execution.
3664 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3665 regardless of command-line options.
3667 @item -fno-local-ivars
3668 @opindex fno-local-ivars
3669 @opindex flocal-ivars
3670 By default instance variables in Objective-C can be accessed as if
3671 they were local variables from within the methods of the class they're
3672 declared in. This can lead to shadowing between instance variables
3673 and other variables declared either locally inside a class method or
3674 globally with the same name. Specifying the @option{-fno-local-ivars}
3675 flag disables this behavior thus avoiding variable shadowing issues.
3677 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3678 @opindex fivar-visibility
3679 Set the default instance variable visibility to the specified option
3680 so that instance variables declared outside the scope of any access
3681 modifier directives default to the specified visibility.
3685 Dump interface declarations for all classes seen in the source file to a
3686 file named @file{@var{sourcename}.decl}.
3688 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3689 @opindex Wassign-intercept
3690 @opindex Wno-assign-intercept
3691 Warn whenever an Objective-C assignment is being intercepted by the
3694 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3695 @opindex Wno-protocol
3697 If a class is declared to implement a protocol, a warning is issued for
3698 every method in the protocol that is not implemented by the class. The
3699 default behavior is to issue a warning for every method not explicitly
3700 implemented in the class, even if a method implementation is inherited
3701 from the superclass. If you use the @option{-Wno-protocol} option, then
3702 methods inherited from the superclass are considered to be implemented,
3703 and no warning is issued for them.
3705 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3707 @opindex Wno-selector
3708 Warn if multiple methods of different types for the same selector are
3709 found during compilation. The check is performed on the list of methods
3710 in the final stage of compilation. Additionally, a check is performed
3711 for each selector appearing in a @code{@@selector(@dots{})}
3712 expression, and a corresponding method for that selector has been found
3713 during compilation. Because these checks scan the method table only at
3714 the end of compilation, these warnings are not produced if the final
3715 stage of compilation is not reached, for example because an error is
3716 found during compilation, or because the @option{-fsyntax-only} option is
3719 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3720 @opindex Wstrict-selector-match
3721 @opindex Wno-strict-selector-match
3722 Warn if multiple methods with differing argument and/or return types are
3723 found for a given selector when attempting to send a message using this
3724 selector to a receiver of type @code{id} or @code{Class}. When this flag
3725 is off (which is the default behavior), the compiler omits such warnings
3726 if any differences found are confined to types that share the same size
3729 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3730 @opindex Wundeclared-selector
3731 @opindex Wno-undeclared-selector
3732 Warn if a @code{@@selector(@dots{})} expression referring to an
3733 undeclared selector is found. A selector is considered undeclared if no
3734 method with that name has been declared before the
3735 @code{@@selector(@dots{})} expression, either explicitly in an
3736 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3737 an @code{@@implementation} section. This option always performs its
3738 checks as soon as a @code{@@selector(@dots{})} expression is found,
3739 while @option{-Wselector} only performs its checks in the final stage of
3740 compilation. This also enforces the coding style convention
3741 that methods and selectors must be declared before being used.
3743 @item -print-objc-runtime-info
3744 @opindex print-objc-runtime-info
3745 Generate C header describing the largest structure that is passed by
3750 @node Diagnostic Message Formatting Options
3751 @section Options to Control Diagnostic Messages Formatting
3752 @cindex options to control diagnostics formatting
3753 @cindex diagnostic messages
3754 @cindex message formatting
3756 Traditionally, diagnostic messages have been formatted irrespective of
3757 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3758 options described below
3759 to control the formatting algorithm for diagnostic messages,
3760 e.g.@: how many characters per line, how often source location
3761 information should be reported. Note that some language front ends may not
3762 honor these options.
3765 @item -fmessage-length=@var{n}
3766 @opindex fmessage-length
3767 Try to format error messages so that they fit on lines of about
3768 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3769 done; each error message appears on a single line. This is the
3770 default for all front ends.
3772 Note - this option also affects the display of the @samp{#error} and
3773 @samp{#warning} pre-processor directives, and the @samp{deprecated}
3774 function/type/variable attribute. It does not however affect the
3775 @samp{pragma GCC warning} and @samp{pragma GCC error} pragmas.
3777 @item -fdiagnostics-show-location=once
3778 @opindex fdiagnostics-show-location
3779 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3780 reporter to emit source location information @emph{once}; that is, in
3781 case the message is too long to fit on a single physical line and has to
3782 be wrapped, the source location won't be emitted (as prefix) again,
3783 over and over, in subsequent continuation lines. This is the default
3786 @item -fdiagnostics-show-location=every-line
3787 Only meaningful in line-wrapping mode. Instructs the diagnostic
3788 messages reporter to emit the same source location information (as
3789 prefix) for physical lines that result from the process of breaking
3790 a message which is too long to fit on a single line.
3792 @item -fdiagnostics-color[=@var{WHEN}]
3793 @itemx -fno-diagnostics-color
3794 @opindex fdiagnostics-color
3795 @cindex highlight, color
3796 @vindex GCC_COLORS @r{environment variable}
3797 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3798 or @samp{auto}. The default depends on how the compiler has been configured,
3799 it can be any of the above @var{WHEN} options or also @samp{never}
3800 if @env{GCC_COLORS} environment variable isn't present in the environment,
3801 and @samp{auto} otherwise.
3802 @samp{auto} means to use color only when the standard error is a terminal.
3803 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3804 aliases for @option{-fdiagnostics-color=always} and
3805 @option{-fdiagnostics-color=never}, respectively.
3807 The colors are defined by the environment variable @env{GCC_COLORS}.
3808 Its value is a colon-separated list of capabilities and Select Graphic
3809 Rendition (SGR) substrings. SGR commands are interpreted by the
3810 terminal or terminal emulator. (See the section in the documentation
3811 of your text terminal for permitted values and their meanings as
3812 character attributes.) These substring values are integers in decimal
3813 representation and can be concatenated with semicolons.
3814 Common values to concatenate include
3816 @samp{4} for underline,
3818 @samp{7} for inverse,
3819 @samp{39} for default foreground color,
3820 @samp{30} to @samp{37} for foreground colors,
3821 @samp{90} to @samp{97} for 16-color mode foreground colors,
3822 @samp{38;5;0} to @samp{38;5;255}
3823 for 88-color and 256-color modes foreground colors,
3824 @samp{49} for default background color,
3825 @samp{40} to @samp{47} for background colors,
3826 @samp{100} to @samp{107} for 16-color mode background colors,
3827 and @samp{48;5;0} to @samp{48;5;255}
3828 for 88-color and 256-color modes background colors.
3830 The default @env{GCC_COLORS} is
3832 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
3833 quote=01:fixit-insert=32:fixit-delete=31:\
3834 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
3838 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3839 @samp{01;36} is bold cyan, @samp{32} is green, @samp{34} is blue,
3840 @samp{01} is bold, and @samp{31} is red.
3841 Setting @env{GCC_COLORS} to the empty string disables colors.
3842 Supported capabilities are as follows.
3846 @vindex error GCC_COLORS @r{capability}
3847 SGR substring for error: markers.
3850 @vindex warning GCC_COLORS @r{capability}
3851 SGR substring for warning: markers.
3854 @vindex note GCC_COLORS @r{capability}
3855 SGR substring for note: markers.
3858 @vindex range1 GCC_COLORS @r{capability}
3859 SGR substring for first additional range.
3862 @vindex range2 GCC_COLORS @r{capability}
3863 SGR substring for second additional range.
3866 @vindex locus GCC_COLORS @r{capability}
3867 SGR substring for location information, @samp{file:line} or
3868 @samp{file:line:column} etc.
3871 @vindex quote GCC_COLORS @r{capability}
3872 SGR substring for information printed within quotes.
3875 @vindex fixit-insert GCC_COLORS @r{capability}
3876 SGR substring for fix-it hints suggesting text to
3877 be inserted or replaced.
3880 @vindex fixit-delete GCC_COLORS @r{capability}
3881 SGR substring for fix-it hints suggesting text to
3884 @item diff-filename=
3885 @vindex diff-filename GCC_COLORS @r{capability}
3886 SGR substring for filename headers within generated patches.
3889 @vindex diff-hunk GCC_COLORS @r{capability}
3890 SGR substring for the starts of hunks within generated patches.
3893 @vindex diff-delete GCC_COLORS @r{capability}
3894 SGR substring for deleted lines within generated patches.
3897 @vindex diff-insert GCC_COLORS @r{capability}
3898 SGR substring for inserted lines within generated patches.
3901 @vindex type-diff GCC_COLORS @r{capability}
3902 SGR substring for highlighting mismatching types within template
3903 arguments in the C++ frontend.
3906 @item -fno-diagnostics-show-option
3907 @opindex fno-diagnostics-show-option
3908 @opindex fdiagnostics-show-option
3909 By default, each diagnostic emitted includes text indicating the
3910 command-line option that directly controls the diagnostic (if such an
3911 option is known to the diagnostic machinery). Specifying the
3912 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3914 @item -fno-diagnostics-show-caret
3915 @opindex fno-diagnostics-show-caret
3916 @opindex fdiagnostics-show-caret
3917 By default, each diagnostic emitted includes the original source line
3918 and a caret @samp{^} indicating the column. This option suppresses this
3919 information. The source line is truncated to @var{n} characters, if
3920 the @option{-fmessage-length=n} option is given. When the output is done
3921 to the terminal, the width is limited to the width given by the
3922 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3924 @item -fno-diagnostics-show-labels
3925 @opindex fno-diagnostics-show-labels
3926 @opindex fdiagnostics-show-labels
3927 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3928 diagnostics can label ranges of source code with pertinent information, such
3929 as the types of expressions:
3932 printf ("foo %s bar", long_i + long_j);
3938 This option suppresses the printing of these labels (in the example above,
3939 the vertical bars and the ``char *'' and ``long int'' text).
3941 @item -fno-diagnostics-show-line-numbers
3942 @opindex fno-diagnostics-show-line-numbers
3943 @opindex fdiagnostics-show-line-numbers
3944 By default, when printing source code (via @option{-fdiagnostics-show-caret}),
3945 a left margin is printed, showing line numbers. This option suppresses this
3948 @item -fdiagnostics-minimum-margin-width=@var{width}
3949 @opindex fdiagnostics-minimum-margin-width
3950 This option controls the minimum width of the left margin printed by
3951 @option{-fdiagnostics-show-line-numbers}. It defaults to 6.
3953 @item -fdiagnostics-parseable-fixits
3954 @opindex fdiagnostics-parseable-fixits
3955 Emit fix-it hints in a machine-parseable format, suitable for consumption
3956 by IDEs. For each fix-it, a line will be printed after the relevant
3957 diagnostic, starting with the string ``fix-it:''. For example:
3960 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3963 The location is expressed as a half-open range, expressed as a count of
3964 bytes, starting at byte 1 for the initial column. In the above example,
3965 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3969 00000000011111111112222222222
3970 12345678901234567890123456789
3971 gtk_widget_showall (dlg);
3976 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3977 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3978 (e.g. vertical tab as ``\013'').
3980 An empty replacement string indicates that the given range is to be removed.
3981 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3982 be inserted at the given position.
3984 @item -fdiagnostics-generate-patch
3985 @opindex fdiagnostics-generate-patch
3986 Print fix-it hints to stderr in unified diff format, after any diagnostics
3987 are printed. For example:
3994 void show_cb(GtkDialog *dlg)
3996 - gtk_widget_showall(dlg);
3997 + gtk_widget_show_all(dlg);
4002 The diff may or may not be colorized, following the same rules
4003 as for diagnostics (see @option{-fdiagnostics-color}).
4005 @item -fdiagnostics-show-template-tree
4006 @opindex fdiagnostics-show-template-tree
4008 In the C++ frontend, when printing diagnostics showing mismatching
4009 template types, such as:
4012 could not convert 'std::map<int, std::vector<double> >()'
4013 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4016 the @option{-fdiagnostics-show-template-tree} flag enables printing a
4017 tree-like structure showing the common and differing parts of the types,
4027 The parts that differ are highlighted with color (``double'' and
4028 ``float'' in this case).
4030 @item -fno-elide-type
4031 @opindex fno-elide-type
4032 @opindex felide-type
4033 By default when the C++ frontend prints diagnostics showing mismatching
4034 template types, common parts of the types are printed as ``[...]'' to
4035 simplify the error message. For example:
4038 could not convert 'std::map<int, std::vector<double> >()'
4039 from 'map<[...],vector<double>>' to 'map<[...],vector<float>>
4042 Specifying the @option{-fno-elide-type} flag suppresses that behavior.
4043 This flag also affects the output of the
4044 @option{-fdiagnostics-show-template-tree} flag.
4046 @item -fno-show-column
4047 @opindex fno-show-column
4048 @opindex fshow-column
4049 Do not print column numbers in diagnostics. This may be necessary if
4050 diagnostics are being scanned by a program that does not understand the
4051 column numbers, such as @command{dejagnu}.
4053 @item -fdiagnostics-format=@var{FORMAT}
4054 @opindex fdiagnostics-format
4055 Select a different format for printing diagnostics.
4056 @var{FORMAT} is @samp{text} or @samp{json}.
4057 The default is @samp{text}.
4059 The @samp{json} format consists of a top-level JSON array containing JSON
4060 objects representing the diagnostics.
4062 The JSON is emitted as one line, without formatting; the examples below
4063 have been formatted for clarity.
4065 Diagnostics can have child diagnostics. For example, this error and note:
4068 misleading-indentation.c:15:3: warning: this 'if' clause does not
4069 guard... [-Wmisleading-indentation]
4072 misleading-indentation.c:17:5: note: ...this statement, but the latter
4073 is misleadingly indented as if it were guarded by the 'if'
4079 might be printed in JSON form (after formatting) like this:
4089 "file": "misleading-indentation.c",
4094 "file": "misleading-indentation.c",
4099 "message": "this \u2018if\u2019 clause does not guard...",
4100 "option": "-Wmisleading-indentation",
4108 "file": "misleading-indentation.c",
4113 "message": "...this statement, but the latter is @dots{}"
4122 where the @code{note} is a child of the @code{warning}.
4124 A diagnostic has a @code{kind}. If this is @code{warning}, then there is
4125 an @code{option} key describing the command-line option controlling the
4128 A diagnostic can contain zero or more locations. Each location has up
4129 to three positions within it: a @code{caret} position and optional
4130 @code{start} and @code{finish} positions. A location can also have
4131 an optional @code{label} string. For example, this error:
4134 bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' @{aka
4135 'struct s'@} and 'T' @{aka 'struct t'@})
4136 64 | return callee_4a () + callee_4b ();
4137 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~
4139 | | T @{aka struct t@}
4140 | S @{aka struct s@}
4144 has three locations. Its primary location is at the ``+'' token at column
4145 23. It has two secondary locations, describing the left and right-hand sides
4146 of the expression, which have labels. It might be printed in JSON form as:
4155 "column": 23, "file": "bad-binary-ops.c", "line": 64
4160 "column": 10, "file": "bad-binary-ops.c", "line": 64
4163 "column": 21, "file": "bad-binary-ops.c", "line": 64
4165 "label": "S @{aka struct s@}"
4169 "column": 25, "file": "bad-binary-ops.c", "line": 64
4172 "column": 36, "file": "bad-binary-ops.c", "line": 64
4174 "label": "T @{aka struct t@}"
4177 "message": "invalid operands to binary + @dots{}"
4181 If a diagnostic contains fix-it hints, it has a @code{fixits} array,
4182 consisting of half-open intervals, similar to the output of
4183 @option{-fdiagnostics-parseable-fixits}. For example, this diagnostic
4184 with a replacement fix-it hint:
4187 demo.c:8:15: error: 'struct s' has no member named 'colour'; did you
4189 8 | return ptr->colour;
4195 might be printed in JSON form as:
4230 "message": "\u2018struct s\u2019 has no member named @dots{}"
4235 where the fix-it hint suggests replacing the text from @code{start} up
4236 to but not including @code{next} with @code{string}'s value. Deletions
4237 are expressed via an empty value for @code{string}, insertions by
4238 having @code{start} equal @code{next}.
4242 @node Warning Options
4243 @section Options to Request or Suppress Warnings
4244 @cindex options to control warnings
4245 @cindex warning messages
4246 @cindex messages, warning
4247 @cindex suppressing warnings
4249 Warnings are diagnostic messages that report constructions that
4250 are not inherently erroneous but that are risky or suggest there
4251 may have been an error.
4253 The following language-independent options do not enable specific
4254 warnings but control the kinds of diagnostics produced by GCC@.
4257 @cindex syntax checking
4259 @opindex fsyntax-only
4260 Check the code for syntax errors, but don't do anything beyond that.
4262 @item -fmax-errors=@var{n}
4263 @opindex fmax-errors
4264 Limits the maximum number of error messages to @var{n}, at which point
4265 GCC bails out rather than attempting to continue processing the source
4266 code. If @var{n} is 0 (the default), there is no limit on the number
4267 of error messages produced. If @option{-Wfatal-errors} is also
4268 specified, then @option{-Wfatal-errors} takes precedence over this
4273 Inhibit all warning messages.
4278 Make all warnings into errors.
4283 Make the specified warning into an error. The specifier for a warning
4284 is appended; for example @option{-Werror=switch} turns the warnings
4285 controlled by @option{-Wswitch} into errors. This switch takes a
4286 negative form, to be used to negate @option{-Werror} for specific
4287 warnings; for example @option{-Wno-error=switch} makes
4288 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
4291 The warning message for each controllable warning includes the
4292 option that controls the warning. That option can then be used with
4293 @option{-Werror=} and @option{-Wno-error=} as described above.
4294 (Printing of the option in the warning message can be disabled using the
4295 @option{-fno-diagnostics-show-option} flag.)
4297 Note that specifying @option{-Werror=}@var{foo} automatically implies
4298 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
4301 @item -Wfatal-errors
4302 @opindex Wfatal-errors
4303 @opindex Wno-fatal-errors
4304 This option causes the compiler to abort compilation on the first error
4305 occurred rather than trying to keep going and printing further error
4310 You can request many specific warnings with options beginning with
4311 @samp{-W}, for example @option{-Wimplicit} to request warnings on
4312 implicit declarations. Each of these specific warning options also
4313 has a negative form beginning @samp{-Wno-} to turn off warnings; for
4314 example, @option{-Wno-implicit}. This manual lists only one of the
4315 two forms, whichever is not the default. For further
4316 language-specific options also refer to @ref{C++ Dialect Options} and
4317 @ref{Objective-C and Objective-C++ Dialect Options}.
4319 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
4320 options, such as @option{-Wunused}, which may turn on further options,
4321 such as @option{-Wunused-value}. The combined effect of positive and
4322 negative forms is that more specific options have priority over less
4323 specific ones, independently of their position in the command-line. For
4324 options of the same specificity, the last one takes effect. Options
4325 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
4326 as if they appeared at the end of the command-line.
4328 When an unrecognized warning option is requested (e.g.,
4329 @option{-Wunknown-warning}), GCC emits a diagnostic stating
4330 that the option is not recognized. However, if the @option{-Wno-} form
4331 is used, the behavior is slightly different: no diagnostic is
4332 produced for @option{-Wno-unknown-warning} unless other diagnostics
4333 are being produced. This allows the use of new @option{-Wno-} options
4334 with old compilers, but if something goes wrong, the compiler
4335 warns that an unrecognized option is present.
4342 @opindex Wno-pedantic
4343 Issue all the warnings demanded by strict ISO C and ISO C++;
4344 reject all programs that use forbidden extensions, and some other
4345 programs that do not follow ISO C and ISO C++. For ISO C, follows the
4346 version of the ISO C standard specified by any @option{-std} option used.
4348 Valid ISO C and ISO C++ programs should compile properly with or without
4349 this option (though a rare few require @option{-ansi} or a
4350 @option{-std} option specifying the required version of ISO C)@. However,
4351 without this option, certain GNU extensions and traditional C and C++
4352 features are supported as well. With this option, they are rejected.
4354 @option{-Wpedantic} does not cause warning messages for use of the
4355 alternate keywords whose names begin and end with @samp{__}. This alternate
4356 format can also be used to disable warnings for non-ISO @samp{__intN} types,
4357 i.e. @samp{__intN__}.
4358 Pedantic warnings are also disabled in the expression that follows
4359 @code{__extension__}. However, only system header files should use
4360 these escape routes; application programs should avoid them.
4361 @xref{Alternate Keywords}.
4363 Some users try to use @option{-Wpedantic} to check programs for strict ISO
4364 C conformance. They soon find that it does not do quite what they want:
4365 it finds some non-ISO practices, but not all---only those for which
4366 ISO C @emph{requires} a diagnostic, and some others for which
4367 diagnostics have been added.
4369 A feature to report any failure to conform to ISO C might be useful in
4370 some instances, but would require considerable additional work and would
4371 be quite different from @option{-Wpedantic}. We don't have plans to
4372 support such a feature in the near future.
4374 Where the standard specified with @option{-std} represents a GNU
4375 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
4376 corresponding @dfn{base standard}, the version of ISO C on which the GNU
4377 extended dialect is based. Warnings from @option{-Wpedantic} are given
4378 where they are required by the base standard. (It does not make sense
4379 for such warnings to be given only for features not in the specified GNU
4380 C dialect, since by definition the GNU dialects of C include all
4381 features the compiler supports with the given option, and there would be
4382 nothing to warn about.)
4384 @item -pedantic-errors
4385 @opindex pedantic-errors
4386 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
4387 requires a diagnostic, in some cases where there is undefined behavior
4388 at compile-time and in some other cases that do not prevent compilation
4389 of programs that are valid according to the standard. This is not
4390 equivalent to @option{-Werror=pedantic}, since there are errors enabled
4391 by this option and not enabled by the latter and vice versa.
4396 This enables all the warnings about constructions that some users
4397 consider questionable, and that are easy to avoid (or modify to
4398 prevent the warning), even in conjunction with macros. This also
4399 enables some language-specific warnings described in @ref{C++ Dialect
4400 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
4402 @option{-Wall} turns on the following warning flags:
4404 @gccoptlist{-Waddress @gol
4405 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
4407 -Wbool-operation @gol
4408 -Wc++11-compat -Wc++14-compat @gol
4409 -Wcatch-value @r{(C++ and Objective-C++ only)} @gol
4410 -Wchar-subscripts @gol
4412 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
4413 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
4414 -Wenum-conversion @r{in C/ObjC;} @gol
4416 -Wint-in-bool-context @gol
4417 -Wimplicit @r{(C and Objective-C only)} @gol
4418 -Wimplicit-int @r{(C and Objective-C only)} @gol
4419 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
4420 -Winit-self @r{(only for C++)} @gol
4421 -Wlogical-not-parentheses @gol
4422 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
4423 -Wmaybe-uninitialized @gol
4424 -Wmemset-elt-size @gol
4425 -Wmemset-transposed-args @gol
4426 -Wmisleading-indentation @r{(only for C/C++)} @gol
4427 -Wmissing-attributes @gol
4428 -Wmissing-braces @r{(only for C/ObjC)} @gol
4429 -Wmultistatement-macros @gol
4430 -Wnarrowing @r{(only for C++)} @gol
4432 -Wnonnull-compare @gol
4435 -Wpessimizing-move @r{(only for C++)} @gol
4440 -Wsequence-point @gol
4441 -Wsign-compare @r{(only in C++)} @gol
4442 -Wsizeof-pointer-div @gol
4443 -Wsizeof-pointer-memaccess @gol
4444 -Wstrict-aliasing @gol
4445 -Wstrict-overflow=1 @gol
4447 -Wtautological-compare @gol
4449 -Wuninitialized @gol
4450 -Wunknown-pragmas @gol
4451 -Wunused-function @gol
4454 -Wunused-variable @gol
4455 -Wvolatile-register-var}
4457 Note that some warning flags are not implied by @option{-Wall}. Some of
4458 them warn about constructions that users generally do not consider
4459 questionable, but which occasionally you might wish to check for;
4460 others warn about constructions that are necessary or hard to avoid in
4461 some cases, and there is no simple way to modify the code to suppress
4462 the warning. Some of them are enabled by @option{-Wextra} but many of
4463 them must be enabled individually.
4469 This enables some extra warning flags that are not enabled by
4470 @option{-Wall}. (This option used to be called @option{-W}. The older
4471 name is still supported, but the newer name is more descriptive.)
4473 @gccoptlist{-Wclobbered @gol
4474 -Wcast-function-type @gol
4475 -Wdeprecated-copy @r{(C++ only)} @gol
4477 -Wignored-qualifiers @gol
4478 -Wimplicit-fallthrough=3 @gol
4479 -Wmissing-field-initializers @gol
4480 -Wmissing-parameter-type @r{(C only)} @gol
4481 -Wold-style-declaration @r{(C only)} @gol
4482 -Woverride-init @gol
4483 -Wsign-compare @r{(C only)} @gol
4484 -Wredundant-move @r{(only for C++)} @gol
4486 -Wuninitialized @gol
4487 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
4488 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
4489 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)}}
4492 The option @option{-Wextra} also prints warning messages for the
4498 A pointer is compared against integer zero with @code{<}, @code{<=},
4499 @code{>}, or @code{>=}.
4502 (C++ only) An enumerator and a non-enumerator both appear in a
4503 conditional expression.
4506 (C++ only) Ambiguous virtual bases.
4509 (C++ only) Subscripting an array that has been declared @code{register}.
4512 (C++ only) Taking the address of a variable that has been declared
4516 (C++ only) A base class is not initialized in the copy constructor
4521 @item -Wchar-subscripts
4522 @opindex Wchar-subscripts
4523 @opindex Wno-char-subscripts
4524 Warn if an array subscript has type @code{char}. This is a common cause
4525 of error, as programmers often forget that this type is signed on some
4527 This warning is enabled by @option{-Wall}.
4529 @item -Wno-coverage-mismatch
4530 @opindex Wno-coverage-mismatch
4531 @opindex Wcoverage-mismatch
4532 Warn if feedback profiles do not match when using the
4533 @option{-fprofile-use} option.
4534 If a source file is changed between compiling with @option{-fprofile-generate}
4535 and with @option{-fprofile-use}, the files with the profile feedback can fail
4536 to match the source file and GCC cannot use the profile feedback
4537 information. By default, this warning is enabled and is treated as an
4538 error. @option{-Wno-coverage-mismatch} can be used to disable the
4539 warning or @option{-Wno-error=coverage-mismatch} can be used to
4540 disable the error. Disabling the error for this warning can result in
4541 poorly optimized code and is useful only in the
4542 case of very minor changes such as bug fixes to an existing code-base.
4543 Completely disabling the warning is not recommended.
4546 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
4548 Suppress warning messages emitted by @code{#warning} directives.
4550 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
4551 @opindex Wdouble-promotion
4552 @opindex Wno-double-promotion
4553 Give a warning when a value of type @code{float} is implicitly
4554 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
4555 floating-point unit implement @code{float} in hardware, but emulate
4556 @code{double} in software. On such a machine, doing computations
4557 using @code{double} values is much more expensive because of the
4558 overhead required for software emulation.
4560 It is easy to accidentally do computations with @code{double} because
4561 floating-point literals are implicitly of type @code{double}. For
4565 float area(float radius)
4567 return 3.14159 * radius * radius;
4571 the compiler performs the entire computation with @code{double}
4572 because the floating-point literal is a @code{double}.
4574 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
4575 @opindex Wduplicate-decl-specifier
4576 @opindex Wno-duplicate-decl-specifier
4577 Warn if a declaration has duplicate @code{const}, @code{volatile},
4578 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
4582 @itemx -Wformat=@var{n}
4585 @opindex ffreestanding
4586 @opindex fno-builtin
4588 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
4589 the arguments supplied have types appropriate to the format string
4590 specified, and that the conversions specified in the format string make
4591 sense. This includes standard functions, and others specified by format
4592 attributes (@pxref{Function Attributes}), in the @code{printf},
4593 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
4594 not in the C standard) families (or other target-specific families).
4595 Which functions are checked without format attributes having been
4596 specified depends on the standard version selected, and such checks of
4597 functions without the attribute specified are disabled by
4598 @option{-ffreestanding} or @option{-fno-builtin}.
4600 The formats are checked against the format features supported by GNU
4601 libc version 2.2. These include all ISO C90 and C99 features, as well
4602 as features from the Single Unix Specification and some BSD and GNU
4603 extensions. Other library implementations may not support all these
4604 features; GCC does not support warning about features that go beyond a
4605 particular library's limitations. However, if @option{-Wpedantic} is used
4606 with @option{-Wformat}, warnings are given about format features not
4607 in the selected standard version (but not for @code{strfmon} formats,
4608 since those are not in any version of the C standard). @xref{C Dialect
4609 Options,,Options Controlling C Dialect}.
4616 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
4617 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
4618 @option{-Wformat} also checks for null format arguments for several
4619 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
4620 aspects of this level of format checking can be disabled by the
4621 options: @option{-Wno-format-contains-nul},
4622 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
4623 @option{-Wformat} is enabled by @option{-Wall}.
4625 @item -Wno-format-contains-nul
4626 @opindex Wno-format-contains-nul
4627 @opindex Wformat-contains-nul
4628 If @option{-Wformat} is specified, do not warn about format strings that
4631 @item -Wno-format-extra-args
4632 @opindex Wno-format-extra-args
4633 @opindex Wformat-extra-args
4634 If @option{-Wformat} is specified, do not warn about excess arguments to a
4635 @code{printf} or @code{scanf} format function. The C standard specifies
4636 that such arguments are ignored.
4638 Where the unused arguments lie between used arguments that are
4639 specified with @samp{$} operand number specifications, normally
4640 warnings are still given, since the implementation could not know what
4641 type to pass to @code{va_arg} to skip the unused arguments. However,
4642 in the case of @code{scanf} formats, this option suppresses the
4643 warning if the unused arguments are all pointers, since the Single
4644 Unix Specification says that such unused arguments are allowed.
4646 @item -Wformat-overflow
4647 @itemx -Wformat-overflow=@var{level}
4648 @opindex Wformat-overflow
4649 @opindex Wno-format-overflow
4650 Warn about calls to formatted input/output functions such as @code{sprintf}
4651 and @code{vsprintf} that might overflow the destination buffer. When the
4652 exact number of bytes written by a format directive cannot be determined
4653 at compile-time it is estimated based on heuristics that depend on the
4654 @var{level} argument and on optimization. While enabling optimization
4655 will in most cases improve the accuracy of the warning, it may also
4656 result in false positives.
4659 @item -Wformat-overflow
4660 @itemx -Wformat-overflow=1
4661 @opindex Wformat-overflow
4662 @opindex Wno-format-overflow
4663 Level @var{1} of @option{-Wformat-overflow} enabled by @option{-Wformat}
4664 employs a conservative approach that warns only about calls that most
4665 likely overflow the buffer. At this level, numeric arguments to format
4666 directives with unknown values are assumed to have the value of one, and
4667 strings of unknown length to be empty. Numeric arguments that are known
4668 to be bounded to a subrange of their type, or string arguments whose output
4669 is bounded either by their directive's precision or by a finite set of
4670 string literals, are assumed to take on the value within the range that
4671 results in the most bytes on output. For example, the call to @code{sprintf}
4672 below is diagnosed because even with both @var{a} and @var{b} equal to zero,
4673 the terminating NUL character (@code{'\0'}) appended by the function
4674 to the destination buffer will be written past its end. Increasing
4675 the size of the buffer by a single byte is sufficient to avoid the
4676 warning, though it may not be sufficient to avoid the overflow.
4679 void f (int a, int b)
4682 sprintf (buf, "a = %i, b = %i\n", a, b);
4686 @item -Wformat-overflow=2
4687 Level @var{2} warns also about calls that might overflow the destination
4688 buffer given an argument of sufficient length or magnitude. At level
4689 @var{2}, unknown numeric arguments are assumed to have the minimum
4690 representable value for signed types with a precision greater than 1, and
4691 the maximum representable value otherwise. Unknown string arguments whose
4692 length cannot be assumed to be bounded either by the directive's precision,
4693 or by a finite set of string literals they may evaluate to, or the character
4694 array they may point to, are assumed to be 1 character long.
4696 At level @var{2}, the call in the example above is again diagnosed, but
4697 this time because with @var{a} equal to a 32-bit @code{INT_MIN} the first
4698 @code{%i} directive will write some of its digits beyond the end of
4699 the destination buffer. To make the call safe regardless of the values
4700 of the two variables, the size of the destination buffer must be increased
4701 to at least 34 bytes. GCC includes the minimum size of the buffer in
4702 an informational note following the warning.
4704 An alternative to increasing the size of the destination buffer is to
4705 constrain the range of formatted values. The maximum length of string
4706 arguments can be bounded by specifying the precision in the format
4707 directive. When numeric arguments of format directives can be assumed
4708 to be bounded by less than the precision of their type, choosing
4709 an appropriate length modifier to the format specifier will reduce
4710 the required buffer size. For example, if @var{a} and @var{b} in the
4711 example above can be assumed to be within the precision of
4712 the @code{short int} type then using either the @code{%hi} format
4713 directive or casting the argument to @code{short} reduces the maximum
4714 required size of the buffer to 24 bytes.
4717 void f (int a, int b)
4720 sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
4725 @item -Wno-format-zero-length
4726 @opindex Wno-format-zero-length
4727 @opindex Wformat-zero-length
4728 If @option{-Wformat} is specified, do not warn about zero-length formats.
4729 The C standard specifies that zero-length formats are allowed.
4734 Enable @option{-Wformat} plus additional format checks. Currently
4735 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
4738 @item -Wformat-nonliteral
4739 @opindex Wformat-nonliteral
4740 @opindex Wno-format-nonliteral
4741 If @option{-Wformat} is specified, also warn if the format string is not a
4742 string literal and so cannot be checked, unless the format function
4743 takes its format arguments as a @code{va_list}.
4745 @item -Wformat-security
4746 @opindex Wformat-security
4747 @opindex Wno-format-security
4748 If @option{-Wformat} is specified, also warn about uses of format
4749 functions that represent possible security problems. At present, this
4750 warns about calls to @code{printf} and @code{scanf} functions where the
4751 format string is not a string literal and there are no format arguments,
4752 as in @code{printf (foo);}. This may be a security hole if the format
4753 string came from untrusted input and contains @samp{%n}. (This is
4754 currently a subset of what @option{-Wformat-nonliteral} warns about, but
4755 in future warnings may be added to @option{-Wformat-security} that are not
4756 included in @option{-Wformat-nonliteral}.)
4758 @item -Wformat-signedness
4759 @opindex Wformat-signedness
4760 @opindex Wno-format-signedness
4761 If @option{-Wformat} is specified, also warn if the format string
4762 requires an unsigned argument and the argument is signed and vice versa.
4764 @item -Wformat-truncation
4765 @itemx -Wformat-truncation=@var{level}
4766 @opindex Wformat-truncation
4767 @opindex Wno-format-truncation
4768 Warn about calls to formatted input/output functions such as @code{snprintf}
4769 and @code{vsnprintf} that might result in output truncation. When the exact
4770 number of bytes written by a format directive cannot be determined at
4771 compile-time it is estimated based on heuristics that depend on
4772 the @var{level} argument and on optimization. While enabling optimization
4773 will in most cases improve the accuracy of the warning, it may also result
4774 in false positives. Except as noted otherwise, the option uses the same
4775 logic @option{-Wformat-overflow}.
4778 @item -Wformat-truncation
4779 @itemx -Wformat-truncation=1
4780 @opindex Wformat-truncation
4781 @opindex Wno-format-truncation
4782 Level @var{1} of @option{-Wformat-truncation} enabled by @option{-Wformat}
4783 employs a conservative approach that warns only about calls to bounded
4784 functions whose return value is unused and that will most likely result
4785 in output truncation.
4787 @item -Wformat-truncation=2
4788 Level @var{2} warns also about calls to bounded functions whose return
4789 value is used and that might result in truncation given an argument of
4790 sufficient length or magnitude.
4794 @opindex Wformat-y2k
4795 @opindex Wno-format-y2k
4796 If @option{-Wformat} is specified, also warn about @code{strftime}
4797 formats that may yield only a two-digit year.
4802 @opindex Wno-nonnull
4803 Warn about passing a null pointer for arguments marked as
4804 requiring a non-null value by the @code{nonnull} function attribute.
4806 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
4807 can be disabled with the @option{-Wno-nonnull} option.
4809 @item -Wnonnull-compare
4810 @opindex Wnonnull-compare
4811 @opindex Wno-nonnull-compare
4812 Warn when comparing an argument marked with the @code{nonnull}
4813 function attribute against null inside the function.
4815 @option{-Wnonnull-compare} is included in @option{-Wall}. It
4816 can be disabled with the @option{-Wno-nonnull-compare} option.
4818 @item -Wnull-dereference
4819 @opindex Wnull-dereference
4820 @opindex Wno-null-dereference
4821 Warn if the compiler detects paths that trigger erroneous or
4822 undefined behavior due to dereferencing a null pointer. This option
4823 is only active when @option{-fdelete-null-pointer-checks} is active,
4824 which is enabled by optimizations in most targets. The precision of
4825 the warnings depends on the optimization options used.
4827 @item -Winaccessible-base @r{(C++, Objective-C++ only)}
4828 @opindex Winaccessible-base
4829 @opindex Wno-inaccessible-base
4830 Warn when a base class is inaccessible in a class derived from it due to
4831 ambiguity. The warning is enabled by default. Note the warning for virtual
4832 bases is enabled by the @option{-Wextra} option.
4835 struct A @{ int a; @};
4839 struct C : B, A @{ @};
4843 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
4845 @opindex Wno-init-self
4846 Warn about uninitialized variables that are initialized with themselves.
4847 Note this option can only be used with the @option{-Wuninitialized} option.
4849 For example, GCC warns about @code{i} being uninitialized in the
4850 following snippet only when @option{-Winit-self} has been specified:
4861 This warning is enabled by @option{-Wall} in C++.
4863 @item -Wimplicit-int @r{(C and Objective-C only)}
4864 @opindex Wimplicit-int
4865 @opindex Wno-implicit-int
4866 Warn when a declaration does not specify a type.
4867 This warning is enabled by @option{-Wall}.
4869 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
4870 @opindex Wimplicit-function-declaration
4871 @opindex Wno-implicit-function-declaration
4872 Give a warning whenever a function is used before being declared. In
4873 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
4874 enabled by default and it is made into an error by
4875 @option{-pedantic-errors}. This warning is also enabled by
4878 @item -Wimplicit @r{(C and Objective-C only)}
4880 @opindex Wno-implicit
4881 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
4882 This warning is enabled by @option{-Wall}.
4884 @item -Wimplicit-fallthrough
4885 @opindex Wimplicit-fallthrough
4886 @opindex Wno-implicit-fallthrough
4887 @option{-Wimplicit-fallthrough} is the same as @option{-Wimplicit-fallthrough=3}
4888 and @option{-Wno-implicit-fallthrough} is the same as
4889 @option{-Wimplicit-fallthrough=0}.
4891 @item -Wimplicit-fallthrough=@var{n}
4892 @opindex Wimplicit-fallthrough=
4893 Warn when a switch case falls through. For example:
4911 This warning does not warn when the last statement of a case cannot
4912 fall through, e.g. when there is a return statement or a call to function
4913 declared with the noreturn attribute. @option{-Wimplicit-fallthrough=}
4914 also takes into account control flow statements, such as ifs, and only
4915 warns when appropriate. E.g.@:
4925 @} else if (i < 1) @{
4935 Since there are occasions where a switch case fall through is desirable,
4936 GCC provides an attribute, @code{__attribute__ ((fallthrough))}, that is
4937 to be used along with a null statement to suppress this warning that
4938 would normally occur:
4946 __attribute__ ((fallthrough));
4953 C++17 provides a standard way to suppress the @option{-Wimplicit-fallthrough}
4954 warning using @code{[[fallthrough]];} instead of the GNU attribute. In C++11
4955 or C++14 users can use @code{[[gnu::fallthrough]];}, which is a GNU extension.
4956 Instead of these attributes, it is also possible to add a fallthrough comment
4957 to silence the warning. The whole body of the C or C++ style comment should
4958 match the given regular expressions listed below. The option argument @var{n}
4959 specifies what kind of comments are accepted:
4963 @item @option{-Wimplicit-fallthrough=0} disables the warning altogether.
4965 @item @option{-Wimplicit-fallthrough=1} matches @code{.*} regular
4966 expression, any comment is used as fallthrough comment.
4968 @item @option{-Wimplicit-fallthrough=2} case insensitively matches
4969 @code{.*falls?[ \t-]*thr(ough|u).*} regular expression.
4971 @item @option{-Wimplicit-fallthrough=3} case sensitively matches one of the
4972 following regular expressions:
4976 @item @code{-fallthrough}
4978 @item @code{@@fallthrough@@}
4980 @item @code{lint -fallthrough[ \t]*}
4982 @item @code{[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?@*FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?}
4984 @item @code{[ \t.!]*(Else,? |Intentional(ly)? )?@*Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4986 @item @code{[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?@*fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?}
4990 @item @option{-Wimplicit-fallthrough=4} case sensitively matches one of the
4991 following regular expressions:
4995 @item @code{-fallthrough}
4997 @item @code{@@fallthrough@@}
4999 @item @code{lint -fallthrough[ \t]*}
5001 @item @code{[ \t]*FALLTHR(OUGH|U)[ \t]*}
5005 @item @option{-Wimplicit-fallthrough=5} doesn't recognize any comments as
5006 fallthrough comments, only attributes disable the warning.
5010 The comment needs to be followed after optional whitespace and other comments
5011 by @code{case} or @code{default} keywords or by a user label that precedes some
5012 @code{case} or @code{default} label.
5027 The @option{-Wimplicit-fallthrough=3} warning is enabled by @option{-Wextra}.
5029 @item -Wif-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
5030 @opindex Wif-not-aligned
5031 @opindex Wno-if-not-aligned
5032 Control if warning triggered by the @code{warn_if_not_aligned} attribute
5033 should be issued. This is enabled by default.
5034 Use @option{-Wno-if-not-aligned} to disable it.
5036 @item -Wignored-qualifiers @r{(C and C++ only)}
5037 @opindex Wignored-qualifiers
5038 @opindex Wno-ignored-qualifiers
5039 Warn if the return type of a function has a type qualifier
5040 such as @code{const}. For ISO C such a type qualifier has no effect,
5041 since the value returned by a function is not an lvalue.
5042 For C++, the warning is only emitted for scalar types or @code{void}.
5043 ISO C prohibits qualified @code{void} return types on function
5044 definitions, so such return types always receive a warning
5045 even without this option.
5047 This warning is also enabled by @option{-Wextra}.
5049 @item -Wignored-attributes @r{(C and C++ only)}
5050 @opindex Wignored-attributes
5051 @opindex Wno-ignored-attributes
5052 Warn when an attribute is ignored. This is different from the
5053 @option{-Wattributes} option in that it warns whenever the compiler decides
5054 to drop an attribute, not that the attribute is either unknown, used in a
5055 wrong place, etc. This warning is enabled by default.
5060 Warn if the type of @code{main} is suspicious. @code{main} should be
5061 a function with external linkage, returning int, taking either zero
5062 arguments, two, or three arguments of appropriate types. This warning
5063 is enabled by default in C++ and is enabled by either @option{-Wall}
5064 or @option{-Wpedantic}.
5066 @item -Wmisleading-indentation @r{(C and C++ only)}
5067 @opindex Wmisleading-indentation
5068 @opindex Wno-misleading-indentation
5069 Warn when the indentation of the code does not reflect the block structure.
5070 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
5071 @code{for} clauses with a guarded statement that does not use braces,
5072 followed by an unguarded statement with the same indentation.
5074 In the following example, the call to ``bar'' is misleadingly indented as
5075 if it were guarded by the ``if'' conditional.
5078 if (some_condition ())
5080 bar (); /* Gotcha: this is not guarded by the "if". */
5083 In the case of mixed tabs and spaces, the warning uses the
5084 @option{-ftabstop=} option to determine if the statements line up
5087 The warning is not issued for code involving multiline preprocessor logic
5088 such as the following example.
5093 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
5099 The warning is not issued after a @code{#line} directive, since this
5100 typically indicates autogenerated code, and no assumptions can be made
5101 about the layout of the file that the directive references.
5103 This warning is enabled by @option{-Wall} in C and C++.
5105 @item -Wmissing-attributes
5106 @opindex Wmissing-attributes
5107 @opindex Wno-missing-attributes
5108 Warn when a declaration of a function is missing one or more attributes
5109 that a related function is declared with and whose absence may adversely
5110 affect the correctness or efficiency of generated code. For example,
5111 the warning is issued for declarations of aliases that use attributes
5112 to specify less restrictive requirements than those of their targets.
5113 This typically represents a potential optimization opportunity.
5114 By contrast, the @option{-Wattribute-alias=2} option controls warnings
5115 issued when the alias is more restrictive than the target, which could
5116 lead to incorrect code generation.
5117 Attributes considered include @code{alloc_align}, @code{alloc_size},
5118 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
5119 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
5120 @code{returns_nonnull}, and @code{returns_twice}.
5122 In C++, the warning is issued when an explicit specialization of a primary
5123 template declared with attribute @code{alloc_align}, @code{alloc_size},
5124 @code{assume_aligned}, @code{format}, @code{format_arg}, @code{malloc},
5125 or @code{nonnull} is declared without it. Attributes @code{deprecated},
5126 @code{error}, and @code{warning} suppress the warning.
5127 (@pxref{Function Attributes}).
5129 You can use the @code{copy} attribute to apply the same
5130 set of attributes to a declaration as that on another declaration without
5131 explicitly enumerating the attributes. This attribute can be applied
5132 to declarations of functions (@pxref{Common Function Attributes}),
5133 variables (@pxref{Common Variable Attributes}), or types
5134 (@pxref{Common Type Attributes}).
5136 @option{-Wmissing-attributes} is enabled by @option{-Wall}.
5138 For example, since the declaration of the primary function template
5139 below makes use of both attribute @code{malloc} and @code{alloc_size}
5140 the declaration of the explicit specialization of the template is
5141 diagnosed because it is missing one of the attributes.
5145 T* __attribute__ ((malloc, alloc_size (1)))
5149 void* __attribute__ ((malloc)) // missing alloc_size
5150 allocate<void> (size_t);
5153 @item -Wmissing-braces
5154 @opindex Wmissing-braces
5155 @opindex Wno-missing-braces
5156 Warn if an aggregate or union initializer is not fully bracketed. In
5157 the following example, the initializer for @code{a} is not fully
5158 bracketed, but that for @code{b} is fully bracketed. This warning is
5159 enabled by @option{-Wall} in C.
5162 int a[2][2] = @{ 0, 1, 2, 3 @};
5163 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
5166 This warning is enabled by @option{-Wall}.
5168 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
5169 @opindex Wmissing-include-dirs
5170 @opindex Wno-missing-include-dirs
5171 Warn if a user-supplied include directory does not exist.
5173 @item -Wmissing-profile
5174 @opindex Wmissing-profile
5175 @opindex Wno-missing-profile
5176 Warn if feedback profiles are missing when using the
5177 @option{-fprofile-use} option.
5178 This option diagnoses those cases where a new function or a new file is added
5179 to the user code between compiling with @option{-fprofile-generate} and with
5180 @option{-fprofile-use}, without regenerating the profiles. In these cases, the
5181 profile feedback data files do not contain any profile feedback information for
5182 the newly added function or file respectively. Also, in the case when profile
5183 count data (.gcda) files are removed, GCC cannot use any profile feedback
5184 information. In all these cases, warnings are issued to inform the user that a
5185 profile generation step is due. @option{-Wno-missing-profile} can be used to
5186 disable the warning. Ignoring the warning can result in poorly optimized code.
5187 Completely disabling the warning is not recommended and should be done only
5188 when non-existent profile data is justified.
5190 @item -Wmultistatement-macros
5191 @opindex Wmultistatement-macros
5192 @opindex Wno-multistatement-macros
5193 Warn about unsafe multiple statement macros that appear to be guarded
5194 by a clause such as @code{if}, @code{else}, @code{for}, @code{switch}, or
5195 @code{while}, in which only the first statement is actually guarded after
5196 the macro is expanded.
5201 #define DOIT x++; y++
5206 will increment @code{y} unconditionally, not just when @code{c} holds.
5207 The can usually be fixed by wrapping the macro in a do-while loop:
5209 #define DOIT do @{ x++; y++; @} while (0)
5214 This warning is enabled by @option{-Wall} in C and C++.
5217 @opindex Wparentheses
5218 @opindex Wno-parentheses
5219 Warn if parentheses are omitted in certain contexts, such
5220 as when there is an assignment in a context where a truth value
5221 is expected, or when operators are nested whose precedence people
5222 often get confused about.
5224 Also warn if a comparison like @code{x<=y<=z} appears; this is
5225 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
5226 interpretation from that of ordinary mathematical notation.
5228 Also warn for dangerous uses of the GNU extension to
5229 @code{?:} with omitted middle operand. When the condition
5230 in the @code{?}: operator is a boolean expression, the omitted value is
5231 always 1. Often programmers expect it to be a value computed
5232 inside the conditional expression instead.
5234 For C++ this also warns for some cases of unnecessary parentheses in
5235 declarations, which can indicate an attempt at a function call instead
5239 // Declares a local variable called mymutex.
5240 std::unique_lock<std::mutex> (mymutex);
5241 // User meant std::unique_lock<std::mutex> lock (mymutex);
5245 This warning is enabled by @option{-Wall}.
5247 @item -Wsequence-point
5248 @opindex Wsequence-point
5249 @opindex Wno-sequence-point
5250 Warn about code that may have undefined semantics because of violations
5251 of sequence point rules in the C and C++ standards.
5253 The C and C++ standards define the order in which expressions in a C/C++
5254 program are evaluated in terms of @dfn{sequence points}, which represent
5255 a partial ordering between the execution of parts of the program: those
5256 executed before the sequence point, and those executed after it. These
5257 occur after the evaluation of a full expression (one which is not part
5258 of a larger expression), after the evaluation of the first operand of a
5259 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
5260 function is called (but after the evaluation of its arguments and the
5261 expression denoting the called function), and in certain other places.
5262 Other than as expressed by the sequence point rules, the order of
5263 evaluation of subexpressions of an expression is not specified. All
5264 these rules describe only a partial order rather than a total order,
5265 since, for example, if two functions are called within one expression
5266 with no sequence point between them, the order in which the functions
5267 are called is not specified. However, the standards committee have
5268 ruled that function calls do not overlap.
5270 It is not specified when between sequence points modifications to the
5271 values of objects take effect. Programs whose behavior depends on this
5272 have undefined behavior; the C and C++ standards specify that ``Between
5273 the previous and next sequence point an object shall have its stored
5274 value modified at most once by the evaluation of an expression.
5275 Furthermore, the prior value shall be read only to determine the value
5276 to be stored.''. If a program breaks these rules, the results on any
5277 particular implementation are entirely unpredictable.
5279 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
5280 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
5281 diagnosed by this option, and it may give an occasional false positive
5282 result, but in general it has been found fairly effective at detecting
5283 this sort of problem in programs.
5285 The C++17 standard will define the order of evaluation of operands in
5286 more cases: in particular it requires that the right-hand side of an
5287 assignment be evaluated before the left-hand side, so the above
5288 examples are no longer undefined. But this warning will still warn
5289 about them, to help people avoid writing code that is undefined in C
5290 and earlier revisions of C++.
5292 The standard is worded confusingly, therefore there is some debate
5293 over the precise meaning of the sequence point rules in subtle cases.
5294 Links to discussions of the problem, including proposed formal
5295 definitions, may be found on the GCC readings page, at
5296 @uref{http://gcc.gnu.org/@/readings.html}.
5298 This warning is enabled by @option{-Wall} for C and C++.
5300 @item -Wno-return-local-addr
5301 @opindex Wno-return-local-addr
5302 @opindex Wreturn-local-addr
5303 Do not warn about returning a pointer (or in C++, a reference) to a
5304 variable that goes out of scope after the function returns.
5307 @opindex Wreturn-type
5308 @opindex Wno-return-type
5309 Warn whenever a function is defined with a return type that defaults
5310 to @code{int}. Also warn about any @code{return} statement with no
5311 return value in a function whose return type is not @code{void}
5312 (falling off the end of the function body is considered returning
5315 For C only, warn about a @code{return} statement with an expression in a
5316 function whose return type is @code{void}, unless the expression type is
5317 also @code{void}. As a GNU extension, the latter case is accepted
5318 without a warning unless @option{-Wpedantic} is used. Attempting
5319 to use the return value of a non-@code{void} function other than @code{main}
5320 that flows off the end by reaching the closing curly brace that terminates
5321 the function is undefined.
5323 Unlike in C, in C++, flowing off the end of a non-@code{void} function other
5324 than @code{main} results in undefined behavior even when the value of
5325 the function is not used.
5327 This warning is enabled by default in C++ and by @option{-Wall} otherwise.
5329 @item -Wshift-count-negative
5330 @opindex Wshift-count-negative
5331 @opindex Wno-shift-count-negative
5332 Warn if shift count is negative. This warning is enabled by default.
5334 @item -Wshift-count-overflow
5335 @opindex Wshift-count-overflow
5336 @opindex Wno-shift-count-overflow
5337 Warn if shift count >= width of type. This warning is enabled by default.
5339 @item -Wshift-negative-value
5340 @opindex Wshift-negative-value
5341 @opindex Wno-shift-negative-value
5342 Warn if left shifting a negative value. This warning is enabled by
5343 @option{-Wextra} in C99 and C++11 modes (and newer).
5345 @item -Wshift-overflow
5346 @itemx -Wshift-overflow=@var{n}
5347 @opindex Wshift-overflow
5348 @opindex Wno-shift-overflow
5349 Warn about left shift overflows. This warning is enabled by
5350 default in C99 and C++11 modes (and newer).
5353 @item -Wshift-overflow=1
5354 This is the warning level of @option{-Wshift-overflow} and is enabled
5355 by default in C99 and C++11 modes (and newer). This warning level does
5356 not warn about left-shifting 1 into the sign bit. (However, in C, such
5357 an overflow is still rejected in contexts where an integer constant expression
5358 is required.) No warning is emitted in C++2A mode (and newer), as signed left
5361 @item -Wshift-overflow=2
5362 This warning level also warns about left-shifting 1 into the sign bit,
5363 unless C++14 mode (or newer) is active.
5369 Warn whenever a @code{switch} statement has an index of enumerated type
5370 and lacks a @code{case} for one or more of the named codes of that
5371 enumeration. (The presence of a @code{default} label prevents this
5372 warning.) @code{case} labels outside the enumeration range also
5373 provoke warnings when this option is used (even if there is a
5374 @code{default} label).
5375 This warning is enabled by @option{-Wall}.
5377 @item -Wswitch-default
5378 @opindex Wswitch-default
5379 @opindex Wno-switch-default
5380 Warn whenever a @code{switch} statement does not have a @code{default}
5384 @opindex Wswitch-enum
5385 @opindex Wno-switch-enum
5386 Warn whenever a @code{switch} statement has an index of enumerated type
5387 and lacks a @code{case} for one or more of the named codes of that
5388 enumeration. @code{case} labels outside the enumeration range also
5389 provoke warnings when this option is used. The only difference
5390 between @option{-Wswitch} and this option is that this option gives a
5391 warning about an omitted enumeration code even if there is a
5392 @code{default} label.
5395 @opindex Wswitch-bool
5396 @opindex Wno-switch-bool
5397 Warn whenever a @code{switch} statement has an index of boolean type
5398 and the case values are outside the range of a boolean type.
5399 It is possible to suppress this warning by casting the controlling
5400 expression to a type other than @code{bool}. For example:
5403 switch ((int) (a == 4))
5409 This warning is enabled by default for C and C++ programs.
5411 @item -Wswitch-outside-range
5412 @opindex Wswitch-outside-range
5413 @opindex Wno-switch-outside-range
5414 Warn whenever a @code{switch} case has a value that is outside of its
5415 respective type range. This warning is enabled by default for
5418 @item -Wswitch-unreachable
5419 @opindex Wswitch-unreachable
5420 @opindex Wno-switch-unreachable
5421 Warn whenever a @code{switch} statement contains statements between the
5422 controlling expression and the first case label, which will never be
5423 executed. For example:
5435 @option{-Wswitch-unreachable} does not warn if the statement between the
5436 controlling expression and the first case label is just a declaration:
5449 This warning is enabled by default for C and C++ programs.
5451 @item -Wsync-nand @r{(C and C++ only)}
5453 @opindex Wno-sync-nand
5454 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
5455 built-in functions are used. These functions changed semantics in GCC 4.4.
5457 @item -Wunused-but-set-parameter
5458 @opindex Wunused-but-set-parameter
5459 @opindex Wno-unused-but-set-parameter
5460 Warn whenever a function parameter is assigned to, but otherwise unused
5461 (aside from its declaration).
5463 To suppress this warning use the @code{unused} attribute
5464 (@pxref{Variable Attributes}).
5466 This warning is also enabled by @option{-Wunused} together with
5469 @item -Wunused-but-set-variable
5470 @opindex Wunused-but-set-variable
5471 @opindex Wno-unused-but-set-variable
5472 Warn whenever a local variable is assigned to, but otherwise unused
5473 (aside from its declaration).
5474 This warning is enabled by @option{-Wall}.
5476 To suppress this warning use the @code{unused} attribute
5477 (@pxref{Variable Attributes}).
5479 This warning is also enabled by @option{-Wunused}, which is enabled
5482 @item -Wunused-function
5483 @opindex Wunused-function
5484 @opindex Wno-unused-function
5485 Warn whenever a static function is declared but not defined or a
5486 non-inline static function is unused.
5487 This warning is enabled by @option{-Wall}.
5489 @item -Wunused-label
5490 @opindex Wunused-label
5491 @opindex Wno-unused-label
5492 Warn whenever a label is declared but not used.
5493 This warning is enabled by @option{-Wall}.
5495 To suppress this warning use the @code{unused} attribute
5496 (@pxref{Variable Attributes}).
5498 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
5499 @opindex Wunused-local-typedefs
5500 @opindex Wno-unused-local-typedefs
5501 Warn when a typedef locally defined in a function is not used.
5502 This warning is enabled by @option{-Wall}.
5504 @item -Wunused-parameter
5505 @opindex Wunused-parameter
5506 @opindex Wno-unused-parameter
5507 Warn whenever a function parameter is unused aside from its declaration.
5509 To suppress this warning use the @code{unused} attribute
5510 (@pxref{Variable Attributes}).
5512 @item -Wno-unused-result
5513 @opindex Wunused-result
5514 @opindex Wno-unused-result
5515 Do not warn if a caller of a function marked with attribute
5516 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
5517 its return value. The default is @option{-Wunused-result}.
5519 @item -Wunused-variable
5520 @opindex Wunused-variable
5521 @opindex Wno-unused-variable
5522 Warn whenever a local or static variable is unused aside from its
5523 declaration. This option implies @option{-Wunused-const-variable=1} for C,
5524 but not for C++. This warning is enabled by @option{-Wall}.
5526 To suppress this warning use the @code{unused} attribute
5527 (@pxref{Variable Attributes}).
5529 @item -Wunused-const-variable
5530 @itemx -Wunused-const-variable=@var{n}
5531 @opindex Wunused-const-variable
5532 @opindex Wno-unused-const-variable
5533 Warn whenever a constant static variable is unused aside from its declaration.
5534 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
5535 for C, but not for C++. In C this declares variable storage, but in C++ this
5536 is not an error since const variables take the place of @code{#define}s.
5538 To suppress this warning use the @code{unused} attribute
5539 (@pxref{Variable Attributes}).
5542 @item -Wunused-const-variable=1
5543 This is the warning level that is enabled by @option{-Wunused-variable} for
5544 C. It warns only about unused static const variables defined in the main
5545 compilation unit, but not about static const variables declared in any
5548 @item -Wunused-const-variable=2
5549 This warning level also warns for unused constant static variables in
5550 headers (excluding system headers). This is the warning level of
5551 @option{-Wunused-const-variable} and must be explicitly requested since
5552 in C++ this isn't an error and in C it might be harder to clean up all
5556 @item -Wunused-value
5557 @opindex Wunused-value
5558 @opindex Wno-unused-value
5559 Warn whenever a statement computes a result that is explicitly not
5560 used. To suppress this warning cast the unused expression to
5561 @code{void}. This includes an expression-statement or the left-hand
5562 side of a comma expression that contains no side effects. For example,
5563 an expression such as @code{x[i,j]} causes a warning, while
5564 @code{x[(void)i,j]} does not.
5566 This warning is enabled by @option{-Wall}.
5571 All the above @option{-Wunused} options combined.
5573 In order to get a warning about an unused function parameter, you must
5574 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
5575 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
5577 @item -Wuninitialized
5578 @opindex Wuninitialized
5579 @opindex Wno-uninitialized
5580 Warn if an automatic variable is used without first being initialized
5581 or if a variable may be clobbered by a @code{setjmp} call. In C++,
5582 warn if a non-static reference or non-static @code{const} member
5583 appears in a class without constructors.
5585 If you want to warn about code that uses the uninitialized value of the
5586 variable in its own initializer, use the @option{-Winit-self} option.
5588 These warnings occur for individual uninitialized or clobbered
5589 elements of structure, union or array variables as well as for
5590 variables that are uninitialized or clobbered as a whole. They do
5591 not occur for variables or elements declared @code{volatile}. Because
5592 these warnings depend on optimization, the exact variables or elements
5593 for which there are warnings depends on the precise optimization
5594 options and version of GCC used.
5596 Note that there may be no warning about a variable that is used only
5597 to compute a value that itself is never used, because such
5598 computations may be deleted by data flow analysis before the warnings
5601 @item -Winvalid-memory-model
5602 @opindex Winvalid-memory-model
5603 @opindex Wno-invalid-memory-model
5604 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
5605 and the C11 atomic generic functions with a memory consistency argument
5606 that is either invalid for the operation or outside the range of values
5607 of the @code{memory_order} enumeration. For example, since the
5608 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
5609 defined for the relaxed, release, and sequentially consistent memory
5610 orders the following code is diagnosed:
5615 __atomic_store_n (i, 0, memory_order_consume);
5619 @option{-Winvalid-memory-model} is enabled by default.
5621 @item -Wmaybe-uninitialized
5622 @opindex Wmaybe-uninitialized
5623 @opindex Wno-maybe-uninitialized
5624 For an automatic (i.e.@: local) variable, if there exists a path from the
5625 function entry to a use of the variable that is initialized, but there exist
5626 some other paths for which the variable is not initialized, the compiler
5627 emits a warning if it cannot prove the uninitialized paths are not
5628 executed at run time.
5630 These warnings are only possible in optimizing compilation, because otherwise
5631 GCC does not keep track of the state of variables.
5633 These warnings are made optional because GCC may not be able to determine when
5634 the code is correct in spite of appearing to have an error. Here is one
5635 example of how this can happen:
5655 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
5656 always initialized, but GCC doesn't know this. To suppress the
5657 warning, you need to provide a default case with assert(0) or
5660 @cindex @code{longjmp} warnings
5661 This option also warns when a non-volatile automatic variable might be
5662 changed by a call to @code{longjmp}.
5663 The compiler sees only the calls to @code{setjmp}. It cannot know
5664 where @code{longjmp} will be called; in fact, a signal handler could
5665 call it at any point in the code. As a result, you may get a warning
5666 even when there is in fact no problem because @code{longjmp} cannot
5667 in fact be called at the place that would cause a problem.
5669 Some spurious warnings can be avoided if you declare all the functions
5670 you use that never return as @code{noreturn}. @xref{Function
5673 This warning is enabled by @option{-Wall} or @option{-Wextra}.
5675 @item -Wunknown-pragmas
5676 @opindex Wunknown-pragmas
5677 @opindex Wno-unknown-pragmas
5678 @cindex warning for unknown pragmas
5679 @cindex unknown pragmas, warning
5680 @cindex pragmas, warning of unknown
5681 Warn when a @code{#pragma} directive is encountered that is not understood by
5682 GCC@. If this command-line option is used, warnings are even issued
5683 for unknown pragmas in system header files. This is not the case if
5684 the warnings are only enabled by the @option{-Wall} command-line option.
5687 @opindex Wno-pragmas
5689 Do not warn about misuses of pragmas, such as incorrect parameters,
5690 invalid syntax, or conflicts between pragmas. See also
5691 @option{-Wunknown-pragmas}.
5693 @item -Wno-prio-ctor-dtor
5694 @opindex Wno-prio-ctor-dtor
5695 @opindex Wprio-ctor-dtor
5696 Do not warn if a priority from 0 to 100 is used for constructor or destructor.
5697 The use of constructor and destructor attributes allow you to assign a
5698 priority to the constructor/destructor to control its order of execution
5699 before @code{main} is called or after it returns. The priority values must be
5700 greater than 100 as the compiler reserves priority values between 0--100 for
5703 @item -Wstrict-aliasing
5704 @opindex Wstrict-aliasing
5705 @opindex Wno-strict-aliasing
5706 This option is only active when @option{-fstrict-aliasing} is active.
5707 It warns about code that might break the strict aliasing rules that the
5708 compiler is using for optimization. The warning does not catch all
5709 cases, but does attempt to catch the more common pitfalls. It is
5710 included in @option{-Wall}.
5711 It is equivalent to @option{-Wstrict-aliasing=3}
5713 @item -Wstrict-aliasing=n
5714 @opindex Wstrict-aliasing=n
5715 This option is only active when @option{-fstrict-aliasing} is active.
5716 It warns about code that might break the strict aliasing rules that the
5717 compiler is using for optimization.
5718 Higher levels correspond to higher accuracy (fewer false positives).
5719 Higher levels also correspond to more effort, similar to the way @option{-O}
5721 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
5723 Level 1: Most aggressive, quick, least accurate.
5724 Possibly useful when higher levels
5725 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
5726 false negatives. However, it has many false positives.
5727 Warns for all pointer conversions between possibly incompatible types,
5728 even if never dereferenced. Runs in the front end only.
5730 Level 2: Aggressive, quick, not too precise.
5731 May still have many false positives (not as many as level 1 though),
5732 and few false negatives (but possibly more than level 1).
5733 Unlike level 1, it only warns when an address is taken. Warns about
5734 incomplete types. Runs in the front end only.
5736 Level 3 (default for @option{-Wstrict-aliasing}):
5737 Should have very few false positives and few false
5738 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
5739 Takes care of the common pun+dereference pattern in the front end:
5740 @code{*(int*)&some_float}.
5741 If optimization is enabled, it also runs in the back end, where it deals
5742 with multiple statement cases using flow-sensitive points-to information.
5743 Only warns when the converted pointer is dereferenced.
5744 Does not warn about incomplete types.
5746 @item -Wstrict-overflow
5747 @itemx -Wstrict-overflow=@var{n}
5748 @opindex Wstrict-overflow
5749 @opindex Wno-strict-overflow
5750 This option is only active when signed overflow is undefined.
5751 It warns about cases where the compiler optimizes based on the
5752 assumption that signed overflow does not occur. Note that it does not
5753 warn about all cases where the code might overflow: it only warns
5754 about cases where the compiler implements some optimization. Thus
5755 this warning depends on the optimization level.
5757 An optimization that assumes that signed overflow does not occur is
5758 perfectly safe if the values of the variables involved are such that
5759 overflow never does, in fact, occur. Therefore this warning can
5760 easily give a false positive: a warning about code that is not
5761 actually a problem. To help focus on important issues, several
5762 warning levels are defined. No warnings are issued for the use of
5763 undefined signed overflow when estimating how many iterations a loop
5764 requires, in particular when determining whether a loop will be
5768 @item -Wstrict-overflow=1
5769 Warn about cases that are both questionable and easy to avoid. For
5770 example the compiler simplifies
5771 @code{x + 1 > x} to @code{1}. This level of
5772 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
5773 are not, and must be explicitly requested.
5775 @item -Wstrict-overflow=2
5776 Also warn about other cases where a comparison is simplified to a
5777 constant. For example: @code{abs (x) >= 0}. This can only be
5778 simplified when signed integer overflow is undefined, because
5779 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
5780 zero. @option{-Wstrict-overflow} (with no level) is the same as
5781 @option{-Wstrict-overflow=2}.
5783 @item -Wstrict-overflow=3
5784 Also warn about other cases where a comparison is simplified. For
5785 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
5787 @item -Wstrict-overflow=4
5788 Also warn about other simplifications not covered by the above cases.
5789 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
5791 @item -Wstrict-overflow=5
5792 Also warn about cases where the compiler reduces the magnitude of a
5793 constant involved in a comparison. For example: @code{x + 2 > y} is
5794 simplified to @code{x + 1 >= y}. This is reported only at the
5795 highest warning level because this simplification applies to many
5796 comparisons, so this warning level gives a very large number of
5800 @item -Wstringop-overflow
5801 @itemx -Wstringop-overflow=@var{type}
5802 @opindex Wstringop-overflow
5803 @opindex Wno-stringop-overflow
5804 Warn for calls to string manipulation functions such as @code{memcpy} and
5805 @code{strcpy} that are determined to overflow the destination buffer. The
5806 optional argument is one greater than the type of Object Size Checking to
5807 perform to determine the size of the destination. @xref{Object Size Checking}.
5808 The argument is meaningful only for functions that operate on character arrays
5809 but not for raw memory functions like @code{memcpy} which always make use
5810 of Object Size type-0. The option also warns for calls that specify a size
5811 in excess of the largest possible object or at most @code{SIZE_MAX / 2} bytes.
5812 The option produces the best results with optimization enabled but can detect
5813 a small subset of simple buffer overflows even without optimization in
5814 calls to the GCC built-in functions like @code{__builtin_memcpy} that
5815 correspond to the standard functions. In any case, the option warns about
5816 just a subset of buffer overflows detected by the corresponding overflow
5817 checking built-ins. For example, the option will issue a warning for
5818 the @code{strcpy} call below because it copies at least 5 characters
5819 (the string @code{"blue"} including the terminating NUL) into the buffer
5823 enum Color @{ blue, purple, yellow @};
5824 const char* f (enum Color clr)
5826 static char buf [4];
5830 case blue: str = "blue"; break;
5831 case purple: str = "purple"; break;
5832 case yellow: str = "yellow"; break;
5835 return strcpy (buf, str); // warning here
5839 Option @option{-Wstringop-overflow=2} is enabled by default.
5842 @item -Wstringop-overflow
5843 @itemx -Wstringop-overflow=1
5844 @opindex Wstringop-overflow
5845 @opindex Wno-stringop-overflow
5846 The @option{-Wstringop-overflow=1} option uses type-zero Object Size Checking
5847 to determine the sizes of destination objects. This is the default setting
5848 of the option. At this setting the option will not warn for writes past
5849 the end of subobjects of larger objects accessed by pointers unless the
5850 size of the largest surrounding object is known. When the destination may
5851 be one of several objects it is assumed to be the largest one of them. On
5852 Linux systems, when optimization is enabled at this setting the option warns
5853 for the same code as when the @code{_FORTIFY_SOURCE} macro is defined to
5856 @item -Wstringop-overflow=2
5857 The @option{-Wstringop-overflow=2} option uses type-one Object Size Checking
5858 to determine the sizes of destination objects. At this setting the option
5859 will warn about overflows when writing to members of the largest complete
5860 objects whose exact size is known. It will, however, not warn for excessive
5861 writes to the same members of unknown objects referenced by pointers since
5862 they may point to arrays containing unknown numbers of elements.
5864 @item -Wstringop-overflow=3
5865 The @option{-Wstringop-overflow=3} option uses type-two Object Size Checking
5866 to determine the sizes of destination objects. At this setting the option
5867 warns about overflowing the smallest object or data member. This is the
5868 most restrictive setting of the option that may result in warnings for safe
5871 @item -Wstringop-overflow=4
5872 The @option{-Wstringop-overflow=4} option uses type-three Object Size Checking
5873 to determine the sizes of destination objects. At this setting the option
5874 will warn about overflowing any data members, and when the destination is
5875 one of several objects it uses the size of the largest of them to decide
5876 whether to issue a warning. Similarly to @option{-Wstringop-overflow=3} this
5877 setting of the option may result in warnings for benign code.
5880 @item -Wstringop-truncation
5881 @opindex Wstringop-truncation
5882 @opindex Wno-stringop-truncation
5883 Warn for calls to bounded string manipulation functions such as @code{strncat},
5884 @code{strncpy}, and @code{stpncpy} that may either truncate the copied string
5885 or leave the destination unchanged.
5887 In the following example, the call to @code{strncat} specifies a bound that
5888 is less than the length of the source string. As a result, the copy of
5889 the source will be truncated and so the call is diagnosed. To avoid the
5890 warning use @code{bufsize - strlen (buf) - 1)} as the bound.
5893 void append (char *buf, size_t bufsize)
5895 strncat (buf, ".txt", 3);
5899 As another example, the following call to @code{strncpy} results in copying
5900 to @code{d} just the characters preceding the terminating NUL, without
5901 appending the NUL to the end. Assuming the result of @code{strncpy} is
5902 necessarily a NUL-terminated string is a common mistake, and so the call
5903 is diagnosed. To avoid the warning when the result is not expected to be
5904 NUL-terminated, call @code{memcpy} instead.
5907 void copy (char *d, const char *s)
5909 strncpy (d, s, strlen (s));
5913 In the following example, the call to @code{strncpy} specifies the size
5914 of the destination buffer as the bound. If the length of the source
5915 string is equal to or greater than this size the result of the copy will
5916 not be NUL-terminated. Therefore, the call is also diagnosed. To avoid
5917 the warning, specify @code{sizeof buf - 1} as the bound and set the last
5918 element of the buffer to @code{NUL}.
5921 void copy (const char *s)
5924 strncpy (buf, s, sizeof buf);
5929 In situations where a character array is intended to store a sequence
5930 of bytes with no terminating @code{NUL} such an array may be annotated
5931 with attribute @code{nonstring} to avoid this warning. Such arrays,
5932 however, are not suitable arguments to functions that expect
5933 @code{NUL}-terminated strings. To help detect accidental misuses of
5934 such arrays GCC issues warnings unless it can prove that the use is
5935 safe. @xref{Common Variable Attributes}.
5937 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{|}cold@r{|}malloc@r{]}
5938 @opindex Wsuggest-attribute=
5939 @opindex Wno-suggest-attribute=
5940 Warn for cases where adding an attribute may be beneficial. The
5941 attributes currently supported are listed below.
5944 @item -Wsuggest-attribute=pure
5945 @itemx -Wsuggest-attribute=const
5946 @itemx -Wsuggest-attribute=noreturn
5947 @itemx -Wmissing-noreturn
5948 @itemx -Wsuggest-attribute=malloc
5949 @opindex Wsuggest-attribute=pure
5950 @opindex Wno-suggest-attribute=pure
5951 @opindex Wsuggest-attribute=const
5952 @opindex Wno-suggest-attribute=const
5953 @opindex Wsuggest-attribute=noreturn
5954 @opindex Wno-suggest-attribute=noreturn
5955 @opindex Wmissing-noreturn
5956 @opindex Wno-missing-noreturn
5957 @opindex Wsuggest-attribute=malloc
5958 @opindex Wno-suggest-attribute=malloc
5960 Warn about functions that might be candidates for attributes
5961 @code{pure}, @code{const} or @code{noreturn} or @code{malloc}. The compiler
5962 only warns for functions visible in other compilation units or (in the case of
5963 @code{pure} and @code{const}) if it cannot prove that the function returns
5964 normally. A function returns normally if it doesn't contain an infinite loop or
5965 return abnormally by throwing, calling @code{abort} or trapping. This analysis
5966 requires option @option{-fipa-pure-const}, which is enabled by default at
5967 @option{-O} and higher. Higher optimization levels improve the accuracy
5970 @item -Wsuggest-attribute=format
5971 @itemx -Wmissing-format-attribute
5972 @opindex Wsuggest-attribute=format
5973 @opindex Wmissing-format-attribute
5974 @opindex Wno-suggest-attribute=format
5975 @opindex Wno-missing-format-attribute
5979 Warn about function pointers that might be candidates for @code{format}
5980 attributes. Note these are only possible candidates, not absolute ones.
5981 GCC guesses that function pointers with @code{format} attributes that
5982 are used in assignment, initialization, parameter passing or return
5983 statements should have a corresponding @code{format} attribute in the
5984 resulting type. I.e.@: the left-hand side of the assignment or
5985 initialization, the type of the parameter variable, or the return type
5986 of the containing function respectively should also have a @code{format}
5987 attribute to avoid the warning.
5989 GCC also warns about function definitions that might be
5990 candidates for @code{format} attributes. Again, these are only
5991 possible candidates. GCC guesses that @code{format} attributes
5992 might be appropriate for any function that calls a function like
5993 @code{vprintf} or @code{vscanf}, but this might not always be the
5994 case, and some functions for which @code{format} attributes are
5995 appropriate may not be detected.
5997 @item -Wsuggest-attribute=cold
5998 @opindex Wsuggest-attribute=cold
5999 @opindex Wno-suggest-attribute=cold
6001 Warn about functions that might be candidates for @code{cold} attribute. This
6002 is based on static detection and generally will only warn about functions which
6003 always leads to a call to another @code{cold} function such as wrappers of
6004 C++ @code{throw} or fatal error reporting functions leading to @code{abort}.
6007 @item -Wsuggest-final-types
6008 @opindex Wno-suggest-final-types
6009 @opindex Wsuggest-final-types
6010 Warn about types with virtual methods where code quality would be improved
6011 if the type were declared with the C++11 @code{final} specifier,
6013 declared in an anonymous namespace. This allows GCC to more aggressively
6014 devirtualize the polymorphic calls. This warning is more effective with
6015 link-time optimization,
6016 where the information about the class hierarchy graph is
6019 @item -Wsuggest-final-methods
6020 @opindex Wno-suggest-final-methods
6021 @opindex Wsuggest-final-methods
6022 Warn about virtual methods where code quality would be improved if the method
6023 were declared with the C++11 @code{final} specifier,
6024 or, if possible, its type were
6025 declared in an anonymous namespace or with the @code{final} specifier.
6027 more effective with link-time optimization, where the information about the
6028 class hierarchy graph is more complete. It is recommended to first consider
6029 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
6032 @item -Wsuggest-override
6033 Warn about overriding virtual functions that are not marked with the override
6037 @opindex Wno-alloc-zero
6038 @opindex Walloc-zero
6039 Warn about calls to allocation functions decorated with attribute
6040 @code{alloc_size} that specify zero bytes, including those to the built-in
6041 forms of the functions @code{aligned_alloc}, @code{alloca}, @code{calloc},
6042 @code{malloc}, and @code{realloc}. Because the behavior of these functions
6043 when called with a zero size differs among implementations (and in the case
6044 of @code{realloc} has been deprecated) relying on it may result in subtle
6045 portability bugs and should be avoided.
6047 @item -Walloc-size-larger-than=@var{byte-size}
6048 @opindex Walloc-size-larger-than=
6049 @opindex Wno-alloc-size-larger-than
6050 Warn about calls to functions decorated with attribute @code{alloc_size}
6051 that attempt to allocate objects larger than the specified number of bytes,
6052 or where the result of the size computation in an integer type with infinite
6053 precision would exceed the value of @samp{PTRDIFF_MAX} on the target.
6054 @option{-Walloc-size-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6055 Warnings controlled by the option can be disabled either by specifying
6056 @var{byte-size} of @samp{SIZE_MAX} or more or by
6057 @option{-Wno-alloc-size-larger-than}.
6058 @xref{Function Attributes}.
6060 @item -Wno-alloc-size-larger-than
6061 @opindex Wno-alloc-size-larger-than
6062 Disable @option{-Walloc-size-larger-than=} warnings. The option is
6063 equivalent to @option{-Walloc-size-larger-than=}@samp{SIZE_MAX} or
6069 This option warns on all uses of @code{alloca} in the source.
6071 @item -Walloca-larger-than=@var{byte-size}
6072 @opindex Walloca-larger-than=
6073 @opindex Wno-alloca-larger-than
6074 This option warns on calls to @code{alloca} with an integer argument whose
6075 value is either zero, or that is not bounded by a controlling predicate
6076 that limits its value to at most @var{byte-size}. It also warns for calls
6077 to @code{alloca} where the bound value is unknown. Arguments of non-integer
6078 types are considered unbounded even if they appear to be constrained to
6081 For example, a bounded case of @code{alloca} could be:
6084 void func (size_t n)
6095 In the above example, passing @code{-Walloca-larger-than=1000} would not
6096 issue a warning because the call to @code{alloca} is known to be at most
6097 1000 bytes. However, if @code{-Walloca-larger-than=500} were passed,
6098 the compiler would emit a warning.
6100 Unbounded uses, on the other hand, are uses of @code{alloca} with no
6101 controlling predicate constraining its integer argument. For example:
6106 void *p = alloca (n);
6111 If @code{-Walloca-larger-than=500} were passed, the above would trigger
6112 a warning, but this time because of the lack of bounds checking.
6114 Note, that even seemingly correct code involving signed integers could
6118 void func (signed int n)
6128 In the above example, @var{n} could be negative, causing a larger than
6129 expected argument to be implicitly cast into the @code{alloca} call.
6131 This option also warns when @code{alloca} is used in a loop.
6133 @option{-Walloca-larger-than=}@samp{PTRDIFF_MAX} is enabled by default
6134 but is usually only effective when @option{-ftree-vrp} is active (default
6135 for @option{-O2} and above).
6137 See also @option{-Wvla-larger-than=}@samp{byte-size}.
6139 @item -Wno-alloca-larger-than
6140 @opindex Wno-alloca-larger-than
6141 Disable @option{-Walloca-larger-than=} warnings. The option is
6142 equivalent to @option{-Walloca-larger-than=}@samp{SIZE_MAX} or larger.
6144 @item -Warray-bounds
6145 @itemx -Warray-bounds=@var{n}
6146 @opindex Wno-array-bounds
6147 @opindex Warray-bounds
6148 This option is only active when @option{-ftree-vrp} is active
6149 (default for @option{-O2} and above). It warns about subscripts to arrays
6150 that are always out of bounds. This warning is enabled by @option{-Wall}.
6153 @item -Warray-bounds=1
6154 This is the warning level of @option{-Warray-bounds} and is enabled
6155 by @option{-Wall}; higher levels are not, and must be explicitly requested.
6157 @item -Warray-bounds=2
6158 This warning level also warns about out of bounds access for
6159 arrays at the end of a struct and for arrays accessed through
6160 pointers. This warning level may give a larger number of
6161 false positives and is deactivated by default.
6164 @item -Wattribute-alias=@var{n}
6165 @itemx -Wno-attribute-alias
6166 @opindex Wattribute-alias
6167 @opindex Wno-attribute-alias
6168 Warn about declarations using the @code{alias} and similar attributes whose
6169 target is incompatible with the type of the alias.
6170 @xref{Function Attributes,,Declaring Attributes of Functions}.
6173 @item -Wattribute-alias=1
6174 The default warning level of the @option{-Wattribute-alias} option diagnoses
6175 incompatibilities between the type of the alias declaration and that of its
6176 target. Such incompatibilities are typically indicative of bugs.
6178 @item -Wattribute-alias=2
6180 At this level @option{-Wattribute-alias} also diagnoses cases where
6181 the attributes of the alias declaration are more restrictive than the
6182 attributes applied to its target. These mismatches can potentially
6183 result in incorrect code generation. In other cases they may be
6184 benign and could be resolved simply by adding the missing attribute to
6185 the target. For comparison, see the @option{-Wmissing-attributes}
6186 option, which controls diagnostics when the alias declaration is less
6187 restrictive than the target, rather than more restrictive.
6189 Attributes considered include @code{alloc_align}, @code{alloc_size},
6190 @code{cold}, @code{const}, @code{hot}, @code{leaf}, @code{malloc},
6191 @code{nonnull}, @code{noreturn}, @code{nothrow}, @code{pure},
6192 @code{returns_nonnull}, and @code{returns_twice}.
6195 @option{-Wattribute-alias} is equivalent to @option{-Wattribute-alias=1}.
6196 This is the default. You can disable these warnings with either
6197 @option{-Wno-attribute-alias} or @option{-Wattribute-alias=0}.
6199 @item -Wbool-compare
6200 @opindex Wno-bool-compare
6201 @opindex Wbool-compare
6202 Warn about boolean expression compared with an integer value different from
6203 @code{true}/@code{false}. For instance, the following comparison is
6208 if ((n > 1) == 2) @{ @dots{} @}
6210 This warning is enabled by @option{-Wall}.
6212 @item -Wbool-operation
6213 @opindex Wno-bool-operation
6214 @opindex Wbool-operation
6215 Warn about suspicious operations on expressions of a boolean type. For
6216 instance, bitwise negation of a boolean is very likely a bug in the program.
6217 For C, this warning also warns about incrementing or decrementing a boolean,
6218 which rarely makes sense. (In C++, decrementing a boolean is always invalid.
6219 Incrementing a boolean is invalid in C++17, and deprecated otherwise.)
6221 This warning is enabled by @option{-Wall}.
6223 @item -Wduplicated-branches
6224 @opindex Wno-duplicated-branches
6225 @opindex Wduplicated-branches
6226 Warn when an if-else has identical branches. This warning detects cases like
6233 It doesn't warn when both branches contain just a null statement. This warning
6234 also warn for conditional operators:
6236 int i = x ? *p : *p;
6239 @item -Wduplicated-cond
6240 @opindex Wno-duplicated-cond
6241 @opindex Wduplicated-cond
6242 Warn about duplicated conditions in an if-else-if chain. For instance,
6243 warn for the following code:
6245 if (p->q != NULL) @{ @dots{} @}
6246 else if (p->q != NULL) @{ @dots{} @}
6249 @item -Wframe-address
6250 @opindex Wno-frame-address
6251 @opindex Wframe-address
6252 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
6253 is called with an argument greater than 0. Such calls may return indeterminate
6254 values or crash the program. The warning is included in @option{-Wall}.
6256 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
6257 @opindex Wno-discarded-qualifiers
6258 @opindex Wdiscarded-qualifiers
6259 Do not warn if type qualifiers on pointers are being discarded.
6260 Typically, the compiler warns if a @code{const char *} variable is
6261 passed to a function that takes a @code{char *} parameter. This option
6262 can be used to suppress such a warning.
6264 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
6265 @opindex Wno-discarded-array-qualifiers
6266 @opindex Wdiscarded-array-qualifiers
6267 Do not warn if type qualifiers on arrays which are pointer targets
6268 are being discarded. Typically, the compiler warns if a
6269 @code{const int (*)[]} variable is passed to a function that
6270 takes a @code{int (*)[]} parameter. This option can be used to
6271 suppress such a warning.
6273 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
6274 @opindex Wno-incompatible-pointer-types
6275 @opindex Wincompatible-pointer-types
6276 Do not warn when there is a conversion between pointers that have incompatible
6277 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
6278 which warns for pointer argument passing or assignment with different
6281 @item -Wno-int-conversion @r{(C and Objective-C only)}
6282 @opindex Wno-int-conversion
6283 @opindex Wint-conversion
6284 Do not warn about incompatible integer to pointer and pointer to integer
6285 conversions. This warning is about implicit conversions; for explicit
6286 conversions the warnings @option{-Wno-int-to-pointer-cast} and
6287 @option{-Wno-pointer-to-int-cast} may be used.
6289 @item -Wno-div-by-zero
6290 @opindex Wno-div-by-zero
6291 @opindex Wdiv-by-zero
6292 Do not warn about compile-time integer division by zero. Floating-point
6293 division by zero is not warned about, as it can be a legitimate way of
6294 obtaining infinities and NaNs.
6296 @item -Wsystem-headers
6297 @opindex Wsystem-headers
6298 @opindex Wno-system-headers
6299 @cindex warnings from system headers
6300 @cindex system headers, warnings from
6301 Print warning messages for constructs found in system header files.
6302 Warnings from system headers are normally suppressed, on the assumption
6303 that they usually do not indicate real problems and would only make the
6304 compiler output harder to read. Using this command-line option tells
6305 GCC to emit warnings from system headers as if they occurred in user
6306 code. However, note that using @option{-Wall} in conjunction with this
6307 option does @emph{not} warn about unknown pragmas in system
6308 headers---for that, @option{-Wunknown-pragmas} must also be used.
6310 @item -Wtautological-compare
6311 @opindex Wtautological-compare
6312 @opindex Wno-tautological-compare
6313 Warn if a self-comparison always evaluates to true or false. This
6314 warning detects various mistakes such as:
6318 if (i > i) @{ @dots{} @}
6321 This warning also warns about bitwise comparisons that always evaluate
6322 to true or false, for instance:
6324 if ((a & 16) == 10) @{ @dots{} @}
6326 will always be false.
6328 This warning is enabled by @option{-Wall}.
6331 @opindex Wtrampolines
6332 @opindex Wno-trampolines
6333 Warn about trampolines generated for pointers to nested functions.
6334 A trampoline is a small piece of data or code that is created at run
6335 time on the stack when the address of a nested function is taken, and is
6336 used to call the nested function indirectly. For some targets, it is
6337 made up of data only and thus requires no special treatment. But, for
6338 most targets, it is made up of code and thus requires the stack to be
6339 made executable in order for the program to work properly.
6342 @opindex Wfloat-equal
6343 @opindex Wno-float-equal
6344 Warn if floating-point values are used in equality comparisons.
6346 The idea behind this is that sometimes it is convenient (for the
6347 programmer) to consider floating-point values as approximations to
6348 infinitely precise real numbers. If you are doing this, then you need
6349 to compute (by analyzing the code, or in some other way) the maximum or
6350 likely maximum error that the computation introduces, and allow for it
6351 when performing comparisons (and when producing output, but that's a
6352 different problem). In particular, instead of testing for equality, you
6353 should check to see whether the two values have ranges that overlap; and
6354 this is done with the relational operators, so equality comparisons are
6357 @item -Wtraditional @r{(C and Objective-C only)}
6358 @opindex Wtraditional
6359 @opindex Wno-traditional
6360 Warn about certain constructs that behave differently in traditional and
6361 ISO C@. Also warn about ISO C constructs that have no traditional C
6362 equivalent, and/or problematic constructs that should be avoided.
6366 Macro parameters that appear within string literals in the macro body.
6367 In traditional C macro replacement takes place within string literals,
6368 but in ISO C it does not.
6371 In traditional C, some preprocessor directives did not exist.
6372 Traditional preprocessors only considered a line to be a directive
6373 if the @samp{#} appeared in column 1 on the line. Therefore
6374 @option{-Wtraditional} warns about directives that traditional C
6375 understands but ignores because the @samp{#} does not appear as the
6376 first character on the line. It also suggests you hide directives like
6377 @code{#pragma} not understood by traditional C by indenting them. Some
6378 traditional implementations do not recognize @code{#elif}, so this option
6379 suggests avoiding it altogether.
6382 A function-like macro that appears without arguments.
6385 The unary plus operator.
6388 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
6389 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
6390 constants.) Note, these suffixes appear in macros defined in the system
6391 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
6392 Use of these macros in user code might normally lead to spurious
6393 warnings, however GCC's integrated preprocessor has enough context to
6394 avoid warning in these cases.
6397 A function declared external in one block and then used after the end of
6401 A @code{switch} statement has an operand of type @code{long}.
6404 A non-@code{static} function declaration follows a @code{static} one.
6405 This construct is not accepted by some traditional C compilers.
6408 The ISO type of an integer constant has a different width or
6409 signedness from its traditional type. This warning is only issued if
6410 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
6411 typically represent bit patterns, are not warned about.
6414 Usage of ISO string concatenation is detected.
6417 Initialization of automatic aggregates.
6420 Identifier conflicts with labels. Traditional C lacks a separate
6421 namespace for labels.
6424 Initialization of unions. If the initializer is zero, the warning is
6425 omitted. This is done under the assumption that the zero initializer in
6426 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
6427 initializer warnings and relies on default initialization to zero in the
6431 Conversions by prototypes between fixed/floating-point values and vice
6432 versa. The absence of these prototypes when compiling with traditional
6433 C causes serious problems. This is a subset of the possible
6434 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
6437 Use of ISO C style function definitions. This warning intentionally is
6438 @emph{not} issued for prototype declarations or variadic functions
6439 because these ISO C features appear in your code when using
6440 libiberty's traditional C compatibility macros, @code{PARAMS} and
6441 @code{VPARAMS}. This warning is also bypassed for nested functions
6442 because that feature is already a GCC extension and thus not relevant to
6443 traditional C compatibility.
6446 @item -Wtraditional-conversion @r{(C and Objective-C only)}
6447 @opindex Wtraditional-conversion
6448 @opindex Wno-traditional-conversion
6449 Warn if a prototype causes a type conversion that is different from what
6450 would happen to the same argument in the absence of a prototype. This
6451 includes conversions of fixed point to floating and vice versa, and
6452 conversions changing the width or signedness of a fixed-point argument
6453 except when the same as the default promotion.
6455 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
6456 @opindex Wdeclaration-after-statement
6457 @opindex Wno-declaration-after-statement
6458 Warn when a declaration is found after a statement in a block. This
6459 construct, known from C++, was introduced with ISO C99 and is by default
6460 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
6465 Warn whenever a local variable or type declaration shadows another
6466 variable, parameter, type, class member (in C++), or instance variable
6467 (in Objective-C) or whenever a built-in function is shadowed. Note
6468 that in C++, the compiler warns if a local variable shadows an
6469 explicit typedef, but not if it shadows a struct/class/enum.
6470 Same as @option{-Wshadow=global}.
6472 @item -Wno-shadow-ivar @r{(Objective-C only)}
6473 @opindex Wno-shadow-ivar
6474 @opindex Wshadow-ivar
6475 Do not warn whenever a local variable shadows an instance variable in an
6478 @item -Wshadow=global
6479 @opindex Wshadow=local
6480 The default for @option{-Wshadow}. Warns for any (global) shadowing.
6482 @item -Wshadow=local
6483 @opindex Wshadow=local
6484 Warn when a local variable shadows another local variable or parameter.
6485 This warning is enabled by @option{-Wshadow=global}.
6487 @item -Wshadow=compatible-local
6488 @opindex Wshadow=compatible-local
6489 Warn when a local variable shadows another local variable or parameter
6490 whose type is compatible with that of the shadowing variable. In C++,
6491 type compatibility here means the type of the shadowing variable can be
6492 converted to that of the shadowed variable. The creation of this flag
6493 (in addition to @option{-Wshadow=local}) is based on the idea that when
6494 a local variable shadows another one of incompatible type, it is most
6495 likely intentional, not a bug or typo, as shown in the following example:
6499 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
6501 for (int i = 0; i < N; ++i)
6510 Since the two variable @code{i} in the example above have incompatible types,
6511 enabling only @option{-Wshadow=compatible-local} will not emit a warning.
6512 Because their types are incompatible, if a programmer accidentally uses one
6513 in place of the other, type checking will catch that and emit an error or
6514 warning. So not warning (about shadowing) in this case will not lead to
6515 undetected bugs. Use of this flag instead of @option{-Wshadow=local} can
6516 possibly reduce the number of warnings triggered by intentional shadowing.
6518 This warning is enabled by @option{-Wshadow=local}.
6520 @item -Wlarger-than=@var{byte-size}
6521 @opindex Wlarger-than=
6522 @opindex Wlarger-than-@var{byte-size}
6523 Warn whenever an object is defined whose size exceeds @var{byte-size}.
6524 @option{-Wlarger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6525 Warnings controlled by the option can be disabled either by specifying
6526 @var{byte-size} of @samp{SIZE_MAX} or more or by
6527 @option{-Wno-larger-than}.
6529 @item -Wno-larger-than
6530 @opindex Wno-larger-than
6531 Disable @option{-Wlarger-than=} warnings. The option is equivalent
6532 to @option{-Wlarger-than=}@samp{SIZE_MAX} or larger.
6534 @item -Wframe-larger-than=@var{byte-size}
6535 @opindex Wframe-larger-than=
6536 @opindex Wno-frame-larger-than
6537 Warn if the size of a function frame exceeds @var{byte-size}.
6538 The computation done to determine the stack frame size is approximate
6539 and not conservative.
6540 The actual requirements may be somewhat greater than @var{byte-size}
6541 even if you do not get a warning. In addition, any space allocated
6542 via @code{alloca}, variable-length arrays, or related constructs
6543 is not included by the compiler when determining
6544 whether or not to issue a warning.
6545 @option{-Wframe-larger-than=}@samp{PTRDIFF_MAX} is enabled by default.
6546 Warnings controlled by the option can be disabled either by specifying
6547 @var{byte-size} of @samp{SIZE_MAX} or more or by
6548 @option{-Wno-frame-larger-than}.
6550 @item -Wno-frame-larger-than
6551 @opindex Wno-frame-larger-than
6552 Disable @option{-Wframe-larger-than=} warnings. The option is equivalent
6553 to @option{-Wframe-larger-than=}@samp{SIZE_MAX} or larger.
6555 @item -Wno-free-nonheap-object
6556 @opindex Wno-free-nonheap-object
6557 @opindex Wfree-nonheap-object
6558 Do not warn when attempting to free an object that was not allocated
6561 @item -Wstack-usage=@var{byte-size}
6562 @opindex Wstack-usage
6563 @opindex Wno-stack-usage
6564 Warn if the stack usage of a function might exceed @var{byte-size}.
6565 The computation done to determine the stack usage is conservative.
6566 Any space allocated via @code{alloca}, variable-length arrays, or related
6567 constructs is included by the compiler when determining whether or not to
6570 The message is in keeping with the output of @option{-fstack-usage}.
6574 If the stack usage is fully static but exceeds the specified amount, it's:
6577 warning: stack usage is 1120 bytes
6580 If the stack usage is (partly) dynamic but bounded, it's:
6583 warning: stack usage might be 1648 bytes
6586 If the stack usage is (partly) dynamic and not bounded, it's:
6589 warning: stack usage might be unbounded
6593 @option{-Wstack-usage=}@samp{PTRDIFF_MAX} is enabled by default.
6594 Warnings controlled by the option can be disabled either by specifying
6595 @var{byte-size} of @samp{SIZE_MAX} or more or by
6596 @option{-Wno-stack-usage}.
6598 @item -Wno-stack-usage
6599 @opindex Wno-stack-usage
6600 Disable @option{-Wstack-usage=} warnings. The option is equivalent
6601 to @option{-Wstack-usage=}@samp{SIZE_MAX} or larger.
6603 @item -Wunsafe-loop-optimizations
6604 @opindex Wunsafe-loop-optimizations
6605 @opindex Wno-unsafe-loop-optimizations
6606 Warn if the loop cannot be optimized because the compiler cannot
6607 assume anything on the bounds of the loop indices. With
6608 @option{-funsafe-loop-optimizations} warn if the compiler makes
6611 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
6612 @opindex Wno-pedantic-ms-format
6613 @opindex Wpedantic-ms-format
6614 When used in combination with @option{-Wformat}
6615 and @option{-pedantic} without GNU extensions, this option
6616 disables the warnings about non-ISO @code{printf} / @code{scanf} format
6617 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
6618 which depend on the MS runtime.
6621 @opindex Waligned-new
6622 @opindex Wno-aligned-new
6623 Warn about a new-expression of a type that requires greater alignment
6624 than the @code{alignof(std::max_align_t)} but uses an allocation
6625 function without an explicit alignment parameter. This option is
6626 enabled by @option{-Wall}.
6628 Normally this only warns about global allocation functions, but
6629 @option{-Waligned-new=all} also warns about class member allocation
6632 @item -Wplacement-new
6633 @itemx -Wplacement-new=@var{n}
6634 @opindex Wplacement-new
6635 @opindex Wno-placement-new
6636 Warn about placement new expressions with undefined behavior, such as
6637 constructing an object in a buffer that is smaller than the type of
6638 the object. For example, the placement new expression below is diagnosed
6639 because it attempts to construct an array of 64 integers in a buffer only
6645 This warning is enabled by default.
6648 @item -Wplacement-new=1
6649 This is the default warning level of @option{-Wplacement-new}. At this
6650 level the warning is not issued for some strictly undefined constructs that
6651 GCC allows as extensions for compatibility with legacy code. For example,
6652 the following @code{new} expression is not diagnosed at this level even
6653 though it has undefined behavior according to the C++ standard because
6654 it writes past the end of the one-element array.
6656 struct S @{ int n, a[1]; @};
6657 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
6658 new (s->a)int [32]();
6661 @item -Wplacement-new=2
6662 At this level, in addition to diagnosing all the same constructs as at level
6663 1, a diagnostic is also issued for placement new expressions that construct
6664 an object in the last member of structure whose type is an array of a single
6665 element and whose size is less than the size of the object being constructed.
6666 While the previous example would be diagnosed, the following construct makes
6667 use of the flexible member array extension to avoid the warning at level 2.
6669 struct S @{ int n, a[]; @};
6670 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
6671 new (s->a)int [32]();
6676 @item -Wpointer-arith
6677 @opindex Wpointer-arith
6678 @opindex Wno-pointer-arith
6679 Warn about anything that depends on the ``size of'' a function type or
6680 of @code{void}. GNU C assigns these types a size of 1, for
6681 convenience in calculations with @code{void *} pointers and pointers
6682 to functions. In C++, warn also when an arithmetic operation involves
6683 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
6685 @item -Wpointer-compare
6686 @opindex Wpointer-compare
6687 @opindex Wno-pointer-compare
6688 Warn if a pointer is compared with a zero character constant. This usually
6689 means that the pointer was meant to be dereferenced. For example:
6692 const char *p = foo ();
6697 Note that the code above is invalid in C++11.
6699 This warning is enabled by default.
6702 @opindex Wtype-limits
6703 @opindex Wno-type-limits
6704 Warn if a comparison is always true or always false due to the limited
6705 range of the data type, but do not warn for constant expressions. For
6706 example, warn if an unsigned variable is compared against zero with
6707 @code{<} or @code{>=}. This warning is also enabled by
6710 @item -Wabsolute-value @r{(C and Objective-C only)}
6711 @opindex Wabsolute-value
6712 @opindex Wno-absolute-value
6713 Warn for calls to standard functions that compute the absolute value
6714 of an argument when a more appropriate standard function is available.
6715 For example, calling @code{abs(3.14)} triggers the warning because the
6716 appropriate function to call to compute the absolute value of a double
6717 argument is @code{fabs}. The option also triggers warnings when the
6718 argument in a call to such a function has an unsigned type. This
6719 warning can be suppressed with an explicit type cast and it is also
6720 enabled by @option{-Wextra}.
6722 @include cppwarnopts.texi
6724 @item -Wbad-function-cast @r{(C and Objective-C only)}
6725 @opindex Wbad-function-cast
6726 @opindex Wno-bad-function-cast
6727 Warn when a function call is cast to a non-matching type.
6728 For example, warn if a call to a function returning an integer type
6729 is cast to a pointer type.
6731 @item -Wc90-c99-compat @r{(C and Objective-C only)}
6732 @opindex Wc90-c99-compat
6733 @opindex Wno-c90-c99-compat
6734 Warn about features not present in ISO C90, but present in ISO C99.
6735 For instance, warn about use of variable length arrays, @code{long long}
6736 type, @code{bool} type, compound literals, designated initializers, and so
6737 on. This option is independent of the standards mode. Warnings are disabled
6738 in the expression that follows @code{__extension__}.
6740 @item -Wc99-c11-compat @r{(C and Objective-C only)}
6741 @opindex Wc99-c11-compat
6742 @opindex Wno-c99-c11-compat
6743 Warn about features not present in ISO C99, but present in ISO C11.
6744 For instance, warn about use of anonymous structures and unions,
6745 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
6746 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
6747 and so on. This option is independent of the standards mode. Warnings are
6748 disabled in the expression that follows @code{__extension__}.
6750 @item -Wc++-compat @r{(C and Objective-C only)}
6751 @opindex Wc++-compat
6752 @opindex Wno-c++-compat
6753 Warn about ISO C constructs that are outside of the common subset of
6754 ISO C and ISO C++, e.g.@: request for implicit conversion from
6755 @code{void *} to a pointer to non-@code{void} type.
6757 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
6758 @opindex Wc++11-compat
6759 @opindex Wno-c++11-compat
6760 Warn about C++ constructs whose meaning differs between ISO C++ 1998
6761 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
6762 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
6763 enabled by @option{-Wall}.
6765 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
6766 @opindex Wc++14-compat
6767 @opindex Wno-c++14-compat
6768 Warn about C++ constructs whose meaning differs between ISO C++ 2011
6769 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
6771 @item -Wc++17-compat @r{(C++ and Objective-C++ only)}
6772 @opindex Wc++17-compat
6773 @opindex Wno-c++17-compat
6774 Warn about C++ constructs whose meaning differs between ISO C++ 2014
6775 and ISO C++ 2017. This warning is enabled by @option{-Wall}.
6777 @item -Wc++20-compat @r{(C++ and Objective-C++ only)}
6778 @opindex Wc++20-compat
6779 @opindex Wno-c++20-compat
6780 Warn about C++ constructs whose meaning differs between ISO C++ 2017
6781 and ISO C++ 2020. This warning is enabled by @option{-Wall}.
6785 @opindex Wno-cast-qual
6786 Warn whenever a pointer is cast so as to remove a type qualifier from
6787 the target type. For example, warn if a @code{const char *} is cast
6788 to an ordinary @code{char *}.
6790 Also warn when making a cast that introduces a type qualifier in an
6791 unsafe way. For example, casting @code{char **} to @code{const char **}
6792 is unsafe, as in this example:
6795 /* p is char ** value. */
6796 const char **q = (const char **) p;
6797 /* Assignment of readonly string to const char * is OK. */
6799 /* Now char** pointer points to read-only memory. */
6804 @opindex Wcast-align
6805 @opindex Wno-cast-align
6806 Warn whenever a pointer is cast such that the required alignment of the
6807 target is increased. For example, warn if a @code{char *} is cast to
6808 an @code{int *} on machines where integers can only be accessed at
6809 two- or four-byte boundaries.
6811 @item -Wcast-align=strict
6812 @opindex Wcast-align=strict
6813 Warn whenever a pointer is cast such that the required alignment of the
6814 target is increased. For example, warn if a @code{char *} is cast to
6815 an @code{int *} regardless of the target machine.
6817 @item -Wcast-function-type
6818 @opindex Wcast-function-type
6819 @opindex Wno-cast-function-type
6820 Warn when a function pointer is cast to an incompatible function pointer.
6821 In a cast involving function types with a variable argument list only
6822 the types of initial arguments that are provided are considered.
6823 Any parameter of pointer-type matches any other pointer-type. Any benign
6824 differences in integral types are ignored, like @code{int} vs.@: @code{long}
6825 on ILP32 targets. Likewise type qualifiers are ignored. The function
6826 type @code{void (*) (void)} is special and matches everything, which can
6827 be used to suppress this warning.
6828 In a cast involving pointer to member types this warning warns whenever
6829 the type cast is changing the pointer to member type.
6830 This warning is enabled by @option{-Wextra}.
6832 @item -Wwrite-strings
6833 @opindex Wwrite-strings
6834 @opindex Wno-write-strings
6835 When compiling C, give string constants the type @code{const
6836 char[@var{length}]} so that copying the address of one into a
6837 non-@code{const} @code{char *} pointer produces a warning. These
6838 warnings help you find at compile time code that can try to write
6839 into a string constant, but only if you have been very careful about
6840 using @code{const} in declarations and prototypes. Otherwise, it is
6841 just a nuisance. This is why we did not make @option{-Wall} request
6844 When compiling C++, warn about the deprecated conversion from string
6845 literals to @code{char *}. This warning is enabled by default for C++
6849 @itemx -Wcatch-value=@var{n} @r{(C++ and Objective-C++ only)}
6850 @opindex Wcatch-value
6851 @opindex Wno-catch-value
6852 Warn about catch handlers that do not catch via reference.
6853 With @option{-Wcatch-value=1} (or @option{-Wcatch-value} for short)
6854 warn about polymorphic class types that are caught by value.
6855 With @option{-Wcatch-value=2} warn about all class types that are caught
6856 by value. With @option{-Wcatch-value=3} warn about all types that are
6857 not caught by reference. @option{-Wcatch-value} is enabled by @option{-Wall}.
6861 @opindex Wno-clobbered
6862 Warn for variables that might be changed by @code{longjmp} or
6863 @code{vfork}. This warning is also enabled by @option{-Wextra}.
6865 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
6866 @opindex Wconditionally-supported
6867 @opindex Wno-conditionally-supported
6868 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
6871 @opindex Wconversion
6872 @opindex Wno-conversion
6873 Warn for implicit conversions that may alter a value. This includes
6874 conversions between real and integer, like @code{abs (x)} when
6875 @code{x} is @code{double}; conversions between signed and unsigned,
6876 like @code{unsigned ui = -1}; and conversions to smaller types, like
6877 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
6878 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
6879 changed by the conversion like in @code{abs (2.0)}. Warnings about
6880 conversions between signed and unsigned integers can be disabled by
6881 using @option{-Wno-sign-conversion}.
6883 For C++, also warn for confusing overload resolution for user-defined
6884 conversions; and conversions that never use a type conversion
6885 operator: conversions to @code{void}, the same type, a base class or a
6886 reference to them. Warnings about conversions between signed and
6887 unsigned integers are disabled by default in C++ unless
6888 @option{-Wsign-conversion} is explicitly enabled.
6890 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
6891 @opindex Wconversion-null
6892 @opindex Wno-conversion-null
6893 Do not warn for conversions between @code{NULL} and non-pointer
6894 types. @option{-Wconversion-null} is enabled by default.
6896 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
6897 @opindex Wzero-as-null-pointer-constant
6898 @opindex Wno-zero-as-null-pointer-constant
6899 Warn when a literal @samp{0} is used as null pointer constant. This can
6900 be useful to facilitate the conversion to @code{nullptr} in C++11.
6902 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
6903 @opindex Wsubobject-linkage
6904 @opindex Wno-subobject-linkage
6905 Warn if a class type has a base or a field whose type uses the anonymous
6906 namespace or depends on a type with no linkage. If a type A depends on
6907 a type B with no or internal linkage, defining it in multiple
6908 translation units would be an ODR violation because the meaning of B
6909 is different in each translation unit. If A only appears in a single
6910 translation unit, the best way to silence the warning is to give it
6911 internal linkage by putting it in an anonymous namespace as well. The
6912 compiler doesn't give this warning for types defined in the main .C
6913 file, as those are unlikely to have multiple definitions.
6914 @option{-Wsubobject-linkage} is enabled by default.
6916 @item -Wdangling-else
6917 @opindex Wdangling-else
6918 @opindex Wno-dangling-else
6919 Warn about constructions where there may be confusion to which
6920 @code{if} statement an @code{else} branch belongs. Here is an example of
6935 In C/C++, every @code{else} branch belongs to the innermost possible
6936 @code{if} statement, which in this example is @code{if (b)}. This is
6937 often not what the programmer expected, as illustrated in the above
6938 example by indentation the programmer chose. When there is the
6939 potential for this confusion, GCC issues a warning when this flag
6940 is specified. To eliminate the warning, add explicit braces around
6941 the innermost @code{if} statement so there is no way the @code{else}
6942 can belong to the enclosing @code{if}. The resulting code
6959 This warning is enabled by @option{-Wparentheses}.
6963 @opindex Wno-date-time
6964 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
6965 are encountered as they might prevent bit-wise-identical reproducible
6968 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
6969 @opindex Wdelete-incomplete
6970 @opindex Wno-delete-incomplete
6971 Warn when deleting a pointer to incomplete type, which may cause
6972 undefined behavior at runtime. This warning is enabled by default.
6974 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
6975 @opindex Wuseless-cast
6976 @opindex Wno-useless-cast
6977 Warn when an expression is casted to its own type.
6980 @opindex Wempty-body
6981 @opindex Wno-empty-body
6982 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
6983 while} statement. This warning is also enabled by @option{-Wextra}.
6985 @item -Wenum-compare
6986 @opindex Wenum-compare
6987 @opindex Wno-enum-compare
6988 Warn about a comparison between values of different enumerated types.
6989 In C++ enumerated type mismatches in conditional expressions are also
6990 diagnosed and the warning is enabled by default. In C this warning is
6991 enabled by @option{-Wall}.
6993 @item -Wenum-conversion @r{(C, Objective-C only)}
6994 @opindex Wenum-conversion
6995 @opindex Wno-enum-conversion
6996 Warn when a value of enumerated type is implicitly converted to a
6997 different enumerated type. This warning is enabled by @option{-Wextra}.
6999 @item -Wextra-semi @r{(C++, Objective-C++ only)}
7000 @opindex Wextra-semi
7001 @opindex Wno-extra-semi
7002 Warn about redundant semicolon after in-class function definition.
7004 @item -Wjump-misses-init @r{(C, Objective-C only)}
7005 @opindex Wjump-misses-init
7006 @opindex Wno-jump-misses-init
7007 Warn if a @code{goto} statement or a @code{switch} statement jumps
7008 forward across the initialization of a variable, or jumps backward to a
7009 label after the variable has been initialized. This only warns about
7010 variables that are initialized when they are declared. This warning is
7011 only supported for C and Objective-C; in C++ this sort of branch is an
7014 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
7015 can be disabled with the @option{-Wno-jump-misses-init} option.
7017 @item -Wsign-compare
7018 @opindex Wsign-compare
7019 @opindex Wno-sign-compare
7020 @cindex warning for comparison of signed and unsigned values
7021 @cindex comparison of signed and unsigned values, warning
7022 @cindex signed and unsigned values, comparison warning
7023 Warn when a comparison between signed and unsigned values could produce
7024 an incorrect result when the signed value is converted to unsigned.
7025 In C++, this warning is also enabled by @option{-Wall}. In C, it is
7026 also enabled by @option{-Wextra}.
7028 @item -Wsign-conversion
7029 @opindex Wsign-conversion
7030 @opindex Wno-sign-conversion
7031 Warn for implicit conversions that may change the sign of an integer
7032 value, like assigning a signed integer expression to an unsigned
7033 integer variable. An explicit cast silences the warning. In C, this
7034 option is enabled also by @option{-Wconversion}.
7036 @item -Wfloat-conversion
7037 @opindex Wfloat-conversion
7038 @opindex Wno-float-conversion
7039 Warn for implicit conversions that reduce the precision of a real value.
7040 This includes conversions from real to integer, and from higher precision
7041 real to lower precision real values. This option is also enabled by
7042 @option{-Wconversion}.
7044 @item -Wno-scalar-storage-order
7045 @opindex Wno-scalar-storage-order
7046 @opindex Wscalar-storage-order
7047 Do not warn on suspicious constructs involving reverse scalar storage order.
7049 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
7050 @opindex Wsized-deallocation
7051 @opindex Wno-sized-deallocation
7052 Warn about a definition of an unsized deallocation function
7054 void operator delete (void *) noexcept;
7055 void operator delete[] (void *) noexcept;
7057 without a definition of the corresponding sized deallocation function
7059 void operator delete (void *, std::size_t) noexcept;
7060 void operator delete[] (void *, std::size_t) noexcept;
7062 or vice versa. Enabled by @option{-Wextra} along with
7063 @option{-fsized-deallocation}.
7065 @item -Wsizeof-pointer-div
7066 @opindex Wsizeof-pointer-div
7067 @opindex Wno-sizeof-pointer-div
7068 Warn for suspicious divisions of two sizeof expressions that divide
7069 the pointer size by the element size, which is the usual way to compute
7070 the array size but won't work out correctly with pointers. This warning
7071 warns e.g.@: about @code{sizeof (ptr) / sizeof (ptr[0])} if @code{ptr} is
7072 not an array, but a pointer. This warning is enabled by @option{-Wall}.
7074 @item -Wsizeof-pointer-memaccess
7075 @opindex Wsizeof-pointer-memaccess
7076 @opindex Wno-sizeof-pointer-memaccess
7077 Warn for suspicious length parameters to certain string and memory built-in
7078 functions if the argument uses @code{sizeof}. This warning triggers for
7079 example for @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not
7080 an array, but a pointer, and suggests a possible fix, or about
7081 @code{memcpy (&foo, ptr, sizeof (&foo));}. @option{-Wsizeof-pointer-memaccess}
7082 also warns about calls to bounded string copy functions like @code{strncat}
7083 or @code{strncpy} that specify as the bound a @code{sizeof} expression of
7084 the source array. For example, in the following function the call to
7085 @code{strncat} specifies the size of the source string as the bound. That
7086 is almost certainly a mistake and so the call is diagnosed.
7088 void make_file (const char *name)
7090 char path[PATH_MAX];
7091 strncpy (path, name, sizeof path - 1);
7092 strncat (path, ".text", sizeof ".text");
7097 The @option{-Wsizeof-pointer-memaccess} option is enabled by @option{-Wall}.
7099 @item -Wsizeof-array-argument
7100 @opindex Wsizeof-array-argument
7101 @opindex Wno-sizeof-array-argument
7102 Warn when the @code{sizeof} operator is applied to a parameter that is
7103 declared as an array in a function definition. This warning is enabled by
7104 default for C and C++ programs.
7106 @item -Wmemset-elt-size
7107 @opindex Wmemset-elt-size
7108 @opindex Wno-memset-elt-size
7109 Warn for suspicious calls to the @code{memset} built-in function, if the
7110 first argument references an array, and the third argument is a number
7111 equal to the number of elements, but not equal to the size of the array
7112 in memory. This indicates that the user has omitted a multiplication by
7113 the element size. This warning is enabled by @option{-Wall}.
7115 @item -Wmemset-transposed-args
7116 @opindex Wmemset-transposed-args
7117 @opindex Wno-memset-transposed-args
7118 Warn for suspicious calls to the @code{memset} built-in function where
7119 the second argument is not zero and the third argument is zero. For
7120 example, the call @code{memset (buf, sizeof buf, 0)} is diagnosed because
7121 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostic
7122 is only emitted if the third argument is a literal zero. Otherwise, if
7123 it is an expression that is folded to zero, or a cast of zero to some
7124 type, it is far less likely that the arguments have been mistakenly
7125 transposed and no warning is emitted. This warning is enabled
7130 @opindex Wno-address
7131 Warn about suspicious uses of memory addresses. These include using
7132 the address of a function in a conditional expression, such as
7133 @code{void func(void); if (func)}, and comparisons against the memory
7134 address of a string literal, such as @code{if (x == "abc")}. Such
7135 uses typically indicate a programmer error: the address of a function
7136 always evaluates to true, so their use in a conditional usually
7137 indicate that the programmer forgot the parentheses in a function
7138 call; and comparisons against string literals result in unspecified
7139 behavior and are not portable in C, so they usually indicate that the
7140 programmer intended to use @code{strcmp}. This warning is enabled by
7143 @item -Waddress-of-packed-member
7144 @opindex Waddress-of-packed-member
7145 @opindex Wno-address-of-packed-member
7146 Warn when the address of packed member of struct or union is taken,
7147 which usually results in an unaligned pointer value. This is
7151 @opindex Wlogical-op
7152 @opindex Wno-logical-op
7153 Warn about suspicious uses of logical operators in expressions.
7154 This includes using logical operators in contexts where a
7155 bit-wise operator is likely to be expected. Also warns when
7156 the operands of a logical operator are the same:
7159 if (a < 0 && a < 0) @{ @dots{} @}
7162 @item -Wlogical-not-parentheses
7163 @opindex Wlogical-not-parentheses
7164 @opindex Wno-logical-not-parentheses
7165 Warn about logical not used on the left hand side operand of a comparison.
7166 This option does not warn if the right operand is considered to be a boolean
7167 expression. Its purpose is to detect suspicious code like the following:
7171 if (!a > 1) @{ @dots{} @}
7174 It is possible to suppress the warning by wrapping the LHS into
7177 if ((!a) > 1) @{ @dots{} @}
7180 This warning is enabled by @option{-Wall}.
7182 @item -Waggregate-return
7183 @opindex Waggregate-return
7184 @opindex Wno-aggregate-return
7185 Warn if any functions that return structures or unions are defined or
7186 called. (In languages where you can return an array, this also elicits
7189 @item -Wno-aggressive-loop-optimizations
7190 @opindex Wno-aggressive-loop-optimizations
7191 @opindex Waggressive-loop-optimizations
7192 Warn if in a loop with constant number of iterations the compiler detects
7193 undefined behavior in some statement during one or more of the iterations.
7195 @item -Wno-attributes
7196 @opindex Wno-attributes
7197 @opindex Wattributes
7198 Do not warn if an unexpected @code{__attribute__} is used, such as
7199 unrecognized attributes, function attributes applied to variables,
7200 etc. This does not stop errors for incorrect use of supported
7203 @item -Wno-builtin-declaration-mismatch
7204 @opindex Wno-builtin-declaration-mismatch
7205 @opindex Wbuiltin-declaration-mismatch
7206 Warn if a built-in function is declared with an incompatible signature
7207 or as a non-function, or when a built-in function declared with a type
7208 that does not include a prototype is called with arguments whose promoted
7209 types do not match those expected by the function. When @option{-Wextra}
7210 is specified, also warn when a built-in function that takes arguments is
7211 declared without a prototype. The @option{-Wno-builtin-declaration-mismatch}
7212 warning is enabled by default. To avoid the warning include the appropriate
7213 header to bring the prototypes of built-in functions into scope.
7215 For example, the call to @code{memset} below is diagnosed by the warning
7216 because the function expects a value of type @code{size_t} as its argument
7217 but the type of @code{32} is @code{int}. With @option{-Wextra},
7218 the declaration of the function is diagnosed as well.
7220 extern void* memset ();
7223 memset (d, '\0', 32);
7227 @item -Wno-builtin-macro-redefined
7228 @opindex Wno-builtin-macro-redefined
7229 @opindex Wbuiltin-macro-redefined
7230 Do not warn if certain built-in macros are redefined. This suppresses
7231 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
7232 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
7234 @item -Wstrict-prototypes @r{(C and Objective-C only)}
7235 @opindex Wstrict-prototypes
7236 @opindex Wno-strict-prototypes
7237 Warn if a function is declared or defined without specifying the
7238 argument types. (An old-style function definition is permitted without
7239 a warning if preceded by a declaration that specifies the argument
7242 @item -Wold-style-declaration @r{(C and Objective-C only)}
7243 @opindex Wold-style-declaration
7244 @opindex Wno-old-style-declaration
7245 Warn for obsolescent usages, according to the C Standard, in a
7246 declaration. For example, warn if storage-class specifiers like
7247 @code{static} are not the first things in a declaration. This warning
7248 is also enabled by @option{-Wextra}.
7250 @item -Wold-style-definition @r{(C and Objective-C only)}
7251 @opindex Wold-style-definition
7252 @opindex Wno-old-style-definition
7253 Warn if an old-style function definition is used. A warning is given
7254 even if there is a previous prototype.
7256 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
7257 @opindex Wmissing-parameter-type
7258 @opindex Wno-missing-parameter-type
7259 A function parameter is declared without a type specifier in K&R-style
7266 This warning is also enabled by @option{-Wextra}.
7268 @item -Wmissing-prototypes @r{(C and Objective-C only)}
7269 @opindex Wmissing-prototypes
7270 @opindex Wno-missing-prototypes
7271 Warn if a global function is defined without a previous prototype
7272 declaration. This warning is issued even if the definition itself
7273 provides a prototype. Use this option to detect global functions
7274 that do not have a matching prototype declaration in a header file.
7275 This option is not valid for C++ because all function declarations
7276 provide prototypes and a non-matching declaration declares an
7277 overload rather than conflict with an earlier declaration.
7278 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
7280 @item -Wmissing-declarations
7281 @opindex Wmissing-declarations
7282 @opindex Wno-missing-declarations
7283 Warn if a global function is defined without a previous declaration.
7284 Do so even if the definition itself provides a prototype.
7285 Use this option to detect global functions that are not declared in
7286 header files. In C, no warnings are issued for functions with previous
7287 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
7288 missing prototypes. In C++, no warnings are issued for function templates,
7289 or for inline functions, or for functions in anonymous namespaces.
7291 @item -Wmissing-field-initializers
7292 @opindex Wmissing-field-initializers
7293 @opindex Wno-missing-field-initializers
7297 Warn if a structure's initializer has some fields missing. For
7298 example, the following code causes such a warning, because
7299 @code{x.h} is implicitly zero:
7302 struct s @{ int f, g, h; @};
7303 struct s x = @{ 3, 4 @};
7306 This option does not warn about designated initializers, so the following
7307 modification does not trigger a warning:
7310 struct s @{ int f, g, h; @};
7311 struct s x = @{ .f = 3, .g = 4 @};
7314 In C this option does not warn about the universal zero initializer
7318 struct s @{ int f, g, h; @};
7319 struct s x = @{ 0 @};
7322 Likewise, in C++ this option does not warn about the empty @{ @}
7323 initializer, for example:
7326 struct s @{ int f, g, h; @};
7330 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
7331 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
7333 @item -Wno-multichar
7334 @opindex Wno-multichar
7336 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
7337 Usually they indicate a typo in the user's code, as they have
7338 implementation-defined values, and should not be used in portable code.
7340 @item -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]}
7341 @opindex Wnormalized=
7342 @opindex Wnormalized
7343 @opindex Wno-normalized
7346 @cindex character set, input normalization
7347 In ISO C and ISO C++, two identifiers are different if they are
7348 different sequences of characters. However, sometimes when characters
7349 outside the basic ASCII character set are used, you can have two
7350 different character sequences that look the same. To avoid confusion,
7351 the ISO 10646 standard sets out some @dfn{normalization rules} which
7352 when applied ensure that two sequences that look the same are turned into
7353 the same sequence. GCC can warn you if you are using identifiers that
7354 have not been normalized; this option controls that warning.
7356 There are four levels of warning supported by GCC@. The default is
7357 @option{-Wnormalized=nfc}, which warns about any identifier that is
7358 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
7359 recommended form for most uses. It is equivalent to
7360 @option{-Wnormalized}.
7362 Unfortunately, there are some characters allowed in identifiers by
7363 ISO C and ISO C++ that, when turned into NFC, are not allowed in
7364 identifiers. That is, there's no way to use these symbols in portable
7365 ISO C or C++ and have all your identifiers in NFC@.
7366 @option{-Wnormalized=id} suppresses the warning for these characters.
7367 It is hoped that future versions of the standards involved will correct
7368 this, which is why this option is not the default.
7370 You can switch the warning off for all characters by writing
7371 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
7372 only do this if you are using some other normalization scheme (like
7373 ``D''), because otherwise you can easily create bugs that are
7374 literally impossible to see.
7376 Some characters in ISO 10646 have distinct meanings but look identical
7377 in some fonts or display methodologies, especially once formatting has
7378 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
7379 LETTER N'', displays just like a regular @code{n} that has been
7380 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
7381 normalization scheme to convert all these into a standard form as
7382 well, and GCC warns if your code is not in NFKC if you use
7383 @option{-Wnormalized=nfkc}. This warning is comparable to warning
7384 about every identifier that contains the letter O because it might be
7385 confused with the digit 0, and so is not the default, but may be
7386 useful as a local coding convention if the programming environment
7387 cannot be fixed to display these characters distinctly.
7389 @item -Wno-attribute-warning
7390 @opindex Wno-attribute-warning
7391 @opindex Wattribute-warning
7392 Do not warn about usage of functions (@pxref{Function Attributes})
7393 declared with @code{warning} attribute. By default, this warning is
7394 enabled. @option{-Wno-attribute-warning} can be used to disable the
7395 warning or @option{-Wno-error=attribute-warning} can be used to
7396 disable the error when compiled with @option{-Werror} flag.
7398 @item -Wno-deprecated
7399 @opindex Wno-deprecated
7400 @opindex Wdeprecated
7401 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
7403 @item -Wno-deprecated-declarations
7404 @opindex Wno-deprecated-declarations
7405 @opindex Wdeprecated-declarations
7406 Do not warn about uses of functions (@pxref{Function Attributes}),
7407 variables (@pxref{Variable Attributes}), and types (@pxref{Type
7408 Attributes}) marked as deprecated by using the @code{deprecated}
7412 @opindex Wno-overflow
7414 Do not warn about compile-time overflow in constant expressions.
7419 Warn about One Definition Rule violations during link-time optimization.
7423 @opindex Wopenmp-simd
7424 @opindex Wno-openmp-simd
7425 Warn if the vectorizer cost model overrides the OpenMP
7426 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
7427 option can be used to relax the cost model.
7429 @item -Woverride-init @r{(C and Objective-C only)}
7430 @opindex Woverride-init
7431 @opindex Wno-override-init
7435 Warn if an initialized field without side effects is overridden when
7436 using designated initializers (@pxref{Designated Inits, , Designated
7439 This warning is included in @option{-Wextra}. To get other
7440 @option{-Wextra} warnings without this one, use @option{-Wextra
7441 -Wno-override-init}.
7443 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
7444 @opindex Woverride-init-side-effects
7445 @opindex Wno-override-init-side-effects
7446 Warn if an initialized field with side effects is overridden when
7447 using designated initializers (@pxref{Designated Inits, , Designated
7448 Initializers}). This warning is enabled by default.
7453 Warn if a structure is given the packed attribute, but the packed
7454 attribute has no effect on the layout or size of the structure.
7455 Such structures may be mis-aligned for little benefit. For
7456 instance, in this code, the variable @code{f.x} in @code{struct bar}
7457 is misaligned even though @code{struct bar} does not itself
7458 have the packed attribute:
7465 @} __attribute__((packed));
7473 @item -Wpacked-bitfield-compat
7474 @opindex Wpacked-bitfield-compat
7475 @opindex Wno-packed-bitfield-compat
7476 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
7477 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
7478 the change can lead to differences in the structure layout. GCC
7479 informs you when the offset of such a field has changed in GCC 4.4.
7480 For example there is no longer a 4-bit padding between field @code{a}
7481 and @code{b} in this structure:
7488 @} __attribute__ ((packed));
7491 This warning is enabled by default. Use
7492 @option{-Wno-packed-bitfield-compat} to disable this warning.
7494 @item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
7495 @opindex Wpacked-not-aligned
7496 @opindex Wno-packed-not-aligned
7497 Warn if a structure field with explicitly specified alignment in a
7498 packed struct or union is misaligned. For example, a warning will
7499 be issued on @code{struct S}, like, @code{warning: alignment 1 of
7500 'struct S' is less than 8}, in this code:
7504 struct __attribute__ ((aligned (8))) S8 @{ char a[8]; @};
7505 struct __attribute__ ((packed)) S @{
7511 This warning is enabled by @option{-Wall}.
7516 Warn if padding is included in a structure, either to align an element
7517 of the structure or to align the whole structure. Sometimes when this
7518 happens it is possible to rearrange the fields of the structure to
7519 reduce the padding and so make the structure smaller.
7521 @item -Wredundant-decls
7522 @opindex Wredundant-decls
7523 @opindex Wno-redundant-decls
7524 Warn if anything is declared more than once in the same scope, even in
7525 cases where multiple declaration is valid and changes nothing.
7529 @opindex Wno-restrict
7530 Warn when an object referenced by a @code{restrict}-qualified parameter
7531 (or, in C++, a @code{__restrict}-qualified parameter) is aliased by another
7532 argument, or when copies between such objects overlap. For example,
7533 the call to the @code{strcpy} function below attempts to truncate the string
7534 by replacing its initial characters with the last four. However, because
7535 the call writes the terminating NUL into @code{a[4]}, the copies overlap and
7536 the call is diagnosed.
7541 char a[] = "abcd1234";
7546 The @option{-Wrestrict} option detects some instances of simple overlap
7547 even without optimization but works best at @option{-O2} and above. It
7548 is included in @option{-Wall}.
7550 @item -Wnested-externs @r{(C and Objective-C only)}
7551 @opindex Wnested-externs
7552 @opindex Wno-nested-externs
7553 Warn if an @code{extern} declaration is encountered within a function.
7555 @item -Wno-inherited-variadic-ctor
7556 @opindex Winherited-variadic-ctor
7557 @opindex Wno-inherited-variadic-ctor
7558 Suppress warnings about use of C++11 inheriting constructors when the
7559 base class inherited from has a C variadic constructor; the warning is
7560 on by default because the ellipsis is not inherited.
7565 Warn if a function that is declared as inline cannot be inlined.
7566 Even with this option, the compiler does not warn about failures to
7567 inline functions declared in system headers.
7569 The compiler uses a variety of heuristics to determine whether or not
7570 to inline a function. For example, the compiler takes into account
7571 the size of the function being inlined and the amount of inlining
7572 that has already been done in the current function. Therefore,
7573 seemingly insignificant changes in the source program can cause the
7574 warnings produced by @option{-Winline} to appear or disappear.
7576 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
7577 @opindex Wno-invalid-offsetof
7578 @opindex Winvalid-offsetof
7579 Suppress warnings from applying the @code{offsetof} macro to a non-POD
7580 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
7581 to a non-standard-layout type is undefined. In existing C++ implementations,
7582 however, @code{offsetof} typically gives meaningful results.
7583 This flag is for users who are aware that they are
7584 writing nonportable code and who have deliberately chosen to ignore the
7587 The restrictions on @code{offsetof} may be relaxed in a future version
7588 of the C++ standard.
7590 @item -Wint-in-bool-context
7591 @opindex Wint-in-bool-context
7592 @opindex Wno-int-in-bool-context
7593 Warn for suspicious use of integer values where boolean values are expected,
7594 such as conditional expressions (?:) using non-boolean integer constants in
7595 boolean context, like @code{if (a <= b ? 2 : 3)}. Or left shifting of signed
7596 integers in boolean context, like @code{for (a = 0; 1 << a; a++);}. Likewise
7597 for all kinds of multiplications regardless of the data type.
7598 This warning is enabled by @option{-Wall}.
7600 @item -Wno-int-to-pointer-cast
7601 @opindex Wno-int-to-pointer-cast
7602 @opindex Wint-to-pointer-cast
7603 Suppress warnings from casts to pointer type of an integer of a
7604 different size. In C++, casting to a pointer type of smaller size is
7605 an error. @option{Wint-to-pointer-cast} is enabled by default.
7608 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
7609 @opindex Wno-pointer-to-int-cast
7610 @opindex Wpointer-to-int-cast
7611 Suppress warnings from casts from a pointer to an integer type of a
7615 @opindex Winvalid-pch
7616 @opindex Wno-invalid-pch
7617 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
7618 the search path but cannot be used.
7622 @opindex Wno-long-long
7623 Warn if @code{long long} type is used. This is enabled by either
7624 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
7625 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
7627 @item -Wvariadic-macros
7628 @opindex Wvariadic-macros
7629 @opindex Wno-variadic-macros
7630 Warn if variadic macros are used in ISO C90 mode, or if the GNU
7631 alternate syntax is used in ISO C99 mode. This is enabled by either
7632 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
7633 messages, use @option{-Wno-variadic-macros}.
7637 @opindex Wno-varargs
7638 Warn upon questionable usage of the macros used to handle variable
7639 arguments like @code{va_start}. This is default. To inhibit the
7640 warning messages, use @option{-Wno-varargs}.
7642 @item -Wvector-operation-performance
7643 @opindex Wvector-operation-performance
7644 @opindex Wno-vector-operation-performance
7645 Warn if vector operation is not implemented via SIMD capabilities of the
7646 architecture. Mainly useful for the performance tuning.
7647 Vector operation can be implemented @code{piecewise}, which means that the
7648 scalar operation is performed on every vector element;
7649 @code{in parallel}, which means that the vector operation is implemented
7650 using scalars of wider type, which normally is more performance efficient;
7651 and @code{as a single scalar}, which means that vector fits into a
7654 @item -Wno-virtual-move-assign
7655 @opindex Wvirtual-move-assign
7656 @opindex Wno-virtual-move-assign
7657 Suppress warnings about inheriting from a virtual base with a
7658 non-trivial C++11 move assignment operator. This is dangerous because
7659 if the virtual base is reachable along more than one path, it is
7660 moved multiple times, which can mean both objects end up in the
7661 moved-from state. If the move assignment operator is written to avoid
7662 moving from a moved-from object, this warning can be disabled.
7667 Warn if a variable-length array is used in the code.
7668 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
7669 the variable-length array.
7671 @item -Wvla-larger-than=@var{byte-size}
7672 @opindex Wvla-larger-than=
7673 @opindex Wno-vla-larger-than
7674 If this option is used, the compiler will warn for declarations of
7675 variable-length arrays whose size is either unbounded, or bounded
7676 by an argument that allows the array size to exceed @var{byte-size}
7677 bytes. This is similar to how @option{-Walloca-larger-than=}@var{byte-size}
7678 works, but with variable-length arrays.
7680 Note that GCC may optimize small variable-length arrays of a known
7681 value into plain arrays, so this warning may not get triggered for
7684 @option{-Wvla-larger-than=}@samp{PTRDIFF_MAX} is enabled by default but
7685 is typically only effective when @option{-ftree-vrp} is active (default
7686 for @option{-O2} and above).
7688 See also @option{-Walloca-larger-than=@var{byte-size}}.
7690 @item -Wno-vla-larger-than
7691 @opindex Wno-vla-larger-than
7692 Disable @option{-Wvla-larger-than=} warnings. The option is equivalent
7693 to @option{-Wvla-larger-than=}@samp{SIZE_MAX} or larger.
7695 @item -Wvolatile-register-var
7696 @opindex Wvolatile-register-var
7697 @opindex Wno-volatile-register-var
7698 Warn if a register variable is declared volatile. The volatile
7699 modifier does not inhibit all optimizations that may eliminate reads
7700 and/or writes to register variables. This warning is enabled by
7703 @item -Wdisabled-optimization
7704 @opindex Wdisabled-optimization
7705 @opindex Wno-disabled-optimization
7706 Warn if a requested optimization pass is disabled. This warning does
7707 not generally indicate that there is anything wrong with your code; it
7708 merely indicates that GCC's optimizers are unable to handle the code
7709 effectively. Often, the problem is that your code is too big or too
7710 complex; GCC refuses to optimize programs when the optimization
7711 itself is likely to take inordinate amounts of time.
7713 @item -Wpointer-sign @r{(C and Objective-C only)}
7714 @opindex Wpointer-sign
7715 @opindex Wno-pointer-sign
7716 Warn for pointer argument passing or assignment with different signedness.
7717 This option is only supported for C and Objective-C@. It is implied by
7718 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
7719 @option{-Wno-pointer-sign}.
7721 @item -Wstack-protector
7722 @opindex Wstack-protector
7723 @opindex Wno-stack-protector
7724 This option is only active when @option{-fstack-protector} is active. It
7725 warns about functions that are not protected against stack smashing.
7727 @item -Woverlength-strings
7728 @opindex Woverlength-strings
7729 @opindex Wno-overlength-strings
7730 Warn about string constants that are longer than the ``minimum
7731 maximum'' length specified in the C standard. Modern compilers
7732 generally allow string constants that are much longer than the
7733 standard's minimum limit, but very portable programs should avoid
7734 using longer strings.
7736 The limit applies @emph{after} string constant concatenation, and does
7737 not count the trailing NUL@. In C90, the limit was 509 characters; in
7738 C99, it was raised to 4095. C++98 does not specify a normative
7739 minimum maximum, so we do not diagnose overlength strings in C++@.
7741 This option is implied by @option{-Wpedantic}, and can be disabled with
7742 @option{-Wno-overlength-strings}.
7744 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
7745 @opindex Wunsuffixed-float-constants
7746 @opindex Wno-unsuffixed-float-constants
7748 Issue a warning for any floating constant that does not have
7749 a suffix. When used together with @option{-Wsystem-headers} it
7750 warns about such constants in system header files. This can be useful
7751 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
7752 from the decimal floating-point extension to C99.
7754 @item -Wno-designated-init @r{(C and Objective-C only)}
7755 Suppress warnings when a positional initializer is used to initialize
7756 a structure that has been marked with the @code{designated_init}
7760 Issue a warning when HSAIL cannot be emitted for the compiled function or
7765 @node Debugging Options
7766 @section Options for Debugging Your Program
7767 @cindex options, debugging
7768 @cindex debugging information options
7770 To tell GCC to emit extra information for use by a debugger, in almost
7771 all cases you need only to add @option{-g} to your other options.
7773 GCC allows you to use @option{-g} with
7774 @option{-O}. The shortcuts taken by optimized code may occasionally
7775 be surprising: some variables you declared may not exist
7776 at all; flow of control may briefly move where you did not expect it;
7777 some statements may not be executed because they compute constant
7778 results or their values are already at hand; some statements may
7779 execute in different places because they have been moved out of loops.
7780 Nevertheless it is possible to debug optimized output. This makes
7781 it reasonable to use the optimizer for programs that might have bugs.
7783 If you are not using some other optimization option, consider
7784 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
7785 With no @option{-O} option at all, some compiler passes that collect
7786 information useful for debugging do not run at all, so that
7787 @option{-Og} may result in a better debugging experience.
7792 Produce debugging information in the operating system's native format
7793 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
7796 On most systems that use stabs format, @option{-g} enables use of extra
7797 debugging information that only GDB can use; this extra information
7798 makes debugging work better in GDB but probably makes other debuggers
7800 refuse to read the program. If you want to control for certain whether
7801 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
7802 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
7806 Produce debugging information for use by GDB@. This means to use the
7807 most expressive format available (DWARF, stabs, or the native format
7808 if neither of those are supported), including GDB extensions if at all
7812 @itemx -gdwarf-@var{version}
7814 Produce debugging information in DWARF format (if that is supported).
7815 The value of @var{version} may be either 2, 3, 4 or 5; the default version
7816 for most targets is 4. DWARF Version 5 is only experimental.
7818 Note that with DWARF Version 2, some ports require and always
7819 use some non-conflicting DWARF 3 extensions in the unwind tables.
7821 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
7822 for maximum benefit.
7824 GCC no longer supports DWARF Version 1, which is substantially
7825 different than Version 2 and later. For historical reasons, some
7826 other DWARF-related options such as
7827 @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
7828 in their names, but apply to all currently-supported versions of DWARF.
7832 Produce debugging information in stabs format (if that is supported),
7833 without GDB extensions. This is the format used by DBX on most BSD
7834 systems. On MIPS, Alpha and System V Release 4 systems this option
7835 produces stabs debugging output that is not understood by DBX@.
7836 On System V Release 4 systems this option requires the GNU assembler.
7840 Produce debugging information in stabs format (if that is supported),
7841 using GNU extensions understood only by the GNU debugger (GDB)@. The
7842 use of these extensions is likely to make other debuggers crash or
7843 refuse to read the program.
7847 Produce debugging information in XCOFF format (if that is supported).
7848 This is the format used by the DBX debugger on IBM RS/6000 systems.
7852 Produce debugging information in XCOFF format (if that is supported),
7853 using GNU extensions understood only by the GNU debugger (GDB)@. The
7854 use of these extensions is likely to make other debuggers crash or
7855 refuse to read the program, and may cause assemblers other than the GNU
7856 assembler (GAS) to fail with an error.
7860 Produce debugging information in Alpha/VMS debug format (if that is
7861 supported). This is the format used by DEBUG on Alpha/VMS systems.
7864 @itemx -ggdb@var{level}
7865 @itemx -gstabs@var{level}
7866 @itemx -gxcoff@var{level}
7867 @itemx -gvms@var{level}
7868 Request debugging information and also use @var{level} to specify how
7869 much information. The default level is 2.
7871 Level 0 produces no debug information at all. Thus, @option{-g0} negates
7874 Level 1 produces minimal information, enough for making backtraces in
7875 parts of the program that you don't plan to debug. This includes
7876 descriptions of functions and external variables, and line number
7877 tables, but no information about local variables.
7879 Level 3 includes extra information, such as all the macro definitions
7880 present in the program. Some debuggers support macro expansion when
7881 you use @option{-g3}.
7883 If you use multiple @option{-g} options, with or without level numbers,
7884 the last such option is the one that is effective.
7886 @option{-gdwarf} does not accept a concatenated debug level, to avoid
7887 confusion with @option{-gdwarf-@var{level}}.
7888 Instead use an additional @option{-g@var{level}} option to change the
7889 debug level for DWARF.
7891 @item -fno-eliminate-unused-debug-symbols
7892 @opindex feliminate-unused-debug-symbols
7893 @opindex fno-eliminate-unused-debug-symbols
7894 By default, no debug information is produced for symbols that are not actually
7895 used. Use this option if you want debug information for all symbols.
7897 @item -femit-class-debug-always
7898 @opindex femit-class-debug-always
7899 Instead of emitting debugging information for a C++ class in only one
7900 object file, emit it in all object files using the class. This option
7901 should be used only with debuggers that are unable to handle the way GCC
7902 normally emits debugging information for classes because using this
7903 option increases the size of debugging information by as much as a
7906 @item -fno-merge-debug-strings
7907 @opindex fmerge-debug-strings
7908 @opindex fno-merge-debug-strings
7909 Direct the linker to not merge together strings in the debugging
7910 information that are identical in different object files. Merging is
7911 not supported by all assemblers or linkers. Merging decreases the size
7912 of the debug information in the output file at the cost of increasing
7913 link processing time. Merging is enabled by default.
7915 @item -fdebug-prefix-map=@var{old}=@var{new}
7916 @opindex fdebug-prefix-map
7917 When compiling files residing in directory @file{@var{old}}, record
7918 debugging information describing them as if the files resided in
7919 directory @file{@var{new}} instead. This can be used to replace a
7920 build-time path with an install-time path in the debug info. It can
7921 also be used to change an absolute path to a relative path by using
7922 @file{.} for @var{new}. This can give more reproducible builds, which
7923 are location independent, but may require an extra command to tell GDB
7924 where to find the source files. See also @option{-ffile-prefix-map}.
7926 @item -fvar-tracking
7927 @opindex fvar-tracking
7928 Run variable tracking pass. It computes where variables are stored at each
7929 position in code. Better debugging information is then generated
7930 (if the debugging information format supports this information).
7932 It is enabled by default when compiling with optimization (@option{-Os},
7933 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7934 the debug info format supports it.
7936 @item -fvar-tracking-assignments
7937 @opindex fvar-tracking-assignments
7938 @opindex fno-var-tracking-assignments
7939 Annotate assignments to user variables early in the compilation and
7940 attempt to carry the annotations over throughout the compilation all the
7941 way to the end, in an attempt to improve debug information while
7942 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7944 It can be enabled even if var-tracking is disabled, in which case
7945 annotations are created and maintained, but discarded at the end.
7946 By default, this flag is enabled together with @option{-fvar-tracking},
7947 except when selective scheduling is enabled.
7950 @opindex gsplit-dwarf
7951 Separate as much DWARF debugging information as possible into a
7952 separate output file with the extension @file{.dwo}. This option allows
7953 the build system to avoid linking files with debug information. To
7954 be useful, this option requires a debugger capable of reading @file{.dwo}
7957 @item -gdescribe-dies
7958 @opindex gdescribe-dies
7959 Add description attributes to some DWARF DIEs that have no name attribute,
7960 such as artificial variables, external references and call site
7965 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
7967 @item -ggnu-pubnames
7968 @opindex ggnu-pubnames
7969 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
7970 suitable for conversion into a GDB@ index. This option is only useful
7971 with a linker that can produce GDB@ index version 7.
7973 @item -fdebug-types-section
7974 @opindex fdebug-types-section
7975 @opindex fno-debug-types-section
7976 When using DWARF Version 4 or higher, type DIEs can be put into
7977 their own @code{.debug_types} section instead of making them part of the
7978 @code{.debug_info} section. It is more efficient to put them in a separate
7979 comdat section since the linker can then remove duplicates.
7980 But not all DWARF consumers support @code{.debug_types} sections yet
7981 and on some objects @code{.debug_types} produces larger instead of smaller
7982 debugging information.
7984 @item -grecord-gcc-switches
7985 @itemx -gno-record-gcc-switches
7986 @opindex grecord-gcc-switches
7987 @opindex gno-record-gcc-switches
7988 This switch causes the command-line options used to invoke the
7989 compiler that may affect code generation to be appended to the
7990 DW_AT_producer attribute in DWARF debugging information. The options
7991 are concatenated with spaces separating them from each other and from
7992 the compiler version.
7993 It is enabled by default.
7994 See also @option{-frecord-gcc-switches} for another
7995 way of storing compiler options into the object file.
7997 @item -gstrict-dwarf
7998 @opindex gstrict-dwarf
7999 Disallow using extensions of later DWARF standard version than selected
8000 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
8001 DWARF extensions from later standard versions is allowed.
8003 @item -gno-strict-dwarf
8004 @opindex gno-strict-dwarf
8005 Allow using extensions of later DWARF standard version than selected with
8006 @option{-gdwarf-@var{version}}.
8008 @item -gas-loc-support
8009 @opindex gas-loc-support
8010 Inform the compiler that the assembler supports @code{.loc} directives.
8011 It may then use them for the assembler to generate DWARF2+ line number
8014 This is generally desirable, because assembler-generated line-number
8015 tables are a lot more compact than those the compiler can generate
8018 This option will be enabled by default if, at GCC configure time, the
8019 assembler was found to support such directives.
8021 @item -gno-as-loc-support
8022 @opindex gno-as-loc-support
8023 Force GCC to generate DWARF2+ line number tables internally, if DWARF2+
8024 line number tables are to be generated.
8026 @item -gas-locview-support
8027 @opindex gas-locview-support
8028 Inform the compiler that the assembler supports @code{view} assignment
8029 and reset assertion checking in @code{.loc} directives.
8031 This option will be enabled by default if, at GCC configure time, the
8032 assembler was found to support them.
8034 @item -gno-as-locview-support
8035 Force GCC to assign view numbers internally, if
8036 @option{-gvariable-location-views} are explicitly requested.
8039 @itemx -gno-column-info
8040 @opindex gcolumn-info
8041 @opindex gno-column-info
8042 Emit location column information into DWARF debugging information, rather
8043 than just file and line.
8044 This option is enabled by default.
8046 @item -gstatement-frontiers
8047 @itemx -gno-statement-frontiers
8048 @opindex gstatement-frontiers
8049 @opindex gno-statement-frontiers
8050 This option causes GCC to create markers in the internal representation
8051 at the beginning of statements, and to keep them roughly in place
8052 throughout compilation, using them to guide the output of @code{is_stmt}
8053 markers in the line number table. This is enabled by default when
8054 compiling with optimization (@option{-Os}, @option{-O}, @option{-O2},
8055 @dots{}), and outputting DWARF 2 debug information at the normal level.
8057 @item -gvariable-location-views
8058 @itemx -gvariable-location-views=incompat5
8059 @itemx -gno-variable-location-views
8060 @opindex gvariable-location-views
8061 @opindex gvariable-location-views=incompat5
8062 @opindex gno-variable-location-views
8063 Augment variable location lists with progressive view numbers implied
8064 from the line number table. This enables debug information consumers to
8065 inspect state at certain points of the program, even if no instructions
8066 associated with the corresponding source locations are present at that
8067 point. If the assembler lacks support for view numbers in line number
8068 tables, this will cause the compiler to emit the line number table,
8069 which generally makes them somewhat less compact. The augmented line
8070 number tables and location lists are fully backward-compatible, so they
8071 can be consumed by debug information consumers that are not aware of
8072 these augmentations, but they won't derive any benefit from them either.
8074 This is enabled by default when outputting DWARF 2 debug information at
8075 the normal level, as long as there is assembler support,
8076 @option{-fvar-tracking-assignments} is enabled and
8077 @option{-gstrict-dwarf} is not. When assembler support is not
8078 available, this may still be enabled, but it will force GCC to output
8079 internal line number tables, and if
8080 @option{-ginternal-reset-location-views} is not enabled, that will most
8081 certainly lead to silently mismatching location views.
8083 There is a proposed representation for view numbers that is not backward
8084 compatible with the location list format introduced in DWARF 5, that can
8085 be enabled with @option{-gvariable-location-views=incompat5}. This
8086 option may be removed in the future, is only provided as a reference
8087 implementation of the proposed representation. Debug information
8088 consumers are not expected to support this extended format, and they
8089 would be rendered unable to decode location lists using it.
8091 @item -ginternal-reset-location-views
8092 @itemx -gnointernal-reset-location-views
8093 @opindex ginternal-reset-location-views
8094 @opindex gno-internal-reset-location-views
8095 Attempt to determine location views that can be omitted from location
8096 view lists. This requires the compiler to have very accurate insn
8097 length estimates, which isn't always the case, and it may cause
8098 incorrect view lists to be generated silently when using an assembler
8099 that does not support location view lists. The GNU assembler will flag
8100 any such error as a @code{view number mismatch}. This is only enabled
8101 on ports that define a reliable estimation function.
8103 @item -ginline-points
8104 @itemx -gno-inline-points
8105 @opindex ginline-points
8106 @opindex gno-inline-points
8107 Generate extended debug information for inlined functions. Location
8108 view tracking markers are inserted at inlined entry points, so that
8109 address and view numbers can be computed and output in debug
8110 information. This can be enabled independently of location views, in
8111 which case the view numbers won't be output, but it can only be enabled
8112 along with statement frontiers, and it is only enabled by default if
8113 location views are enabled.
8115 @item -gz@r{[}=@var{type}@r{]}
8117 Produce compressed debug sections in DWARF format, if that is supported.
8118 If @var{type} is not given, the default type depends on the capabilities
8119 of the assembler and linker used. @var{type} may be one of
8120 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
8121 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
8122 compression in traditional GNU format). If the linker doesn't support
8123 writing compressed debug sections, the option is rejected. Otherwise,
8124 if the assembler does not support them, @option{-gz} is silently ignored
8125 when producing object files.
8127 @item -femit-struct-debug-baseonly
8128 @opindex femit-struct-debug-baseonly
8129 Emit debug information for struct-like types
8130 only when the base name of the compilation source file
8131 matches the base name of file in which the struct is defined.
8133 This option substantially reduces the size of debugging information,
8134 but at significant potential loss in type information to the debugger.
8135 See @option{-femit-struct-debug-reduced} for a less aggressive option.
8136 See @option{-femit-struct-debug-detailed} for more detailed control.
8138 This option works only with DWARF debug output.
8140 @item -femit-struct-debug-reduced
8141 @opindex femit-struct-debug-reduced
8142 Emit debug information for struct-like types
8143 only when the base name of the compilation source file
8144 matches the base name of file in which the type is defined,
8145 unless the struct is a template or defined in a system header.
8147 This option significantly reduces the size of debugging information,
8148 with some potential loss in type information to the debugger.
8149 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
8150 See @option{-femit-struct-debug-detailed} for more detailed control.
8152 This option works only with DWARF debug output.
8154 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
8155 @opindex femit-struct-debug-detailed
8156 Specify the struct-like types
8157 for which the compiler generates debug information.
8158 The intent is to reduce duplicate struct debug information
8159 between different object files within the same program.
8161 This option is a detailed version of
8162 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
8163 which serves for most needs.
8165 A specification has the syntax@*
8166 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
8168 The optional first word limits the specification to
8169 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
8170 A struct type is used directly when it is the type of a variable, member.
8171 Indirect uses arise through pointers to structs.
8172 That is, when use of an incomplete struct is valid, the use is indirect.
8174 @samp{struct one direct; struct two * indirect;}.
8176 The optional second word limits the specification to
8177 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
8178 Generic structs are a bit complicated to explain.
8179 For C++, these are non-explicit specializations of template classes,
8180 or non-template classes within the above.
8181 Other programming languages have generics,
8182 but @option{-femit-struct-debug-detailed} does not yet implement them.
8184 The third word specifies the source files for those
8185 structs for which the compiler should emit debug information.
8186 The values @samp{none} and @samp{any} have the normal meaning.
8187 The value @samp{base} means that
8188 the base of name of the file in which the type declaration appears
8189 must match the base of the name of the main compilation file.
8190 In practice, this means that when compiling @file{foo.c}, debug information
8191 is generated for types declared in that file and @file{foo.h},
8192 but not other header files.
8193 The value @samp{sys} means those types satisfying @samp{base}
8194 or declared in system or compiler headers.
8196 You may need to experiment to determine the best settings for your application.
8198 The default is @option{-femit-struct-debug-detailed=all}.
8200 This option works only with DWARF debug output.
8202 @item -fno-dwarf2-cfi-asm
8203 @opindex fdwarf2-cfi-asm
8204 @opindex fno-dwarf2-cfi-asm
8205 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
8206 instead of using GAS @code{.cfi_*} directives.
8208 @item -fno-eliminate-unused-debug-types
8209 @opindex feliminate-unused-debug-types
8210 @opindex fno-eliminate-unused-debug-types
8211 Normally, when producing DWARF output, GCC avoids producing debug symbol
8212 output for types that are nowhere used in the source file being compiled.
8213 Sometimes it is useful to have GCC emit debugging
8214 information for all types declared in a compilation
8215 unit, regardless of whether or not they are actually used
8216 in that compilation unit, for example
8217 if, in the debugger, you want to cast a value to a type that is
8218 not actually used in your program (but is declared). More often,
8219 however, this results in a significant amount of wasted space.
8222 @node Optimize Options
8223 @section Options That Control Optimization
8224 @cindex optimize options
8225 @cindex options, optimization
8227 These options control various sorts of optimizations.
8229 Without any optimization option, the compiler's goal is to reduce the
8230 cost of compilation and to make debugging produce the expected
8231 results. Statements are independent: if you stop the program with a
8232 breakpoint between statements, you can then assign a new value to any
8233 variable or change the program counter to any other statement in the
8234 function and get exactly the results you expect from the source
8237 Turning on optimization flags makes the compiler attempt to improve
8238 the performance and/or code size at the expense of compilation time
8239 and possibly the ability to debug the program.
8241 The compiler performs optimization based on the knowledge it has of the
8242 program. Compiling multiple files at once to a single output file mode allows
8243 the compiler to use information gained from all of the files when compiling
8246 Not all optimizations are controlled directly by a flag. Only
8247 optimizations that have a flag are listed in this section.
8249 Most optimizations are completely disabled at @option{-O0} or if an
8250 @option{-O} level is not set on the command line, even if individual
8251 optimization flags are specified. Similarly, @option{-Og} suppresses
8252 many optimization passes.
8254 Depending on the target and how GCC was configured, a slightly different
8255 set of optimizations may be enabled at each @option{-O} level than
8256 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
8257 to find out the exact set of optimizations that are enabled at each level.
8258 @xref{Overall Options}, for examples.
8265 Optimize. Optimizing compilation takes somewhat more time, and a lot
8266 more memory for a large function.
8268 With @option{-O}, the compiler tries to reduce code size and execution
8269 time, without performing any optimizations that take a great deal of
8272 @c Note that in addition to the default_options_table list in opts.c,
8273 @c several optimization flags default to true but control optimization
8274 @c passes that are explicitly disabled at -O0.
8276 @option{-O} turns on the following optimization flags:
8278 @c Please keep the following list alphabetized.
8279 @gccoptlist{-fauto-inc-dec @gol
8280 -fbranch-count-reg @gol
8281 -fcombine-stack-adjustments @gol
8283 -fcprop-registers @gol
8286 -fdelayed-branch @gol
8288 -fforward-propagate @gol
8289 -fguess-branch-probability @gol
8290 -fif-conversion @gol
8291 -fif-conversion2 @gol
8292 -finline-functions-called-once @gol
8294 -fipa-pure-const @gol
8295 -fipa-reference @gol
8296 -fipa-reference-addressable @gol
8297 -fmerge-constants @gol
8298 -fmove-loop-invariants @gol
8299 -fomit-frame-pointer @gol
8300 -freorder-blocks @gol
8302 -fshrink-wrap-separate @gol
8303 -fsplit-wide-types @gol
8309 -ftree-coalesce-vars @gol
8310 -ftree-copy-prop @gol
8312 -ftree-dominator-opts @gol
8314 -ftree-forwprop @gol
8318 -ftree-scev-cprop @gol
8327 Optimize even more. GCC performs nearly all supported optimizations
8328 that do not involve a space-speed tradeoff.
8329 As compared to @option{-O}, this option increases both compilation time
8330 and the performance of the generated code.
8332 @option{-O2} turns on all optimization flags specified by @option{-O}. It
8333 also turns on the following optimization flags:
8335 @c Please keep the following list alphabetized!
8336 @gccoptlist{-falign-functions -falign-jumps @gol
8337 -falign-labels -falign-loops @gol
8339 -fcode-hoisting @gol
8341 -fcse-follow-jumps -fcse-skip-blocks @gol
8342 -fdelete-null-pointer-checks @gol
8343 -fdevirtualize -fdevirtualize-speculatively @gol
8344 -fexpensive-optimizations @gol
8346 -fgcse -fgcse-lm @gol
8347 -fhoist-adjacent-loads @gol
8348 -finline-small-functions @gol
8349 -findirect-inlining @gol
8350 -fipa-bit-cp -fipa-cp -fipa-icf @gol
8351 -fipa-ra -fipa-sra -fipa-vrp @gol
8352 -fisolate-erroneous-paths-dereference @gol
8354 -foptimize-sibling-calls @gol
8355 -foptimize-strlen @gol
8356 -fpartial-inlining @gol
8358 -freorder-blocks-algorithm=stc @gol
8359 -freorder-blocks-and-partition -freorder-functions @gol
8360 -frerun-cse-after-loop @gol
8361 -fschedule-insns -fschedule-insns2 @gol
8362 -fsched-interblock -fsched-spec @gol
8363 -fstore-merging @gol
8364 -fstrict-aliasing @gol
8366 -ftree-builtin-call-dce @gol
8368 -ftree-switch-conversion -ftree-tail-merge @gol
8371 Please note the warning under @option{-fgcse} about
8372 invoking @option{-O2} on programs that use computed gotos.
8376 Optimize yet more. @option{-O3} turns on all optimizations specified
8377 by @option{-O2} and also turns on the following optimization flags:
8379 @c Please keep the following list alphabetized!
8380 @gccoptlist{-fgcse-after-reload @gol
8381 -finline-functions @gol
8383 -floop-interchange @gol
8384 -floop-unroll-and-jam @gol
8386 -fpredictive-commoning @gol
8388 -ftree-loop-distribute-patterns @gol
8389 -ftree-loop-distribution @gol
8390 -ftree-loop-vectorize @gol
8391 -ftree-partial-pre @gol
8392 -ftree-slp-vectorize @gol
8393 -funswitch-loops @gol
8394 -fvect-cost-model @gol
8395 -fversion-loops-for-strides}
8399 Reduce compilation time and make debugging produce the expected
8400 results. This is the default.
8404 Optimize for size. @option{-Os} enables all @option{-O2} optimizations
8405 except those that often increase code size:
8407 @gccoptlist{-falign-functions -falign-jumps @gol
8408 -falign-labels -falign-loops @gol
8409 -fprefetch-loop-arrays -freorder-blocks-algorithm=stc}
8411 It also enables @option{-finline-functions}, causes the compiler to tune for
8412 code size rather than execution speed, and performs further optimizations
8413 designed to reduce code size.
8417 Disregard strict standards compliance. @option{-Ofast} enables all
8418 @option{-O3} optimizations. It also enables optimizations that are not
8419 valid for all standard-compliant programs.
8420 It turns on @option{-ffast-math} and the Fortran-specific
8421 @option{-fstack-arrays}, unless @option{-fmax-stack-var-size} is
8422 specified, and @option{-fno-protect-parens}.
8426 Optimize debugging experience. @option{-Og} should be the optimization
8427 level of choice for the standard edit-compile-debug cycle, offering
8428 a reasonable level of optimization while maintaining fast compilation
8429 and a good debugging experience. It is a better choice than @option{-O0}
8430 for producing debuggable code because some compiler passes
8431 that collect debug information are disabled at @option{-O0}.
8433 Like @option{-O0}, @option{-Og} completely disables a number of
8434 optimization passes so that individual options controlling them have
8435 no effect. Otherwise @option{-Og} enables all @option{-O1}
8436 optimization flags except for those that may interfere with debugging:
8438 @gccoptlist{-fbranch-count-reg -fdelayed-branch @gol
8439 -fdse -fif-conversion -fif-conversion2 @gol
8440 -finline-functions-called-once @gol
8441 -fmove-loop-invariants -fssa-phiopt @gol
8442 -ftree-bit-ccp -ftree-dse -ftree-pta -ftree-sra}
8446 If you use multiple @option{-O} options, with or without level numbers,
8447 the last such option is the one that is effective.
8449 Options of the form @option{-f@var{flag}} specify machine-independent
8450 flags. Most flags have both positive and negative forms; the negative
8451 form of @option{-ffoo} is @option{-fno-foo}. In the table
8452 below, only one of the forms is listed---the one you typically
8453 use. You can figure out the other form by either removing @samp{no-}
8456 The following options control specific optimizations. They are either
8457 activated by @option{-O} options or are related to ones that are. You
8458 can use the following flags in the rare cases when ``fine-tuning'' of
8459 optimizations to be performed is desired.
8462 @item -fno-defer-pop
8463 @opindex fno-defer-pop
8465 For machines that must pop arguments after a function call, always pop
8466 the arguments as soon as each function returns.
8467 At levels @option{-O1} and higher, @option{-fdefer-pop} is the default;
8468 this allows the compiler to let arguments accumulate on the stack for several
8469 function calls and pop them all at once.
8471 @item -fforward-propagate
8472 @opindex fforward-propagate
8473 Perform a forward propagation pass on RTL@. The pass tries to combine two
8474 instructions and checks if the result can be simplified. If loop unrolling
8475 is active, two passes are performed and the second is scheduled after
8478 This option is enabled by default at optimization levels @option{-O},
8479 @option{-O2}, @option{-O3}, @option{-Os}.
8481 @item -ffp-contract=@var{style}
8482 @opindex ffp-contract
8483 @option{-ffp-contract=off} disables floating-point expression contraction.
8484 @option{-ffp-contract=fast} enables floating-point expression contraction
8485 such as forming of fused multiply-add operations if the target has
8486 native support for them.
8487 @option{-ffp-contract=on} enables floating-point expression contraction
8488 if allowed by the language standard. This is currently not implemented
8489 and treated equal to @option{-ffp-contract=off}.
8491 The default is @option{-ffp-contract=fast}.
8493 @item -fomit-frame-pointer
8494 @opindex fomit-frame-pointer
8495 Omit the frame pointer in functions that don't need one. This avoids the
8496 instructions to save, set up and restore the frame pointer; on many targets
8497 it also makes an extra register available.
8499 On some targets this flag has no effect because the standard calling sequence
8500 always uses a frame pointer, so it cannot be omitted.
8502 Note that @option{-fno-omit-frame-pointer} doesn't guarantee the frame pointer
8503 is used in all functions. Several targets always omit the frame pointer in
8506 Enabled by default at @option{-O} and higher.
8508 @item -foptimize-sibling-calls
8509 @opindex foptimize-sibling-calls
8510 Optimize sibling and tail recursive calls.
8512 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8514 @item -foptimize-strlen
8515 @opindex foptimize-strlen
8516 Optimize various standard C string functions (e.g.@: @code{strlen},
8517 @code{strchr} or @code{strcpy}) and
8518 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8520 Enabled at levels @option{-O2}, @option{-O3}.
8525 Do not expand any functions inline apart from those marked with
8526 the @code{always_inline} attribute. This is the default when not
8529 Single functions can be exempted from inlining by marking them
8530 with the @code{noinline} attribute.
8532 @item -finline-small-functions
8533 @opindex finline-small-functions
8534 Integrate functions into their callers when their body is smaller than expected
8535 function call code (so overall size of program gets smaller). The compiler
8536 heuristically decides which functions are simple enough to be worth integrating
8537 in this way. This inlining applies to all functions, even those not declared
8540 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8542 @item -findirect-inlining
8543 @opindex findirect-inlining
8544 Inline also indirect calls that are discovered to be known at compile
8545 time thanks to previous inlining. This option has any effect only
8546 when inlining itself is turned on by the @option{-finline-functions}
8547 or @option{-finline-small-functions} options.
8549 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8551 @item -finline-functions
8552 @opindex finline-functions
8553 Consider all functions for inlining, even if they are not declared inline.
8554 The compiler heuristically decides which functions are worth integrating
8557 If all calls to a given function are integrated, and the function is
8558 declared @code{static}, then the function is normally not output as
8559 assembler code in its own right.
8561 Enabled at levels @option{-O3}, @option{-Os}. Also enabled
8562 by @option{-fprofile-use} and @option{-fauto-profile}.
8564 @item -finline-functions-called-once
8565 @opindex finline-functions-called-once
8566 Consider all @code{static} functions called once for inlining into their
8567 caller even if they are not marked @code{inline}. If a call to a given
8568 function is integrated, then the function is not output as assembler code
8571 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os},
8572 but not @option{-Og}.
8574 @item -fearly-inlining
8575 @opindex fearly-inlining
8576 Inline functions marked by @code{always_inline} and functions whose body seems
8577 smaller than the function call overhead early before doing
8578 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8579 makes profiling significantly cheaper and usually inlining faster on programs
8580 having large chains of nested wrapper functions.
8586 Perform interprocedural scalar replacement of aggregates, removal of
8587 unused parameters and replacement of parameters passed by reference
8588 by parameters passed by value.
8590 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8592 @item -finline-limit=@var{n}
8593 @opindex finline-limit
8594 By default, GCC limits the size of functions that can be inlined. This flag
8595 allows coarse control of this limit. @var{n} is the size of functions that
8596 can be inlined in number of pseudo instructions.
8598 Inlining is actually controlled by a number of parameters, which may be
8599 specified individually by using @option{--param @var{name}=@var{value}}.
8600 The @option{-finline-limit=@var{n}} option sets some of these parameters
8604 @item max-inline-insns-single
8605 is set to @var{n}/2.
8606 @item max-inline-insns-auto
8607 is set to @var{n}/2.
8610 See below for a documentation of the individual
8611 parameters controlling inlining and for the defaults of these parameters.
8613 @emph{Note:} there may be no value to @option{-finline-limit} that results
8614 in default behavior.
8616 @emph{Note:} pseudo instruction represents, in this particular context, an
8617 abstract measurement of function's size. In no way does it represent a count
8618 of assembly instructions and as such its exact meaning might change from one
8619 release to an another.
8621 @item -fno-keep-inline-dllexport
8622 @opindex fno-keep-inline-dllexport
8623 @opindex fkeep-inline-dllexport
8624 This is a more fine-grained version of @option{-fkeep-inline-functions},
8625 which applies only to functions that are declared using the @code{dllexport}
8626 attribute or declspec. @xref{Function Attributes,,Declaring Attributes of
8629 @item -fkeep-inline-functions
8630 @opindex fkeep-inline-functions
8631 In C, emit @code{static} functions that are declared @code{inline}
8632 into the object file, even if the function has been inlined into all
8633 of its callers. This switch does not affect functions using the
8634 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8635 inline functions into the object file.
8637 @item -fkeep-static-functions
8638 @opindex fkeep-static-functions
8639 Emit @code{static} functions into the object file, even if the function
8642 @item -fkeep-static-consts
8643 @opindex fkeep-static-consts
8644 Emit variables declared @code{static const} when optimization isn't turned
8645 on, even if the variables aren't referenced.
8647 GCC enables this option by default. If you want to force the compiler to
8648 check if a variable is referenced, regardless of whether or not
8649 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8651 @item -fmerge-constants
8652 @opindex fmerge-constants
8653 Attempt to merge identical constants (string constants and floating-point
8654 constants) across compilation units.
8656 This option is the default for optimized compilation if the assembler and
8657 linker support it. Use @option{-fno-merge-constants} to inhibit this
8660 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8662 @item -fmerge-all-constants
8663 @opindex fmerge-all-constants
8664 Attempt to merge identical constants and identical variables.
8666 This option implies @option{-fmerge-constants}. In addition to
8667 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8668 arrays or initialized constant variables with integral or floating-point
8669 types. Languages like C or C++ require each variable, including multiple
8670 instances of the same variable in recursive calls, to have distinct locations,
8671 so using this option results in non-conforming
8674 @item -fmodulo-sched
8675 @opindex fmodulo-sched
8676 Perform swing modulo scheduling immediately before the first scheduling
8677 pass. This pass looks at innermost loops and reorders their
8678 instructions by overlapping different iterations.
8680 @item -fmodulo-sched-allow-regmoves
8681 @opindex fmodulo-sched-allow-regmoves
8682 Perform more aggressive SMS-based modulo scheduling with register moves
8683 allowed. By setting this flag certain anti-dependences edges are
8684 deleted, which triggers the generation of reg-moves based on the
8685 life-range analysis. This option is effective only with
8686 @option{-fmodulo-sched} enabled.
8688 @item -fno-branch-count-reg
8689 @opindex fno-branch-count-reg
8690 @opindex fbranch-count-reg
8691 Disable the optimization pass that scans for opportunities to use
8692 ``decrement and branch'' instructions on a count register instead of
8693 instruction sequences that decrement a register, compare it against zero, and
8694 then branch based upon the result. This option is only meaningful on
8695 architectures that support such instructions, which include x86, PowerPC,
8696 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
8697 doesn't remove the decrement and branch instructions from the generated
8698 instruction stream introduced by other optimization passes.
8700 The default is @option{-fbranch-count-reg} at @option{-O1} and higher,
8701 except for @option{-Og}.
8703 @item -fno-function-cse
8704 @opindex fno-function-cse
8705 @opindex ffunction-cse
8706 Do not put function addresses in registers; make each instruction that
8707 calls a constant function contain the function's address explicitly.
8709 This option results in less efficient code, but some strange hacks
8710 that alter the assembler output may be confused by the optimizations
8711 performed when this option is not used.
8713 The default is @option{-ffunction-cse}
8715 @item -fno-zero-initialized-in-bss
8716 @opindex fno-zero-initialized-in-bss
8717 @opindex fzero-initialized-in-bss
8718 If the target supports a BSS section, GCC by default puts variables that
8719 are initialized to zero into BSS@. This can save space in the resulting
8722 This option turns off this behavior because some programs explicitly
8723 rely on variables going to the data section---e.g., so that the
8724 resulting executable can find the beginning of that section and/or make
8725 assumptions based on that.
8727 The default is @option{-fzero-initialized-in-bss}.
8729 @item -fthread-jumps
8730 @opindex fthread-jumps
8731 Perform optimizations that check to see if a jump branches to a
8732 location where another comparison subsumed by the first is found. If
8733 so, the first branch is redirected to either the destination of the
8734 second branch or a point immediately following it, depending on whether
8735 the condition is known to be true or false.
8737 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8739 @item -fsplit-wide-types
8740 @opindex fsplit-wide-types
8741 When using a type that occupies multiple registers, such as @code{long
8742 long} on a 32-bit system, split the registers apart and allocate them
8743 independently. This normally generates better code for those types,
8744 but may make debugging more difficult.
8746 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8749 @item -fsplit-wide-types-early
8750 @opindex fsplit-wide-types-early
8751 Fully split wide types early, instead of very late.
8752 This option has no effect unless @option{-fsplit-wide-types} is turned on.
8754 This is the default on some targets.
8756 @item -fcse-follow-jumps
8757 @opindex fcse-follow-jumps
8758 In common subexpression elimination (CSE), scan through jump instructions
8759 when the target of the jump is not reached by any other path. For
8760 example, when CSE encounters an @code{if} statement with an
8761 @code{else} clause, CSE follows the jump when the condition
8764 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8766 @item -fcse-skip-blocks
8767 @opindex fcse-skip-blocks
8768 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8769 follow jumps that conditionally skip over blocks. When CSE
8770 encounters a simple @code{if} statement with no else clause,
8771 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8772 body of the @code{if}.
8774 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8776 @item -frerun-cse-after-loop
8777 @opindex frerun-cse-after-loop
8778 Re-run common subexpression elimination after loop optimizations are
8781 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8785 Perform a global common subexpression elimination pass.
8786 This pass also performs global constant and copy propagation.
8788 @emph{Note:} When compiling a program using computed gotos, a GCC
8789 extension, you may get better run-time performance if you disable
8790 the global common subexpression elimination pass by adding
8791 @option{-fno-gcse} to the command line.
8793 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8797 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8798 attempts to move loads that are only killed by stores into themselves. This
8799 allows a loop containing a load/store sequence to be changed to a load outside
8800 the loop, and a copy/store within the loop.
8802 Enabled by default when @option{-fgcse} is enabled.
8806 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8807 global common subexpression elimination. This pass attempts to move
8808 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8809 loops containing a load/store sequence can be changed to a load before
8810 the loop and a store after the loop.
8812 Not enabled at any optimization level.
8816 When @option{-fgcse-las} is enabled, the global common subexpression
8817 elimination pass eliminates redundant loads that come after stores to the
8818 same memory location (both partial and full redundancies).
8820 Not enabled at any optimization level.
8822 @item -fgcse-after-reload
8823 @opindex fgcse-after-reload
8824 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8825 pass is performed after reload. The purpose of this pass is to clean up
8828 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
8830 @item -faggressive-loop-optimizations
8831 @opindex faggressive-loop-optimizations
8832 This option tells the loop optimizer to use language constraints to
8833 derive bounds for the number of iterations of a loop. This assumes that
8834 loop code does not invoke undefined behavior by for example causing signed
8835 integer overflows or out-of-bound array accesses. The bounds for the
8836 number of iterations of a loop are used to guide loop unrolling and peeling
8837 and loop exit test optimizations.
8838 This option is enabled by default.
8840 @item -funconstrained-commons
8841 @opindex funconstrained-commons
8842 This option tells the compiler that variables declared in common blocks
8843 (e.g.@: Fortran) may later be overridden with longer trailing arrays. This
8844 prevents certain optimizations that depend on knowing the array bounds.
8846 @item -fcrossjumping
8847 @opindex fcrossjumping
8848 Perform cross-jumping transformation.
8849 This transformation unifies equivalent code and saves code size. The
8850 resulting code may or may not perform better than without cross-jumping.
8852 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8854 @item -fauto-inc-dec
8855 @opindex fauto-inc-dec
8856 Combine increments or decrements of addresses with memory accesses.
8857 This pass is always skipped on architectures that do not have
8858 instructions to support this. Enabled by default at @option{-O} and
8859 higher on architectures that support this.
8863 Perform dead code elimination (DCE) on RTL@.
8864 Enabled by default at @option{-O} and higher.
8868 Perform dead store elimination (DSE) on RTL@.
8869 Enabled by default at @option{-O} and higher.
8871 @item -fif-conversion
8872 @opindex fif-conversion
8873 Attempt to transform conditional jumps into branch-less equivalents. This
8874 includes use of conditional moves, min, max, set flags and abs instructions, and
8875 some tricks doable by standard arithmetics. The use of conditional execution
8876 on chips where it is available is controlled by @option{-fif-conversion2}.
8878 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8879 not with @option{-Og}.
8881 @item -fif-conversion2
8882 @opindex fif-conversion2
8883 Use conditional execution (where available) to transform conditional jumps into
8884 branch-less equivalents.
8886 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}, but
8887 not with @option{-Og}.
8889 @item -fdeclone-ctor-dtor
8890 @opindex fdeclone-ctor-dtor
8891 The C++ ABI requires multiple entry points for constructors and
8892 destructors: one for a base subobject, one for a complete object, and
8893 one for a virtual destructor that calls operator delete afterwards.
8894 For a hierarchy with virtual bases, the base and complete variants are
8895 clones, which means two copies of the function. With this option, the
8896 base and complete variants are changed to be thunks that call a common
8899 Enabled by @option{-Os}.
8901 @item -fdelete-null-pointer-checks
8902 @opindex fdelete-null-pointer-checks
8903 Assume that programs cannot safely dereference null pointers, and that
8904 no code or data element resides at address zero.
8905 This option enables simple constant
8906 folding optimizations at all optimization levels. In addition, other
8907 optimization passes in GCC use this flag to control global dataflow
8908 analyses that eliminate useless checks for null pointers; these assume
8909 that a memory access to address zero always results in a trap, so
8910 that if a pointer is checked after it has already been dereferenced,
8913 Note however that in some environments this assumption is not true.
8914 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8915 for programs that depend on that behavior.
8917 This option is enabled by default on most targets. On Nios II ELF, it
8918 defaults to off. On AVR, CR16, and MSP430, this option is completely disabled.
8920 Passes that use the dataflow information
8921 are enabled independently at different optimization levels.
8923 @item -fdevirtualize
8924 @opindex fdevirtualize
8925 Attempt to convert calls to virtual functions to direct calls. This
8926 is done both within a procedure and interprocedurally as part of
8927 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8928 propagation (@option{-fipa-cp}).
8929 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8931 @item -fdevirtualize-speculatively
8932 @opindex fdevirtualize-speculatively
8933 Attempt to convert calls to virtual functions to speculative direct calls.
8934 Based on the analysis of the type inheritance graph, determine for a given call
8935 the set of likely targets. If the set is small, preferably of size 1, change
8936 the call into a conditional deciding between direct and indirect calls. The
8937 speculative calls enable more optimizations, such as inlining. When they seem
8938 useless after further optimization, they are converted back into original form.
8940 @item -fdevirtualize-at-ltrans
8941 @opindex fdevirtualize-at-ltrans
8942 Stream extra information needed for aggressive devirtualization when running
8943 the link-time optimizer in local transformation mode.
8944 This option enables more devirtualization but
8945 significantly increases the size of streamed data. For this reason it is
8946 disabled by default.
8948 @item -fexpensive-optimizations
8949 @opindex fexpensive-optimizations
8950 Perform a number of minor optimizations that are relatively expensive.
8952 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8956 Attempt to remove redundant extension instructions. This is especially
8957 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8958 registers after writing to their lower 32-bit half.
8960 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8961 @option{-O3}, @option{-Os}.
8963 @item -fno-lifetime-dse
8964 @opindex fno-lifetime-dse
8965 @opindex flifetime-dse
8966 In C++ the value of an object is only affected by changes within its
8967 lifetime: when the constructor begins, the object has an indeterminate
8968 value, and any changes during the lifetime of the object are dead when
8969 the object is destroyed. Normally dead store elimination will take
8970 advantage of this; if your code relies on the value of the object
8971 storage persisting beyond the lifetime of the object, you can use this
8972 flag to disable this optimization. To preserve stores before the
8973 constructor starts (e.g.@: because your operator new clears the object
8974 storage) but still treat the object as dead after the destructor you,
8975 can use @option{-flifetime-dse=1}. The default behavior can be
8976 explicitly selected with @option{-flifetime-dse=2}.
8977 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
8979 @item -flive-range-shrinkage
8980 @opindex flive-range-shrinkage
8981 Attempt to decrease register pressure through register live range
8982 shrinkage. This is helpful for fast processors with small or moderate
8985 @item -fira-algorithm=@var{algorithm}
8986 @opindex fira-algorithm
8987 Use the specified coloring algorithm for the integrated register
8988 allocator. The @var{algorithm} argument can be @samp{priority}, which
8989 specifies Chow's priority coloring, or @samp{CB}, which specifies
8990 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8991 for all architectures, but for those targets that do support it, it is
8992 the default because it generates better code.
8994 @item -fira-region=@var{region}
8995 @opindex fira-region
8996 Use specified regions for the integrated register allocator. The
8997 @var{region} argument should be one of the following:
9002 Use all loops as register allocation regions.
9003 This can give the best results for machines with a small and/or
9004 irregular register set.
9007 Use all loops except for loops with small register pressure
9008 as the regions. This value usually gives
9009 the best results in most cases and for most architectures,
9010 and is enabled by default when compiling with optimization for speed
9011 (@option{-O}, @option{-O2}, @dots{}).
9014 Use all functions as a single region.
9015 This typically results in the smallest code size, and is enabled by default for
9016 @option{-Os} or @option{-O0}.
9020 @item -fira-hoist-pressure
9021 @opindex fira-hoist-pressure
9022 Use IRA to evaluate register pressure in the code hoisting pass for
9023 decisions to hoist expressions. This option usually results in smaller
9024 code, but it can slow the compiler down.
9026 This option is enabled at level @option{-Os} for all targets.
9028 @item -fira-loop-pressure
9029 @opindex fira-loop-pressure
9030 Use IRA to evaluate register pressure in loops for decisions to move
9031 loop invariants. This option usually results in generation
9032 of faster and smaller code on machines with large register files (>= 32
9033 registers), but it can slow the compiler down.
9035 This option is enabled at level @option{-O3} for some targets.
9037 @item -fno-ira-share-save-slots
9038 @opindex fno-ira-share-save-slots
9039 @opindex fira-share-save-slots
9040 Disable sharing of stack slots used for saving call-used hard
9041 registers living through a call. Each hard register gets a
9042 separate stack slot, and as a result function stack frames are
9045 @item -fno-ira-share-spill-slots
9046 @opindex fno-ira-share-spill-slots
9047 @opindex fira-share-spill-slots
9048 Disable sharing of stack slots allocated for pseudo-registers. Each
9049 pseudo-register that does not get a hard register gets a separate
9050 stack slot, and as a result function stack frames are larger.
9054 Enable CFG-sensitive rematerialization in LRA. Instead of loading
9055 values of spilled pseudos, LRA tries to rematerialize (recalculate)
9056 values if it is profitable.
9058 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9060 @item -fdelayed-branch
9061 @opindex fdelayed-branch
9062 If supported for the target machine, attempt to reorder instructions
9063 to exploit instruction slots available after delayed branch
9066 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os},
9067 but not at @option{-Og}.
9069 @item -fschedule-insns
9070 @opindex fschedule-insns
9071 If supported for the target machine, attempt to reorder instructions to
9072 eliminate execution stalls due to required data being unavailable. This
9073 helps machines that have slow floating point or memory load instructions
9074 by allowing other instructions to be issued until the result of the load
9075 or floating-point instruction is required.
9077 Enabled at levels @option{-O2}, @option{-O3}.
9079 @item -fschedule-insns2
9080 @opindex fschedule-insns2
9081 Similar to @option{-fschedule-insns}, but requests an additional pass of
9082 instruction scheduling after register allocation has been done. This is
9083 especially useful on machines with a relatively small number of
9084 registers and where memory load instructions take more than one cycle.
9086 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9088 @item -fno-sched-interblock
9089 @opindex fno-sched-interblock
9090 @opindex fsched-interblock
9091 Disable instruction scheduling across basic blocks, which
9092 is normally enabled when scheduling before register allocation, i.e.@:
9093 with @option{-fschedule-insns} or at @option{-O2} or higher.
9095 @item -fno-sched-spec
9096 @opindex fno-sched-spec
9097 @opindex fsched-spec
9098 Disable speculative motion of non-load instructions, which
9099 is normally enabled when scheduling before register allocation, i.e.@:
9100 with @option{-fschedule-insns} or at @option{-O2} or higher.
9102 @item -fsched-pressure
9103 @opindex fsched-pressure
9104 Enable register pressure sensitive insn scheduling before register
9105 allocation. This only makes sense when scheduling before register
9106 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
9107 @option{-O2} or higher. Usage of this option can improve the
9108 generated code and decrease its size by preventing register pressure
9109 increase above the number of available hard registers and subsequent
9110 spills in register allocation.
9112 @item -fsched-spec-load
9113 @opindex fsched-spec-load
9114 Allow speculative motion of some load instructions. This only makes
9115 sense when scheduling before register allocation, i.e.@: with
9116 @option{-fschedule-insns} or at @option{-O2} or higher.
9118 @item -fsched-spec-load-dangerous
9119 @opindex fsched-spec-load-dangerous
9120 Allow speculative motion of more load instructions. This only makes
9121 sense when scheduling before register allocation, i.e.@: with
9122 @option{-fschedule-insns} or at @option{-O2} or higher.
9124 @item -fsched-stalled-insns
9125 @itemx -fsched-stalled-insns=@var{n}
9126 @opindex fsched-stalled-insns
9127 Define how many insns (if any) can be moved prematurely from the queue
9128 of stalled insns into the ready list during the second scheduling pass.
9129 @option{-fno-sched-stalled-insns} means that no insns are moved
9130 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
9131 on how many queued insns can be moved prematurely.
9132 @option{-fsched-stalled-insns} without a value is equivalent to
9133 @option{-fsched-stalled-insns=1}.
9135 @item -fsched-stalled-insns-dep
9136 @itemx -fsched-stalled-insns-dep=@var{n}
9137 @opindex fsched-stalled-insns-dep
9138 Define how many insn groups (cycles) are examined for a dependency
9139 on a stalled insn that is a candidate for premature removal from the queue
9140 of stalled insns. This has an effect only during the second scheduling pass,
9141 and only if @option{-fsched-stalled-insns} is used.
9142 @option{-fno-sched-stalled-insns-dep} is equivalent to
9143 @option{-fsched-stalled-insns-dep=0}.
9144 @option{-fsched-stalled-insns-dep} without a value is equivalent to
9145 @option{-fsched-stalled-insns-dep=1}.
9147 @item -fsched2-use-superblocks
9148 @opindex fsched2-use-superblocks
9149 When scheduling after register allocation, use superblock scheduling.
9150 This allows motion across basic block boundaries,
9151 resulting in faster schedules. This option is experimental, as not all machine
9152 descriptions used by GCC model the CPU closely enough to avoid unreliable
9153 results from the algorithm.
9155 This only makes sense when scheduling after register allocation, i.e.@: with
9156 @option{-fschedule-insns2} or at @option{-O2} or higher.
9158 @item -fsched-group-heuristic
9159 @opindex fsched-group-heuristic
9160 Enable the group heuristic in the scheduler. This heuristic favors
9161 the instruction that belongs to a schedule group. This is enabled
9162 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9163 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9165 @item -fsched-critical-path-heuristic
9166 @opindex fsched-critical-path-heuristic
9167 Enable the critical-path heuristic in the scheduler. This heuristic favors
9168 instructions on the critical path. This is enabled by default when
9169 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
9170 or @option{-fschedule-insns2} or at @option{-O2} or higher.
9172 @item -fsched-spec-insn-heuristic
9173 @opindex fsched-spec-insn-heuristic
9174 Enable the speculative instruction heuristic in the scheduler. This
9175 heuristic favors speculative instructions with greater dependency weakness.
9176 This is enabled by default when scheduling is enabled, i.e.@:
9177 with @option{-fschedule-insns} or @option{-fschedule-insns2}
9178 or at @option{-O2} or higher.
9180 @item -fsched-rank-heuristic
9181 @opindex fsched-rank-heuristic
9182 Enable the rank heuristic in the scheduler. This heuristic favors
9183 the instruction belonging to a basic block with greater size or frequency.
9184 This is enabled by default when scheduling is enabled, i.e.@:
9185 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9186 at @option{-O2} or higher.
9188 @item -fsched-last-insn-heuristic
9189 @opindex fsched-last-insn-heuristic
9190 Enable the last-instruction heuristic in the scheduler. This heuristic
9191 favors the instruction that is less dependent on the last instruction
9192 scheduled. This is enabled by default when scheduling is enabled,
9193 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9194 at @option{-O2} or higher.
9196 @item -fsched-dep-count-heuristic
9197 @opindex fsched-dep-count-heuristic
9198 Enable the dependent-count heuristic in the scheduler. This heuristic
9199 favors the instruction that has more instructions depending on it.
9200 This is enabled by default when scheduling is enabled, i.e.@:
9201 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
9202 at @option{-O2} or higher.
9204 @item -freschedule-modulo-scheduled-loops
9205 @opindex freschedule-modulo-scheduled-loops
9206 Modulo scheduling is performed before traditional scheduling. If a loop
9207 is modulo scheduled, later scheduling passes may change its schedule.
9208 Use this option to control that behavior.
9210 @item -fselective-scheduling
9211 @opindex fselective-scheduling
9212 Schedule instructions using selective scheduling algorithm. Selective
9213 scheduling runs instead of the first scheduler pass.
9215 @item -fselective-scheduling2
9216 @opindex fselective-scheduling2
9217 Schedule instructions using selective scheduling algorithm. Selective
9218 scheduling runs instead of the second scheduler pass.
9220 @item -fsel-sched-pipelining
9221 @opindex fsel-sched-pipelining
9222 Enable software pipelining of innermost loops during selective scheduling.
9223 This option has no effect unless one of @option{-fselective-scheduling} or
9224 @option{-fselective-scheduling2} is turned on.
9226 @item -fsel-sched-pipelining-outer-loops
9227 @opindex fsel-sched-pipelining-outer-loops
9228 When pipelining loops during selective scheduling, also pipeline outer loops.
9229 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
9231 @item -fsemantic-interposition
9232 @opindex fsemantic-interposition
9233 Some object formats, like ELF, allow interposing of symbols by the
9235 This means that for symbols exported from the DSO, the compiler cannot perform
9236 interprocedural propagation, inlining and other optimizations in anticipation
9237 that the function or variable in question may change. While this feature is
9238 useful, for example, to rewrite memory allocation functions by a debugging
9239 implementation, it is expensive in the terms of code quality.
9240 With @option{-fno-semantic-interposition} the compiler assumes that
9241 if interposition happens for functions the overwriting function will have
9242 precisely the same semantics (and side effects).
9243 Similarly if interposition happens
9244 for variables, the constructor of the variable will be the same. The flag
9245 has no effect for functions explicitly declared inline
9246 (where it is never allowed for interposition to change semantics)
9247 and for symbols explicitly declared weak.
9250 @opindex fshrink-wrap
9251 Emit function prologues only before parts of the function that need it,
9252 rather than at the top of the function. This flag is enabled by default at
9253 @option{-O} and higher.
9255 @item -fshrink-wrap-separate
9256 @opindex fshrink-wrap-separate
9257 Shrink-wrap separate parts of the prologue and epilogue separately, so that
9258 those parts are only executed when needed.
9259 This option is on by default, but has no effect unless @option{-fshrink-wrap}
9260 is also turned on and the target supports this.
9262 @item -fcaller-saves
9263 @opindex fcaller-saves
9264 Enable allocation of values to registers that are clobbered by
9265 function calls, by emitting extra instructions to save and restore the
9266 registers around such calls. Such allocation is done only when it
9267 seems to result in better code.
9269 This option is always enabled by default on certain machines, usually
9270 those which have no call-preserved registers to use instead.
9272 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9274 @item -fcombine-stack-adjustments
9275 @opindex fcombine-stack-adjustments
9276 Tracks stack adjustments (pushes and pops) and stack memory references
9277 and then tries to find ways to combine them.
9279 Enabled by default at @option{-O1} and higher.
9283 Use caller save registers for allocation if those registers are not used by
9284 any called function. In that case it is not necessary to save and restore
9285 them around calls. This is only possible if called functions are part of
9286 same compilation unit as current function and they are compiled before it.
9288 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}, however the option
9289 is disabled if generated code will be instrumented for profiling
9290 (@option{-p}, or @option{-pg}) or if callee's register usage cannot be known
9291 exactly (this happens on targets that do not expose prologues
9292 and epilogues in RTL).
9294 @item -fconserve-stack
9295 @opindex fconserve-stack
9296 Attempt to minimize stack usage. The compiler attempts to use less
9297 stack space, even if that makes the program slower. This option
9298 implies setting the @option{large-stack-frame} parameter to 100
9299 and the @option{large-stack-frame-growth} parameter to 400.
9301 @item -ftree-reassoc
9302 @opindex ftree-reassoc
9303 Perform reassociation on trees. This flag is enabled by default
9304 at @option{-O} and higher.
9306 @item -fcode-hoisting
9307 @opindex fcode-hoisting
9308 Perform code hoisting. Code hoisting tries to move the
9309 evaluation of expressions executed on all paths to the function exit
9310 as early as possible. This is especially useful as a code size
9311 optimization, but it often helps for code speed as well.
9312 This flag is enabled by default at @option{-O2} and higher.
9316 Perform partial redundancy elimination (PRE) on trees. This flag is
9317 enabled by default at @option{-O2} and @option{-O3}.
9319 @item -ftree-partial-pre
9320 @opindex ftree-partial-pre
9321 Make partial redundancy elimination (PRE) more aggressive. This flag is
9322 enabled by default at @option{-O3}.
9324 @item -ftree-forwprop
9325 @opindex ftree-forwprop
9326 Perform forward propagation on trees. This flag is enabled by default
9327 at @option{-O} and higher.
9331 Perform full redundancy elimination (FRE) on trees. The difference
9332 between FRE and PRE is that FRE only considers expressions
9333 that are computed on all paths leading to the redundant computation.
9334 This analysis is faster than PRE, though it exposes fewer redundancies.
9335 This flag is enabled by default at @option{-O} and higher.
9337 @item -ftree-phiprop
9338 @opindex ftree-phiprop
9339 Perform hoisting of loads from conditional pointers on trees. This
9340 pass is enabled by default at @option{-O} and higher.
9342 @item -fhoist-adjacent-loads
9343 @opindex fhoist-adjacent-loads
9344 Speculatively hoist loads from both branches of an if-then-else if the
9345 loads are from adjacent locations in the same structure and the target
9346 architecture has a conditional move instruction. This flag is enabled
9347 by default at @option{-O2} and higher.
9349 @item -ftree-copy-prop
9350 @opindex ftree-copy-prop
9351 Perform copy propagation on trees. This pass eliminates unnecessary
9352 copy operations. This flag is enabled by default at @option{-O} and
9355 @item -fipa-pure-const
9356 @opindex fipa-pure-const
9357 Discover which functions are pure or constant.
9358 Enabled by default at @option{-O} and higher.
9360 @item -fipa-reference
9361 @opindex fipa-reference
9362 Discover which static variables do not escape the
9364 Enabled by default at @option{-O} and higher.
9366 @item -fipa-reference-addressable
9367 @opindex fipa-reference-addressable
9368 Discover read-only, write-only and non-addressable static variables.
9369 Enabled by default at @option{-O} and higher.
9371 @item -fipa-stack-alignment
9372 @opindex fipa-stack-alignment
9373 Reduce stack alignment on call sites if possible.
9378 Perform interprocedural pointer analysis and interprocedural modification
9379 and reference analysis. This option can cause excessive memory and
9380 compile-time usage on large compilation units. It is not enabled by
9381 default at any optimization level.
9384 @opindex fipa-profile
9385 Perform interprocedural profile propagation. The functions called only from
9386 cold functions are marked as cold. Also functions executed once (such as
9387 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
9388 functions and loop less parts of functions executed once are then optimized for
9390 Enabled by default at @option{-O} and higher.
9394 Perform interprocedural constant propagation.
9395 This optimization analyzes the program to determine when values passed
9396 to functions are constants and then optimizes accordingly.
9397 This optimization can substantially increase performance
9398 if the application has constants passed to functions.
9399 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
9400 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9402 @item -fipa-cp-clone
9403 @opindex fipa-cp-clone
9404 Perform function cloning to make interprocedural constant propagation stronger.
9405 When enabled, interprocedural constant propagation performs function cloning
9406 when externally visible function can be called with constant arguments.
9407 Because this optimization can create multiple copies of functions,
9408 it may significantly increase code size
9409 (see @option{--param ipcp-unit-growth=@var{value}}).
9410 This flag is enabled by default at @option{-O3}.
9411 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9414 @opindex fipa-bit-cp
9415 When enabled, perform interprocedural bitwise constant
9416 propagation. This flag is enabled by default at @option{-O2} and
9417 by @option{-fprofile-use} and @option{-fauto-profile}.
9418 It requires that @option{-fipa-cp} is enabled.
9422 When enabled, perform interprocedural propagation of value
9423 ranges. This flag is enabled by default at @option{-O2}. It requires
9424 that @option{-fipa-cp} is enabled.
9428 Perform Identical Code Folding for functions and read-only variables.
9429 The optimization reduces code size and may disturb unwind stacks by replacing
9430 a function by equivalent one with a different name. The optimization works
9431 more effectively with link-time optimization enabled.
9433 Although the behavior is similar to the Gold Linker's ICF optimization, GCC ICF
9434 works on different levels and thus the optimizations are not same - there are
9435 equivalences that are found only by GCC and equivalences found only by Gold.
9437 This flag is enabled by default at @option{-O2} and @option{-Os}.
9439 @item -flive-patching=@var{level}
9440 @opindex flive-patching
9441 Control GCC's optimizations to produce output suitable for live-patching.
9443 If the compiler's optimization uses a function's body or information extracted
9444 from its body to optimize/change another function, the latter is called an
9445 impacted function of the former. If a function is patched, its impacted
9446 functions should be patched too.
9448 The impacted functions are determined by the compiler's interprocedural
9449 optimizations. For example, a caller is impacted when inlining a function
9451 cloning a function and changing its caller to call this new clone,
9452 or extracting a function's pureness/constness information to optimize
9453 its direct or indirect callers, etc.
9455 Usually, the more IPA optimizations enabled, the larger the number of
9456 impacted functions for each function. In order to control the number of
9457 impacted functions and more easily compute the list of impacted function,
9458 IPA optimizations can be partially enabled at two different levels.
9460 The @var{level} argument should be one of the following:
9466 Only enable inlining and cloning optimizations, which includes inlining,
9467 cloning, interprocedural scalar replacement of aggregates and partial inlining.
9468 As a result, when patching a function, all its callers and its clones'
9469 callers are impacted, therefore need to be patched as well.
9471 @option{-flive-patching=inline-clone} disables the following optimization flags:
9472 @gccoptlist{-fwhole-program -fipa-pta -fipa-reference -fipa-ra @gol
9473 -fipa-icf -fipa-icf-functions -fipa-icf-variables @gol
9474 -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable @gol
9475 -fipa-stack-alignment}
9477 @item inline-only-static
9479 Only enable inlining of static functions.
9480 As a result, when patching a static function, all its callers are impacted
9481 and so need to be patched as well.
9483 In addition to all the flags that @option{-flive-patching=inline-clone}
9485 @option{-flive-patching=inline-only-static} disables the following additional
9487 @gccoptlist{-fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp}
9491 When @option{-flive-patching} is specified without any value, the default value
9492 is @var{inline-clone}.
9494 This flag is disabled by default.
9496 Note that @option{-flive-patching} is not supported with link-time optimization
9499 @item -fisolate-erroneous-paths-dereference
9500 @opindex fisolate-erroneous-paths-dereference
9501 Detect paths that trigger erroneous or undefined behavior due to
9502 dereferencing a null pointer. Isolate those paths from the main control
9503 flow and turn the statement with erroneous or undefined behavior into a trap.
9504 This flag is enabled by default at @option{-O2} and higher and depends on
9505 @option{-fdelete-null-pointer-checks} also being enabled.
9507 @item -fisolate-erroneous-paths-attribute
9508 @opindex fisolate-erroneous-paths-attribute
9509 Detect paths that trigger erroneous or undefined behavior due to a null value
9510 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
9511 attribute. Isolate those paths from the main control flow and turn the
9512 statement with erroneous or undefined behavior into a trap. This is not
9513 currently enabled, but may be enabled by @option{-O2} in the future.
9517 Perform forward store motion on trees. This flag is
9518 enabled by default at @option{-O} and higher.
9520 @item -ftree-bit-ccp
9521 @opindex ftree-bit-ccp
9522 Perform sparse conditional bit constant propagation on trees and propagate
9523 pointer alignment information.
9524 This pass only operates on local scalar variables and is enabled by default
9525 at @option{-O1} and higher, except for @option{-Og}.
9526 It requires that @option{-ftree-ccp} is enabled.
9530 Perform sparse conditional constant propagation (CCP) on trees. This
9531 pass only operates on local scalar variables and is enabled by default
9532 at @option{-O} and higher.
9534 @item -fssa-backprop
9535 @opindex fssa-backprop
9536 Propagate information about uses of a value up the definition chain
9537 in order to simplify the definitions. For example, this pass strips
9538 sign operations if the sign of a value never matters. The flag is
9539 enabled by default at @option{-O} and higher.
9542 @opindex fssa-phiopt
9543 Perform pattern matching on SSA PHI nodes to optimize conditional
9544 code. This pass is enabled by default at @option{-O1} and higher,
9545 except for @option{-Og}.
9547 @item -ftree-switch-conversion
9548 @opindex ftree-switch-conversion
9549 Perform conversion of simple initializations in a switch to
9550 initializations from a scalar array. This flag is enabled by default
9551 at @option{-O2} and higher.
9553 @item -ftree-tail-merge
9554 @opindex ftree-tail-merge
9555 Look for identical code sequences. When found, replace one with a jump to the
9556 other. This optimization is known as tail merging or cross jumping. This flag
9557 is enabled by default at @option{-O2} and higher. The compilation time
9559 be limited using @option{max-tail-merge-comparisons} parameter and
9560 @option{max-tail-merge-iterations} parameter.
9564 Perform dead code elimination (DCE) on trees. This flag is enabled by
9565 default at @option{-O} and higher.
9567 @item -ftree-builtin-call-dce
9568 @opindex ftree-builtin-call-dce
9569 Perform conditional dead code elimination (DCE) for calls to built-in functions
9570 that may set @code{errno} but are otherwise free of side effects. This flag is
9571 enabled by default at @option{-O2} and higher if @option{-Os} is not also
9574 @item -ffinite-loops
9575 @opindex ffinite-loops
9576 @opindex fno-finite-loops
9577 Assume that a loop with an exit will eventually take the exit and not loop
9578 indefinitely. This allows the compiler to remove loops that otherwise have
9579 no side-effects, not considering eventual endless looping as such.
9581 This option is enabled by default at @option{-O2}.
9583 @item -ftree-dominator-opts
9584 @opindex ftree-dominator-opts
9585 Perform a variety of simple scalar cleanups (constant/copy
9586 propagation, redundancy elimination, range propagation and expression
9587 simplification) based on a dominator tree traversal. This also
9588 performs jump threading (to reduce jumps to jumps). This flag is
9589 enabled by default at @option{-O} and higher.
9593 Perform dead store elimination (DSE) on trees. A dead store is a store into
9594 a memory location that is later overwritten by another store without
9595 any intervening loads. In this case the earlier store can be deleted. This
9596 flag is enabled by default at @option{-O} and higher.
9600 Perform loop header copying on trees. This is beneficial since it increases
9601 effectiveness of code motion optimizations. It also saves one jump. This flag
9602 is enabled by default at @option{-O} and higher. It is not enabled
9603 for @option{-Os}, since it usually increases code size.
9605 @item -ftree-loop-optimize
9606 @opindex ftree-loop-optimize
9607 Perform loop optimizations on trees. This flag is enabled by default
9608 at @option{-O} and higher.
9610 @item -ftree-loop-linear
9611 @itemx -floop-strip-mine
9613 @opindex ftree-loop-linear
9614 @opindex floop-strip-mine
9615 @opindex floop-block
9616 Perform loop nest optimizations. Same as
9617 @option{-floop-nest-optimize}. To use this code transformation, GCC has
9618 to be configured with @option{--with-isl} to enable the Graphite loop
9619 transformation infrastructure.
9621 @item -fgraphite-identity
9622 @opindex fgraphite-identity
9623 Enable the identity transformation for graphite. For every SCoP we generate
9624 the polyhedral representation and transform it back to gimple. Using
9625 @option{-fgraphite-identity} we can check the costs or benefits of the
9626 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
9627 are also performed by the code generator isl, like index splitting and
9628 dead code elimination in loops.
9630 @item -floop-nest-optimize
9631 @opindex floop-nest-optimize
9632 Enable the isl based loop nest optimizer. This is a generic loop nest
9633 optimizer based on the Pluto optimization algorithms. It calculates a loop
9634 structure optimized for data-locality and parallelism. This option
9637 @item -floop-parallelize-all
9638 @opindex floop-parallelize-all
9639 Use the Graphite data dependence analysis to identify loops that can
9640 be parallelized. Parallelize all the loops that can be analyzed to
9641 not contain loop carried dependences without checking that it is
9642 profitable to parallelize the loops.
9644 @item -ftree-coalesce-vars
9645 @opindex ftree-coalesce-vars
9646 While transforming the program out of the SSA representation, attempt to
9647 reduce copying by coalescing versions of different user-defined
9648 variables, instead of just compiler temporaries. This may severely
9649 limit the ability to debug an optimized program compiled with
9650 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9651 prevents SSA coalescing of user variables. This option is enabled by
9652 default if optimization is enabled, and it does very little otherwise.
9654 @item -ftree-loop-if-convert
9655 @opindex ftree-loop-if-convert
9656 Attempt to transform conditional jumps in the innermost loops to
9657 branch-less equivalents. The intent is to remove control-flow from
9658 the innermost loops in order to improve the ability of the
9659 vectorization pass to handle these loops. This is enabled by default
9660 if vectorization is enabled.
9662 @item -ftree-loop-distribution
9663 @opindex ftree-loop-distribution
9664 Perform loop distribution. This flag can improve cache performance on
9665 big loop bodies and allow further loop optimizations, like
9666 parallelization or vectorization, to take place. For example, the loop
9682 This flag is enabled by default at @option{-O3}.
9683 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9685 @item -ftree-loop-distribute-patterns
9686 @opindex ftree-loop-distribute-patterns
9687 Perform loop distribution of patterns that can be code generated with
9688 calls to a library. This flag is enabled by default at @option{-O3}, and
9689 by @option{-fprofile-use} and @option{-fauto-profile}.
9691 This pass distributes the initialization loops and generates a call to
9692 memset zero. For example, the loop
9708 and the initialization loop is transformed into a call to memset zero.
9709 This flag is enabled by default at @option{-O3}.
9710 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9712 @item -floop-interchange
9713 @opindex floop-interchange
9714 Perform loop interchange outside of graphite. This flag can improve cache
9715 performance on loop nest and allow further loop optimizations, like
9716 vectorization, to take place. For example, the loop
9718 for (int i = 0; i < N; i++)
9719 for (int j = 0; j < N; j++)
9720 for (int k = 0; k < N; k++)
9721 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9725 for (int i = 0; i < N; i++)
9726 for (int k = 0; k < N; k++)
9727 for (int j = 0; j < N; j++)
9728 c[i][j] = c[i][j] + a[i][k]*b[k][j];
9730 This flag is enabled by default at @option{-O3}.
9731 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9733 @item -floop-unroll-and-jam
9734 @opindex floop-unroll-and-jam
9735 Apply unroll and jam transformations on feasible loops. In a loop
9736 nest this unrolls the outer loop by some factor and fuses the resulting
9737 multiple inner loops. This flag is enabled by default at @option{-O3}.
9738 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9740 @item -ftree-loop-im
9741 @opindex ftree-loop-im
9742 Perform loop invariant motion on trees. This pass moves only invariants that
9743 are hard to handle at RTL level (function calls, operations that expand to
9744 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9745 operands of conditions that are invariant out of the loop, so that we can use
9746 just trivial invariantness analysis in loop unswitching. The pass also includes
9749 @item -ftree-loop-ivcanon
9750 @opindex ftree-loop-ivcanon
9751 Create a canonical counter for number of iterations in loops for which
9752 determining number of iterations requires complicated analysis. Later
9753 optimizations then may determine the number easily. Useful especially
9754 in connection with unrolling.
9756 @item -ftree-scev-cprop
9757 @opindex ftree-scev-cprop
9758 Perform final value replacement. If a variable is modified in a loop
9759 in such a way that its value when exiting the loop can be determined using
9760 only its initial value and the number of loop iterations, replace uses of
9761 the final value by such a computation, provided it is sufficiently cheap.
9762 This reduces data dependencies and may allow further simplifications.
9763 Enabled by default at @option{-O} and higher.
9767 Perform induction variable optimizations (strength reduction, induction
9768 variable merging and induction variable elimination) on trees.
9770 @item -ftree-parallelize-loops=n
9771 @opindex ftree-parallelize-loops
9772 Parallelize loops, i.e., split their iteration space to run in n threads.
9773 This is only possible for loops whose iterations are independent
9774 and can be arbitrarily reordered. The optimization is only
9775 profitable on multiprocessor machines, for loops that are CPU-intensive,
9776 rather than constrained e.g.@: by memory bandwidth. This option
9777 implies @option{-pthread}, and thus is only supported on targets
9778 that have support for @option{-pthread}.
9782 Perform function-local points-to analysis on trees. This flag is
9783 enabled by default at @option{-O1} and higher, except for @option{-Og}.
9787 Perform scalar replacement of aggregates. This pass replaces structure
9788 references with scalars to prevent committing structures to memory too
9789 early. This flag is enabled by default at @option{-O1} and higher,
9790 except for @option{-Og}.
9792 @item -fstore-merging
9793 @opindex fstore-merging
9794 Perform merging of narrow stores to consecutive memory addresses. This pass
9795 merges contiguous stores of immediate values narrower than a word into fewer
9796 wider stores to reduce the number of instructions. This is enabled by default
9797 at @option{-O2} and higher as well as @option{-Os}.
9801 Perform temporary expression replacement during the SSA->normal phase. Single
9802 use/single def temporaries are replaced at their use location with their
9803 defining expression. This results in non-GIMPLE code, but gives the expanders
9804 much more complex trees to work on resulting in better RTL generation. This is
9805 enabled by default at @option{-O} and higher.
9809 Perform straight-line strength reduction on trees. This recognizes related
9810 expressions involving multiplications and replaces them by less expensive
9811 calculations when possible. This is enabled by default at @option{-O} and
9814 @item -ftree-vectorize
9815 @opindex ftree-vectorize
9816 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9817 and @option{-ftree-slp-vectorize} if not explicitly specified.
9819 @item -ftree-loop-vectorize
9820 @opindex ftree-loop-vectorize
9821 Perform loop vectorization on trees. This flag is enabled by default at
9822 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9823 and @option{-fauto-profile}.
9825 @item -ftree-slp-vectorize
9826 @opindex ftree-slp-vectorize
9827 Perform basic block vectorization on trees. This flag is enabled by default at
9828 @option{-O3} and by @option{-ftree-vectorize}, @option{-fprofile-use},
9829 and @option{-fauto-profile}.
9831 @item -fvect-cost-model=@var{model}
9832 @opindex fvect-cost-model
9833 Alter the cost model used for vectorization. The @var{model} argument
9834 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9835 With the @samp{unlimited} model the vectorized code-path is assumed
9836 to be profitable while with the @samp{dynamic} model a runtime check
9837 guards the vectorized code-path to enable it only for iteration
9838 counts that will likely execute faster than when executing the original
9839 scalar loop. The @samp{cheap} model disables vectorization of
9840 loops where doing so would be cost prohibitive for example due to
9841 required runtime checks for data dependence or alignment but otherwise
9842 is equal to the @samp{dynamic} model.
9843 The default cost model depends on other optimization flags and is
9844 either @samp{dynamic} or @samp{cheap}.
9846 @item -fsimd-cost-model=@var{model}
9847 @opindex fsimd-cost-model
9848 Alter the cost model used for vectorization of loops marked with the OpenMP
9849 simd directive. The @var{model} argument should be one of
9850 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9851 have the same meaning as described in @option{-fvect-cost-model} and by
9852 default a cost model defined with @option{-fvect-cost-model} is used.
9856 Perform Value Range Propagation on trees. This is similar to the
9857 constant propagation pass, but instead of values, ranges of values are
9858 propagated. This allows the optimizers to remove unnecessary range
9859 checks like array bound checks and null pointer checks. This is
9860 enabled by default at @option{-O2} and higher. Null pointer check
9861 elimination is only done if @option{-fdelete-null-pointer-checks} is
9865 @opindex fsplit-paths
9866 Split paths leading to loop backedges. This can improve dead code
9867 elimination and common subexpression elimination. This is enabled by
9868 default at @option{-O3} and above.
9870 @item -fsplit-ivs-in-unroller
9871 @opindex fsplit-ivs-in-unroller
9872 Enables expression of values of induction variables in later iterations
9873 of the unrolled loop using the value in the first iteration. This breaks
9874 long dependency chains, thus improving efficiency of the scheduling passes.
9876 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9877 same effect. However, that is not reliable in cases where the loop body
9878 is more complicated than a single basic block. It also does not work at all
9879 on some architectures due to restrictions in the CSE pass.
9881 This optimization is enabled by default.
9883 @item -fvariable-expansion-in-unroller
9884 @opindex fvariable-expansion-in-unroller
9885 With this option, the compiler creates multiple copies of some
9886 local variables when unrolling a loop, which can result in superior code.
9888 This optimization is enabled by default for PowerPC targets, but disabled
9889 by default otherwise.
9891 @item -fpartial-inlining
9892 @opindex fpartial-inlining
9893 Inline parts of functions. This option has any effect only
9894 when inlining itself is turned on by the @option{-finline-functions}
9895 or @option{-finline-small-functions} options.
9897 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9899 @item -fpredictive-commoning
9900 @opindex fpredictive-commoning
9901 Perform predictive commoning optimization, i.e., reusing computations
9902 (especially memory loads and stores) performed in previous
9903 iterations of loops.
9905 This option is enabled at level @option{-O3}.
9906 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
9908 @item -fprefetch-loop-arrays
9909 @opindex fprefetch-loop-arrays
9910 If supported by the target machine, generate instructions to prefetch
9911 memory to improve the performance of loops that access large arrays.
9913 This option may generate better or worse code; results are highly
9914 dependent on the structure of loops within the source code.
9916 Disabled at level @option{-Os}.
9918 @item -fno-printf-return-value
9919 @opindex fno-printf-return-value
9920 @opindex fprintf-return-value
9921 Do not substitute constants for known return value of formatted output
9922 functions such as @code{sprintf}, @code{snprintf}, @code{vsprintf}, and
9923 @code{vsnprintf} (but not @code{printf} of @code{fprintf}). This
9924 transformation allows GCC to optimize or even eliminate branches based
9925 on the known return value of these functions called with arguments that
9926 are either constant, or whose values are known to be in a range that
9927 makes determining the exact return value possible. For example, when
9928 @option{-fprintf-return-value} is in effect, both the branch and the
9929 body of the @code{if} statement (but not the call to @code{snprint})
9930 can be optimized away when @code{i} is a 32-bit or smaller integer
9931 because the return value is guaranteed to be at most 8.
9935 if (snprintf (buf, "%08x", i) >= sizeof buf)
9939 The @option{-fprintf-return-value} option relies on other optimizations
9940 and yields best results with @option{-O2} and above. It works in tandem
9941 with the @option{-Wformat-overflow} and @option{-Wformat-truncation}
9942 options. The @option{-fprintf-return-value} option is enabled by default.
9945 @itemx -fno-peephole2
9946 @opindex fno-peephole
9948 @opindex fno-peephole2
9950 Disable any machine-specific peephole optimizations. The difference
9951 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9952 are implemented in the compiler; some targets use one, some use the
9953 other, a few use both.
9955 @option{-fpeephole} is enabled by default.
9956 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9958 @item -fno-guess-branch-probability
9959 @opindex fno-guess-branch-probability
9960 @opindex fguess-branch-probability
9961 Do not guess branch probabilities using heuristics.
9963 GCC uses heuristics to guess branch probabilities if they are
9964 not provided by profiling feedback (@option{-fprofile-arcs}). These
9965 heuristics are based on the control flow graph. If some branch probabilities
9966 are specified by @code{__builtin_expect}, then the heuristics are
9967 used to guess branch probabilities for the rest of the control flow graph,
9968 taking the @code{__builtin_expect} info into account. The interactions
9969 between the heuristics and @code{__builtin_expect} can be complex, and in
9970 some cases, it may be useful to disable the heuristics so that the effects
9971 of @code{__builtin_expect} are easier to understand.
9973 It is also possible to specify expected probability of the expression
9974 with @code{__builtin_expect_with_probability} built-in function.
9976 The default is @option{-fguess-branch-probability} at levels
9977 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9979 @item -freorder-blocks
9980 @opindex freorder-blocks
9981 Reorder basic blocks in the compiled function in order to reduce number of
9982 taken branches and improve code locality.
9984 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9986 @item -freorder-blocks-algorithm=@var{algorithm}
9987 @opindex freorder-blocks-algorithm
9988 Use the specified algorithm for basic block reordering. The
9989 @var{algorithm} argument can be @samp{simple}, which does not increase
9990 code size (except sometimes due to secondary effects like alignment),
9991 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9992 put all often executed code together, minimizing the number of branches
9993 executed by making extra copies of code.
9995 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9996 @samp{stc} at levels @option{-O2}, @option{-O3}.
9998 @item -freorder-blocks-and-partition
9999 @opindex freorder-blocks-and-partition
10000 In addition to reordering basic blocks in the compiled function, in order
10001 to reduce number of taken branches, partitions hot and cold basic blocks
10002 into separate sections of the assembly and @file{.o} files, to improve
10003 paging and cache locality performance.
10005 This optimization is automatically turned off in the presence of
10006 exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined
10007 section attribute and on any architecture that does not support named
10008 sections. When @option{-fsplit-stack} is used this option is not
10009 enabled by default (to avoid linker errors), but may be enabled
10010 explicitly (if using a working linker).
10012 Enabled for x86 at levels @option{-O2}, @option{-O3}, @option{-Os}.
10014 @item -freorder-functions
10015 @opindex freorder-functions
10016 Reorder functions in the object file in order to
10017 improve code locality. This is implemented by using special
10018 subsections @code{.text.hot} for most frequently executed functions and
10019 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
10020 the linker so object file format must support named sections and linker must
10021 place them in a reasonable way.
10023 This option isn't effective unless you either provide profile feedback
10024 (see @option{-fprofile-arcs} for details) or manually annotate functions with
10025 @code{hot} or @code{cold} attributes (@pxref{Common Function Attributes}).
10027 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10029 @item -fstrict-aliasing
10030 @opindex fstrict-aliasing
10031 Allow the compiler to assume the strictest aliasing rules applicable to
10032 the language being compiled. For C (and C++), this activates
10033 optimizations based on the type of expressions. In particular, an
10034 object of one type is assumed never to reside at the same address as an
10035 object of a different type, unless the types are almost the same. For
10036 example, an @code{unsigned int} can alias an @code{int}, but not a
10037 @code{void*} or a @code{double}. A character type may alias any other
10040 @anchor{Type-punning}Pay special attention to code like this:
10053 The practice of reading from a different union member than the one most
10054 recently written to (called ``type-punning'') is common. Even with
10055 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
10056 is accessed through the union type. So, the code above works as
10057 expected. @xref{Structures unions enumerations and bit-fields
10058 implementation}. However, this code might not:
10069 Similarly, access by taking the address, casting the resulting pointer
10070 and dereferencing the result has undefined behavior, even if the cast
10071 uses a union type, e.g.:
10075 return ((union a_union *) &d)->i;
10079 The @option{-fstrict-aliasing} option is enabled at levels
10080 @option{-O2}, @option{-O3}, @option{-Os}.
10082 @item -falign-functions
10083 @itemx -falign-functions=@var{n}
10084 @itemx -falign-functions=@var{n}:@var{m}
10085 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}
10086 @itemx -falign-functions=@var{n}:@var{m}:@var{n2}:@var{m2}
10087 @opindex falign-functions
10088 Align the start of functions to the next power-of-two greater than
10089 @var{n}, skipping up to @var{m}-1 bytes. This ensures that at least
10090 the first @var{m} bytes of the function can be fetched by the CPU
10091 without crossing an @var{n}-byte alignment boundary.
10093 If @var{m} is not specified, it defaults to @var{n}.
10095 Examples: @option{-falign-functions=32} aligns functions to the next
10096 32-byte boundary, @option{-falign-functions=24} aligns to the next
10097 32-byte boundary only if this can be done by skipping 23 bytes or less,
10098 @option{-falign-functions=32:7} aligns to the next
10099 32-byte boundary only if this can be done by skipping 6 bytes or less.
10101 The second pair of @var{n2}:@var{m2} values allows you to specify
10102 a secondary alignment: @option{-falign-functions=64:7:32:3} aligns to
10103 the next 64-byte boundary if this can be done by skipping 6 bytes or less,
10104 otherwise aligns to the next 32-byte boundary if this can be done
10105 by skipping 2 bytes or less.
10106 If @var{m2} is not specified, it defaults to @var{n2}.
10108 Some assemblers only support this flag when @var{n} is a power of two;
10109 in that case, it is rounded up.
10111 @option{-fno-align-functions} and @option{-falign-functions=1} are
10112 equivalent and mean that functions are not aligned.
10114 If @var{n} is not specified or is zero, use a machine-dependent default.
10115 The maximum allowed @var{n} option value is 65536.
10117 Enabled at levels @option{-O2}, @option{-O3}.
10119 @item -flimit-function-alignment
10120 If this option is enabled, the compiler tries to avoid unnecessarily
10121 overaligning functions. It attempts to instruct the assembler to align
10122 by the amount specified by @option{-falign-functions}, but not to
10123 skip more bytes than the size of the function.
10125 @item -falign-labels
10126 @itemx -falign-labels=@var{n}
10127 @itemx -falign-labels=@var{n}:@var{m}
10128 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}
10129 @itemx -falign-labels=@var{n}:@var{m}:@var{n2}:@var{m2}
10130 @opindex falign-labels
10131 Align all branch targets to a power-of-two boundary.
10133 Parameters of this option are analogous to the @option{-falign-functions} option.
10134 @option{-fno-align-labels} and @option{-falign-labels=1} are
10135 equivalent and mean that labels are not aligned.
10137 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
10138 are greater than this value, then their values are used instead.
10140 If @var{n} is not specified or is zero, use a machine-dependent default
10141 which is very likely to be @samp{1}, meaning no alignment.
10142 The maximum allowed @var{n} option value is 65536.
10144 Enabled at levels @option{-O2}, @option{-O3}.
10146 @item -falign-loops
10147 @itemx -falign-loops=@var{n}
10148 @itemx -falign-loops=@var{n}:@var{m}
10149 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}
10150 @itemx -falign-loops=@var{n}:@var{m}:@var{n2}:@var{m2}
10151 @opindex falign-loops
10152 Align loops to a power-of-two boundary. If the loops are executed
10153 many times, this makes up for any execution of the dummy padding
10156 Parameters of this option are analogous to the @option{-falign-functions} option.
10157 @option{-fno-align-loops} and @option{-falign-loops=1} are
10158 equivalent and mean that loops are not aligned.
10159 The maximum allowed @var{n} option value is 65536.
10161 If @var{n} is not specified or is zero, use a machine-dependent default.
10163 Enabled at levels @option{-O2}, @option{-O3}.
10165 @item -falign-jumps
10166 @itemx -falign-jumps=@var{n}
10167 @itemx -falign-jumps=@var{n}:@var{m}
10168 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}
10169 @itemx -falign-jumps=@var{n}:@var{m}:@var{n2}:@var{m2}
10170 @opindex falign-jumps
10171 Align branch targets to a power-of-two boundary, for branch targets
10172 where the targets can only be reached by jumping. In this case,
10173 no dummy operations need be executed.
10175 Parameters of this option are analogous to the @option{-falign-functions} option.
10176 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
10177 equivalent and mean that loops are not aligned.
10179 If @var{n} is not specified or is zero, use a machine-dependent default.
10180 The maximum allowed @var{n} option value is 65536.
10182 Enabled at levels @option{-O2}, @option{-O3}.
10184 @item -funit-at-a-time
10185 @opindex funit-at-a-time
10186 This option is left for compatibility reasons. @option{-funit-at-a-time}
10187 has no effect, while @option{-fno-unit-at-a-time} implies
10188 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
10190 Enabled by default.
10192 @item -fno-toplevel-reorder
10193 @opindex fno-toplevel-reorder
10194 @opindex ftoplevel-reorder
10195 Do not reorder top-level functions, variables, and @code{asm}
10196 statements. Output them in the same order that they appear in the
10197 input file. When this option is used, unreferenced static variables
10198 are not removed. This option is intended to support existing code
10199 that relies on a particular ordering. For new code, it is better to
10200 use attributes when possible.
10202 @option{-ftoplevel-reorder} is the default at @option{-O1} and higher, and
10203 also at @option{-O0} if @option{-fsection-anchors} is explicitly requested.
10204 Additionally @option{-fno-toplevel-reorder} implies
10205 @option{-fno-section-anchors}.
10209 Constructs webs as commonly used for register allocation purposes and assign
10210 each web individual pseudo register. This allows the register allocation pass
10211 to operate on pseudos directly, but also strengthens several other optimization
10212 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
10213 however, make debugging impossible, since variables no longer stay in a
10216 Enabled by default with @option{-funroll-loops}.
10218 @item -fwhole-program
10219 @opindex fwhole-program
10220 Assume that the current compilation unit represents the whole program being
10221 compiled. All public functions and variables with the exception of @code{main}
10222 and those merged by attribute @code{externally_visible} become static functions
10223 and in effect are optimized more aggressively by interprocedural optimizers.
10225 This option should not be used in combination with @option{-flto}.
10226 Instead relying on a linker plugin should provide safer and more precise
10229 @item -flto[=@var{n}]
10231 This option runs the standard link-time optimizer. When invoked
10232 with source code, it generates GIMPLE (one of GCC's internal
10233 representations) and writes it to special ELF sections in the object
10234 file. When the object files are linked together, all the function
10235 bodies are read from these ELF sections and instantiated as if they
10236 had been part of the same translation unit.
10238 To use the link-time optimizer, @option{-flto} and optimization
10239 options should be specified at compile time and during the final link.
10240 It is recommended that you compile all the files participating in the
10241 same link with the same options and also specify those options at
10246 gcc -c -O2 -flto foo.c
10247 gcc -c -O2 -flto bar.c
10248 gcc -o myprog -flto -O2 foo.o bar.o
10251 The first two invocations to GCC save a bytecode representation
10252 of GIMPLE into special ELF sections inside @file{foo.o} and
10253 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
10254 @file{foo.o} and @file{bar.o}, merges the two files into a single
10255 internal image, and compiles the result as usual. Since both
10256 @file{foo.o} and @file{bar.o} are merged into a single image, this
10257 causes all the interprocedural analyses and optimizations in GCC to
10258 work across the two files as if they were a single one. This means,
10259 for example, that the inliner is able to inline functions in
10260 @file{bar.o} into functions in @file{foo.o} and vice-versa.
10262 Another (simpler) way to enable link-time optimization is:
10265 gcc -o myprog -flto -O2 foo.c bar.c
10268 The above generates bytecode for @file{foo.c} and @file{bar.c},
10269 merges them together into a single GIMPLE representation and optimizes
10270 them as usual to produce @file{myprog}.
10272 The important thing to keep in mind is that to enable link-time
10273 optimizations you need to use the GCC driver to perform the link step.
10274 GCC automatically performs link-time optimization if any of the
10275 objects involved were compiled with the @option{-flto} command-line option.
10276 You can always override
10277 the automatic decision to do link-time optimization
10278 by passing @option{-fno-lto} to the link command.
10280 To make whole program optimization effective, it is necessary to make
10281 certain whole program assumptions. The compiler needs to know
10282 what functions and variables can be accessed by libraries and runtime
10283 outside of the link-time optimized unit. When supported by the linker,
10284 the linker plugin (see @option{-fuse-linker-plugin}) passes information
10285 to the compiler about used and externally visible symbols. When
10286 the linker plugin is not available, @option{-fwhole-program} should be
10287 used to allow the compiler to make these assumptions, which leads
10288 to more aggressive optimization decisions.
10290 When a file is compiled with @option{-flto} without
10291 @option{-fuse-linker-plugin}, the generated object file is larger than
10292 a regular object file because it contains GIMPLE bytecodes and the usual
10293 final code (see @option{-ffat-lto-objects}. This means that
10294 object files with LTO information can be linked as normal object
10295 files; if @option{-fno-lto} is passed to the linker, no
10296 interprocedural optimizations are applied. Note that when
10297 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
10298 but you cannot perform a regular, non-LTO link on them.
10300 When producing the final binary, GCC only
10301 applies link-time optimizations to those files that contain bytecode.
10302 Therefore, you can mix and match object files and libraries with
10303 GIMPLE bytecodes and final object code. GCC automatically selects
10304 which files to optimize in LTO mode and which files to link without
10305 further processing.
10307 Generally, options specified at link time override those
10308 specified at compile time, although in some cases GCC attempts to infer
10309 link-time options from the settings used to compile the input files.
10311 If you do not specify an optimization level option @option{-O} at
10312 link time, then GCC uses the highest optimization level
10313 used when compiling the object files. Note that it is generally
10314 ineffective to specify an optimization level option only at link time and
10315 not at compile time, for two reasons. First, compiling without
10316 optimization suppresses compiler passes that gather information
10317 needed for effective optimization at link time. Second, some early
10318 optimization passes can be performed only at compile time and
10321 There are some code generation flags preserved by GCC when
10322 generating bytecodes, as they need to be used during the final link.
10323 Currently, the following options and their settings are taken from
10324 the first object file that explicitly specifies them:
10325 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
10326 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
10327 and all the @option{-m} target flags.
10329 Certain ABI-changing flags are required to match in all compilation units,
10330 and trying to override this at link time with a conflicting value
10331 is ignored. This includes options such as @option{-freg-struct-return}
10332 and @option{-fpcc-struct-return}.
10334 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
10335 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
10336 are passed through to the link stage and merged conservatively for
10337 conflicting translation units. Specifically
10338 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
10339 precedence; and for example @option{-ffp-contract=off} takes precedence
10340 over @option{-ffp-contract=fast}. You can override them at link time.
10342 To enable debug info generation you need to supply @option{-g} at
10343 compile-time. If any of the input files at link time were built
10344 with debug info generation enabled the link will enable debug info
10345 generation as well. Any elaborate debug info settings
10346 like the dwarf level @option{-gdwarf-5} need to be explicitly repeated
10347 at the linker command line and mixing different settings in different
10348 translation units is discouraged.
10350 If LTO encounters objects with C linkage declared with incompatible
10351 types in separate translation units to be linked together (undefined
10352 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
10353 issued. The behavior is still undefined at run time. Similar
10354 diagnostics may be raised for other languages.
10356 Another feature of LTO is that it is possible to apply interprocedural
10357 optimizations on files written in different languages:
10361 g++ -c -flto bar.cc
10362 gfortran -c -flto baz.f90
10363 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
10366 Notice that the final link is done with @command{g++} to get the C++
10367 runtime libraries and @option{-lgfortran} is added to get the Fortran
10368 runtime libraries. In general, when mixing languages in LTO mode, you
10369 should use the same link command options as when mixing languages in a
10370 regular (non-LTO) compilation.
10372 If object files containing GIMPLE bytecode are stored in a library archive, say
10373 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
10374 are using a linker with plugin support. To create static libraries suitable
10375 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
10376 and @command{ranlib};
10377 to show the symbols of object files with GIMPLE bytecode, use
10378 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
10379 and @command{nm} have been compiled with plugin support. At link time, use the
10380 flag @option{-fuse-linker-plugin} to ensure that the library participates in
10381 the LTO optimization process:
10384 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
10387 With the linker plugin enabled, the linker extracts the needed
10388 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
10389 to make them part of the aggregated GIMPLE image to be optimized.
10391 If you are not using a linker with plugin support and/or do not
10392 enable the linker plugin, then the objects inside @file{libfoo.a}
10393 are extracted and linked as usual, but they do not participate
10394 in the LTO optimization process. In order to make a static library suitable
10395 for both LTO optimization and usual linkage, compile its object files with
10396 @option{-flto} @option{-ffat-lto-objects}.
10398 Link-time optimizations do not require the presence of the whole program to
10399 operate. If the program does not require any symbols to be exported, it is
10400 possible to combine @option{-flto} and @option{-fwhole-program} to allow
10401 the interprocedural optimizers to use more aggressive assumptions which may
10402 lead to improved optimization opportunities.
10403 Use of @option{-fwhole-program} is not needed when linker plugin is
10404 active (see @option{-fuse-linker-plugin}).
10406 The current implementation of LTO makes no
10407 attempt to generate bytecode that is portable between different
10408 types of hosts. The bytecode files are versioned and there is a
10409 strict version check, so bytecode files generated in one version of
10410 GCC do not work with an older or newer version of GCC.
10412 Link-time optimization does not work well with generation of debugging
10413 information on systems other than those using a combination of ELF and
10416 If you specify the optional @var{n}, the optimization and code
10417 generation done at link time is executed in parallel using @var{n}
10418 parallel jobs by utilizing an installed @command{make} program. The
10419 environment variable @env{MAKE} may be used to override the program
10422 You can also specify @option{-flto=jobserver} to use GNU make's
10423 job server mode to determine the number of parallel jobs. This
10424 is useful when the Makefile calling GCC is already executing in parallel.
10425 You must prepend a @samp{+} to the command recipe in the parent Makefile
10426 for this to work. This option likely only works if @env{MAKE} is
10427 GNU make. Even without the option value, GCC tries to automatically
10428 detect a running GNU make's job server.
10430 Use @option{-flto=auto} to use GNU make's job server, if available,
10431 or otherwise fall back to autodetection of the number of CPU threads
10432 present in your system.
10434 @item -flto-partition=@var{alg}
10435 @opindex flto-partition
10436 Specify the partitioning algorithm used by the link-time optimizer.
10437 The value is either @samp{1to1} to specify a partitioning mirroring
10438 the original source files or @samp{balanced} to specify partitioning
10439 into equally sized chunks (whenever possible) or @samp{max} to create
10440 new partition for every symbol where possible. Specifying @samp{none}
10441 as an algorithm disables partitioning and streaming completely.
10442 The default value is @samp{balanced}. While @samp{1to1} can be used
10443 as an workaround for various code ordering issues, the @samp{max}
10444 partitioning is intended for internal testing only.
10445 The value @samp{one} specifies that exactly one partition should be
10446 used while the value @samp{none} bypasses partitioning and executes
10447 the link-time optimization step directly from the WPA phase.
10449 @item -flto-compression-level=@var{n}
10450 @opindex flto-compression-level
10451 This option specifies the level of compression used for intermediate
10452 language written to LTO object files, and is only meaningful in
10453 conjunction with LTO mode (@option{-flto}). Valid
10454 values are 0 (no compression) to 9 (maximum compression). Values
10455 outside this range are clamped to either 0 or 9. If the option is not
10456 given, a default balanced compression setting is used.
10458 @item -fuse-linker-plugin
10459 @opindex fuse-linker-plugin
10460 Enables the use of a linker plugin during link-time optimization. This
10461 option relies on plugin support in the linker, which is available in gold
10462 or in GNU ld 2.21 or newer.
10464 This option enables the extraction of object files with GIMPLE bytecode out
10465 of library archives. This improves the quality of optimization by exposing
10466 more code to the link-time optimizer. This information specifies what
10467 symbols can be accessed externally (by non-LTO object or during dynamic
10468 linking). Resulting code quality improvements on binaries (and shared
10469 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
10470 See @option{-flto} for a description of the effect of this flag and how to
10473 This option is enabled by default when LTO support in GCC is enabled
10474 and GCC was configured for use with
10475 a linker supporting plugins (GNU ld 2.21 or newer or gold).
10477 @item -ffat-lto-objects
10478 @opindex ffat-lto-objects
10479 Fat LTO objects are object files that contain both the intermediate language
10480 and the object code. This makes them usable for both LTO linking and normal
10481 linking. This option is effective only when compiling with @option{-flto}
10482 and is ignored at link time.
10484 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
10485 requires the complete toolchain to be aware of LTO. It requires a linker with
10486 linker plugin support for basic functionality. Additionally,
10487 @command{nm}, @command{ar} and @command{ranlib}
10488 need to support linker plugins to allow a full-featured build environment
10489 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
10490 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
10491 to these tools. With non fat LTO makefiles need to be modified to use them.
10493 Note that modern binutils provide plugin auto-load mechanism.
10494 Installing the linker plugin into @file{$libdir/bfd-plugins} has the same
10495 effect as usage of the command wrappers (@command{gcc-ar}, @command{gcc-nm} and
10496 @command{gcc-ranlib}).
10498 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
10501 @item -fcompare-elim
10502 @opindex fcompare-elim
10503 After register allocation and post-register allocation instruction splitting,
10504 identify arithmetic instructions that compute processor flags similar to a
10505 comparison operation based on that arithmetic. If possible, eliminate the
10506 explicit comparison operation.
10508 This pass only applies to certain targets that cannot explicitly represent
10509 the comparison operation before register allocation is complete.
10511 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10513 @item -fcprop-registers
10514 @opindex fcprop-registers
10515 After register allocation and post-register allocation instruction splitting,
10516 perform a copy-propagation pass to try to reduce scheduling dependencies
10517 and occasionally eliminate the copy.
10519 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
10521 @item -fprofile-correction
10522 @opindex fprofile-correction
10523 Profiles collected using an instrumented binary for multi-threaded programs may
10524 be inconsistent due to missed counter updates. When this option is specified,
10525 GCC uses heuristics to correct or smooth out such inconsistencies. By
10526 default, GCC emits an error message when an inconsistent profile is detected.
10528 This option is enabled by @option{-fauto-profile}.
10530 @item -fprofile-use
10531 @itemx -fprofile-use=@var{path}
10532 @opindex fprofile-use
10533 Enable profile feedback-directed optimizations,
10534 and the following optimizations, many of which
10535 are generally profitable only with profile feedback available:
10537 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10538 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10539 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10540 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10541 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10542 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10543 -fprofile-reorder-functions}
10545 Before you can use this option, you must first generate profiling information.
10546 @xref{Instrumentation Options}, for information about the
10547 @option{-fprofile-generate} option.
10549 By default, GCC emits an error message if the feedback profiles do not
10550 match the source code. This error can be turned into a warning by using
10551 @option{-Wno-error=coverage-mismatch}. Note this may result in poorly
10552 optimized code. Additionally, by default, GCC also emits a warning message if
10553 the feedback profiles do not exist (see @option{-Wmissing-profile}).
10555 If @var{path} is specified, GCC looks at the @var{path} to find
10556 the profile feedback data files. See @option{-fprofile-dir}.
10558 @item -fauto-profile
10559 @itemx -fauto-profile=@var{path}
10560 @opindex fauto-profile
10561 Enable sampling-based feedback-directed optimizations,
10562 and the following optimizations,
10563 many of which are generally profitable only with profile feedback available:
10565 @gccoptlist{-fbranch-probabilities -fprofile-values @gol
10566 -funroll-loops -fpeel-loops -ftracer -fvpt @gol
10567 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp @gol
10568 -fpredictive-commoning -fsplit-loops -funswitch-loops @gol
10569 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize @gol
10570 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns @gol
10571 -fprofile-correction}
10573 @var{path} is the name of a file containing AutoFDO profile information.
10574 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
10576 Producing an AutoFDO profile data file requires running your program
10577 with the @command{perf} utility on a supported GNU/Linux target system.
10578 For more information, see @uref{https://perf.wiki.kernel.org/}.
10582 perf record -e br_inst_retired:near_taken -b -o perf.data \
10586 Then use the @command{create_gcov} tool to convert the raw profile data
10587 to a format that can be used by GCC.@ You must also supply the
10588 unstripped binary for your program to this tool.
10589 See @uref{https://github.com/google/autofdo}.
10593 create_gcov --binary=your_program.unstripped --profile=perf.data \
10594 --gcov=profile.afdo
10598 The following options control compiler behavior regarding floating-point
10599 arithmetic. These options trade off between speed and
10600 correctness. All must be specifically enabled.
10603 @item -ffloat-store
10604 @opindex ffloat-store
10605 Do not store floating-point variables in registers, and inhibit other
10606 options that might change whether a floating-point value is taken from a
10607 register or memory.
10609 @cindex floating-point precision
10610 This option prevents undesirable excess precision on machines such as
10611 the 68000 where the floating registers (of the 68881) keep more
10612 precision than a @code{double} is supposed to have. Similarly for the
10613 x86 architecture. For most programs, the excess precision does only
10614 good, but a few programs rely on the precise definition of IEEE floating
10615 point. Use @option{-ffloat-store} for such programs, after modifying
10616 them to store all pertinent intermediate computations into variables.
10618 @item -fexcess-precision=@var{style}
10619 @opindex fexcess-precision
10620 This option allows further control over excess precision on machines
10621 where floating-point operations occur in a format with more precision or
10622 range than the IEEE standard and interchange floating-point types. By
10623 default, @option{-fexcess-precision=fast} is in effect; this means that
10624 operations may be carried out in a wider precision than the types specified
10625 in the source if that would result in faster code, and it is unpredictable
10626 when rounding to the types specified in the source code takes place.
10627 When compiling C, if @option{-fexcess-precision=standard} is specified then
10628 excess precision follows the rules specified in ISO C99; in particular,
10629 both casts and assignments cause values to be rounded to their
10630 semantic types (whereas @option{-ffloat-store} only affects
10631 assignments). This option is enabled by default for C if a strict
10632 conformance option such as @option{-std=c99} is used.
10633 @option{-ffast-math} enables @option{-fexcess-precision=fast} by default
10634 regardless of whether a strict conformance option is used.
10637 @option{-fexcess-precision=standard} is not implemented for languages
10638 other than C. On the x86, it has no effect if @option{-mfpmath=sse}
10639 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
10640 semantics apply without excess precision, and in the latter, rounding
10644 @opindex ffast-math
10645 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
10646 @option{-ffinite-math-only}, @option{-fno-rounding-math},
10647 @option{-fno-signaling-nans}, @option{-fcx-limited-range} and
10648 @option{-fexcess-precision=fast}.
10650 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
10652 This option is not turned on by any @option{-O} option besides
10653 @option{-Ofast} since it can result in incorrect output for programs
10654 that depend on an exact implementation of IEEE or ISO rules/specifications
10655 for math functions. It may, however, yield faster code for programs
10656 that do not require the guarantees of these specifications.
10658 @item -fno-math-errno
10659 @opindex fno-math-errno
10660 @opindex fmath-errno
10661 Do not set @code{errno} after calling math functions that are executed
10662 with a single instruction, e.g., @code{sqrt}. A program that relies on
10663 IEEE exceptions for math error handling may want to use this flag
10664 for speed while maintaining IEEE arithmetic compatibility.
10666 This option is not turned on by any @option{-O} option since
10667 it can result in incorrect output for programs that depend on
10668 an exact implementation of IEEE or ISO rules/specifications for
10669 math functions. It may, however, yield faster code for programs
10670 that do not require the guarantees of these specifications.
10672 The default is @option{-fmath-errno}.
10674 On Darwin systems, the math library never sets @code{errno}. There is
10675 therefore no reason for the compiler to consider the possibility that
10676 it might, and @option{-fno-math-errno} is the default.
10678 @item -funsafe-math-optimizations
10679 @opindex funsafe-math-optimizations
10681 Allow optimizations for floating-point arithmetic that (a) assume
10682 that arguments and results are valid and (b) may violate IEEE or
10683 ANSI standards. When used at link time, it may include libraries
10684 or startup files that change the default FPU control word or other
10685 similar optimizations.
10687 This option is not turned on by any @option{-O} option since
10688 it can result in incorrect output for programs that depend on
10689 an exact implementation of IEEE or ISO rules/specifications for
10690 math functions. It may, however, yield faster code for programs
10691 that do not require the guarantees of these specifications.
10692 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10693 @option{-fassociative-math} and @option{-freciprocal-math}.
10695 The default is @option{-fno-unsafe-math-optimizations}.
10697 @item -fassociative-math
10698 @opindex fassociative-math
10700 Allow re-association of operands in series of floating-point operations.
10701 This violates the ISO C and C++ language standard by possibly changing
10702 computation result. NOTE: re-ordering may change the sign of zero as
10703 well as ignore NaNs and inhibit or create underflow or overflow (and
10704 thus cannot be used on code that relies on rounding behavior like
10705 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10706 and thus may not be used when ordered comparisons are required.
10707 This option requires that both @option{-fno-signed-zeros} and
10708 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10709 much sense with @option{-frounding-math}. For Fortran the option
10710 is automatically enabled when both @option{-fno-signed-zeros} and
10711 @option{-fno-trapping-math} are in effect.
10713 The default is @option{-fno-associative-math}.
10715 @item -freciprocal-math
10716 @opindex freciprocal-math
10718 Allow the reciprocal of a value to be used instead of dividing by
10719 the value if this enables optimizations. For example @code{x / y}
10720 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10721 is subject to common subexpression elimination. Note that this loses
10722 precision and increases the number of flops operating on the value.
10724 The default is @option{-fno-reciprocal-math}.
10726 @item -ffinite-math-only
10727 @opindex ffinite-math-only
10728 Allow optimizations for floating-point arithmetic that assume
10729 that arguments and results are not NaNs or +-Infs.
10731 This option is not turned on by any @option{-O} option since
10732 it can result in incorrect output for programs that depend on
10733 an exact implementation of IEEE or ISO rules/specifications for
10734 math functions. It may, however, yield faster code for programs
10735 that do not require the guarantees of these specifications.
10737 The default is @option{-fno-finite-math-only}.
10739 @item -fno-signed-zeros
10740 @opindex fno-signed-zeros
10741 @opindex fsigned-zeros
10742 Allow optimizations for floating-point arithmetic that ignore the
10743 signedness of zero. IEEE arithmetic specifies the behavior of
10744 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10745 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10746 This option implies that the sign of a zero result isn't significant.
10748 The default is @option{-fsigned-zeros}.
10750 @item -fno-trapping-math
10751 @opindex fno-trapping-math
10752 @opindex ftrapping-math
10753 Compile code assuming that floating-point operations cannot generate
10754 user-visible traps. These traps include division by zero, overflow,
10755 underflow, inexact result and invalid operation. This option requires
10756 that @option{-fno-signaling-nans} be in effect. Setting this option may
10757 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10759 This option should never be turned on by any @option{-O} option since
10760 it can result in incorrect output for programs that depend on
10761 an exact implementation of IEEE or ISO rules/specifications for
10764 The default is @option{-ftrapping-math}.
10766 @item -frounding-math
10767 @opindex frounding-math
10768 Disable transformations and optimizations that assume default floating-point
10769 rounding behavior. This is round-to-zero for all floating point
10770 to integer conversions, and round-to-nearest for all other arithmetic
10771 truncations. This option should be specified for programs that change
10772 the FP rounding mode dynamically, or that may be executed with a
10773 non-default rounding mode. This option disables constant folding of
10774 floating-point expressions at compile time (which may be affected by
10775 rounding mode) and arithmetic transformations that are unsafe in the
10776 presence of sign-dependent rounding modes.
10778 The default is @option{-fno-rounding-math}.
10780 This option is experimental and does not currently guarantee to
10781 disable all GCC optimizations that are affected by rounding mode.
10782 Future versions of GCC may provide finer control of this setting
10783 using C99's @code{FENV_ACCESS} pragma. This command-line option
10784 will be used to specify the default state for @code{FENV_ACCESS}.
10786 @item -fsignaling-nans
10787 @opindex fsignaling-nans
10788 Compile code assuming that IEEE signaling NaNs may generate user-visible
10789 traps during floating-point operations. Setting this option disables
10790 optimizations that may change the number of exceptions visible with
10791 signaling NaNs. This option implies @option{-ftrapping-math}.
10793 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10796 The default is @option{-fno-signaling-nans}.
10798 This option is experimental and does not currently guarantee to
10799 disable all GCC optimizations that affect signaling NaN behavior.
10801 @item -fno-fp-int-builtin-inexact
10802 @opindex fno-fp-int-builtin-inexact
10803 @opindex ffp-int-builtin-inexact
10804 Do not allow the built-in functions @code{ceil}, @code{floor},
10805 @code{round} and @code{trunc}, and their @code{float} and @code{long
10806 double} variants, to generate code that raises the ``inexact''
10807 floating-point exception for noninteger arguments. ISO C99 and C11
10808 allow these functions to raise the ``inexact'' exception, but ISO/IEC
10809 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
10810 functions to do so.
10812 The default is @option{-ffp-int-builtin-inexact}, allowing the
10813 exception to be raised. This option does nothing unless
10814 @option{-ftrapping-math} is in effect.
10816 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
10817 generate a call to a library function then the ``inexact'' exception
10818 may be raised if the library implementation does not follow TS 18661.
10820 @item -fsingle-precision-constant
10821 @opindex fsingle-precision-constant
10822 Treat floating-point constants as single precision instead of
10823 implicitly converting them to double-precision constants.
10825 @item -fcx-limited-range
10826 @opindex fcx-limited-range
10827 When enabled, this option states that a range reduction step is not
10828 needed when performing complex division. Also, there is no checking
10829 whether the result of a complex multiplication or division is @code{NaN
10830 + I*NaN}, with an attempt to rescue the situation in that case. The
10831 default is @option{-fno-cx-limited-range}, but is enabled by
10832 @option{-ffast-math}.
10834 This option controls the default setting of the ISO C99
10835 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10838 @item -fcx-fortran-rules
10839 @opindex fcx-fortran-rules
10840 Complex multiplication and division follow Fortran rules. Range
10841 reduction is done as part of complex division, but there is no checking
10842 whether the result of a complex multiplication or division is @code{NaN
10843 + I*NaN}, with an attempt to rescue the situation in that case.
10845 The default is @option{-fno-cx-fortran-rules}.
10849 The following options control optimizations that may improve
10850 performance, but are not enabled by any @option{-O} options. This
10851 section includes experimental options that may produce broken code.
10854 @item -fbranch-probabilities
10855 @opindex fbranch-probabilities
10856 After running a program compiled with @option{-fprofile-arcs}
10857 (@pxref{Instrumentation Options}),
10858 you can compile it a second time using
10859 @option{-fbranch-probabilities}, to improve optimizations based on
10860 the number of times each branch was taken. When a program
10861 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10862 counts to a file called @file{@var{sourcename}.gcda} for each source
10863 file. The information in this data file is very dependent on the
10864 structure of the generated code, so you must use the same source code
10865 and the same optimization options for both compilations.
10867 With @option{-fbranch-probabilities}, GCC puts a
10868 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10869 These can be used to improve optimization. Currently, they are only
10870 used in one place: in @file{reorg.c}, instead of guessing which path a
10871 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10872 exactly determine which path is taken more often.
10874 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10876 @item -fprofile-values
10877 @opindex fprofile-values
10878 If combined with @option{-fprofile-arcs}, it adds code so that some
10879 data about values of expressions in the program is gathered.
10881 With @option{-fbranch-probabilities}, it reads back the data gathered
10882 from profiling values of expressions for usage in optimizations.
10884 Enabled by @option{-fprofile-generate}, @option{-fprofile-use}, and
10885 @option{-fauto-profile}.
10887 @item -fprofile-reorder-functions
10888 @opindex fprofile-reorder-functions
10889 Function reordering based on profile instrumentation collects
10890 first time of execution of a function and orders these functions
10891 in ascending order.
10893 Enabled with @option{-fprofile-use}.
10897 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10898 to add code to gather information about values of expressions.
10900 With @option{-fbranch-probabilities}, it reads back the data gathered
10901 and actually performs the optimizations based on them.
10902 Currently the optimizations include specialization of division operations
10903 using the knowledge about the value of the denominator.
10905 Enabled with @option{-fprofile-use} and @option{-fauto-profile}.
10907 @item -frename-registers
10908 @opindex frename-registers
10909 Attempt to avoid false dependencies in scheduled code by making use
10910 of registers left over after register allocation. This optimization
10911 most benefits processors with lots of registers. Depending on the
10912 debug information format adopted by the target, however, it can
10913 make debugging impossible, since variables no longer stay in
10914 a ``home register''.
10916 Enabled by default with @option{-funroll-loops}.
10918 @item -fschedule-fusion
10919 @opindex fschedule-fusion
10920 Performs a target dependent pass over the instruction stream to schedule
10921 instructions of same type together because target machine can execute them
10922 more efficiently if they are adjacent to each other in the instruction flow.
10924 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10928 Perform tail duplication to enlarge superblock size. This transformation
10929 simplifies the control flow of the function allowing other optimizations to do
10932 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10934 @item -funroll-loops
10935 @opindex funroll-loops
10936 Unroll loops whose number of iterations can be determined at compile time or
10937 upon entry to the loop. @option{-funroll-loops} implies
10938 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10939 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10940 a small constant number of iterations). This option makes code larger, and may
10941 or may not make it run faster.
10943 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10945 @item -funroll-all-loops
10946 @opindex funroll-all-loops
10947 Unroll all loops, even if their number of iterations is uncertain when
10948 the loop is entered. This usually makes programs run more slowly.
10949 @option{-funroll-all-loops} implies the same options as
10950 @option{-funroll-loops}.
10953 @opindex fpeel-loops
10954 Peels loops for which there is enough information that they do not
10955 roll much (from profile feedback or static analysis). It also turns on
10956 complete loop peeling (i.e.@: complete removal of loops with small constant
10957 number of iterations).
10959 Enabled by @option{-O3}, @option{-fprofile-use}, and @option{-fauto-profile}.
10961 @item -fmove-loop-invariants
10962 @opindex fmove-loop-invariants
10963 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10964 at level @option{-O1} and higher, except for @option{-Og}.
10966 @item -fsplit-loops
10967 @opindex fsplit-loops
10968 Split a loop into two if it contains a condition that's always true
10969 for one side of the iteration space and false for the other.
10971 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10973 @item -funswitch-loops
10974 @opindex funswitch-loops
10975 Move branches with loop invariant conditions out of the loop, with duplicates
10976 of the loop on both branches (modified according to result of the condition).
10978 Enabled by @option{-fprofile-use} and @option{-fauto-profile}.
10980 @item -fversion-loops-for-strides
10981 @opindex fversion-loops-for-strides
10982 If a loop iterates over an array with a variable stride, create another
10983 version of the loop that assumes the stride is always one. For example:
10986 for (int i = 0; i < n; ++i)
10987 x[i * stride] = @dots{};
10994 for (int i = 0; i < n; ++i)
10997 for (int i = 0; i < n; ++i)
10998 x[i * stride] = @dots{};
11001 This is particularly useful for assumed-shape arrays in Fortran where
11002 (for example) it allows better vectorization assuming contiguous accesses.
11003 This flag is enabled by default at @option{-O3}.
11004 It is also enabled by @option{-fprofile-use} and @option{-fauto-profile}.
11006 @item -ffunction-sections
11007 @itemx -fdata-sections
11008 @opindex ffunction-sections
11009 @opindex fdata-sections
11010 Place each function or data item into its own section in the output
11011 file if the target supports arbitrary sections. The name of the
11012 function or the name of the data item determines the section's name
11013 in the output file.
11015 Use these options on systems where the linker can perform optimizations to
11016 improve locality of reference in the instruction space. Most systems using the
11017 ELF object format have linkers with such optimizations. On AIX, the linker
11018 rearranges sections (CSECTs) based on the call graph. The performance impact
11021 Together with a linker garbage collection (linker @option{--gc-sections}
11022 option) these options may lead to smaller statically-linked executables (after
11025 On ELF/DWARF systems these options do not degenerate the quality of the debug
11026 information. There could be issues with other object files/debug info formats.
11028 Only use these options when there are significant benefits from doing so. When
11029 you specify these options, the assembler and linker create larger object and
11030 executable files and are also slower. These options affect code generation.
11031 They prevent optimizations by the compiler and assembler using relative
11032 locations inside a translation unit since the locations are unknown until
11033 link time. An example of such an optimization is relaxing calls to short call
11037 @opindex fstdarg-opt
11038 Optimize the prologue of variadic argument functions with respect to usage of
11041 @item -fsection-anchors
11042 @opindex fsection-anchors
11043 Try to reduce the number of symbolic address calculations by using
11044 shared ``anchor'' symbols to address nearby objects. This transformation
11045 can help to reduce the number of GOT entries and GOT accesses on some
11048 For example, the implementation of the following function @code{foo}:
11051 static int a, b, c;
11052 int foo (void) @{ return a + b + c; @}
11056 usually calculates the addresses of all three variables, but if you
11057 compile it with @option{-fsection-anchors}, it accesses the variables
11058 from a common anchor point instead. The effect is similar to the
11059 following pseudocode (which isn't valid C):
11064 register int *xr = &x;
11065 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
11069 Not all targets support this option.
11071 @item --param @var{name}=@var{value}
11073 In some places, GCC uses various constants to control the amount of
11074 optimization that is done. For example, GCC does not inline functions
11075 that contain more than a certain number of instructions. You can
11076 control some of these constants on the command line using the
11077 @option{--param} option.
11079 The names of specific parameters, and the meaning of the values, are
11080 tied to the internals of the compiler, and are subject to change
11081 without notice in future releases.
11083 In order to get minimal, maximal and default value of a parameter,
11084 one can use @option{--help=param -Q} options.
11086 In each case, the @var{value} is an integer. The allowable choices for
11090 @item predictable-branch-outcome
11091 When branch is predicted to be taken with probability lower than this threshold
11092 (in percent), then it is considered well predictable.
11094 @item max-rtl-if-conversion-insns
11095 RTL if-conversion tries to remove conditional branches around a block and
11096 replace them with conditionally executed instructions. This parameter
11097 gives the maximum number of instructions in a block which should be
11098 considered for if-conversion. The compiler will
11099 also use other heuristics to decide whether if-conversion is likely to be
11102 @item max-rtl-if-conversion-predictable-cost
11103 @itemx max-rtl-if-conversion-unpredictable-cost
11104 RTL if-conversion will try to remove conditional branches around a block
11105 and replace them with conditionally executed instructions. These parameters
11106 give the maximum permissible cost for the sequence that would be generated
11107 by if-conversion depending on whether the branch is statically determined
11108 to be predictable or not. The units for this parameter are the same as
11109 those for the GCC internal seq_cost metric. The compiler will try to
11110 provide a reasonable default for this parameter using the BRANCH_COST
11113 @item max-crossjump-edges
11114 The maximum number of incoming edges to consider for cross-jumping.
11115 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
11116 the number of edges incoming to each block. Increasing values mean
11117 more aggressive optimization, making the compilation time increase with
11118 probably small improvement in executable size.
11120 @item min-crossjump-insns
11121 The minimum number of instructions that must be matched at the end
11122 of two blocks before cross-jumping is performed on them. This
11123 value is ignored in the case where all instructions in the block being
11124 cross-jumped from are matched.
11126 @item max-grow-copy-bb-insns
11127 The maximum code size expansion factor when copying basic blocks
11128 instead of jumping. The expansion is relative to a jump instruction.
11130 @item max-goto-duplication-insns
11131 The maximum number of instructions to duplicate to a block that jumps
11132 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
11133 passes, GCC factors computed gotos early in the compilation process,
11134 and unfactors them as late as possible. Only computed jumps at the
11135 end of a basic blocks with no more than max-goto-duplication-insns are
11138 @item max-delay-slot-insn-search
11139 The maximum number of instructions to consider when looking for an
11140 instruction to fill a delay slot. If more than this arbitrary number of
11141 instructions are searched, the time savings from filling the delay slot
11142 are minimal, so stop searching. Increasing values mean more
11143 aggressive optimization, making the compilation time increase with probably
11144 small improvement in execution time.
11146 @item max-delay-slot-live-search
11147 When trying to fill delay slots, the maximum number of instructions to
11148 consider when searching for a block with valid live register
11149 information. Increasing this arbitrarily chosen value means more
11150 aggressive optimization, increasing the compilation time. This parameter
11151 should be removed when the delay slot code is rewritten to maintain the
11152 control-flow graph.
11154 @item max-gcse-memory
11155 The approximate maximum amount of memory that can be allocated in
11156 order to perform the global common subexpression elimination
11157 optimization. If more memory than specified is required, the
11158 optimization is not done.
11160 @item max-gcse-insertion-ratio
11161 If the ratio of expression insertions to deletions is larger than this value
11162 for any expression, then RTL PRE inserts or removes the expression and thus
11163 leaves partially redundant computations in the instruction stream.
11165 @item max-pending-list-length
11166 The maximum number of pending dependencies scheduling allows
11167 before flushing the current state and starting over. Large functions
11168 with few branches or calls can create excessively large lists which
11169 needlessly consume memory and resources.
11171 @item max-modulo-backtrack-attempts
11172 The maximum number of backtrack attempts the scheduler should make
11173 when modulo scheduling a loop. Larger values can exponentially increase
11176 @item max-inline-insns-single
11177 Several parameters control the tree inliner used in GCC@.
11178 This number sets the maximum number of instructions (counted in GCC's
11179 internal representation) in a single function that the tree inliner
11180 considers for inlining. This only affects functions declared
11181 inline and methods implemented in a class declaration (C++).
11183 @item max-inline-insns-auto
11184 When you use @option{-finline-functions} (included in @option{-O3}),
11185 a lot of functions that would otherwise not be considered for inlining
11186 by the compiler are investigated. To those functions, a different
11187 (more restrictive) limit compared to functions declared inline can
11190 @item max-inline-insns-small
11191 This is bound applied to calls which are considered relevant with
11192 @option{-finline-small-functions}.
11194 @item max-inline-insns-size
11195 This is bound applied to calls which are optimized for size. Small growth
11196 may be desirable to anticipate optimization oppurtunities exposed by inlining.
11198 @item uninlined-function-insns
11199 Number of instructions accounted by inliner for function overhead such as
11200 function prologue and epilogue.
11202 @item uninlined-function-time
11203 Extra time accounted by inliner for function overhead such as time needed to
11204 execute function prologue and epilogue
11206 @item uninlined-thunk-insns
11207 @item uninlined-thunk-time
11208 Same as @option{--param uninlined-function-insns} and
11209 @option{--param uninlined-function-time} but applied to function thunks
11211 @item inline-min-speedup
11212 When estimated performance improvement of caller + callee runtime exceeds this
11213 threshold (in percent), the function can be inlined regardless of the limit on
11214 @option{--param max-inline-insns-single} and @option{--param
11215 max-inline-insns-auto}.
11217 @item large-function-insns
11218 The limit specifying really large functions. For functions larger than this
11219 limit after inlining, inlining is constrained by
11220 @option{--param large-function-growth}. This parameter is useful primarily
11221 to avoid extreme compilation time caused by non-linear algorithms used by the
11224 @item large-function-growth
11225 Specifies maximal growth of large function caused by inlining in percents.
11226 For example, parameter value 100 limits large function growth to 2.0 times
11229 @item large-unit-insns
11230 The limit specifying large translation unit. Growth caused by inlining of
11231 units larger than this limit is limited by @option{--param inline-unit-growth}.
11232 For small units this might be too tight.
11233 For example, consider a unit consisting of function A
11234 that is inline and B that just calls A three times. If B is small relative to
11235 A, the growth of unit is 300\% and yet such inlining is very sane. For very
11236 large units consisting of small inlineable functions, however, the overall unit
11237 growth limit is needed to avoid exponential explosion of code size. Thus for
11238 smaller units, the size is increased to @option{--param large-unit-insns}
11239 before applying @option{--param inline-unit-growth}.
11241 @item inline-unit-growth
11242 Specifies maximal overall growth of the compilation unit caused by inlining.
11243 For example, parameter value 20 limits unit growth to 1.2 times the original
11244 size. Cold functions (either marked cold via an attribute or by profile
11245 feedback) are not accounted into the unit size.
11247 @item ipcp-unit-growth
11248 Specifies maximal overall growth of the compilation unit caused by
11249 interprocedural constant propagation. For example, parameter value 10 limits
11250 unit growth to 1.1 times the original size.
11252 @item large-stack-frame
11253 The limit specifying large stack frames. While inlining the algorithm is trying
11254 to not grow past this limit too much.
11256 @item large-stack-frame-growth
11257 Specifies maximal growth of large stack frames caused by inlining in percents.
11258 For example, parameter value 1000 limits large stack frame growth to 11 times
11261 @item max-inline-insns-recursive
11262 @itemx max-inline-insns-recursive-auto
11263 Specifies the maximum number of instructions an out-of-line copy of a
11264 self-recursive inline
11265 function can grow into by performing recursive inlining.
11267 @option{--param max-inline-insns-recursive} applies to functions
11269 For functions not declared inline, recursive inlining
11270 happens only when @option{-finline-functions} (included in @option{-O3}) is
11271 enabled; @option{--param max-inline-insns-recursive-auto} applies instead.
11273 @item max-inline-recursive-depth
11274 @itemx max-inline-recursive-depth-auto
11275 Specifies the maximum recursion depth used for recursive inlining.
11277 @option{--param max-inline-recursive-depth} applies to functions
11278 declared inline. For functions not declared inline, recursive inlining
11279 happens only when @option{-finline-functions} (included in @option{-O3}) is
11280 enabled; @option{--param max-inline-recursive-depth-auto} applies instead.
11282 @item min-inline-recursive-probability
11283 Recursive inlining is profitable only for function having deep recursion
11284 in average and can hurt for function having little recursion depth by
11285 increasing the prologue size or complexity of function body to other
11288 When profile feedback is available (see @option{-fprofile-generate}) the actual
11289 recursion depth can be guessed from the probability that function recurses
11290 via a given call expression. This parameter limits inlining only to call
11291 expressions whose probability exceeds the given threshold (in percents).
11293 @item early-inlining-insns
11294 Specify growth that the early inliner can make. In effect it increases
11295 the amount of inlining for code having a large abstraction penalty.
11297 @item max-early-inliner-iterations
11298 Limit of iterations of the early inliner. This basically bounds
11299 the number of nested indirect calls the early inliner can resolve.
11300 Deeper chains are still handled by late inlining.
11302 @item comdat-sharing-probability
11303 Probability (in percent) that C++ inline function with comdat visibility
11304 are shared across multiple compilation units.
11306 @item profile-func-internal-id
11307 A parameter to control whether to use function internal id in profile
11308 database lookup. If the value is 0, the compiler uses an id that
11309 is based on function assembler name and filename, which makes old profile
11310 data more tolerant to source changes such as function reordering etc.
11312 @item min-vect-loop-bound
11313 The minimum number of iterations under which loops are not vectorized
11314 when @option{-ftree-vectorize} is used. The number of iterations after
11315 vectorization needs to be greater than the value specified by this option
11316 to allow vectorization.
11318 @item gcse-cost-distance-ratio
11319 Scaling factor in calculation of maximum distance an expression
11320 can be moved by GCSE optimizations. This is currently supported only in the
11321 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
11322 is with simple expressions, i.e., the expressions that have cost
11323 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
11324 hoisting of simple expressions.
11326 @item gcse-unrestricted-cost
11327 Cost, roughly measured as the cost of a single typical machine
11328 instruction, at which GCSE optimizations do not constrain
11329 the distance an expression can travel. This is currently
11330 supported only in the code hoisting pass. The lesser the cost,
11331 the more aggressive code hoisting is. Specifying 0
11332 allows all expressions to travel unrestricted distances.
11334 @item max-hoist-depth
11335 The depth of search in the dominator tree for expressions to hoist.
11336 This is used to avoid quadratic behavior in hoisting algorithm.
11337 The value of 0 does not limit on the search, but may slow down compilation
11340 @item max-tail-merge-comparisons
11341 The maximum amount of similar bbs to compare a bb with. This is used to
11342 avoid quadratic behavior in tree tail merging.
11344 @item max-tail-merge-iterations
11345 The maximum amount of iterations of the pass over the function. This is used to
11346 limit compilation time in tree tail merging.
11348 @item store-merging-allow-unaligned
11349 Allow the store merging pass to introduce unaligned stores if it is legal to
11352 @item max-stores-to-merge
11353 The maximum number of stores to attempt to merge into wider stores in the store
11356 @item max-unrolled-insns
11357 The maximum number of instructions that a loop may have to be unrolled.
11358 If a loop is unrolled, this parameter also determines how many times
11359 the loop code is unrolled.
11361 @item max-average-unrolled-insns
11362 The maximum number of instructions biased by probabilities of their execution
11363 that a loop may have to be unrolled. If a loop is unrolled,
11364 this parameter also determines how many times the loop code is unrolled.
11366 @item max-unroll-times
11367 The maximum number of unrollings of a single loop.
11369 @item max-peeled-insns
11370 The maximum number of instructions that a loop may have to be peeled.
11371 If a loop is peeled, this parameter also determines how many times
11372 the loop code is peeled.
11374 @item max-peel-times
11375 The maximum number of peelings of a single loop.
11377 @item max-peel-branches
11378 The maximum number of branches on the hot path through the peeled sequence.
11380 @item max-completely-peeled-insns
11381 The maximum number of insns of a completely peeled loop.
11383 @item max-completely-peel-times
11384 The maximum number of iterations of a loop to be suitable for complete peeling.
11386 @item max-completely-peel-loop-nest-depth
11387 The maximum depth of a loop nest suitable for complete peeling.
11389 @item max-unswitch-insns
11390 The maximum number of insns of an unswitched loop.
11392 @item max-unswitch-level
11393 The maximum number of branches unswitched in a single loop.
11395 @item lim-expensive
11396 The minimum cost of an expensive expression in the loop invariant motion.
11398 @item iv-consider-all-candidates-bound
11399 Bound on number of candidates for induction variables, below which
11400 all candidates are considered for each use in induction variable
11401 optimizations. If there are more candidates than this,
11402 only the most relevant ones are considered to avoid quadratic time complexity.
11404 @item iv-max-considered-uses
11405 The induction variable optimizations give up on loops that contain more
11406 induction variable uses.
11408 @item iv-always-prune-cand-set-bound
11409 If the number of candidates in the set is smaller than this value,
11410 always try to remove unnecessary ivs from the set
11411 when adding a new one.
11413 @item avg-loop-niter
11414 Average number of iterations of a loop.
11416 @item dse-max-object-size
11417 Maximum size (in bytes) of objects tracked bytewise by dead store elimination.
11418 Larger values may result in larger compilation times.
11420 @item dse-max-alias-queries-per-store
11421 Maximum number of queries into the alias oracle per store.
11422 Larger values result in larger compilation times and may result in more
11423 removed dead stores.
11425 @item scev-max-expr-size
11426 Bound on size of expressions used in the scalar evolutions analyzer.
11427 Large expressions slow the analyzer.
11429 @item scev-max-expr-complexity
11430 Bound on the complexity of the expressions in the scalar evolutions analyzer.
11431 Complex expressions slow the analyzer.
11433 @item max-tree-if-conversion-phi-args
11434 Maximum number of arguments in a PHI supported by TREE if conversion
11435 unless the loop is marked with simd pragma.
11437 @item vect-max-version-for-alignment-checks
11438 The maximum number of run-time checks that can be performed when
11439 doing loop versioning for alignment in the vectorizer.
11441 @item vect-max-version-for-alias-checks
11442 The maximum number of run-time checks that can be performed when
11443 doing loop versioning for alias in the vectorizer.
11445 @item vect-max-peeling-for-alignment
11446 The maximum number of loop peels to enhance access alignment
11447 for vectorizer. Value -1 means no limit.
11449 @item max-iterations-to-track
11450 The maximum number of iterations of a loop the brute-force algorithm
11451 for analysis of the number of iterations of the loop tries to evaluate.
11453 @item hot-bb-count-fraction
11454 The denominator n of fraction 1/n of the maximal execution count of a
11455 basic block in the entire program that a basic block needs to at least
11456 have in order to be considered hot. The default is 10000, which means
11457 that a basic block is considered hot if its execution count is greater
11458 than 1/10000 of the maximal execution count. 0 means that it is never
11459 considered hot. Used in non-LTO mode.
11461 @item hot-bb-count-ws-permille
11462 The number of most executed permilles, ranging from 0 to 1000, of the
11463 profiled execution of the entire program to which the execution count
11464 of a basic block must be part of in order to be considered hot. The
11465 default is 990, which means that a basic block is considered hot if
11466 its execution count contributes to the upper 990 permilles, or 99.0%,
11467 of the profiled execution of the entire program. 0 means that it is
11468 never considered hot. Used in LTO mode.
11470 @item hot-bb-frequency-fraction
11471 The denominator n of fraction 1/n of the execution frequency of the
11472 entry block of a function that a basic block of this function needs
11473 to at least have in order to be considered hot. The default is 1000,
11474 which means that a basic block is considered hot in a function if it
11475 is executed more frequently than 1/1000 of the frequency of the entry
11476 block of the function. 0 means that it is never considered hot.
11478 @item unlikely-bb-count-fraction
11479 The denominator n of fraction 1/n of the number of profiled runs of
11480 the entire program below which the execution count of a basic block
11481 must be in order for the basic block to be considered unlikely executed.
11482 The default is 20, which means that a basic block is considered unlikely
11483 executed if it is executed in fewer than 1/20, or 5%, of the runs of
11484 the program. 0 means that it is always considered unlikely executed.
11486 @item max-predicted-iterations
11487 The maximum number of loop iterations we predict statically. This is useful
11488 in cases where a function contains a single loop with known bound and
11489 another loop with unknown bound.
11490 The known number of iterations is predicted correctly, while
11491 the unknown number of iterations average to roughly 10. This means that the
11492 loop without bounds appears artificially cold relative to the other one.
11494 @item builtin-expect-probability
11495 Control the probability of the expression having the specified value. This
11496 parameter takes a percentage (i.e.@: 0 ... 100) as input.
11498 @item builtin-string-cmp-inline-length
11499 The maximum length of a constant string for a builtin string cmp call
11500 eligible for inlining.
11502 @item align-threshold
11504 Select fraction of the maximal frequency of executions of a basic block in
11505 a function to align the basic block.
11507 @item align-loop-iterations
11509 A loop expected to iterate at least the selected number of iterations is
11512 @item tracer-dynamic-coverage
11513 @itemx tracer-dynamic-coverage-feedback
11515 This value is used to limit superblock formation once the given percentage of
11516 executed instructions is covered. This limits unnecessary code size
11519 The @option{tracer-dynamic-coverage-feedback} parameter
11520 is used only when profile
11521 feedback is available. The real profiles (as opposed to statically estimated
11522 ones) are much less balanced allowing the threshold to be larger value.
11524 @item tracer-max-code-growth
11525 Stop tail duplication once code growth has reached given percentage. This is
11526 a rather artificial limit, as most of the duplicates are eliminated later in
11527 cross jumping, so it may be set to much higher values than is the desired code
11530 @item tracer-min-branch-ratio
11532 Stop reverse growth when the reverse probability of best edge is less than this
11533 threshold (in percent).
11535 @item tracer-min-branch-probability
11536 @itemx tracer-min-branch-probability-feedback
11538 Stop forward growth if the best edge has probability lower than this
11541 Similarly to @option{tracer-dynamic-coverage} two parameters are
11542 provided. @option{tracer-min-branch-probability-feedback} is used for
11543 compilation with profile feedback and @option{tracer-min-branch-probability}
11544 compilation without. The value for compilation with profile feedback
11545 needs to be more conservative (higher) in order to make tracer
11548 @item stack-clash-protection-guard-size
11549 Specify the size of the operating system provided stack guard as
11550 2 raised to @var{num} bytes. Higher values may reduce the
11551 number of explicit probes, but a value larger than the operating system
11552 provided guard will leave code vulnerable to stack clash style attacks.
11554 @item stack-clash-protection-probe-interval
11555 Stack clash protection involves probing stack space as it is allocated. This
11556 param controls the maximum distance between probes into the stack as 2 raised
11557 to @var{num} bytes. Higher values may reduce the number of explicit probes, but a value
11558 larger than the operating system provided guard will leave code vulnerable to
11559 stack clash style attacks.
11561 @item max-cse-path-length
11563 The maximum number of basic blocks on path that CSE considers.
11565 @item max-cse-insns
11566 The maximum number of instructions CSE processes before flushing.
11568 @item ggc-min-expand
11570 GCC uses a garbage collector to manage its own memory allocation. This
11571 parameter specifies the minimum percentage by which the garbage
11572 collector's heap should be allowed to expand between collections.
11573 Tuning this may improve compilation speed; it has no effect on code
11576 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
11577 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
11578 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
11579 GCC is not able to calculate RAM on a particular platform, the lower
11580 bound of 30% is used. Setting this parameter and
11581 @option{ggc-min-heapsize} to zero causes a full collection to occur at
11582 every opportunity. This is extremely slow, but can be useful for
11585 @item ggc-min-heapsize
11587 Minimum size of the garbage collector's heap before it begins bothering
11588 to collect garbage. The first collection occurs after the heap expands
11589 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
11590 tuning this may improve compilation speed, and has no effect on code
11593 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
11594 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
11595 with a lower bound of 4096 (four megabytes) and an upper bound of
11596 131072 (128 megabytes). If GCC is not able to calculate RAM on a
11597 particular platform, the lower bound is used. Setting this parameter
11598 very large effectively disables garbage collection. Setting this
11599 parameter and @option{ggc-min-expand} to zero causes a full collection
11600 to occur at every opportunity.
11602 @item max-reload-search-insns
11603 The maximum number of instruction reload should look backward for equivalent
11604 register. Increasing values mean more aggressive optimization, making the
11605 compilation time increase with probably slightly better performance.
11607 @item max-cselib-memory-locations
11608 The maximum number of memory locations cselib should take into account.
11609 Increasing values mean more aggressive optimization, making the compilation time
11610 increase with probably slightly better performance.
11612 @item max-sched-ready-insns
11613 The maximum number of instructions ready to be issued the scheduler should
11614 consider at any given time during the first scheduling pass. Increasing
11615 values mean more thorough searches, making the compilation time increase
11616 with probably little benefit.
11618 @item max-sched-region-blocks
11619 The maximum number of blocks in a region to be considered for
11620 interblock scheduling.
11622 @item max-pipeline-region-blocks
11623 The maximum number of blocks in a region to be considered for
11624 pipelining in the selective scheduler.
11626 @item max-sched-region-insns
11627 The maximum number of insns in a region to be considered for
11628 interblock scheduling.
11630 @item max-pipeline-region-insns
11631 The maximum number of insns in a region to be considered for
11632 pipelining in the selective scheduler.
11634 @item min-spec-prob
11635 The minimum probability (in percents) of reaching a source block
11636 for interblock speculative scheduling.
11638 @item max-sched-extend-regions-iters
11639 The maximum number of iterations through CFG to extend regions.
11640 A value of 0 disables region extensions.
11642 @item max-sched-insn-conflict-delay
11643 The maximum conflict delay for an insn to be considered for speculative motion.
11645 @item sched-spec-prob-cutoff
11646 The minimal probability of speculation success (in percents), so that
11647 speculative insns are scheduled.
11649 @item sched-state-edge-prob-cutoff
11650 The minimum probability an edge must have for the scheduler to save its
11653 @item sched-mem-true-dep-cost
11654 Minimal distance (in CPU cycles) between store and load targeting same
11657 @item selsched-max-lookahead
11658 The maximum size of the lookahead window of selective scheduling. It is a
11659 depth of search for available instructions.
11661 @item selsched-max-sched-times
11662 The maximum number of times that an instruction is scheduled during
11663 selective scheduling. This is the limit on the number of iterations
11664 through which the instruction may be pipelined.
11666 @item selsched-insns-to-rename
11667 The maximum number of best instructions in the ready list that are considered
11668 for renaming in the selective scheduler.
11671 The minimum value of stage count that swing modulo scheduler
11674 @item max-last-value-rtl
11675 The maximum size measured as number of RTLs that can be recorded in an expression
11676 in combiner for a pseudo register as last known value of that register.
11678 @item max-combine-insns
11679 The maximum number of instructions the RTL combiner tries to combine.
11681 @item integer-share-limit
11682 Small integer constants can use a shared data structure, reducing the
11683 compiler's memory usage and increasing its speed. This sets the maximum
11684 value of a shared integer constant.
11686 @item ssp-buffer-size
11687 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
11688 protection when @option{-fstack-protection} is used.
11690 @item min-size-for-stack-sharing
11691 The minimum size of variables taking part in stack slot sharing when not
11694 @item max-jump-thread-duplication-stmts
11695 Maximum number of statements allowed in a block that needs to be
11696 duplicated when threading jumps.
11698 @item max-fields-for-field-sensitive
11699 Maximum number of fields in a structure treated in
11700 a field sensitive manner during pointer analysis.
11702 @item prefetch-latency
11703 Estimate on average number of instructions that are executed before
11704 prefetch finishes. The distance prefetched ahead is proportional
11705 to this constant. Increasing this number may also lead to less
11706 streams being prefetched (see @option{simultaneous-prefetches}).
11708 @item simultaneous-prefetches
11709 Maximum number of prefetches that can run at the same time.
11711 @item l1-cache-line-size
11712 The size of cache line in L1 data cache, in bytes.
11714 @item l1-cache-size
11715 The size of L1 data cache, in kilobytes.
11717 @item l2-cache-size
11718 The size of L2 data cache, in kilobytes.
11720 @item prefetch-dynamic-strides
11721 Whether the loop array prefetch pass should issue software prefetch hints
11722 for strides that are non-constant. In some cases this may be
11723 beneficial, though the fact the stride is non-constant may make it
11724 hard to predict when there is clear benefit to issuing these hints.
11726 Set to 1 if the prefetch hints should be issued for non-constant
11727 strides. Set to 0 if prefetch hints should be issued only for strides that
11728 are known to be constant and below @option{prefetch-minimum-stride}.
11730 @item prefetch-minimum-stride
11731 Minimum constant stride, in bytes, to start using prefetch hints for. If
11732 the stride is less than this threshold, prefetch hints will not be issued.
11734 This setting is useful for processors that have hardware prefetchers, in
11735 which case there may be conflicts between the hardware prefetchers and
11736 the software prefetchers. If the hardware prefetchers have a maximum
11737 stride they can handle, it should be used here to improve the use of
11738 software prefetchers.
11740 A value of -1 means we don't have a threshold and therefore
11741 prefetch hints can be issued for any constant stride.
11743 This setting is only useful for strides that are known and constant.
11745 @item loop-interchange-max-num-stmts
11746 The maximum number of stmts in a loop to be interchanged.
11748 @item loop-interchange-stride-ratio
11749 The minimum ratio between stride of two loops for interchange to be profitable.
11751 @item min-insn-to-prefetch-ratio
11752 The minimum ratio between the number of instructions and the
11753 number of prefetches to enable prefetching in a loop.
11755 @item prefetch-min-insn-to-mem-ratio
11756 The minimum ratio between the number of instructions and the
11757 number of memory references to enable prefetching in a loop.
11759 @item use-canonical-types
11760 Whether the compiler should use the ``canonical'' type system.
11761 Should always be 1, which uses a more efficient internal
11762 mechanism for comparing types in C++ and Objective-C++. However, if
11763 bugs in the canonical type system are causing compilation failures,
11764 set this value to 0 to disable canonical types.
11766 @item switch-conversion-max-branch-ratio
11767 Switch initialization conversion refuses to create arrays that are
11768 bigger than @option{switch-conversion-max-branch-ratio} times the number of
11769 branches in the switch.
11771 @item max-partial-antic-length
11772 Maximum length of the partial antic set computed during the tree
11773 partial redundancy elimination optimization (@option{-ftree-pre}) when
11774 optimizing at @option{-O3} and above. For some sorts of source code
11775 the enhanced partial redundancy elimination optimization can run away,
11776 consuming all of the memory available on the host machine. This
11777 parameter sets a limit on the length of the sets that are computed,
11778 which prevents the runaway behavior. Setting a value of 0 for
11779 this parameter allows an unlimited set length.
11781 @item rpo-vn-max-loop-depth
11782 Maximum loop depth that is value-numbered optimistically.
11783 When the limit hits the innermost
11784 @var{rpo-vn-max-loop-depth} loops and the outermost loop in the
11785 loop nest are value-numbered optimistically and the remaining ones not.
11787 @item sccvn-max-alias-queries-per-access
11788 Maximum number of alias-oracle queries we perform when looking for
11789 redundancies for loads and stores. If this limit is hit the search
11790 is aborted and the load or store is not considered redundant. The
11791 number of queries is algorithmically limited to the number of
11792 stores on all paths from the load to the function entry.
11794 @item ira-max-loops-num
11795 IRA uses regional register allocation by default. If a function
11796 contains more loops than the number given by this parameter, only at most
11797 the given number of the most frequently-executed loops form regions
11798 for regional register allocation.
11800 @item ira-max-conflict-table-size
11801 Although IRA uses a sophisticated algorithm to compress the conflict
11802 table, the table can still require excessive amounts of memory for
11803 huge functions. If the conflict table for a function could be more
11804 than the size in MB given by this parameter, the register allocator
11805 instead uses a faster, simpler, and lower-quality
11806 algorithm that does not require building a pseudo-register conflict table.
11808 @item ira-loop-reserved-regs
11809 IRA can be used to evaluate more accurate register pressure in loops
11810 for decisions to move loop invariants (see @option{-O3}). The number
11811 of available registers reserved for some other purposes is given
11812 by this parameter. Default of the parameter
11813 is the best found from numerous experiments.
11815 @item lra-inheritance-ebb-probability-cutoff
11816 LRA tries to reuse values reloaded in registers in subsequent insns.
11817 This optimization is called inheritance. EBB is used as a region to
11818 do this optimization. The parameter defines a minimal fall-through
11819 edge probability in percentage used to add BB to inheritance EBB in
11820 LRA. The default value was chosen
11821 from numerous runs of SPEC2000 on x86-64.
11823 @item loop-invariant-max-bbs-in-loop
11824 Loop invariant motion can be very expensive, both in compilation time and
11825 in amount of needed compile-time memory, with very large loops. Loops
11826 with more basic blocks than this parameter won't have loop invariant
11827 motion optimization performed on them.
11829 @item loop-max-datarefs-for-datadeps
11830 Building data dependencies is expensive for very large loops. This
11831 parameter limits the number of data references in loops that are
11832 considered for data dependence analysis. These large loops are no
11833 handled by the optimizations using loop data dependencies.
11835 @item max-vartrack-size
11836 Sets a maximum number of hash table slots to use during variable
11837 tracking dataflow analysis of any function. If this limit is exceeded
11838 with variable tracking at assignments enabled, analysis for that
11839 function is retried without it, after removing all debug insns from
11840 the function. If the limit is exceeded even without debug insns, var
11841 tracking analysis is completely disabled for the function. Setting
11842 the parameter to zero makes it unlimited.
11844 @item max-vartrack-expr-depth
11845 Sets a maximum number of recursion levels when attempting to map
11846 variable names or debug temporaries to value expressions. This trades
11847 compilation time for more complete debug information. If this is set too
11848 low, value expressions that are available and could be represented in
11849 debug information may end up not being used; setting this higher may
11850 enable the compiler to find more complex debug expressions, but compile
11851 time and memory use may grow.
11853 @item max-debug-marker-count
11854 Sets a threshold on the number of debug markers (e.g.@: begin stmt
11855 markers) to avoid complexity explosion at inlining or expanding to RTL.
11856 If a function has more such gimple stmts than the set limit, such stmts
11857 will be dropped from the inlined copy of a function, and from its RTL
11860 @item min-nondebug-insn-uid
11861 Use uids starting at this parameter for nondebug insns. The range below
11862 the parameter is reserved exclusively for debug insns created by
11863 @option{-fvar-tracking-assignments}, but debug insns may get
11864 (non-overlapping) uids above it if the reserved range is exhausted.
11866 @item ipa-sra-ptr-growth-factor
11867 IPA-SRA replaces a pointer to an aggregate with one or more new
11868 parameters only when their cumulative size is less or equal to
11869 @option{ipa-sra-ptr-growth-factor} times the size of the original
11872 @item sra-max-scalarization-size-Ospeed
11873 @itemx sra-max-scalarization-size-Osize
11874 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11875 replace scalar parts of aggregates with uses of independent scalar
11876 variables. These parameters control the maximum size, in storage units,
11877 of aggregate which is considered for replacement when compiling for
11879 (@option{sra-max-scalarization-size-Ospeed}) or size
11880 (@option{sra-max-scalarization-size-Osize}) respectively.
11882 @item tm-max-aggregate-size
11883 When making copies of thread-local variables in a transaction, this
11884 parameter specifies the size in bytes after which variables are
11885 saved with the logging functions as opposed to save/restore code
11886 sequence pairs. This option only applies when using
11889 @item graphite-max-nb-scop-params
11890 To avoid exponential effects in the Graphite loop transforms, the
11891 number of parameters in a Static Control Part (SCoP) is bounded.
11892 A value of zero can be used to lift
11893 the bound. A variable whose value is unknown at compilation time and
11894 defined outside a SCoP is a parameter of the SCoP.
11896 @item loop-block-tile-size
11897 Loop blocking or strip mining transforms, enabled with
11898 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11899 loop in the loop nest by a given number of iterations. The strip
11900 length can be changed using the @option{loop-block-tile-size}
11903 @item ipa-cp-value-list-size
11904 IPA-CP attempts to track all possible values and types passed to a function's
11905 parameter in order to propagate them and perform devirtualization.
11906 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11907 stores per one formal parameter of a function.
11909 @item ipa-cp-eval-threshold
11910 IPA-CP calculates its own score of cloning profitability heuristics
11911 and performs those cloning opportunities with scores that exceed
11912 @option{ipa-cp-eval-threshold}.
11914 @item ipa-cp-recursion-penalty
11915 Percentage penalty the recursive functions will receive when they
11916 are evaluated for cloning.
11918 @item ipa-cp-single-call-penalty
11919 Percentage penalty functions containing a single call to another
11920 function will receive when they are evaluated for cloning.
11922 @item ipa-max-agg-items
11923 IPA-CP is also capable to propagate a number of scalar values passed
11924 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11925 number of such values per one parameter.
11927 @item ipa-cp-loop-hint-bonus
11928 When IPA-CP determines that a cloning candidate would make the number
11929 of iterations of a loop known, it adds a bonus of
11930 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11933 @item ipa-max-aa-steps
11934 During its analysis of function bodies, IPA-CP employs alias analysis
11935 in order to track values pointed to by function parameters. In order
11936 not spend too much time analyzing huge functions, it gives up and
11937 consider all memory clobbered after examining
11938 @option{ipa-max-aa-steps} statements modifying memory.
11940 @item ipa-max-switch-predicate-bounds
11941 Maximal number of boundary endpoints of case ranges of switch statement.
11942 For switch exceeding this limit, IPA-CP will not construct cloning cost
11943 predicate, which is used to estimate cloning benefit, for default case
11944 of the switch statement.
11946 @item lto-partitions
11947 Specify desired number of partitions produced during WHOPR compilation.
11948 The number of partitions should exceed the number of CPUs used for compilation.
11950 @item lto-min-partition
11951 Size of minimal partition for WHOPR (in estimated instructions).
11952 This prevents expenses of splitting very small programs into too many
11955 @item lto-max-partition
11956 Size of max partition for WHOPR (in estimated instructions).
11957 to provide an upper bound for individual size of partition.
11958 Meant to be used only with balanced partitioning.
11960 @item lto-max-streaming-parallelism
11961 Maximal number of parallel processes used for LTO streaming.
11963 @item cxx-max-namespaces-for-diagnostic-help
11964 The maximum number of namespaces to consult for suggestions when C++
11965 name lookup fails for an identifier.
11967 @item sink-frequency-threshold
11968 The maximum relative execution frequency (in percents) of the target block
11969 relative to a statement's original block to allow statement sinking of a
11970 statement. Larger numbers result in more aggressive statement sinking.
11971 A small positive adjustment is applied for
11972 statements with memory operands as those are even more profitable so sink.
11974 @item max-stores-to-sink
11975 The maximum number of conditional store pairs that can be sunk. Set to 0
11976 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11977 (@option{-ftree-loop-if-convert}) is disabled.
11979 @item allow-store-data-races
11980 Allow optimizers to introduce new data races on stores.
11981 Set to 1 to allow, otherwise to 0.
11983 @item case-values-threshold
11984 The smallest number of different values for which it is best to use a
11985 jump-table instead of a tree of conditional branches. If the value is
11986 0, use the default for the machine.
11988 @item jump-table-max-growth-ratio-for-size
11989 The maximum code size growth ratio when expanding
11990 into a jump table (in percent). The parameter is used when
11991 optimizing for size.
11993 @item jump-table-max-growth-ratio-for-speed
11994 The maximum code size growth ratio when expanding
11995 into a jump table (in percent). The parameter is used when
11996 optimizing for speed.
11998 @item tree-reassoc-width
11999 Set the maximum number of instructions executed in parallel in
12000 reassociated tree. This parameter overrides target dependent
12001 heuristics used by default if has non zero value.
12003 @item sched-pressure-algorithm
12004 Choose between the two available implementations of
12005 @option{-fsched-pressure}. Algorithm 1 is the original implementation
12006 and is the more likely to prevent instructions from being reordered.
12007 Algorithm 2 was designed to be a compromise between the relatively
12008 conservative approach taken by algorithm 1 and the rather aggressive
12009 approach taken by the default scheduler. It relies more heavily on
12010 having a regular register file and accurate register pressure classes.
12011 See @file{haifa-sched.c} in the GCC sources for more details.
12013 The default choice depends on the target.
12015 @item max-slsr-cand-scan
12016 Set the maximum number of existing candidates that are considered when
12017 seeking a basis for a new straight-line strength reduction candidate.
12020 Enable buffer overflow detection for global objects. This kind
12021 of protection is enabled by default if you are using
12022 @option{-fsanitize=address} option.
12023 To disable global objects protection use @option{--param asan-globals=0}.
12026 Enable buffer overflow detection for stack objects. This kind of
12027 protection is enabled by default when using @option{-fsanitize=address}.
12028 To disable stack protection use @option{--param asan-stack=0} option.
12030 @item asan-instrument-reads
12031 Enable buffer overflow detection for memory reads. This kind of
12032 protection is enabled by default when using @option{-fsanitize=address}.
12033 To disable memory reads protection use
12034 @option{--param asan-instrument-reads=0}.
12036 @item asan-instrument-writes
12037 Enable buffer overflow detection for memory writes. This kind of
12038 protection is enabled by default when using @option{-fsanitize=address}.
12039 To disable memory writes protection use
12040 @option{--param asan-instrument-writes=0} option.
12042 @item asan-memintrin
12043 Enable detection for built-in functions. This kind of protection
12044 is enabled by default when using @option{-fsanitize=address}.
12045 To disable built-in functions protection use
12046 @option{--param asan-memintrin=0}.
12048 @item asan-use-after-return
12049 Enable detection of use-after-return. This kind of protection
12050 is enabled by default when using the @option{-fsanitize=address} option.
12051 To disable it use @option{--param asan-use-after-return=0}.
12053 Note: By default the check is disabled at run time. To enable it,
12054 add @code{detect_stack_use_after_return=1} to the environment variable
12055 @env{ASAN_OPTIONS}.
12057 @item asan-instrumentation-with-call-threshold
12058 If number of memory accesses in function being instrumented
12059 is greater or equal to this number, use callbacks instead of inline checks.
12060 E.g. to disable inline code use
12061 @option{--param asan-instrumentation-with-call-threshold=0}.
12063 @item use-after-scope-direct-emission-threshold
12064 If the size of a local variable in bytes is smaller or equal to this
12065 number, directly poison (or unpoison) shadow memory instead of using
12066 run-time callbacks.
12068 @item max-fsm-thread-path-insns
12069 Maximum number of instructions to copy when duplicating blocks on a
12070 finite state automaton jump thread path.
12072 @item max-fsm-thread-length
12073 Maximum number of basic blocks on a finite state automaton jump thread
12076 @item max-fsm-thread-paths
12077 Maximum number of new jump thread paths to create for a finite state
12080 @item parloops-chunk-size
12081 Chunk size of omp schedule for loops parallelized by parloops.
12083 @item parloops-schedule
12084 Schedule type of omp schedule for loops parallelized by parloops (static,
12085 dynamic, guided, auto, runtime).
12087 @item parloops-min-per-thread
12088 The minimum number of iterations per thread of an innermost parallelized
12089 loop for which the parallelized variant is preferred over the single threaded
12090 one. Note that for a parallelized loop nest the
12091 minimum number of iterations of the outermost loop per thread is two.
12093 @item max-ssa-name-query-depth
12094 Maximum depth of recursion when querying properties of SSA names in things
12095 like fold routines. One level of recursion corresponds to following a
12098 @item hsa-gen-debug-stores
12099 Enable emission of special debug stores within HSA kernels which are
12100 then read and reported by libgomp plugin. Generation of these stores
12101 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
12104 @item max-speculative-devirt-maydefs
12105 The maximum number of may-defs we analyze when looking for a must-def
12106 specifying the dynamic type of an object that invokes a virtual call
12107 we may be able to devirtualize speculatively.
12109 @item max-vrp-switch-assertions
12110 The maximum number of assertions to add along the default edge of a switch
12111 statement during VRP.
12113 @item unroll-jam-min-percent
12114 The minimum percentage of memory references that must be optimized
12115 away for the unroll-and-jam transformation to be considered profitable.
12117 @item unroll-jam-max-unroll
12118 The maximum number of times the outer loop should be unrolled by
12119 the unroll-and-jam transformation.
12121 @item max-rtl-if-conversion-unpredictable-cost
12122 Maximum permissible cost for the sequence that would be generated
12123 by the RTL if-conversion pass for a branch that is considered unpredictable.
12125 @item max-variable-expansions-in-unroller
12126 If @option{-fvariable-expansion-in-unroller} is used, the maximum number
12127 of times that an individual variable will be expanded during loop unrolling.
12129 @item tracer-min-branch-probability-feedback
12130 Stop forward growth if the probability of best edge is less than
12131 this threshold (in percent). Used when profile feedback is available.
12133 @item partial-inlining-entry-probability
12134 Maximum probability of the entry BB of split region
12135 (in percent relative to entry BB of the function)
12136 to make partial inlining happen.
12138 @item max-tracked-strlens
12139 Maximum number of strings for which strlen optimization pass will
12140 track string lengths.
12142 @item gcse-after-reload-partial-fraction
12143 The threshold ratio for performing partial redundancy
12144 elimination after reload.
12146 @item gcse-after-reload-critical-fraction
12147 The threshold ratio of critical edges execution count that
12148 permit performing redundancy elimination after reload.
12150 @item max-loop-header-insns
12151 The maximum number of insns in loop header duplicated
12152 by the copy loop headers pass.
12154 @item vect-epilogues-nomask
12155 Enable loop epilogue vectorization using smaller vector size.
12157 @item slp-max-insns-in-bb
12158 Maximum number of instructions in basic block to be
12159 considered for SLP vectorization.
12161 @item avoid-fma-max-bits
12162 Maximum number of bits for which we avoid creating FMAs.
12164 @item sms-loop-average-count-threshold
12165 A threshold on the average loop count considered by the swing modulo scheduler.
12167 @item sms-dfa-history
12168 The number of cycles the swing modulo scheduler considers when checking
12169 conflicts using DFA.
12171 @item max-inline-insns-recursive-auto
12172 The maximum number of instructions non-inline function
12173 can grow to via recursive inlining.
12175 @item graphite-allow-codegen-errors
12176 Whether codegen errors should be ICEs when @option{-fchecking}.
12178 @item sms-max-ii-factor
12179 A factor for tuning the upper bound that swing modulo scheduler
12180 uses for scheduling a loop.
12182 @item lra-max-considered-reload-pseudos
12183 The max number of reload pseudos which are considered during
12184 spilling a non-reload pseudo.
12186 @item max-pow-sqrt-depth
12187 Maximum depth of sqrt chains to use when synthesizing exponentiation
12188 by a real constant.
12190 @item max-dse-active-local-stores
12191 Maximum number of active local stores in RTL dead store elimination.
12193 @item asan-instrument-allocas
12194 Enable asan allocas/VLAs protection.
12196 @item max-iterations-computation-cost
12197 Bound on the cost of an expression to compute the number of iterations.
12199 @item max-isl-operations
12200 Maximum number of isl operations, 0 means unlimited.
12202 @item graphite-max-arrays-per-scop
12203 Maximum number of arrays per scop.
12205 @item max-vartrack-reverse-op-size
12206 Max. size of loc list for which reverse ops should be added.
12208 @item tracer-dynamic-coverage-feedback
12209 The percentage of function, weighted by execution frequency,
12210 that must be covered by trace formation.
12211 Used when profile feedback is available.
12213 @item max-inline-recursive-depth-auto
12214 The maximum depth of recursive inlining for non-inline functions.
12216 @item fsm-scale-path-stmts
12217 Scale factor to apply to the number of statements in a threading path
12218 when comparing to the number of (scaled) blocks.
12220 @item fsm-maximum-phi-arguments
12221 Maximum number of arguments a PHI may have before the FSM threader
12222 will not try to thread through its block.
12224 @item uninit-control-dep-attempts
12225 Maximum number of nested calls to search for control dependencies
12226 during uninitialized variable analysis.
12228 @item max-once-peeled-insns
12229 The maximum number of insns of a peeled loop that rolls only once.
12231 @item sra-max-scalarization-size-Osize
12232 Maximum size, in storage units, of an aggregate
12233 which should be considered for scalarization when compiling for size.
12235 @item fsm-scale-path-blocks
12236 Scale factor to apply to the number of blocks in a threading path
12237 when comparing to the number of (scaled) statements.
12239 @item sched-autopref-queue-depth
12240 Hardware autoprefetcher scheduler model control flag.
12241 Number of lookahead cycles the model looks into; at '
12242 ' only enable instruction sorting heuristic.
12244 @item loop-versioning-max-inner-insns
12245 The maximum number of instructions that an inner loop can have
12246 before the loop versioning pass considers it too big to copy.
12248 @item loop-versioning-max-outer-insns
12249 The maximum number of instructions that an outer loop can have
12250 before the loop versioning pass considers it too big to copy,
12251 discounting any instructions in inner loops that directly benefit
12254 @item ssa-name-def-chain-limit
12255 The maximum number of SSA_NAME assignments to follow in determining
12256 a property of a variable such as its value. This limits the number
12257 of iterations or recursive calls GCC performs when optimizing certain
12258 statements or when determining their validity prior to issuing
12264 @node Instrumentation Options
12265 @section Program Instrumentation Options
12266 @cindex instrumentation options
12267 @cindex program instrumentation options
12268 @cindex run-time error checking options
12269 @cindex profiling options
12270 @cindex options, program instrumentation
12271 @cindex options, run-time error checking
12272 @cindex options, profiling
12274 GCC supports a number of command-line options that control adding
12275 run-time instrumentation to the code it normally generates.
12276 For example, one purpose of instrumentation is collect profiling
12277 statistics for use in finding program hot spots, code coverage
12278 analysis, or profile-guided optimizations.
12279 Another class of program instrumentation is adding run-time checking
12280 to detect programming errors like invalid pointer
12281 dereferences or out-of-bounds array accesses, as well as deliberately
12282 hostile attacks such as stack smashing or C++ vtable hijacking.
12283 There is also a general hook which can be used to implement other
12284 forms of tracing or function-level instrumentation for debug or
12285 program analysis purposes.
12288 @cindex @command{prof}
12289 @cindex @command{gprof}
12294 Generate extra code to write profile information suitable for the
12295 analysis program @command{prof} (for @option{-p}) or @command{gprof}
12296 (for @option{-pg}). You must use this option when compiling
12297 the source files you want data about, and you must also use it when
12300 You can use the function attribute @code{no_instrument_function} to
12301 suppress profiling of individual functions when compiling with these options.
12302 @xref{Common Function Attributes}.
12304 @item -fprofile-arcs
12305 @opindex fprofile-arcs
12306 Add code so that program flow @dfn{arcs} are instrumented. During
12307 execution the program records how many times each branch and call is
12308 executed and how many times it is taken or returns. On targets that support
12309 constructors with priority support, profiling properly handles constructors,
12310 destructors and C++ constructors (and destructors) of classes which are used
12311 as a type of a global variable.
12314 program exits it saves this data to a file called
12315 @file{@var{auxname}.gcda} for each source file. The data may be used for
12316 profile-directed optimizations (@option{-fbranch-probabilities}), or for
12317 test coverage analysis (@option{-ftest-coverage}). Each object file's
12318 @var{auxname} is generated from the name of the output file, if
12319 explicitly specified and it is not the final executable, otherwise it is
12320 the basename of the source file. In both cases any suffix is removed
12321 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
12322 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
12323 @xref{Cross-profiling}.
12325 @cindex @command{gcov}
12329 This option is used to compile and link code instrumented for coverage
12330 analysis. The option is a synonym for @option{-fprofile-arcs}
12331 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
12332 linking). See the documentation for those options for more details.
12337 Compile the source files with @option{-fprofile-arcs} plus optimization
12338 and code generation options. For test coverage analysis, use the
12339 additional @option{-ftest-coverage} option. You do not need to profile
12340 every source file in a program.
12343 Compile the source files additionally with @option{-fprofile-abs-path}
12344 to create absolute path names in the @file{.gcno} files. This allows
12345 @command{gcov} to find the correct sources in projects where compilations
12346 occur with different working directories.
12349 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
12350 (the latter implies the former).
12353 Run the program on a representative workload to generate the arc profile
12354 information. This may be repeated any number of times. You can run
12355 concurrent instances of your program, and provided that the file system
12356 supports locking, the data files will be correctly updated. Unless
12357 a strict ISO C dialect option is in effect, @code{fork} calls are
12358 detected and correctly handled without double counting.
12361 For profile-directed optimizations, compile the source files again with
12362 the same optimization and code generation options plus
12363 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
12364 Control Optimization}).
12367 For test coverage analysis, use @command{gcov} to produce human readable
12368 information from the @file{.gcno} and @file{.gcda} files. Refer to the
12369 @command{gcov} documentation for further information.
12373 With @option{-fprofile-arcs}, for each function of your program GCC
12374 creates a program flow graph, then finds a spanning tree for the graph.
12375 Only arcs that are not on the spanning tree have to be instrumented: the
12376 compiler adds code to count the number of times that these arcs are
12377 executed. When an arc is the only exit or only entrance to a block, the
12378 instrumentation code can be added to the block; otherwise, a new basic
12379 block must be created to hold the instrumentation code.
12382 @item -ftest-coverage
12383 @opindex ftest-coverage
12384 Produce a notes file that the @command{gcov} code-coverage utility
12385 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
12386 show program coverage. Each source file's note file is called
12387 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
12388 above for a description of @var{auxname} and instructions on how to
12389 generate test coverage data. Coverage data matches the source files
12390 more closely if you do not optimize.
12392 @item -fprofile-abs-path
12393 @opindex fprofile-abs-path
12394 Automatically convert relative source file names to absolute path names
12395 in the @file{.gcno} files. This allows @command{gcov} to find the correct
12396 sources in projects where compilations occur with different working
12399 @item -fprofile-dir=@var{path}
12400 @opindex fprofile-dir
12402 Set the directory to search for the profile data files in to @var{path}.
12403 This option affects only the profile data generated by
12404 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
12405 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
12406 and its related options. Both absolute and relative paths can be used.
12407 By default, GCC uses the current directory as @var{path}, thus the
12408 profile data file appears in the same directory as the object file.
12409 In order to prevent the file name clashing, if the object file name is
12410 not an absolute path, we mangle the absolute path of the
12411 @file{@var{sourcename}.gcda} file and use it as the file name of a
12412 @file{.gcda} file. See similar option @option{-fprofile-note}.
12414 When an executable is run in a massive parallel environment, it is recommended
12415 to save profile to different folders. That can be done with variables
12416 in @var{path} that are exported during run-time:
12424 value of environment variable @var{VAR}
12428 @item -fprofile-generate
12429 @itemx -fprofile-generate=@var{path}
12430 @opindex fprofile-generate
12432 Enable options usually used for instrumenting application to produce
12433 profile useful for later recompilation with profile feedback based
12434 optimization. You must use @option{-fprofile-generate} both when
12435 compiling and when linking your program.
12437 The following options are enabled:
12438 @option{-fprofile-arcs}, @option{-fprofile-values},
12439 @option{-finline-functions}, and @option{-fipa-bit-cp}.
12441 If @var{path} is specified, GCC looks at the @var{path} to find
12442 the profile feedback data files. See @option{-fprofile-dir}.
12444 To optimize the program based on the collected profile information, use
12445 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
12447 @item -fprofile-note=@var{path}
12448 @opindex fprofile-note
12450 If @var{path} is specified, GCC saves @file{.gcno} file into @var{path}
12451 location. If you combine the option with multiple source files,
12452 the @file{.gcno} file will be overwritten.
12454 @item -fprofile-update=@var{method}
12455 @opindex fprofile-update
12457 Alter the update method for an application instrumented for profile
12458 feedback based optimization. The @var{method} argument should be one of
12459 @samp{single}, @samp{atomic} or @samp{prefer-atomic}.
12460 The first one is useful for single-threaded applications,
12461 while the second one prevents profile corruption by emitting thread-safe code.
12463 @strong{Warning:} When an application does not properly join all threads
12464 (or creates an detached thread), a profile file can be still corrupted.
12466 Using @samp{prefer-atomic} would be transformed either to @samp{atomic},
12467 when supported by a target, or to @samp{single} otherwise. The GCC driver
12468 automatically selects @samp{prefer-atomic} when @option{-pthread}
12469 is present in the command line.
12471 @item -fprofile-filter-files=@var{regex}
12472 @opindex fprofile-filter-files
12474 Instrument only functions from files where names match
12475 any regular expression (separated by a semi-colon).
12477 For example, @option{-fprofile-filter-files=main.c;module.*.c} will instrument
12478 only @file{main.c} and all C files starting with 'module'.
12480 @item -fprofile-exclude-files=@var{regex}
12481 @opindex fprofile-exclude-files
12483 Instrument only functions from files where names do not match
12484 all the regular expressions (separated by a semi-colon).
12486 For example, @option{-fprofile-exclude-files=/usr/*} will prevent instrumentation
12487 of all files that are located in @file{/usr/} folder.
12489 @item -fsanitize=address
12490 @opindex fsanitize=address
12491 Enable AddressSanitizer, a fast memory error detector.
12492 Memory access instructions are instrumented to detect
12493 out-of-bounds and use-after-free bugs.
12494 The option enables @option{-fsanitize-address-use-after-scope}.
12495 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
12496 more details. The run-time behavior can be influenced using the
12497 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
12498 the available options are shown at startup of the instrumented program. See
12499 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
12500 for a list of supported options.
12501 The option cannot be combined with @option{-fsanitize=thread}.
12503 @item -fsanitize=kernel-address
12504 @opindex fsanitize=kernel-address
12505 Enable AddressSanitizer for Linux kernel.
12506 See @uref{https://github.com/google/kasan/wiki} for more details.
12508 @item -fsanitize=pointer-compare
12509 @opindex fsanitize=pointer-compare
12510 Instrument comparison operation (<, <=, >, >=) with pointer operands.
12511 The option must be combined with either @option{-fsanitize=kernel-address} or
12512 @option{-fsanitize=address}
12513 The option cannot be combined with @option{-fsanitize=thread}.
12514 Note: By default the check is disabled at run time. To enable it,
12515 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12516 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12517 invalid operation only when both pointers are non-null.
12519 @item -fsanitize=pointer-subtract
12520 @opindex fsanitize=pointer-subtract
12521 Instrument subtraction with pointer operands.
12522 The option must be combined with either @option{-fsanitize=kernel-address} or
12523 @option{-fsanitize=address}
12524 The option cannot be combined with @option{-fsanitize=thread}.
12525 Note: By default the check is disabled at run time. To enable it,
12526 add @code{detect_invalid_pointer_pairs=2} to the environment variable
12527 @env{ASAN_OPTIONS}. Using @code{detect_invalid_pointer_pairs=1} detects
12528 invalid operation only when both pointers are non-null.
12530 @item -fsanitize=thread
12531 @opindex fsanitize=thread
12532 Enable ThreadSanitizer, a fast data race detector.
12533 Memory access instructions are instrumented to detect
12534 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
12535 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
12536 environment variable; see
12537 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
12539 The option cannot be combined with @option{-fsanitize=address},
12540 @option{-fsanitize=leak}.
12542 Note that sanitized atomic builtins cannot throw exceptions when
12543 operating on invalid memory addresses with non-call exceptions
12544 (@option{-fnon-call-exceptions}).
12546 @item -fsanitize=leak
12547 @opindex fsanitize=leak
12548 Enable LeakSanitizer, a memory leak detector.
12549 This option only matters for linking of executables and
12550 the executable is linked against a library that overrides @code{malloc}
12551 and other allocator functions. See
12552 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
12553 details. The run-time behavior can be influenced using the
12554 @env{LSAN_OPTIONS} environment variable.
12555 The option cannot be combined with @option{-fsanitize=thread}.
12557 @item -fsanitize=undefined
12558 @opindex fsanitize=undefined
12559 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
12560 Various computations are instrumented to detect undefined behavior
12561 at runtime. Current suboptions are:
12565 @item -fsanitize=shift
12566 @opindex fsanitize=shift
12567 This option enables checking that the result of a shift operation is
12568 not undefined. Note that what exactly is considered undefined differs
12569 slightly between C and C++, as well as between ISO C90 and C99, etc.
12570 This option has two suboptions, @option{-fsanitize=shift-base} and
12571 @option{-fsanitize=shift-exponent}.
12573 @item -fsanitize=shift-exponent
12574 @opindex fsanitize=shift-exponent
12575 This option enables checking that the second argument of a shift operation
12576 is not negative and is smaller than the precision of the promoted first
12579 @item -fsanitize=shift-base
12580 @opindex fsanitize=shift-base
12581 If the second argument of a shift operation is within range, check that the
12582 result of a shift operation is not undefined. Note that what exactly is
12583 considered undefined differs slightly between C and C++, as well as between
12584 ISO C90 and C99, etc.
12586 @item -fsanitize=integer-divide-by-zero
12587 @opindex fsanitize=integer-divide-by-zero
12588 Detect integer division by zero as well as @code{INT_MIN / -1} division.
12590 @item -fsanitize=unreachable
12591 @opindex fsanitize=unreachable
12592 With this option, the compiler turns the @code{__builtin_unreachable}
12593 call into a diagnostics message call instead. When reaching the
12594 @code{__builtin_unreachable} call, the behavior is undefined.
12596 @item -fsanitize=vla-bound
12597 @opindex fsanitize=vla-bound
12598 This option instructs the compiler to check that the size of a variable
12599 length array is positive.
12601 @item -fsanitize=null
12602 @opindex fsanitize=null
12603 This option enables pointer checking. Particularly, the application
12604 built with this option turned on will issue an error message when it
12605 tries to dereference a NULL pointer, or if a reference (possibly an
12606 rvalue reference) is bound to a NULL pointer, or if a method is invoked
12607 on an object pointed by a NULL pointer.
12609 @item -fsanitize=return
12610 @opindex fsanitize=return
12611 This option enables return statement checking. Programs
12612 built with this option turned on will issue an error message
12613 when the end of a non-void function is reached without actually
12614 returning a value. This option works in C++ only.
12616 @item -fsanitize=signed-integer-overflow
12617 @opindex fsanitize=signed-integer-overflow
12618 This option enables signed integer overflow checking. We check that
12619 the result of @code{+}, @code{*}, and both unary and binary @code{-}
12620 does not overflow in the signed arithmetics. Note, integer promotion
12621 rules must be taken into account. That is, the following is not an
12624 signed char a = SCHAR_MAX;
12628 @item -fsanitize=bounds
12629 @opindex fsanitize=bounds
12630 This option enables instrumentation of array bounds. Various out of bounds
12631 accesses are detected. Flexible array members, flexible array member-like
12632 arrays, and initializers of variables with static storage are not instrumented.
12634 @item -fsanitize=bounds-strict
12635 @opindex fsanitize=bounds-strict
12636 This option enables strict instrumentation of array bounds. Most out of bounds
12637 accesses are detected, including flexible array members and flexible array
12638 member-like arrays. Initializers of variables with static storage are not
12641 @item -fsanitize=alignment
12642 @opindex fsanitize=alignment
12644 This option enables checking of alignment of pointers when they are
12645 dereferenced, or when a reference is bound to insufficiently aligned target,
12646 or when a method or constructor is invoked on insufficiently aligned object.
12648 @item -fsanitize=object-size
12649 @opindex fsanitize=object-size
12650 This option enables instrumentation of memory references using the
12651 @code{__builtin_object_size} function. Various out of bounds pointer
12652 accesses are detected.
12654 @item -fsanitize=float-divide-by-zero
12655 @opindex fsanitize=float-divide-by-zero
12656 Detect floating-point division by zero. Unlike other similar options,
12657 @option{-fsanitize=float-divide-by-zero} is not enabled by
12658 @option{-fsanitize=undefined}, since floating-point division by zero can
12659 be a legitimate way of obtaining infinities and NaNs.
12661 @item -fsanitize=float-cast-overflow
12662 @opindex fsanitize=float-cast-overflow
12663 This option enables floating-point type to integer conversion checking.
12664 We check that the result of the conversion does not overflow.
12665 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
12666 not enabled by @option{-fsanitize=undefined}.
12667 This option does not work well with @code{FE_INVALID} exceptions enabled.
12669 @item -fsanitize=nonnull-attribute
12670 @opindex fsanitize=nonnull-attribute
12672 This option enables instrumentation of calls, checking whether null values
12673 are not passed to arguments marked as requiring a non-null value by the
12674 @code{nonnull} function attribute.
12676 @item -fsanitize=returns-nonnull-attribute
12677 @opindex fsanitize=returns-nonnull-attribute
12679 This option enables instrumentation of return statements in functions
12680 marked with @code{returns_nonnull} function attribute, to detect returning
12681 of null values from such functions.
12683 @item -fsanitize=bool
12684 @opindex fsanitize=bool
12686 This option enables instrumentation of loads from bool. If a value other
12687 than 0/1 is loaded, a run-time error is issued.
12689 @item -fsanitize=enum
12690 @opindex fsanitize=enum
12692 This option enables instrumentation of loads from an enum type. If
12693 a value outside the range of values for the enum type is loaded,
12694 a run-time error is issued.
12696 @item -fsanitize=vptr
12697 @opindex fsanitize=vptr
12699 This option enables instrumentation of C++ member function calls, member
12700 accesses and some conversions between pointers to base and derived classes,
12701 to verify the referenced object has the correct dynamic type.
12703 @item -fsanitize=pointer-overflow
12704 @opindex fsanitize=pointer-overflow
12706 This option enables instrumentation of pointer arithmetics. If the pointer
12707 arithmetics overflows, a run-time error is issued.
12709 @item -fsanitize=builtin
12710 @opindex fsanitize=builtin
12712 This option enables instrumentation of arguments to selected builtin
12713 functions. If an invalid value is passed to such arguments, a run-time
12714 error is issued. E.g.@ passing 0 as the argument to @code{__builtin_ctz}
12715 or @code{__builtin_clz} invokes undefined behavior and is diagnosed
12720 While @option{-ftrapv} causes traps for signed overflows to be emitted,
12721 @option{-fsanitize=undefined} gives a diagnostic message.
12722 This currently works only for the C family of languages.
12724 @item -fno-sanitize=all
12725 @opindex fno-sanitize=all
12727 This option disables all previously enabled sanitizers.
12728 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
12731 @item -fasan-shadow-offset=@var{number}
12732 @opindex fasan-shadow-offset
12733 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
12734 It is useful for experimenting with different shadow memory layouts in
12735 Kernel AddressSanitizer.
12737 @item -fsanitize-sections=@var{s1},@var{s2},...
12738 @opindex fsanitize-sections
12739 Sanitize global variables in selected user-defined sections. @var{si} may
12742 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
12743 @opindex fsanitize-recover
12744 @opindex fno-sanitize-recover
12745 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
12746 mentioned in comma-separated list of @var{opts}. Enabling this option
12747 for a sanitizer component causes it to attempt to continue
12748 running the program as if no error happened. This means multiple
12749 runtime errors can be reported in a single program run, and the exit
12750 code of the program may indicate success even when errors
12751 have been reported. The @option{-fno-sanitize-recover=} option
12752 can be used to alter
12753 this behavior: only the first detected error is reported
12754 and program then exits with a non-zero exit code.
12756 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
12757 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
12758 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
12759 @option{-fsanitize=bounds-strict},
12760 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
12761 For these sanitizers error recovery is turned on by default,
12762 except @option{-fsanitize=address}, for which this feature is experimental.
12763 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
12764 accepted, the former enables recovery for all sanitizers that support it,
12765 the latter disables recovery for all sanitizers that support it.
12767 Even if a recovery mode is turned on the compiler side, it needs to be also
12768 enabled on the runtime library side, otherwise the failures are still fatal.
12769 The runtime library defaults to @code{halt_on_error=0} for
12770 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
12771 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
12772 setting the @code{halt_on_error} flag in the corresponding environment variable.
12774 Syntax without an explicit @var{opts} parameter is deprecated. It is
12775 equivalent to specifying an @var{opts} list of:
12778 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict
12781 @item -fsanitize-address-use-after-scope
12782 @opindex fsanitize-address-use-after-scope
12783 Enable sanitization of local variables to detect use-after-scope bugs.
12784 The option sets @option{-fstack-reuse} to @samp{none}.
12786 @item -fsanitize-undefined-trap-on-error
12787 @opindex fsanitize-undefined-trap-on-error
12788 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
12789 report undefined behavior using @code{__builtin_trap} rather than
12790 a @code{libubsan} library routine. The advantage of this is that the
12791 @code{libubsan} library is not needed and is not linked in, so this
12792 is usable even in freestanding environments.
12794 @item -fsanitize-coverage=trace-pc
12795 @opindex fsanitize-coverage=trace-pc
12796 Enable coverage-guided fuzzing code instrumentation.
12797 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
12799 @item -fsanitize-coverage=trace-cmp
12800 @opindex fsanitize-coverage=trace-cmp
12801 Enable dataflow guided fuzzing code instrumentation.
12802 Inserts a call to @code{__sanitizer_cov_trace_cmp1},
12803 @code{__sanitizer_cov_trace_cmp2}, @code{__sanitizer_cov_trace_cmp4} or
12804 @code{__sanitizer_cov_trace_cmp8} for integral comparison with both operands
12805 variable or @code{__sanitizer_cov_trace_const_cmp1},
12806 @code{__sanitizer_cov_trace_const_cmp2},
12807 @code{__sanitizer_cov_trace_const_cmp4} or
12808 @code{__sanitizer_cov_trace_const_cmp8} for integral comparison with one
12809 operand constant, @code{__sanitizer_cov_trace_cmpf} or
12810 @code{__sanitizer_cov_trace_cmpd} for float or double comparisons and
12811 @code{__sanitizer_cov_trace_switch} for switch statements.
12813 @item -fcf-protection=@r{[}full@r{|}branch@r{|}return@r{|}none@r{]}
12814 @opindex fcf-protection
12815 Enable code instrumentation of control-flow transfers to increase
12816 program security by checking that target addresses of control-flow
12817 transfer instructions (such as indirect function call, function return,
12818 indirect jump) are valid. This prevents diverting the flow of control
12819 to an unexpected target. This is intended to protect against such
12820 threats as Return-oriented Programming (ROP), and similarly
12821 call/jmp-oriented programming (COP/JOP).
12823 The value @code{branch} tells the compiler to implement checking of
12824 validity of control-flow transfer at the point of indirect branch
12825 instructions, i.e.@: call/jmp instructions. The value @code{return}
12826 implements checking of validity at the point of returning from a
12827 function. The value @code{full} is an alias for specifying both
12828 @code{branch} and @code{return}. The value @code{none} turns off
12831 The macro @code{__CET__} is defined when @option{-fcf-protection} is
12832 used. The first bit of @code{__CET__} is set to 1 for the value
12833 @code{branch} and the second bit of @code{__CET__} is set to 1 for
12836 You can also use the @code{nocf_check} attribute to identify
12837 which functions and calls should be skipped from instrumentation
12838 (@pxref{Function Attributes}).
12840 Currently the x86 GNU/Linux target provides an implementation based
12841 on Intel Control-flow Enforcement Technology (CET).
12843 @item -fstack-protector
12844 @opindex fstack-protector
12845 Emit extra code to check for buffer overflows, such as stack smashing
12846 attacks. This is done by adding a guard variable to functions with
12847 vulnerable objects. This includes functions that call @code{alloca}, and
12848 functions with buffers larger than 8 bytes. The guards are initialized
12849 when a function is entered and then checked when the function exits.
12850 If a guard check fails, an error message is printed and the program exits.
12852 @item -fstack-protector-all
12853 @opindex fstack-protector-all
12854 Like @option{-fstack-protector} except that all functions are protected.
12856 @item -fstack-protector-strong
12857 @opindex fstack-protector-strong
12858 Like @option{-fstack-protector} but includes additional functions to
12859 be protected --- those that have local array definitions, or have
12860 references to local frame addresses.
12862 @item -fstack-protector-explicit
12863 @opindex fstack-protector-explicit
12864 Like @option{-fstack-protector} but only protects those functions which
12865 have the @code{stack_protect} attribute.
12867 @item -fstack-check
12868 @opindex fstack-check
12869 Generate code to verify that you do not go beyond the boundary of the
12870 stack. You should specify this flag if you are running in an
12871 environment with multiple threads, but you only rarely need to specify it in
12872 a single-threaded environment since stack overflow is automatically
12873 detected on nearly all systems if there is only one stack.
12875 Note that this switch does not actually cause checking to be done; the
12876 operating system or the language runtime must do that. The switch causes
12877 generation of code to ensure that they see the stack being extended.
12879 You can additionally specify a string parameter: @samp{no} means no
12880 checking, @samp{generic} means force the use of old-style checking,
12881 @samp{specific} means use the best checking method and is equivalent
12882 to bare @option{-fstack-check}.
12884 Old-style checking is a generic mechanism that requires no specific
12885 target support in the compiler but comes with the following drawbacks:
12889 Modified allocation strategy for large objects: they are always
12890 allocated dynamically if their size exceeds a fixed threshold. Note this
12891 may change the semantics of some code.
12894 Fixed limit on the size of the static frame of functions: when it is
12895 topped by a particular function, stack checking is not reliable and
12896 a warning is issued by the compiler.
12899 Inefficiency: because of both the modified allocation strategy and the
12900 generic implementation, code performance is hampered.
12903 Note that old-style stack checking is also the fallback method for
12904 @samp{specific} if no target support has been added in the compiler.
12906 @samp{-fstack-check=} is designed for Ada's needs to detect infinite recursion
12907 and stack overflows. @samp{specific} is an excellent choice when compiling
12908 Ada code. It is not generally sufficient to protect against stack-clash
12909 attacks. To protect against those you want @samp{-fstack-clash-protection}.
12911 @item -fstack-clash-protection
12912 @opindex fstack-clash-protection
12913 Generate code to prevent stack clash style attacks. When this option is
12914 enabled, the compiler will only allocate one page of stack space at a time
12915 and each page is accessed immediately after allocation. Thus, it prevents
12916 allocations from jumping over any stack guard page provided by the
12919 Most targets do not fully support stack clash protection. However, on
12920 those targets @option{-fstack-clash-protection} will protect dynamic stack
12921 allocations. @option{-fstack-clash-protection} may also provide limited
12922 protection for static stack allocations if the target supports
12923 @option{-fstack-check=specific}.
12925 @item -fstack-limit-register=@var{reg}
12926 @itemx -fstack-limit-symbol=@var{sym}
12927 @itemx -fno-stack-limit
12928 @opindex fstack-limit-register
12929 @opindex fstack-limit-symbol
12930 @opindex fno-stack-limit
12931 Generate code to ensure that the stack does not grow beyond a certain value,
12932 either the value of a register or the address of a symbol. If a larger
12933 stack is required, a signal is raised at run time. For most targets,
12934 the signal is raised before the stack overruns the boundary, so
12935 it is possible to catch the signal without taking special precautions.
12937 For instance, if the stack starts at absolute address @samp{0x80000000}
12938 and grows downwards, you can use the flags
12939 @option{-fstack-limit-symbol=__stack_limit} and
12940 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12941 of 128KB@. Note that this may only work with the GNU linker.
12943 You can locally override stack limit checking by using the
12944 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
12946 @item -fsplit-stack
12947 @opindex fsplit-stack
12948 Generate code to automatically split the stack before it overflows.
12949 The resulting program has a discontiguous stack which can only
12950 overflow if the program is unable to allocate any more memory. This
12951 is most useful when running threaded programs, as it is no longer
12952 necessary to calculate a good stack size to use for each thread. This
12953 is currently only implemented for the x86 targets running
12956 When code compiled with @option{-fsplit-stack} calls code compiled
12957 without @option{-fsplit-stack}, there may not be much stack space
12958 available for the latter code to run. If compiling all code,
12959 including library code, with @option{-fsplit-stack} is not an option,
12960 then the linker can fix up these calls so that the code compiled
12961 without @option{-fsplit-stack} always has a large stack. Support for
12962 this is implemented in the gold linker in GNU binutils release 2.21
12965 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
12966 @opindex fvtable-verify
12967 This option is only available when compiling C++ code.
12968 It turns on (or off, if using @option{-fvtable-verify=none}) the security
12969 feature that verifies at run time, for every virtual call, that
12970 the vtable pointer through which the call is made is valid for the type of
12971 the object, and has not been corrupted or overwritten. If an invalid vtable
12972 pointer is detected at run time, an error is reported and execution of the
12973 program is immediately halted.
12975 This option causes run-time data structures to be built at program startup,
12976 which are used for verifying the vtable pointers.
12977 The options @samp{std} and @samp{preinit}
12978 control the timing of when these data structures are built. In both cases the
12979 data structures are built before execution reaches @code{main}. Using
12980 @option{-fvtable-verify=std} causes the data structures to be built after
12981 shared libraries have been loaded and initialized.
12982 @option{-fvtable-verify=preinit} causes them to be built before shared
12983 libraries have been loaded and initialized.
12985 If this option appears multiple times in the command line with different
12986 values specified, @samp{none} takes highest priority over both @samp{std} and
12987 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
12990 @opindex fvtv-debug
12991 When used in conjunction with @option{-fvtable-verify=std} or
12992 @option{-fvtable-verify=preinit}, causes debug versions of the
12993 runtime functions for the vtable verification feature to be called.
12994 This flag also causes the compiler to log information about which
12995 vtable pointers it finds for each class.
12996 This information is written to a file named @file{vtv_set_ptr_data.log}
12997 in the directory named by the environment variable @env{VTV_LOGS_DIR}
12998 if that is defined or the current working directory otherwise.
13000 Note: This feature @emph{appends} data to the log file. If you want a fresh log
13001 file, be sure to delete any existing one.
13004 @opindex fvtv-counts
13005 This is a debugging flag. When used in conjunction with
13006 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
13007 causes the compiler to keep track of the total number of virtual calls
13008 it encounters and the number of verifications it inserts. It also
13009 counts the number of calls to certain run-time library functions
13010 that it inserts and logs this information for each compilation unit.
13011 The compiler writes this information to a file named
13012 @file{vtv_count_data.log} in the directory named by the environment
13013 variable @env{VTV_LOGS_DIR} if that is defined or the current working
13014 directory otherwise. It also counts the size of the vtable pointer sets
13015 for each class, and writes this information to @file{vtv_class_set_sizes.log}
13016 in the same directory.
13018 Note: This feature @emph{appends} data to the log files. To get fresh log
13019 files, be sure to delete any existing ones.
13021 @item -finstrument-functions
13022 @opindex finstrument-functions
13023 Generate instrumentation calls for entry and exit to functions. Just
13024 after function entry and just before function exit, the following
13025 profiling functions are called with the address of the current
13026 function and its call site. (On some platforms,
13027 @code{__builtin_return_address} does not work beyond the current
13028 function, so the call site information may not be available to the
13029 profiling functions otherwise.)
13032 void __cyg_profile_func_enter (void *this_fn,
13034 void __cyg_profile_func_exit (void *this_fn,
13038 The first argument is the address of the start of the current function,
13039 which may be looked up exactly in the symbol table.
13041 This instrumentation is also done for functions expanded inline in other
13042 functions. The profiling calls indicate where, conceptually, the
13043 inline function is entered and exited. This means that addressable
13044 versions of such functions must be available. If all your uses of a
13045 function are expanded inline, this may mean an additional expansion of
13046 code size. If you use @code{extern inline} in your C code, an
13047 addressable version of such functions must be provided. (This is
13048 normally the case anyway, but if you get lucky and the optimizer always
13049 expands the functions inline, you might have gotten away without
13050 providing static copies.)
13052 A function may be given the attribute @code{no_instrument_function}, in
13053 which case this instrumentation is not done. This can be used, for
13054 example, for the profiling functions listed above, high-priority
13055 interrupt routines, and any functions from which the profiling functions
13056 cannot safely be called (perhaps signal handlers, if the profiling
13057 routines generate output or allocate memory).
13058 @xref{Common Function Attributes}.
13060 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
13061 @opindex finstrument-functions-exclude-file-list
13063 Set the list of functions that are excluded from instrumentation (see
13064 the description of @option{-finstrument-functions}). If the file that
13065 contains a function definition matches with one of @var{file}, then
13066 that function is not instrumented. The match is done on substrings:
13067 if the @var{file} parameter is a substring of the file name, it is
13068 considered to be a match.
13073 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
13077 excludes any inline function defined in files whose pathnames
13078 contain @file{/bits/stl} or @file{include/sys}.
13080 If, for some reason, you want to include letter @samp{,} in one of
13081 @var{sym}, write @samp{\,}. For example,
13082 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
13083 (note the single quote surrounding the option).
13085 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
13086 @opindex finstrument-functions-exclude-function-list
13088 This is similar to @option{-finstrument-functions-exclude-file-list},
13089 but this option sets the list of function names to be excluded from
13090 instrumentation. The function name to be matched is its user-visible
13091 name, such as @code{vector<int> blah(const vector<int> &)}, not the
13092 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
13093 match is done on substrings: if the @var{sym} parameter is a substring
13094 of the function name, it is considered to be a match. For C99 and C++
13095 extended identifiers, the function name must be given in UTF-8, not
13096 using universal character names.
13098 @item -fpatchable-function-entry=@var{N}[,@var{M}]
13099 @opindex fpatchable-function-entry
13100 Generate @var{N} NOPs right at the beginning
13101 of each function, with the function entry point before the @var{M}th NOP.
13102 If @var{M} is omitted, it defaults to @code{0} so the
13103 function entry points to the address just at the first NOP.
13104 The NOP instructions reserve extra space which can be used to patch in
13105 any desired instrumentation at run time, provided that the code segment
13106 is writable. The amount of space is controllable indirectly via
13107 the number of NOPs; the NOP instruction used corresponds to the instruction
13108 emitted by the internal GCC back-end interface @code{gen_nop}. This behavior
13109 is target-specific and may also depend on the architecture variant and/or
13110 other compilation options.
13112 For run-time identification, the starting addresses of these areas,
13113 which correspond to their respective function entries minus @var{M},
13114 are additionally collected in the @code{__patchable_function_entries}
13115 section of the resulting binary.
13117 Note that the value of @code{__attribute__ ((patchable_function_entry
13118 (N,M)))} takes precedence over command-line option
13119 @option{-fpatchable-function-entry=N,M}. This can be used to increase
13120 the area size or to remove it completely on a single function.
13121 If @code{N=0}, no pad location is recorded.
13123 The NOP instructions are inserted at---and maybe before, depending on
13124 @var{M}---the function entry address, even before the prologue.
13129 @node Preprocessor Options
13130 @section Options Controlling the Preprocessor
13131 @cindex preprocessor options
13132 @cindex options, preprocessor
13134 These options control the C preprocessor, which is run on each C source
13135 file before actual compilation.
13137 If you use the @option{-E} option, nothing is done except preprocessing.
13138 Some of these options make sense only together with @option{-E} because
13139 they cause the preprocessor output to be unsuitable for actual
13142 In addition to the options listed here, there are a number of options
13143 to control search paths for include files documented in
13144 @ref{Directory Options}.
13145 Options to control preprocessor diagnostics are listed in
13146 @ref{Warning Options}.
13149 @include cppopts.texi
13151 @item -Wp,@var{option}
13153 You can use @option{-Wp,@var{option}} to bypass the compiler driver
13154 and pass @var{option} directly through to the preprocessor. If
13155 @var{option} contains commas, it is split into multiple options at the
13156 commas. However, many options are modified, translated or interpreted
13157 by the compiler driver before being passed to the preprocessor, and
13158 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
13159 interface is undocumented and subject to change, so whenever possible
13160 you should avoid using @option{-Wp} and let the driver handle the
13163 @item -Xpreprocessor @var{option}
13164 @opindex Xpreprocessor
13165 Pass @var{option} as an option to the preprocessor. You can use this to
13166 supply system-specific preprocessor options that GCC does not
13169 If you want to pass an option that takes an argument, you must use
13170 @option{-Xpreprocessor} twice, once for the option and once for the argument.
13172 @item -no-integrated-cpp
13173 @opindex no-integrated-cpp
13174 Perform preprocessing as a separate pass before compilation.
13175 By default, GCC performs preprocessing as an integrated part of
13176 input tokenization and parsing.
13177 If this option is provided, the appropriate language front end
13178 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
13179 and Objective-C, respectively) is instead invoked twice,
13180 once for preprocessing only and once for actual compilation
13181 of the preprocessed input.
13182 This option may be useful in conjunction with the @option{-B} or
13183 @option{-wrapper} options to specify an alternate preprocessor or
13184 perform additional processing of the program source between
13185 normal preprocessing and compilation.
13189 @node Assembler Options
13190 @section Passing Options to the Assembler
13192 @c prevent bad page break with this line
13193 You can pass options to the assembler.
13196 @item -Wa,@var{option}
13198 Pass @var{option} as an option to the assembler. If @var{option}
13199 contains commas, it is split into multiple options at the commas.
13201 @item -Xassembler @var{option}
13202 @opindex Xassembler
13203 Pass @var{option} as an option to the assembler. You can use this to
13204 supply system-specific assembler options that GCC does not
13207 If you want to pass an option that takes an argument, you must use
13208 @option{-Xassembler} twice, once for the option and once for the argument.
13213 @section Options for Linking
13214 @cindex link options
13215 @cindex options, linking
13217 These options come into play when the compiler links object files into
13218 an executable output file. They are meaningless if the compiler is
13219 not doing a link step.
13223 @item @var{object-file-name}
13224 A file name that does not end in a special recognized suffix is
13225 considered to name an object file or library. (Object files are
13226 distinguished from libraries by the linker according to the file
13227 contents.) If linking is done, these object files are used as input
13236 If any of these options is used, then the linker is not run, and
13237 object file names should not be used as arguments. @xref{Overall
13240 @item -flinker-output=@var{type}
13241 @opindex flinker-output
13242 This option controls code generation of the link-time optimizer. By
13243 default the linker output is automatically determined by the linker
13244 plugin. For debugging the compiler and if incremental linking with a
13245 non-LTO object file is desired, it may be useful to control the type
13248 If @var{type} is @samp{exec}, code generation produces a static
13249 binary. In this case @option{-fpic} and @option{-fpie} are both
13252 If @var{type} is @samp{dyn}, code generation produces a shared
13253 library. In this case @option{-fpic} or @option{-fPIC} is preserved,
13254 but not enabled automatically. This allows to build shared libraries
13255 without position-independent code on architectures where this is
13256 possible, i.e.@: on x86.
13258 If @var{type} is @samp{pie}, code generation produces an @option{-fpie}
13259 executable. This results in similar optimizations as @samp{exec}
13260 except that @option{-fpie} is not disabled if specified at compilation
13263 If @var{type} is @samp{rel}, the compiler assumes that incremental linking is
13264 done. The sections containing intermediate code for link-time optimization are
13265 merged, pre-optimized, and output to the resulting object file. In addition, if
13266 @option{-ffat-lto-objects} is specified, binary code is produced for future
13267 non-LTO linking. The object file produced by incremental linking is smaller
13268 than a static library produced from the same object files. At link time the
13269 result of incremental linking also loads faster than a static
13270 library assuming that the majority of objects in the library are used.
13272 Finally @samp{nolto-rel} configures the compiler for incremental linking where
13273 code generation is forced, a final binary is produced, and the intermediate
13274 code for later link-time optimization is stripped. When multiple object files
13275 are linked together the resulting code is better optimized than with
13276 link-time optimizations disabled (for example, cross-module inlining
13277 happens), but most of benefits of whole program optimizations are lost.
13279 During the incremental link (by @option{-r}) the linker plugin defaults to
13280 @option{rel}. With current interfaces to GNU Binutils it is however not
13281 possible to incrementally link LTO objects and non-LTO objects into a single
13282 mixed object file. If any of object files in incremental link cannot
13283 be used for link-time optimization, the linker plugin issues a warning and
13284 uses @samp{nolto-rel}. To maintain whole program optimization, it is
13285 recommended to link such objects into static library instead. Alternatively it
13286 is possible to use H.J. Lu's binutils with support for mixed objects.
13289 @opindex fuse-ld=bfd
13290 Use the @command{bfd} linker instead of the default linker.
13292 @item -fuse-ld=gold
13293 @opindex fuse-ld=gold
13294 Use the @command{gold} linker instead of the default linker.
13297 @opindex fuse-ld=lld
13298 Use the LLVM @command{lld} linker instead of the default linker.
13301 @item -l@var{library}
13302 @itemx -l @var{library}
13304 Search the library named @var{library} when linking. (The second
13305 alternative with the library as a separate argument is only for
13306 POSIX compliance and is not recommended.)
13308 The @option{-l} option is passed directly to the linker by GCC. Refer
13309 to your linker documentation for exact details. The general
13310 description below applies to the GNU linker.
13312 The linker searches a standard list of directories for the library.
13313 The directories searched include several standard system directories
13314 plus any that you specify with @option{-L}.
13316 Static libraries are archives of object files, and have file names
13317 like @file{lib@var{library}.a}. Some targets also support shared
13318 libraries, which typically have names like @file{lib@var{library}.so}.
13319 If both static and shared libraries are found, the linker gives
13320 preference to linking with the shared library unless the
13321 @option{-static} option is used.
13323 It makes a difference where in the command you write this option; the
13324 linker searches and processes libraries and object files in the order they
13325 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
13326 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
13327 to functions in @samp{z}, those functions may not be loaded.
13331 You need this special case of the @option{-l} option in order to
13332 link an Objective-C or Objective-C++ program.
13334 @item -nostartfiles
13335 @opindex nostartfiles
13336 Do not use the standard system startup files when linking.
13337 The standard system libraries are used normally, unless @option{-nostdlib},
13338 @option{-nolibc}, or @option{-nodefaultlibs} is used.
13340 @item -nodefaultlibs
13341 @opindex nodefaultlibs
13342 Do not use the standard system libraries when linking.
13343 Only the libraries you specify are passed to the linker, and options
13344 specifying linkage of the system libraries, such as @option{-static-libgcc}
13345 or @option{-shared-libgcc}, are ignored.
13346 The standard startup files are used normally, unless @option{-nostartfiles}
13349 The compiler may generate calls to @code{memcmp},
13350 @code{memset}, @code{memcpy} and @code{memmove}.
13351 These entries are usually resolved by entries in
13352 libc. These entry points should be supplied through some other
13353 mechanism when this option is specified.
13357 Do not use the C library or system libraries tightly coupled with it when
13358 linking. Still link with the startup files, @file{libgcc} or toolchain
13359 provided language support libraries such as @file{libgnat}, @file{libgfortran}
13360 or @file{libstdc++} unless options preventing their inclusion are used as
13361 well. This typically removes @option{-lc} from the link command line, as well
13362 as system libraries that normally go with it and become meaningless when
13363 absence of a C library is assumed, for example @option{-lpthread} or
13364 @option{-lm} in some configurations. This is intended for bare-board
13365 targets when there is indeed no C library available.
13369 Do not use the standard system startup files or libraries when linking.
13370 No startup files and only the libraries you specify are passed to
13371 the linker, and options specifying linkage of the system libraries, such as
13372 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
13374 The compiler may generate calls to @code{memcmp}, @code{memset},
13375 @code{memcpy} and @code{memmove}.
13376 These entries are usually resolved by entries in
13377 libc. These entry points should be supplied through some other
13378 mechanism when this option is specified.
13380 @cindex @option{-lgcc}, use with @option{-nostdlib}
13381 @cindex @option{-nostdlib} and unresolved references
13382 @cindex unresolved references and @option{-nostdlib}
13383 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
13384 @cindex @option{-nodefaultlibs} and unresolved references
13385 @cindex unresolved references and @option{-nodefaultlibs}
13386 One of the standard libraries bypassed by @option{-nostdlib} and
13387 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
13388 which GCC uses to overcome shortcomings of particular machines, or special
13389 needs for some languages.
13390 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
13391 Collection (GCC) Internals},
13392 for more discussion of @file{libgcc.a}.)
13393 In most cases, you need @file{libgcc.a} even when you want to avoid
13394 other standard libraries. In other words, when you specify @option{-nostdlib}
13395 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
13396 This ensures that you have no unresolved references to internal GCC
13397 library subroutines.
13398 (An example of such an internal subroutine is @code{__main}, used to ensure C++
13399 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
13400 GNU Compiler Collection (GCC) Internals}.)
13402 @item -e @var{entry}
13403 @itemx --entry=@var{entry}
13407 Specify that the program entry point is @var{entry}. The argument is
13408 interpreted by the linker; the GNU linker accepts either a symbol name
13413 Produce a dynamically linked position independent executable on targets
13414 that support it. For predictable results, you must also specify the same
13415 set of options used for compilation (@option{-fpie}, @option{-fPIE},
13416 or model suboptions) when you specify this linker option.
13420 Don't produce a dynamically linked position independent executable.
13423 @opindex static-pie
13424 Produce a static position independent executable on targets that support
13425 it. A static position independent executable is similar to a static
13426 executable, but can be loaded at any address without a dynamic linker.
13427 For predictable results, you must also specify the same set of options
13428 used for compilation (@option{-fpie}, @option{-fPIE}, or model
13429 suboptions) when you specify this linker option.
13433 Link with the POSIX threads library. This option is supported on
13434 GNU/Linux targets, most other Unix derivatives, and also on
13435 x86 Cygwin and MinGW targets. On some targets this option also sets
13436 flags for the preprocessor, so it should be used consistently for both
13437 compilation and linking.
13441 Produce a relocatable object as output. This is also known as partial
13446 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
13447 that support it. This instructs the linker to add all symbols, not
13448 only used ones, to the dynamic symbol table. This option is needed
13449 for some uses of @code{dlopen} or to allow obtaining backtraces
13450 from within a program.
13454 Remove all symbol table and relocation information from the executable.
13458 On systems that support dynamic linking, this overrides @option{-pie}
13459 and prevents linking with the shared libraries. On other systems, this
13460 option has no effect.
13464 Produce a shared object which can then be linked with other objects to
13465 form an executable. Not all systems support this option. For predictable
13466 results, you must also specify the same set of options used for compilation
13467 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
13468 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
13469 needs to build supplementary stub code for constructors to work. On
13470 multi-libbed systems, @samp{gcc -shared} must select the correct support
13471 libraries to link against. Failing to supply the correct flags may lead
13472 to subtle defects. Supplying them in cases where they are not necessary
13475 @item -shared-libgcc
13476 @itemx -static-libgcc
13477 @opindex shared-libgcc
13478 @opindex static-libgcc
13479 On systems that provide @file{libgcc} as a shared library, these options
13480 force the use of either the shared or static version, respectively.
13481 If no shared version of @file{libgcc} was built when the compiler was
13482 configured, these options have no effect.
13484 There are several situations in which an application should use the
13485 shared @file{libgcc} instead of the static version. The most common
13486 of these is when the application wishes to throw and catch exceptions
13487 across different shared libraries. In that case, each of the libraries
13488 as well as the application itself should use the shared @file{libgcc}.
13490 Therefore, the G++ driver automatically adds @option{-shared-libgcc}
13491 whenever you build a shared library or a main executable, because C++
13492 programs typically use exceptions, so this is the right thing to do.
13494 If, instead, you use the GCC driver to create shared libraries, you may
13495 find that they are not always linked with the shared @file{libgcc}.
13496 If GCC finds, at its configuration time, that you have a non-GNU linker
13497 or a GNU linker that does not support option @option{--eh-frame-hdr},
13498 it links the shared version of @file{libgcc} into shared libraries
13499 by default. Otherwise, it takes advantage of the linker and optimizes
13500 away the linking with the shared version of @file{libgcc}, linking with
13501 the static version of libgcc by default. This allows exceptions to
13502 propagate through such shared libraries, without incurring relocation
13503 costs at library load time.
13505 However, if a library or main executable is supposed to throw or catch
13506 exceptions, you must link it using the G++ driver, or using the option
13507 @option{-shared-libgcc}, such that it is linked with the shared
13510 @item -static-libasan
13511 @opindex static-libasan
13512 When the @option{-fsanitize=address} option is used to link a program,
13513 the GCC driver automatically links against @option{libasan}. If
13514 @file{libasan} is available as a shared library, and the @option{-static}
13515 option is not used, then this links against the shared version of
13516 @file{libasan}. The @option{-static-libasan} option directs the GCC
13517 driver to link @file{libasan} statically, without necessarily linking
13518 other libraries statically.
13520 @item -static-libtsan
13521 @opindex static-libtsan
13522 When the @option{-fsanitize=thread} option is used to link a program,
13523 the GCC driver automatically links against @option{libtsan}. If
13524 @file{libtsan} is available as a shared library, and the @option{-static}
13525 option is not used, then this links against the shared version of
13526 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
13527 driver to link @file{libtsan} statically, without necessarily linking
13528 other libraries statically.
13530 @item -static-liblsan
13531 @opindex static-liblsan
13532 When the @option{-fsanitize=leak} option is used to link a program,
13533 the GCC driver automatically links against @option{liblsan}. If
13534 @file{liblsan} is available as a shared library, and the @option{-static}
13535 option is not used, then this links against the shared version of
13536 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
13537 driver to link @file{liblsan} statically, without necessarily linking
13538 other libraries statically.
13540 @item -static-libubsan
13541 @opindex static-libubsan
13542 When the @option{-fsanitize=undefined} option is used to link a program,
13543 the GCC driver automatically links against @option{libubsan}. If
13544 @file{libubsan} is available as a shared library, and the @option{-static}
13545 option is not used, then this links against the shared version of
13546 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
13547 driver to link @file{libubsan} statically, without necessarily linking
13548 other libraries statically.
13550 @item -static-libstdc++
13551 @opindex static-libstdc++
13552 When the @command{g++} program is used to link a C++ program, it
13553 normally automatically links against @option{libstdc++}. If
13554 @file{libstdc++} is available as a shared library, and the
13555 @option{-static} option is not used, then this links against the
13556 shared version of @file{libstdc++}. That is normally fine. However, it
13557 is sometimes useful to freeze the version of @file{libstdc++} used by
13558 the program without going all the way to a fully static link. The
13559 @option{-static-libstdc++} option directs the @command{g++} driver to
13560 link @file{libstdc++} statically, without necessarily linking other
13561 libraries statically.
13565 Bind references to global symbols when building a shared object. Warn
13566 about any unresolved references (unless overridden by the link editor
13567 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
13570 @item -T @var{script}
13572 @cindex linker script
13573 Use @var{script} as the linker script. This option is supported by most
13574 systems using the GNU linker. On some targets, such as bare-board
13575 targets without an operating system, the @option{-T} option may be required
13576 when linking to avoid references to undefined symbols.
13578 @item -Xlinker @var{option}
13580 Pass @var{option} as an option to the linker. You can use this to
13581 supply system-specific linker options that GCC does not recognize.
13583 If you want to pass an option that takes a separate argument, you must use
13584 @option{-Xlinker} twice, once for the option and once for the argument.
13585 For example, to pass @option{-assert definitions}, you must write
13586 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
13587 @option{-Xlinker "-assert definitions"}, because this passes the entire
13588 string as a single argument, which is not what the linker expects.
13590 When using the GNU linker, it is usually more convenient to pass
13591 arguments to linker options using the @option{@var{option}=@var{value}}
13592 syntax than as separate arguments. For example, you can specify
13593 @option{-Xlinker -Map=output.map} rather than
13594 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
13595 this syntax for command-line options.
13597 @item -Wl,@var{option}
13599 Pass @var{option} as an option to the linker. If @var{option} contains
13600 commas, it is split into multiple options at the commas. You can use this
13601 syntax to pass an argument to the option.
13602 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
13603 linker. When using the GNU linker, you can also get the same effect with
13604 @option{-Wl,-Map=output.map}.
13606 @item -u @var{symbol}
13608 Pretend the symbol @var{symbol} is undefined, to force linking of
13609 library modules to define it. You can use @option{-u} multiple times with
13610 different symbols to force loading of additional library modules.
13612 @item -z @var{keyword}
13614 @option{-z} is passed directly on to the linker along with the keyword
13615 @var{keyword}. See the section in the documentation of your linker for
13616 permitted values and their meanings.
13619 @node Directory Options
13620 @section Options for Directory Search
13621 @cindex directory options
13622 @cindex options, directory search
13623 @cindex search path
13625 These options specify directories to search for header files, for
13626 libraries and for parts of the compiler:
13629 @include cppdiropts.texi
13631 @item -iplugindir=@var{dir}
13632 @opindex iplugindir=
13633 Set the directory to search for plugins that are passed
13634 by @option{-fplugin=@var{name}} instead of
13635 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
13636 to be used by the user, but only passed by the driver.
13640 Add directory @var{dir} to the list of directories to be searched
13643 @item -B@var{prefix}
13645 This option specifies where to find the executables, libraries,
13646 include files, and data files of the compiler itself.
13648 The compiler driver program runs one or more of the subprograms
13649 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
13650 @var{prefix} as a prefix for each program it tries to run, both with and
13651 without @samp{@var{machine}/@var{version}/} for the corresponding target
13652 machine and compiler version.
13654 For each subprogram to be run, the compiler driver first tries the
13655 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
13656 is not specified, the driver tries two standard prefixes,
13657 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
13658 those results in a file name that is found, the unmodified program
13659 name is searched for using the directories specified in your
13660 @env{PATH} environment variable.
13662 The compiler checks to see if the path provided by @option{-B}
13663 refers to a directory, and if necessary it adds a directory
13664 separator character at the end of the path.
13666 @option{-B} prefixes that effectively specify directory names also apply
13667 to libraries in the linker, because the compiler translates these
13668 options into @option{-L} options for the linker. They also apply to
13669 include files in the preprocessor, because the compiler translates these
13670 options into @option{-isystem} options for the preprocessor. In this case,
13671 the compiler appends @samp{include} to the prefix.
13673 The runtime support file @file{libgcc.a} can also be searched for using
13674 the @option{-B} prefix, if needed. If it is not found there, the two
13675 standard prefixes above are tried, and that is all. The file is left
13676 out of the link if it is not found by those means.
13678 Another way to specify a prefix much like the @option{-B} prefix is to use
13679 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
13682 As a special kludge, if the path provided by @option{-B} is
13683 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
13684 9, then it is replaced by @file{[dir/]include}. This is to help
13685 with boot-strapping the compiler.
13687 @item -no-canonical-prefixes
13688 @opindex no-canonical-prefixes
13689 Do not expand any symbolic links, resolve references to @samp{/../}
13690 or @samp{/./}, or make the path absolute when generating a relative
13693 @item --sysroot=@var{dir}
13695 Use @var{dir} as the logical root directory for headers and libraries.
13696 For example, if the compiler normally searches for headers in
13697 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
13698 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
13700 If you use both this option and the @option{-isysroot} option, then
13701 the @option{--sysroot} option applies to libraries, but the
13702 @option{-isysroot} option applies to header files.
13704 The GNU linker (beginning with version 2.16) has the necessary support
13705 for this option. If your linker does not support this option, the
13706 header file aspect of @option{--sysroot} still works, but the
13707 library aspect does not.
13709 @item --no-sysroot-suffix
13710 @opindex no-sysroot-suffix
13711 For some targets, a suffix is added to the root directory specified
13712 with @option{--sysroot}, depending on the other options used, so that
13713 headers may for example be found in
13714 @file{@var{dir}/@var{suffix}/usr/include} instead of
13715 @file{@var{dir}/usr/include}. This option disables the addition of
13720 @node Code Gen Options
13721 @section Options for Code Generation Conventions
13722 @cindex code generation conventions
13723 @cindex options, code generation
13724 @cindex run-time options
13726 These machine-independent options control the interface conventions
13727 used in code generation.
13729 Most of them have both positive and negative forms; the negative form
13730 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
13731 one of the forms is listed---the one that is not the default. You
13732 can figure out the other form by either removing @samp{no-} or adding
13736 @item -fstack-reuse=@var{reuse-level}
13737 @opindex fstack_reuse
13738 This option controls stack space reuse for user declared local/auto variables
13739 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
13740 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
13741 local variables and temporaries, @samp{named_vars} enables the reuse only for
13742 user defined local variables with names, and @samp{none} disables stack reuse
13743 completely. The default value is @samp{all}. The option is needed when the
13744 program extends the lifetime of a scoped local variable or a compiler generated
13745 temporary beyond the end point defined by the language. When a lifetime of
13746 a variable ends, and if the variable lives in memory, the optimizing compiler
13747 has the freedom to reuse its stack space with other temporaries or scoped
13748 local variables whose live range does not overlap with it. Legacy code extending
13749 local lifetime is likely to break with the stack reuse optimization.
13768 if (*p == 10) // out of scope use of local1
13779 A(int k) : i(k), j(k) @{ @}
13786 void foo(const A& ar)
13793 foo(A(10)); // temp object's lifetime ends when foo returns
13799 ap->i+= 10; // ap references out of scope temp whose space
13800 // is reused with a. What is the value of ap->i?
13805 The lifetime of a compiler generated temporary is well defined by the C++
13806 standard. When a lifetime of a temporary ends, and if the temporary lives
13807 in memory, the optimizing compiler has the freedom to reuse its stack
13808 space with other temporaries or scoped local variables whose live range
13809 does not overlap with it. However some of the legacy code relies on
13810 the behavior of older compilers in which temporaries' stack space is
13811 not reused, the aggressive stack reuse can lead to runtime errors. This
13812 option is used to control the temporary stack reuse optimization.
13816 This option generates traps for signed overflow on addition, subtraction,
13817 multiplication operations.
13818 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13819 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13820 @option{-fwrapv} being effective. Note that only active options override, so
13821 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13822 results in @option{-ftrapv} being effective.
13826 This option instructs the compiler to assume that signed arithmetic
13827 overflow of addition, subtraction and multiplication wraps around
13828 using twos-complement representation. This flag enables some optimizations
13829 and disables others.
13830 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
13831 @option{-ftrapv} @option{-fwrapv} on the command-line results in
13832 @option{-fwrapv} being effective. Note that only active options override, so
13833 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
13834 results in @option{-ftrapv} being effective.
13836 @item -fwrapv-pointer
13837 @opindex fwrapv-pointer
13838 This option instructs the compiler to assume that pointer arithmetic
13839 overflow on addition and subtraction wraps around using twos-complement
13840 representation. This flag disables some optimizations which assume
13841 pointer overflow is invalid.
13843 @item -fstrict-overflow
13844 @opindex fstrict-overflow
13845 This option implies @option{-fno-wrapv} @option{-fno-wrapv-pointer} and when
13846 negated implies @option{-fwrapv} @option{-fwrapv-pointer}.
13849 @opindex fexceptions
13850 Enable exception handling. Generates extra code needed to propagate
13851 exceptions. For some targets, this implies GCC generates frame
13852 unwind information for all functions, which can produce significant data
13853 size overhead, although it does not affect execution. If you do not
13854 specify this option, GCC enables it by default for languages like
13855 C++ that normally require exception handling, and disables it for
13856 languages like C that do not normally require it. However, you may need
13857 to enable this option when compiling C code that needs to interoperate
13858 properly with exception handlers written in C++. You may also wish to
13859 disable this option if you are compiling older C++ programs that don't
13860 use exception handling.
13862 @item -fnon-call-exceptions
13863 @opindex fnon-call-exceptions
13864 Generate code that allows trapping instructions to throw exceptions.
13865 Note that this requires platform-specific runtime support that does
13866 not exist everywhere. Moreover, it only allows @emph{trapping}
13867 instructions to throw exceptions, i.e.@: memory references or floating-point
13868 instructions. It does not allow exceptions to be thrown from
13869 arbitrary signal handlers such as @code{SIGALRM}.
13871 @item -fdelete-dead-exceptions
13872 @opindex fdelete-dead-exceptions
13873 Consider that instructions that may throw exceptions but don't otherwise
13874 contribute to the execution of the program can be optimized away.
13875 This option is enabled by default for the Ada front end, as permitted by
13876 the Ada language specification.
13877 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
13879 @item -funwind-tables
13880 @opindex funwind-tables
13881 Similar to @option{-fexceptions}, except that it just generates any needed
13882 static data, but does not affect the generated code in any other way.
13883 You normally do not need to enable this option; instead, a language processor
13884 that needs this handling enables it on your behalf.
13886 @item -fasynchronous-unwind-tables
13887 @opindex fasynchronous-unwind-tables
13888 Generate unwind table in DWARF format, if supported by target machine. The
13889 table is exact at each instruction boundary, so it can be used for stack
13890 unwinding from asynchronous events (such as debugger or garbage collector).
13892 @item -fno-gnu-unique
13893 @opindex fno-gnu-unique
13894 @opindex fgnu-unique
13895 On systems with recent GNU assembler and C library, the C++ compiler
13896 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
13897 of template static data members and static local variables in inline
13898 functions are unique even in the presence of @code{RTLD_LOCAL}; this
13899 is necessary to avoid problems with a library used by two different
13900 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
13901 therefore disagreeing with the other one about the binding of the
13902 symbol. But this causes @code{dlclose} to be ignored for affected
13903 DSOs; if your program relies on reinitialization of a DSO via
13904 @code{dlclose} and @code{dlopen}, you can use
13905 @option{-fno-gnu-unique}.
13907 @item -fpcc-struct-return
13908 @opindex fpcc-struct-return
13909 Return ``short'' @code{struct} and @code{union} values in memory like
13910 longer ones, rather than in registers. This convention is less
13911 efficient, but it has the advantage of allowing intercallability between
13912 GCC-compiled files and files compiled with other compilers, particularly
13913 the Portable C Compiler (pcc).
13915 The precise convention for returning structures in memory depends
13916 on the target configuration macros.
13918 Short structures and unions are those whose size and alignment match
13919 that of some integer type.
13921 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13922 switch is not binary compatible with code compiled with the
13923 @option{-freg-struct-return} switch.
13924 Use it to conform to a non-default application binary interface.
13926 @item -freg-struct-return
13927 @opindex freg-struct-return
13928 Return @code{struct} and @code{union} values in registers when possible.
13929 This is more efficient for small structures than
13930 @option{-fpcc-struct-return}.
13932 If you specify neither @option{-fpcc-struct-return} nor
13933 @option{-freg-struct-return}, GCC defaults to whichever convention is
13934 standard for the target. If there is no standard convention, GCC
13935 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13936 the principal compiler. In those cases, we can choose the standard, and
13937 we chose the more efficient register return alternative.
13939 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13940 switch is not binary compatible with code compiled with the
13941 @option{-fpcc-struct-return} switch.
13942 Use it to conform to a non-default application binary interface.
13944 @item -fshort-enums
13945 @opindex fshort-enums
13946 Allocate to an @code{enum} type only as many bytes as it needs for the
13947 declared range of possible values. Specifically, the @code{enum} type
13948 is equivalent to the smallest integer type that has enough room.
13950 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13951 code that is not binary compatible with code generated without that switch.
13952 Use it to conform to a non-default application binary interface.
13954 @item -fshort-wchar
13955 @opindex fshort-wchar
13956 Override the underlying type for @code{wchar_t} to be @code{short
13957 unsigned int} instead of the default for the target. This option is
13958 useful for building programs to run under WINE@.
13960 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13961 code that is not binary compatible with code generated without that switch.
13962 Use it to conform to a non-default application binary interface.
13965 @opindex fno-common
13967 @cindex tentative definitions
13968 In C code, this option controls the placement of global variables
13969 defined without an initializer, known as @dfn{tentative definitions}
13970 in the C standard. Tentative definitions are distinct from declarations
13971 of a variable with the @code{extern} keyword, which do not allocate storage.
13973 Unix C compilers have traditionally allocated storage for
13974 uninitialized global variables in a common block. This allows the
13975 linker to resolve all tentative definitions of the same variable
13976 in different compilation units to the same object, or to a non-tentative
13978 This is the behavior specified by @option{-fcommon}, and is the default for
13979 GCC on most targets.
13980 On the other hand, this behavior is not required by ISO
13981 C, and on some targets may carry a speed or code size penalty on
13982 variable references.
13984 The @option{-fno-common} option specifies that the compiler should instead
13985 place uninitialized global variables in the BSS section of the object file.
13986 This inhibits the merging of tentative definitions by the linker so
13987 you get a multiple-definition error if the same
13988 variable is defined in more than one compilation unit.
13989 Compiling with @option{-fno-common} is useful on targets for which
13990 it provides better performance, or if you wish to verify that the
13991 program will work on other systems that always treat uninitialized
13992 variable definitions this way.
13997 Ignore the @code{#ident} directive.
13999 @item -finhibit-size-directive
14000 @opindex finhibit-size-directive
14001 Don't output a @code{.size} assembler directive, or anything else that
14002 would cause trouble if the function is split in the middle, and the
14003 two halves are placed at locations far apart in memory. This option is
14004 used when compiling @file{crtstuff.c}; you should not need to use it
14007 @item -fverbose-asm
14008 @opindex fverbose-asm
14009 Put extra commentary information in the generated assembly code to
14010 make it more readable. This option is generally only of use to those
14011 who actually need to read the generated assembly code (perhaps while
14012 debugging the compiler itself).
14014 @option{-fno-verbose-asm}, the default, causes the
14015 extra information to be omitted and is useful when comparing two assembler
14018 The added comments include:
14023 information on the compiler version and command-line options,
14026 the source code lines associated with the assembly instructions,
14027 in the form FILENAME:LINENUMBER:CONTENT OF LINE,
14030 hints on which high-level expressions correspond to
14031 the various assembly instruction operands.
14035 For example, given this C source file:
14043 for (i = 0; i < n; i++)
14050 compiling to (x86_64) assembly via @option{-S} and emitting the result
14051 direct to stdout via @option{-o} @option{-}
14054 gcc -S test.c -fverbose-asm -Os -o -
14057 gives output similar to this:
14061 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
14068 .type test, @@function
14072 # test.c:4: int total = 0;
14073 xorl %eax, %eax # <retval>
14074 # test.c:6: for (i = 0; i < n; i++)
14075 xorl %edx, %edx # i
14077 # test.c:6: for (i = 0; i < n; i++)
14078 cmpl %edi, %edx # n, i
14080 # test.c:7: total += i * i;
14081 movl %edx, %ecx # i, tmp92
14082 imull %edx, %ecx # i, tmp92
14083 # test.c:6: for (i = 0; i < n; i++)
14085 # test.c:7: total += i * i;
14086 addl %ecx, %eax # tmp92, <retval>
14094 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)"
14095 .section .note.GNU-stack,"",@@progbits
14098 The comments are intended for humans rather than machines and hence the
14099 precise format of the comments is subject to change.
14101 @item -frecord-gcc-switches
14102 @opindex frecord-gcc-switches
14103 This switch causes the command line used to invoke the
14104 compiler to be recorded into the object file that is being created.
14105 This switch is only implemented on some targets and the exact format
14106 of the recording is target and binary file format dependent, but it
14107 usually takes the form of a section containing ASCII text. This
14108 switch is related to the @option{-fverbose-asm} switch, but that
14109 switch only records information in the assembler output file as
14110 comments, so it never reaches the object file.
14111 See also @option{-grecord-gcc-switches} for another
14112 way of storing compiler options into the object file.
14116 @cindex global offset table
14118 Generate position-independent code (PIC) suitable for use in a shared
14119 library, if supported for the target machine. Such code accesses all
14120 constant addresses through a global offset table (GOT)@. The dynamic
14121 loader resolves the GOT entries when the program starts (the dynamic
14122 loader is not part of GCC; it is part of the operating system). If
14123 the GOT size for the linked executable exceeds a machine-specific
14124 maximum size, you get an error message from the linker indicating that
14125 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14126 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
14127 on the m68k and RS/6000. The x86 has no such limit.)
14129 Position-independent code requires special support, and therefore works
14130 only on certain machines. For the x86, GCC supports PIC for System V
14131 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14132 position-independent.
14134 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14139 If supported for the target machine, emit position-independent code,
14140 suitable for dynamic linking and avoiding any limit on the size of the
14141 global offset table. This option makes a difference on AArch64, m68k,
14142 PowerPC and SPARC@.
14144 Position-independent code requires special support, and therefore works
14145 only on certain machines.
14147 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14154 These options are similar to @option{-fpic} and @option{-fPIC}, but the
14155 generated position-independent code can be only linked into executables.
14156 Usually these options are used to compile code that will be linked using
14157 the @option{-pie} GCC option.
14159 @option{-fpie} and @option{-fPIE} both define the macros
14160 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14161 for @option{-fpie} and 2 for @option{-fPIE}.
14166 Do not use the PLT for external function calls in position-independent code.
14167 Instead, load the callee address at call sites from the GOT and branch to it.
14168 This leads to more efficient code by eliminating PLT stubs and exposing
14169 GOT loads to optimizations. On architectures such as 32-bit x86 where
14170 PLT stubs expect the GOT pointer in a specific register, this gives more
14171 register allocation freedom to the compiler.
14172 Lazy binding requires use of the PLT;
14173 with @option{-fno-plt} all external symbols are resolved at load time.
14175 Alternatively, the function attribute @code{noplt} can be used to avoid calls
14176 through the PLT for specific external functions.
14178 In position-dependent code, a few targets also convert calls to
14179 functions that are marked to not use the PLT to use the GOT instead.
14181 @item -fno-jump-tables
14182 @opindex fno-jump-tables
14183 @opindex fjump-tables
14184 Do not use jump tables for switch statements even where it would be
14185 more efficient than other code generation strategies. This option is
14186 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14187 building code that forms part of a dynamic linker and cannot
14188 reference the address of a jump table. On some targets, jump tables
14189 do not require a GOT and this option is not needed.
14191 @item -ffixed-@var{reg}
14193 Treat the register named @var{reg} as a fixed register; generated code
14194 should never refer to it (except perhaps as a stack pointer, frame
14195 pointer or in some other fixed role).
14197 @var{reg} must be the name of a register. The register names accepted
14198 are machine-specific and are defined in the @code{REGISTER_NAMES}
14199 macro in the machine description macro file.
14201 This flag does not have a negative form, because it specifies a
14204 @item -fcall-used-@var{reg}
14205 @opindex fcall-used
14206 Treat the register named @var{reg} as an allocable register that is
14207 clobbered by function calls. It may be allocated for temporaries or
14208 variables that do not live across a call. Functions compiled this way
14209 do not save and restore the register @var{reg}.
14211 It is an error to use this flag with the frame pointer or stack pointer.
14212 Use of this flag for other registers that have fixed pervasive roles in
14213 the machine's execution model produces disastrous results.
14215 This flag does not have a negative form, because it specifies a
14218 @item -fcall-saved-@var{reg}
14219 @opindex fcall-saved
14220 Treat the register named @var{reg} as an allocable register saved by
14221 functions. It may be allocated even for temporaries or variables that
14222 live across a call. Functions compiled this way save and restore
14223 the register @var{reg} if they use it.
14225 It is an error to use this flag with the frame pointer or stack pointer.
14226 Use of this flag for other registers that have fixed pervasive roles in
14227 the machine's execution model produces disastrous results.
14229 A different sort of disaster results from the use of this flag for
14230 a register in which function values may be returned.
14232 This flag does not have a negative form, because it specifies a
14235 @item -fpack-struct[=@var{n}]
14236 @opindex fpack-struct
14237 Without a value specified, pack all structure members together without
14238 holes. When a value is specified (which must be a small power of two), pack
14239 structure members according to this value, representing the maximum
14240 alignment (that is, objects with default alignment requirements larger than
14241 this are output potentially unaligned at the next fitting location.
14243 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14244 code that is not binary compatible with code generated without that switch.
14245 Additionally, it makes the code suboptimal.
14246 Use it to conform to a non-default application binary interface.
14248 @item -fleading-underscore
14249 @opindex fleading-underscore
14250 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14251 change the way C symbols are represented in the object file. One use
14252 is to help link with legacy assembly code.
14254 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14255 generate code that is not binary compatible with code generated without that
14256 switch. Use it to conform to a non-default application binary interface.
14257 Not all targets provide complete support for this switch.
14259 @item -ftls-model=@var{model}
14260 @opindex ftls-model
14261 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14262 The @var{model} argument should be one of @samp{global-dynamic},
14263 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
14264 Note that the choice is subject to optimization: the compiler may use
14265 a more efficient model for symbols not visible outside of the translation
14266 unit, or if @option{-fpic} is not given on the command line.
14268 The default without @option{-fpic} is @samp{initial-exec}; with
14269 @option{-fpic} the default is @samp{global-dynamic}.
14271 @item -ftrampolines
14272 @opindex ftrampolines
14273 For targets that normally need trampolines for nested functions, always
14274 generate them instead of using descriptors. Otherwise, for targets that
14275 do not need them, like for example HP-PA or IA-64, do nothing.
14277 A trampoline is a small piece of code that is created at run time on the
14278 stack when the address of a nested function is taken, and is used to call
14279 the nested function indirectly. Therefore, it requires the stack to be
14280 made executable in order for the program to work properly.
14282 @option{-fno-trampolines} is enabled by default on a language by language
14283 basis to let the compiler avoid generating them, if it computes that this
14284 is safe, and replace them with descriptors. Descriptors are made up of data
14285 only, but the generated code must be prepared to deal with them. As of this
14286 writing, @option{-fno-trampolines} is enabled by default only for Ada.
14288 Moreover, code compiled with @option{-ftrampolines} and code compiled with
14289 @option{-fno-trampolines} are not binary compatible if nested functions are
14290 present. This option must therefore be used on a program-wide basis and be
14291 manipulated with extreme care.
14293 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
14294 @opindex fvisibility
14295 Set the default ELF image symbol visibility to the specified option---all
14296 symbols are marked with this unless overridden within the code.
14297 Using this feature can very substantially improve linking and
14298 load times of shared object libraries, produce more optimized
14299 code, provide near-perfect API export and prevent symbol clashes.
14300 It is @strong{strongly} recommended that you use this in any shared objects
14303 Despite the nomenclature, @samp{default} always means public; i.e.,
14304 available to be linked against from outside the shared object.
14305 @samp{protected} and @samp{internal} are pretty useless in real-world
14306 usage so the only other commonly used option is @samp{hidden}.
14307 The default if @option{-fvisibility} isn't specified is
14308 @samp{default}, i.e., make every symbol public.
14310 A good explanation of the benefits offered by ensuring ELF
14311 symbols have the correct visibility is given by ``How To Write
14312 Shared Libraries'' by Ulrich Drepper (which can be found at
14313 @w{@uref{https://www.akkadia.org/drepper/}})---however a superior
14314 solution made possible by this option to marking things hidden when
14315 the default is public is to make the default hidden and mark things
14316 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
14317 and @code{__attribute__ ((visibility("default")))} instead of
14318 @code{__declspec(dllexport)} you get almost identical semantics with
14319 identical syntax. This is a great boon to those working with
14320 cross-platform projects.
14322 For those adding visibility support to existing code, you may find
14323 @code{#pragma GCC visibility} of use. This works by you enclosing
14324 the declarations you wish to set visibility for with (for example)
14325 @code{#pragma GCC visibility push(hidden)} and
14326 @code{#pragma GCC visibility pop}.
14327 Bear in mind that symbol visibility should be viewed @strong{as
14328 part of the API interface contract} and thus all new code should
14329 always specify visibility when it is not the default; i.e., declarations
14330 only for use within the local DSO should @strong{always} be marked explicitly
14331 as hidden as so to avoid PLT indirection overheads---making this
14332 abundantly clear also aids readability and self-documentation of the code.
14333 Note that due to ISO C++ specification requirements, @code{operator new} and
14334 @code{operator delete} must always be of default visibility.
14336 Be aware that headers from outside your project, in particular system
14337 headers and headers from any other library you use, may not be
14338 expecting to be compiled with visibility other than the default. You
14339 may need to explicitly say @code{#pragma GCC visibility push(default)}
14340 before including any such headers.
14342 @code{extern} declarations are not affected by @option{-fvisibility}, so
14343 a lot of code can be recompiled with @option{-fvisibility=hidden} with
14344 no modifications. However, this means that calls to @code{extern}
14345 functions with no explicit visibility use the PLT, so it is more
14346 effective to use @code{__attribute ((visibility))} and/or
14347 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
14348 declarations should be treated as hidden.
14350 Note that @option{-fvisibility} does affect C++ vague linkage
14351 entities. This means that, for instance, an exception class that is
14352 be thrown between DSOs must be explicitly marked with default
14353 visibility so that the @samp{type_info} nodes are unified between
14356 An overview of these techniques, their benefits and how to use them
14357 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
14359 @item -fstrict-volatile-bitfields
14360 @opindex fstrict-volatile-bitfields
14361 This option should be used if accesses to volatile bit-fields (or other
14362 structure fields, although the compiler usually honors those types
14363 anyway) should use a single access of the width of the
14364 field's type, aligned to a natural alignment if possible. For
14365 example, targets with memory-mapped peripheral registers might require
14366 all such accesses to be 16 bits wide; with this flag you can
14367 declare all peripheral bit-fields as @code{unsigned short} (assuming short
14368 is 16 bits on these targets) to force GCC to use 16-bit accesses
14369 instead of, perhaps, a more efficient 32-bit access.
14371 If this option is disabled, the compiler uses the most efficient
14372 instruction. In the previous example, that might be a 32-bit load
14373 instruction, even though that accesses bytes that do not contain
14374 any portion of the bit-field, or memory-mapped registers unrelated to
14375 the one being updated.
14377 In some cases, such as when the @code{packed} attribute is applied to a
14378 structure field, it may not be possible to access the field with a single
14379 read or write that is correctly aligned for the target machine. In this
14380 case GCC falls back to generating multiple accesses rather than code that
14381 will fault or truncate the result at run time.
14383 Note: Due to restrictions of the C/C++11 memory model, write accesses are
14384 not allowed to touch non bit-field members. It is therefore recommended
14385 to define all bits of the field's type as bit-field members.
14387 The default value of this option is determined by the application binary
14388 interface for the target processor.
14390 @item -fsync-libcalls
14391 @opindex fsync-libcalls
14392 This option controls whether any out-of-line instance of the @code{__sync}
14393 family of functions may be used to implement the C++11 @code{__atomic}
14394 family of functions.
14396 The default value of this option is enabled, thus the only useful form
14397 of the option is @option{-fno-sync-libcalls}. This option is used in
14398 the implementation of the @file{libatomic} runtime library.
14402 @node Developer Options
14403 @section GCC Developer Options
14404 @cindex developer options
14405 @cindex debugging GCC
14406 @cindex debug dump options
14407 @cindex dump options
14408 @cindex compilation statistics
14410 This section describes command-line options that are primarily of
14411 interest to GCC developers, including options to support compiler
14412 testing and investigation of compiler bugs and compile-time
14413 performance problems. This includes options that produce debug dumps
14414 at various points in the compilation; that print statistics such as
14415 memory use and execution time; and that print information about GCC's
14416 configuration, such as where it searches for libraries. You should
14417 rarely need to use any of these options for ordinary compilation and
14420 Many developer options that cause GCC to dump output to a file take an
14421 optional @samp{=@var{filename}} suffix. You can specify @samp{stdout}
14422 or @samp{-} to dump to standard output, and @samp{stderr} for standard
14425 If @samp{=@var{filename}} is omitted, a default dump file name is
14426 constructed by concatenating the base dump file name, a pass number,
14427 phase letter, and pass name. The base dump file name is the name of
14428 output file produced by the compiler if explicitly specified and not
14429 an executable; otherwise it is the source file name.
14430 The pass number is determined by the order passes are registered with
14431 the compiler's pass manager.
14432 This is generally the same as the order of execution, but passes
14433 registered by plugins, target-specific passes, or passes that are
14434 otherwise registered late are numbered higher than the pass named
14435 @samp{final}, even if they are executed earlier. The phase letter is
14436 one of @samp{i} (inter-procedural analysis), @samp{l}
14437 (language-specific), @samp{r} (RTL), or @samp{t} (tree).
14438 The files are created in the directory of the output file.
14442 @item -d@var{letters}
14443 @itemx -fdump-rtl-@var{pass}
14444 @itemx -fdump-rtl-@var{pass}=@var{filename}
14446 @opindex fdump-rtl-@var{pass}
14447 Says to make debugging dumps during compilation at times specified by
14448 @var{letters}. This is used for debugging the RTL-based passes of the
14451 Some @option{-d@var{letters}} switches have different meaning when
14452 @option{-E} is used for preprocessing. @xref{Preprocessor Options},
14453 for information about preprocessor-specific dump options.
14455 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
14456 @option{-d} option @var{letters}. Here are the possible
14457 letters for use in @var{pass} and @var{letters}, and their meanings:
14461 @item -fdump-rtl-alignments
14462 @opindex fdump-rtl-alignments
14463 Dump after branch alignments have been computed.
14465 @item -fdump-rtl-asmcons
14466 @opindex fdump-rtl-asmcons
14467 Dump after fixing rtl statements that have unsatisfied in/out constraints.
14469 @item -fdump-rtl-auto_inc_dec
14470 @opindex fdump-rtl-auto_inc_dec
14471 Dump after auto-inc-dec discovery. This pass is only run on
14472 architectures that have auto inc or auto dec instructions.
14474 @item -fdump-rtl-barriers
14475 @opindex fdump-rtl-barriers
14476 Dump after cleaning up the barrier instructions.
14478 @item -fdump-rtl-bbpart
14479 @opindex fdump-rtl-bbpart
14480 Dump after partitioning hot and cold basic blocks.
14482 @item -fdump-rtl-bbro
14483 @opindex fdump-rtl-bbro
14484 Dump after block reordering.
14486 @item -fdump-rtl-btl1
14487 @itemx -fdump-rtl-btl2
14488 @opindex fdump-rtl-btl2
14489 @opindex fdump-rtl-btl2
14490 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
14491 after the two branch
14492 target load optimization passes.
14494 @item -fdump-rtl-bypass
14495 @opindex fdump-rtl-bypass
14496 Dump after jump bypassing and control flow optimizations.
14498 @item -fdump-rtl-combine
14499 @opindex fdump-rtl-combine
14500 Dump after the RTL instruction combination pass.
14502 @item -fdump-rtl-compgotos
14503 @opindex fdump-rtl-compgotos
14504 Dump after duplicating the computed gotos.
14506 @item -fdump-rtl-ce1
14507 @itemx -fdump-rtl-ce2
14508 @itemx -fdump-rtl-ce3
14509 @opindex fdump-rtl-ce1
14510 @opindex fdump-rtl-ce2
14511 @opindex fdump-rtl-ce3
14512 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
14513 @option{-fdump-rtl-ce3} enable dumping after the three
14514 if conversion passes.
14516 @item -fdump-rtl-cprop_hardreg
14517 @opindex fdump-rtl-cprop_hardreg
14518 Dump after hard register copy propagation.
14520 @item -fdump-rtl-csa
14521 @opindex fdump-rtl-csa
14522 Dump after combining stack adjustments.
14524 @item -fdump-rtl-cse1
14525 @itemx -fdump-rtl-cse2
14526 @opindex fdump-rtl-cse1
14527 @opindex fdump-rtl-cse2
14528 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
14529 the two common subexpression elimination passes.
14531 @item -fdump-rtl-dce
14532 @opindex fdump-rtl-dce
14533 Dump after the standalone dead code elimination passes.
14535 @item -fdump-rtl-dbr
14536 @opindex fdump-rtl-dbr
14537 Dump after delayed branch scheduling.
14539 @item -fdump-rtl-dce1
14540 @itemx -fdump-rtl-dce2
14541 @opindex fdump-rtl-dce1
14542 @opindex fdump-rtl-dce2
14543 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
14544 the two dead store elimination passes.
14546 @item -fdump-rtl-eh
14547 @opindex fdump-rtl-eh
14548 Dump after finalization of EH handling code.
14550 @item -fdump-rtl-eh_ranges
14551 @opindex fdump-rtl-eh_ranges
14552 Dump after conversion of EH handling range regions.
14554 @item -fdump-rtl-expand
14555 @opindex fdump-rtl-expand
14556 Dump after RTL generation.
14558 @item -fdump-rtl-fwprop1
14559 @itemx -fdump-rtl-fwprop2
14560 @opindex fdump-rtl-fwprop1
14561 @opindex fdump-rtl-fwprop2
14562 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
14563 dumping after the two forward propagation passes.
14565 @item -fdump-rtl-gcse1
14566 @itemx -fdump-rtl-gcse2
14567 @opindex fdump-rtl-gcse1
14568 @opindex fdump-rtl-gcse2
14569 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
14570 after global common subexpression elimination.
14572 @item -fdump-rtl-init-regs
14573 @opindex fdump-rtl-init-regs
14574 Dump after the initialization of the registers.
14576 @item -fdump-rtl-initvals
14577 @opindex fdump-rtl-initvals
14578 Dump after the computation of the initial value sets.
14580 @item -fdump-rtl-into_cfglayout
14581 @opindex fdump-rtl-into_cfglayout
14582 Dump after converting to cfglayout mode.
14584 @item -fdump-rtl-ira
14585 @opindex fdump-rtl-ira
14586 Dump after iterated register allocation.
14588 @item -fdump-rtl-jump
14589 @opindex fdump-rtl-jump
14590 Dump after the second jump optimization.
14592 @item -fdump-rtl-loop2
14593 @opindex fdump-rtl-loop2
14594 @option{-fdump-rtl-loop2} enables dumping after the rtl
14595 loop optimization passes.
14597 @item -fdump-rtl-mach
14598 @opindex fdump-rtl-mach
14599 Dump after performing the machine dependent reorganization pass, if that
14602 @item -fdump-rtl-mode_sw
14603 @opindex fdump-rtl-mode_sw
14604 Dump after removing redundant mode switches.
14606 @item -fdump-rtl-rnreg
14607 @opindex fdump-rtl-rnreg
14608 Dump after register renumbering.
14610 @item -fdump-rtl-outof_cfglayout
14611 @opindex fdump-rtl-outof_cfglayout
14612 Dump after converting from cfglayout mode.
14614 @item -fdump-rtl-peephole2
14615 @opindex fdump-rtl-peephole2
14616 Dump after the peephole pass.
14618 @item -fdump-rtl-postreload
14619 @opindex fdump-rtl-postreload
14620 Dump after post-reload optimizations.
14622 @item -fdump-rtl-pro_and_epilogue
14623 @opindex fdump-rtl-pro_and_epilogue
14624 Dump after generating the function prologues and epilogues.
14626 @item -fdump-rtl-sched1
14627 @itemx -fdump-rtl-sched2
14628 @opindex fdump-rtl-sched1
14629 @opindex fdump-rtl-sched2
14630 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
14631 after the basic block scheduling passes.
14633 @item -fdump-rtl-ree
14634 @opindex fdump-rtl-ree
14635 Dump after sign/zero extension elimination.
14637 @item -fdump-rtl-seqabstr
14638 @opindex fdump-rtl-seqabstr
14639 Dump after common sequence discovery.
14641 @item -fdump-rtl-shorten
14642 @opindex fdump-rtl-shorten
14643 Dump after shortening branches.
14645 @item -fdump-rtl-sibling
14646 @opindex fdump-rtl-sibling
14647 Dump after sibling call optimizations.
14649 @item -fdump-rtl-split1
14650 @itemx -fdump-rtl-split2
14651 @itemx -fdump-rtl-split3
14652 @itemx -fdump-rtl-split4
14653 @itemx -fdump-rtl-split5
14654 @opindex fdump-rtl-split1
14655 @opindex fdump-rtl-split2
14656 @opindex fdump-rtl-split3
14657 @opindex fdump-rtl-split4
14658 @opindex fdump-rtl-split5
14659 These options enable dumping after five rounds of
14660 instruction splitting.
14662 @item -fdump-rtl-sms
14663 @opindex fdump-rtl-sms
14664 Dump after modulo scheduling. This pass is only run on some
14667 @item -fdump-rtl-stack
14668 @opindex fdump-rtl-stack
14669 Dump after conversion from GCC's ``flat register file'' registers to the
14670 x87's stack-like registers. This pass is only run on x86 variants.
14672 @item -fdump-rtl-subreg1
14673 @itemx -fdump-rtl-subreg2
14674 @opindex fdump-rtl-subreg1
14675 @opindex fdump-rtl-subreg2
14676 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
14677 the two subreg expansion passes.
14679 @item -fdump-rtl-unshare
14680 @opindex fdump-rtl-unshare
14681 Dump after all rtl has been unshared.
14683 @item -fdump-rtl-vartrack
14684 @opindex fdump-rtl-vartrack
14685 Dump after variable tracking.
14687 @item -fdump-rtl-vregs
14688 @opindex fdump-rtl-vregs
14689 Dump after converting virtual registers to hard registers.
14691 @item -fdump-rtl-web
14692 @opindex fdump-rtl-web
14693 Dump after live range splitting.
14695 @item -fdump-rtl-regclass
14696 @itemx -fdump-rtl-subregs_of_mode_init
14697 @itemx -fdump-rtl-subregs_of_mode_finish
14698 @itemx -fdump-rtl-dfinit
14699 @itemx -fdump-rtl-dfinish
14700 @opindex fdump-rtl-regclass
14701 @opindex fdump-rtl-subregs_of_mode_init
14702 @opindex fdump-rtl-subregs_of_mode_finish
14703 @opindex fdump-rtl-dfinit
14704 @opindex fdump-rtl-dfinish
14705 These dumps are defined but always produce empty files.
14708 @itemx -fdump-rtl-all
14710 @opindex fdump-rtl-all
14711 Produce all the dumps listed above.
14715 Annotate the assembler output with miscellaneous debugging information.
14719 Dump all macro definitions, at the end of preprocessing, in addition to
14724 Produce a core dump whenever an error occurs.
14728 Annotate the assembler output with a comment indicating which
14729 pattern and alternative is used. The length and cost of each instruction are
14734 Dump the RTL in the assembler output as a comment before each instruction.
14735 Also turns on @option{-dp} annotation.
14739 Just generate RTL for a function instead of compiling it. Usually used
14740 with @option{-fdump-rtl-expand}.
14744 @opindex fdump-debug
14745 Dump debugging information generated during the debug
14748 @item -fdump-earlydebug
14749 @opindex fdump-earlydebug
14750 Dump debugging information generated during the early debug
14753 @item -fdump-noaddr
14754 @opindex fdump-noaddr
14755 When doing debugging dumps, suppress address output. This makes it more
14756 feasible to use diff on debugging dumps for compiler invocations with
14757 different compiler binaries and/or different
14758 text / bss / data / heap / stack / dso start locations.
14761 @opindex freport-bug
14762 Collect and dump debug information into a temporary file if an
14763 internal compiler error (ICE) occurs.
14765 @item -fdump-unnumbered
14766 @opindex fdump-unnumbered
14767 When doing debugging dumps, suppress instruction numbers and address output.
14768 This makes it more feasible to use diff on debugging dumps for compiler
14769 invocations with different options, in particular with and without
14772 @item -fdump-unnumbered-links
14773 @opindex fdump-unnumbered-links
14774 When doing debugging dumps (see @option{-d} option above), suppress
14775 instruction numbers for the links to the previous and next instructions
14778 @item -fdump-ipa-@var{switch}
14779 @itemx -fdump-ipa-@var{switch}-@var{options}
14781 Control the dumping at various stages of inter-procedural analysis
14782 language tree to a file. The file name is generated by appending a
14783 switch specific suffix to the source file name, and the file is created
14784 in the same directory as the output file. The following dumps are
14789 Enables all inter-procedural analysis dumps.
14792 Dumps information about call-graph optimization, unused function removal,
14793 and inlining decisions.
14796 Dump after function inlining.
14800 Additionally, the options @option{-optimized}, @option{-missed},
14801 @option{-note}, and @option{-all} can be provided, with the same meaning
14802 as for @option{-fopt-info}, defaulting to @option{-optimized}.
14804 For example, @option{-fdump-ipa-inline-optimized-missed} will emit
14805 information on callsites that were inlined, along with callsites
14806 that were not inlined.
14808 By default, the dump will contain messages about successful
14809 optimizations (equivalent to @option{-optimized}) together with
14810 low-level details about the analysis.
14812 @item -fdump-lang-all
14813 @itemx -fdump-lang-@var{switch}
14814 @itemx -fdump-lang-@var{switch}-@var{options}
14815 @itemx -fdump-lang-@var{switch}-@var{options}=@var{filename}
14816 @opindex fdump-lang-all
14817 @opindex fdump-lang
14818 Control the dumping of language-specific information. The @var{options}
14819 and @var{filename} portions behave as described in the
14820 @option{-fdump-tree} option. The following @var{switch} values are
14826 Enable all language-specific dumps.
14829 Dump class hierarchy information. Virtual table information is emitted
14830 unless '@option{slim}' is specified. This option is applicable to C++ only.
14833 Dump the raw internal tree data. This option is applicable to C++ only.
14837 @item -fdump-passes
14838 @opindex fdump-passes
14839 Print on @file{stderr} the list of optimization passes that are turned
14840 on and off by the current command-line options.
14842 @item -fdump-statistics-@var{option}
14843 @opindex fdump-statistics
14844 Enable and control dumping of pass statistics in a separate file. The
14845 file name is generated by appending a suffix ending in
14846 @samp{.statistics} to the source file name, and the file is created in
14847 the same directory as the output file. If the @samp{-@var{option}}
14848 form is used, @samp{-stats} causes counters to be summed over the
14849 whole compilation unit while @samp{-details} dumps every event as
14850 the passes generate them. The default with no option is to sum
14851 counters for each function compiled.
14853 @item -fdump-tree-all
14854 @itemx -fdump-tree-@var{switch}
14855 @itemx -fdump-tree-@var{switch}-@var{options}
14856 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
14857 @opindex fdump-tree-all
14858 @opindex fdump-tree
14859 Control the dumping at various stages of processing the intermediate
14860 language tree to a file. If the @samp{-@var{options}}
14861 form is used, @var{options} is a list of @samp{-} separated options
14862 which control the details of the dump. Not all options are applicable
14863 to all dumps; those that are not meaningful are ignored. The
14864 following options are available
14868 Print the address of each node. Usually this is not meaningful as it
14869 changes according to the environment and source file. Its primary use
14870 is for tying up a dump file with a debug environment.
14872 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
14873 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
14874 use working backward from mangled names in the assembly file.
14876 When dumping front-end intermediate representations, inhibit dumping
14877 of members of a scope or body of a function merely because that scope
14878 has been reached. Only dump such items when they are directly reachable
14879 by some other path.
14881 When dumping pretty-printed trees, this option inhibits dumping the
14882 bodies of control structures.
14884 When dumping RTL, print the RTL in slim (condensed) form instead of
14885 the default LISP-like representation.
14887 Print a raw representation of the tree. By default, trees are
14888 pretty-printed into a C-like representation.
14890 Enable more detailed dumps (not honored by every dump option). Also
14891 include information from the optimization passes.
14893 Enable dumping various statistics about the pass (not honored by every dump
14896 Enable showing basic block boundaries (disabled in raw dumps).
14898 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
14899 dump a representation of the control flow graph suitable for viewing with
14900 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
14901 the file is pretty-printed as a subgraph, so that GraphViz can render them
14902 all in a single plot.
14904 This option currently only works for RTL dumps, and the RTL is always
14905 dumped in slim form.
14907 Enable showing virtual operands for every statement.
14909 Enable showing line numbers for statements.
14911 Enable showing the unique ID (@code{DECL_UID}) for each variable.
14913 Enable showing the tree dump for each statement.
14915 Enable showing the EH region number holding each statement.
14917 Enable showing scalar evolution analysis details.
14919 Enable showing optimization information (only available in certain
14922 Enable showing missed optimization information (only available in certain
14925 Enable other detailed optimization information (only available in
14928 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
14929 and @option{lineno}.
14931 Turn on all optimization options, i.e., @option{optimized},
14932 @option{missed}, and @option{note}.
14935 To determine what tree dumps are available or find the dump for a pass
14936 of interest follow the steps below.
14940 Invoke GCC with @option{-fdump-passes} and in the @file{stderr} output
14941 look for a code that corresponds to the pass you are interested in.
14942 For example, the codes @code{tree-evrp}, @code{tree-vrp1}, and
14943 @code{tree-vrp2} correspond to the three Value Range Propagation passes.
14944 The number at the end distinguishes distinct invocations of the same pass.
14946 To enable the creation of the dump file, append the pass code to
14947 the @option{-fdump-} option prefix and invoke GCC with it. For example,
14948 to enable the dump from the Early Value Range Propagation pass, invoke
14949 GCC with the @option{-fdump-tree-evrp} option. Optionally, you may
14950 specify the name of the dump file. If you don't specify one, GCC
14951 creates as described below.
14953 Find the pass dump in a file whose name is composed of three components
14954 separated by a period: the name of the source file GCC was invoked to
14955 compile, a numeric suffix indicating the pass number followed by the
14956 letter @samp{t} for tree passes (and the letter @samp{r} for RTL passes),
14957 and finally the pass code. For example, the Early VRP pass dump might
14958 be in a file named @file{myfile.c.038t.evrp} in the current working
14959 directory. Note that the numeric codes are not stable and may change
14960 from one version of GCC to another.
14964 @itemx -fopt-info-@var{options}
14965 @itemx -fopt-info-@var{options}=@var{filename}
14967 Controls optimization dumps from various optimization passes. If the
14968 @samp{-@var{options}} form is used, @var{options} is a list of
14969 @samp{-} separated option keywords to select the dump details and
14972 The @var{options} can be divided into three groups:
14975 options describing what kinds of messages should be emitted,
14977 options describing the verbosity of the dump, and
14979 options describing which optimizations should be included.
14981 The options from each group can be freely mixed as they are
14982 non-overlapping. However, in case of any conflicts,
14983 the later options override the earlier options on the command
14986 The following options control which kinds of messages should be emitted:
14990 Print information when an optimization is successfully applied. It is
14991 up to a pass to decide which information is relevant. For example, the
14992 vectorizer passes print the source location of loops which are
14993 successfully vectorized.
14995 Print information about missed optimizations. Individual passes
14996 control which information to include in the output.
14998 Print verbose information about optimizations, such as certain
14999 transformations, more detailed messages about decisions etc.
15001 Print detailed optimization information. This includes
15002 @samp{optimized}, @samp{missed}, and @samp{note}.
15005 The following option controls the dump verbosity:
15009 By default, only ``high-level'' messages are emitted. This option enables
15010 additional, more detailed, messages, which are likely to only be of interest
15014 One or more of the following option keywords can be used to describe a
15015 group of optimizations:
15019 Enable dumps from all interprocedural optimizations.
15021 Enable dumps from all loop optimizations.
15023 Enable dumps from all inlining optimizations.
15025 Enable dumps from all OMP (Offloading and Multi Processing) optimizations.
15027 Enable dumps from all vectorization optimizations.
15029 Enable dumps from all optimizations. This is a superset of
15030 the optimization groups listed above.
15033 If @var{options} is
15034 omitted, it defaults to @samp{optimized-optall}, which means to dump messages
15035 about successful optimizations from all the passes, omitting messages
15036 that are treated as ``internals''.
15038 If the @var{filename} is provided, then the dumps from all the
15039 applicable optimizations are concatenated into the @var{filename}.
15040 Otherwise the dump is output onto @file{stderr}. Though multiple
15041 @option{-fopt-info} options are accepted, only one of them can include
15042 a @var{filename}. If other filenames are provided then all but the
15043 first such option are ignored.
15045 Note that the output @var{filename} is overwritten
15046 in case of multiple translation units. If a combined output from
15047 multiple translation units is desired, @file{stderr} should be used
15050 In the following example, the optimization info is output to
15059 gcc -O3 -fopt-info-missed=missed.all
15063 outputs missed optimization report from all the passes into
15064 @file{missed.all}, and this one:
15067 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
15071 prints information about missed optimization opportunities from
15072 vectorization passes on @file{stderr}.
15073 Note that @option{-fopt-info-vec-missed} is equivalent to
15074 @option{-fopt-info-missed-vec}. The order of the optimization group
15075 names and message types listed after @option{-fopt-info} does not matter.
15077 As another example,
15079 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
15083 outputs information about missed optimizations as well as
15084 optimized locations from all the inlining passes into
15090 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
15094 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
15095 in conflict since only one output file is allowed. In this case, only
15096 the first option takes effect and the subsequent options are
15097 ignored. Thus only @file{vec.miss} is produced which contains
15098 dumps from the vectorizer about missed opportunities.
15100 @item -fsave-optimization-record
15101 @opindex fsave-optimization-record
15102 Write a SRCFILE.opt-record.json.gz file detailing what optimizations
15103 were performed, for those optimizations that support @option{-fopt-info}.
15105 This option is experimental and the format of the data within the
15106 compressed JSON file is subject to change.
15108 It is roughly equivalent to a machine-readable version of
15109 @option{-fopt-info-all}, as a collection of messages with source file,
15110 line number and column number, with the following additional data for
15116 the execution count of the code being optimized, along with metadata about
15117 whether this was from actual profile data, or just an estimate, allowing
15118 consumers to prioritize messages by code hotness,
15121 the function name of the code being optimized, where applicable,
15124 the ``inlining chain'' for the code being optimized, so that when
15125 a function is inlined into several different places (which might
15126 themselves be inlined), the reader can distinguish between the copies,
15129 objects identifying those parts of the message that refer to expressions,
15130 statements or symbol-table nodes, which of these categories they are, and,
15131 when available, their source code location,
15134 the GCC pass that emitted the message, and
15137 the location in GCC's own code from which the message was emitted
15141 Additionally, some messages are logically nested within other
15142 messages, reflecting implementation details of the optimization
15145 @item -fsched-verbose=@var{n}
15146 @opindex fsched-verbose
15147 On targets that use instruction scheduling, this option controls the
15148 amount of debugging output the scheduler prints to the dump files.
15150 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
15151 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
15152 For @var{n} greater than one, it also output basic block probabilities,
15153 detailed ready list information and unit/insn info. For @var{n} greater
15154 than two, it includes RTL at abort point, control-flow and regions info.
15155 And for @var{n} over four, @option{-fsched-verbose} also includes
15160 @item -fenable-@var{kind}-@var{pass}
15161 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
15165 This is a set of options that are used to explicitly disable/enable
15166 optimization passes. These options are intended for use for debugging GCC.
15167 Compiler users should use regular options for enabling/disabling
15172 @item -fdisable-ipa-@var{pass}
15173 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15174 statically invoked in the compiler multiple times, the pass name should be
15175 appended with a sequential number starting from 1.
15177 @item -fdisable-rtl-@var{pass}
15178 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
15179 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
15180 statically invoked in the compiler multiple times, the pass name should be
15181 appended with a sequential number starting from 1. @var{range-list} is a
15182 comma-separated list of function ranges or assembler names. Each range is a number
15183 pair separated by a colon. The range is inclusive in both ends. If the range
15184 is trivial, the number pair can be simplified as a single number. If the
15185 function's call graph node's @var{uid} falls within one of the specified ranges,
15186 the @var{pass} is disabled for that function. The @var{uid} is shown in the
15187 function header of a dump file, and the pass names can be dumped by using
15188 option @option{-fdump-passes}.
15190 @item -fdisable-tree-@var{pass}
15191 @itemx -fdisable-tree-@var{pass}=@var{range-list}
15192 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
15195 @item -fenable-ipa-@var{pass}
15196 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
15197 statically invoked in the compiler multiple times, the pass name should be
15198 appended with a sequential number starting from 1.
15200 @item -fenable-rtl-@var{pass}
15201 @itemx -fenable-rtl-@var{pass}=@var{range-list}
15202 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
15203 description and examples.
15205 @item -fenable-tree-@var{pass}
15206 @itemx -fenable-tree-@var{pass}=@var{range-list}
15207 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
15208 of option arguments.
15212 Here are some examples showing uses of these options.
15216 # disable ccp1 for all functions
15217 -fdisable-tree-ccp1
15218 # disable complete unroll for function whose cgraph node uid is 1
15219 -fenable-tree-cunroll=1
15220 # disable gcse2 for functions at the following ranges [1,1],
15221 # [300,400], and [400,1000]
15222 # disable gcse2 for functions foo and foo2
15223 -fdisable-rtl-gcse2=foo,foo2
15224 # disable early inlining
15225 -fdisable-tree-einline
15226 # disable ipa inlining
15227 -fdisable-ipa-inline
15228 # enable tree full unroll
15229 -fenable-tree-unroll
15234 @itemx -fchecking=@var{n}
15236 @opindex fno-checking
15237 Enable internal consistency checking. The default depends on
15238 the compiler configuration. @option{-fchecking=2} enables further
15239 internal consistency checking that might affect code generation.
15241 @item -frandom-seed=@var{string}
15242 @opindex frandom-seed
15243 This option provides a seed that GCC uses in place of
15244 random numbers in generating certain symbol names
15245 that have to be different in every compiled file. It is also used to
15246 place unique stamps in coverage data files and the object files that
15247 produce them. You can use the @option{-frandom-seed} option to produce
15248 reproducibly identical object files.
15250 The @var{string} can either be a number (decimal, octal or hex) or an
15251 arbitrary string (in which case it's converted to a number by
15254 The @var{string} should be different for every file you compile.
15257 @itemx -save-temps=cwd
15258 @opindex save-temps
15259 Store the usual ``temporary'' intermediate files permanently; place them
15260 in the current directory and name them based on the source file. Thus,
15261 compiling @file{foo.c} with @option{-c -save-temps} produces files
15262 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
15263 preprocessed @file{foo.i} output file even though the compiler now
15264 normally uses an integrated preprocessor.
15266 When used in combination with the @option{-x} command-line option,
15267 @option{-save-temps} is sensible enough to avoid over writing an
15268 input source file with the same extension as an intermediate file.
15269 The corresponding intermediate file may be obtained by renaming the
15270 source file before using @option{-save-temps}.
15272 If you invoke GCC in parallel, compiling several different source
15273 files that share a common base name in different subdirectories or the
15274 same source file compiled for multiple output destinations, it is
15275 likely that the different parallel compilers will interfere with each
15276 other, and overwrite the temporary files. For instance:
15279 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
15280 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
15283 may result in @file{foo.i} and @file{foo.o} being written to
15284 simultaneously by both compilers.
15286 @item -save-temps=obj
15287 @opindex save-temps=obj
15288 Store the usual ``temporary'' intermediate files permanently. If the
15289 @option{-o} option is used, the temporary files are based on the
15290 object file. If the @option{-o} option is not used, the
15291 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
15296 gcc -save-temps=obj -c foo.c
15297 gcc -save-temps=obj -c bar.c -o dir/xbar.o
15298 gcc -save-temps=obj foobar.c -o dir2/yfoobar
15302 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
15303 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
15304 @file{dir2/yfoobar.o}.
15306 @item -time@r{[}=@var{file}@r{]}
15308 Report the CPU time taken by each subprocess in the compilation
15309 sequence. For C source files, this is the compiler proper and assembler
15310 (plus the linker if linking is done).
15312 Without the specification of an output file, the output looks like this:
15319 The first number on each line is the ``user time'', that is time spent
15320 executing the program itself. The second number is ``system time'',
15321 time spent executing operating system routines on behalf of the program.
15322 Both numbers are in seconds.
15324 With the specification of an output file, the output is appended to the
15325 named file, and it looks like this:
15328 0.12 0.01 cc1 @var{options}
15329 0.00 0.01 as @var{options}
15332 The ``user time'' and the ``system time'' are moved before the program
15333 name, and the options passed to the program are displayed, so that one
15334 can later tell what file was being compiled, and with which options.
15336 @item -fdump-final-insns@r{[}=@var{file}@r{]}
15337 @opindex fdump-final-insns
15338 Dump the final internal representation (RTL) to @var{file}. If the
15339 optional argument is omitted (or if @var{file} is @code{.}), the name
15340 of the dump file is determined by appending @code{.gkd} to the
15341 compilation output file name.
15343 @item -fcompare-debug@r{[}=@var{opts}@r{]}
15344 @opindex fcompare-debug
15345 @opindex fno-compare-debug
15346 If no error occurs during compilation, run the compiler a second time,
15347 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
15348 passed to the second compilation. Dump the final internal
15349 representation in both compilations, and print an error if they differ.
15351 If the equal sign is omitted, the default @option{-gtoggle} is used.
15353 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
15354 and nonzero, implicitly enables @option{-fcompare-debug}. If
15355 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
15356 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
15359 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
15360 is equivalent to @option{-fno-compare-debug}, which disables the dumping
15361 of the final representation and the second compilation, preventing even
15362 @env{GCC_COMPARE_DEBUG} from taking effect.
15364 To verify full coverage during @option{-fcompare-debug} testing, set
15365 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
15366 which GCC rejects as an invalid option in any actual compilation
15367 (rather than preprocessing, assembly or linking). To get just a
15368 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
15369 not overridden} will do.
15371 @item -fcompare-debug-second
15372 @opindex fcompare-debug-second
15373 This option is implicitly passed to the compiler for the second
15374 compilation requested by @option{-fcompare-debug}, along with options to
15375 silence warnings, and omitting other options that would cause the compiler
15376 to produce output to files or to standard output as a side effect. Dump
15377 files and preserved temporary files are renamed so as to contain the
15378 @code{.gk} additional extension during the second compilation, to avoid
15379 overwriting those generated by the first.
15381 When this option is passed to the compiler driver, it causes the
15382 @emph{first} compilation to be skipped, which makes it useful for little
15383 other than debugging the compiler proper.
15387 Turn off generation of debug info, if leaving out this option
15388 generates it, or turn it on at level 2 otherwise. The position of this
15389 argument in the command line does not matter; it takes effect after all
15390 other options are processed, and it does so only once, no matter how
15391 many times it is given. This is mainly intended to be used with
15392 @option{-fcompare-debug}.
15394 @item -fvar-tracking-assignments-toggle
15395 @opindex fvar-tracking-assignments-toggle
15396 @opindex fno-var-tracking-assignments-toggle
15397 Toggle @option{-fvar-tracking-assignments}, in the same way that
15398 @option{-gtoggle} toggles @option{-g}.
15402 Makes the compiler print out each function name as it is compiled, and
15403 print some statistics about each pass when it finishes.
15405 @item -ftime-report
15406 @opindex ftime-report
15407 Makes the compiler print some statistics about the time consumed by each
15408 pass when it finishes.
15410 @item -ftime-report-details
15411 @opindex ftime-report-details
15412 Record the time consumed by infrastructure parts separately for each pass.
15414 @item -fira-verbose=@var{n}
15415 @opindex fira-verbose
15416 Control the verbosity of the dump file for the integrated register allocator.
15417 The default value is 5. If the value @var{n} is greater or equal to 10,
15418 the dump output is sent to stderr using the same format as @var{n} minus 10.
15421 @opindex flto-report
15422 Prints a report with internal details on the workings of the link-time
15423 optimizer. The contents of this report vary from version to version.
15424 It is meant to be useful to GCC developers when processing object
15425 files in LTO mode (via @option{-flto}).
15427 Disabled by default.
15429 @item -flto-report-wpa
15430 @opindex flto-report-wpa
15431 Like @option{-flto-report}, but only print for the WPA phase of link-time
15435 @opindex fmem-report
15436 Makes the compiler print some statistics about permanent memory
15437 allocation when it finishes.
15439 @item -fmem-report-wpa
15440 @opindex fmem-report-wpa
15441 Makes the compiler print some statistics about permanent memory
15442 allocation for the WPA phase only.
15444 @item -fpre-ipa-mem-report
15445 @opindex fpre-ipa-mem-report
15446 @item -fpost-ipa-mem-report
15447 @opindex fpost-ipa-mem-report
15448 Makes the compiler print some statistics about permanent memory
15449 allocation before or after interprocedural optimization.
15451 @item -fprofile-report
15452 @opindex fprofile-report
15453 Makes the compiler print some statistics about consistency of the
15454 (estimated) profile and effect of individual passes.
15456 @item -fstack-usage
15457 @opindex fstack-usage
15458 Makes the compiler output stack usage information for the program, on a
15459 per-function basis. The filename for the dump is made by appending
15460 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
15461 the output file, if explicitly specified and it is not an executable,
15462 otherwise it is the basename of the source file. An entry is made up
15467 The name of the function.
15471 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
15474 The qualifier @code{static} means that the function manipulates the stack
15475 statically: a fixed number of bytes are allocated for the frame on function
15476 entry and released on function exit; no stack adjustments are otherwise made
15477 in the function. The second field is this fixed number of bytes.
15479 The qualifier @code{dynamic} means that the function manipulates the stack
15480 dynamically: in addition to the static allocation described above, stack
15481 adjustments are made in the body of the function, for example to push/pop
15482 arguments around function calls. If the qualifier @code{bounded} is also
15483 present, the amount of these adjustments is bounded at compile time and
15484 the second field is an upper bound of the total amount of stack used by
15485 the function. If it is not present, the amount of these adjustments is
15486 not bounded at compile time and the second field only represents the
15491 Emit statistics about front-end processing at the end of the compilation.
15492 This option is supported only by the C++ front end, and
15493 the information is generally only useful to the G++ development team.
15495 @item -fdbg-cnt-list
15496 @opindex fdbg-cnt-list
15497 Print the name and the counter upper bound for all debug counters.
15500 @item -fdbg-cnt=@var{counter-value-list}
15502 Set the internal debug counter lower and upper bound. @var{counter-value-list}
15503 is a comma-separated list of @var{name}:@var{lower_bound}:@var{upper_bound}
15504 tuples which sets the lower and the upper bound of each debug
15505 counter @var{name}. The @var{lower_bound} is optional and is zero
15506 initialized if not set.
15507 All debug counters have the initial upper bound of @code{UINT_MAX};
15508 thus @code{dbg_cnt} returns true always unless the upper bound
15509 is set by this option.
15510 For example, with @option{-fdbg-cnt=dce:2:4,tail_call:10},
15511 @code{dbg_cnt(dce)} returns true only for third and fourth invocation.
15512 For @code{dbg_cnt(tail_call)} true is returned for first 10 invocations.
15514 @item -print-file-name=@var{library}
15515 @opindex print-file-name
15516 Print the full absolute name of the library file @var{library} that
15517 would be used when linking---and don't do anything else. With this
15518 option, GCC does not compile or link anything; it just prints the
15521 @item -print-multi-directory
15522 @opindex print-multi-directory
15523 Print the directory name corresponding to the multilib selected by any
15524 other switches present in the command line. This directory is supposed
15525 to exist in @env{GCC_EXEC_PREFIX}.
15527 @item -print-multi-lib
15528 @opindex print-multi-lib
15529 Print the mapping from multilib directory names to compiler switches
15530 that enable them. The directory name is separated from the switches by
15531 @samp{;}, and each switch starts with an @samp{@@} instead of the
15532 @samp{-}, without spaces between multiple switches. This is supposed to
15533 ease shell processing.
15535 @item -print-multi-os-directory
15536 @opindex print-multi-os-directory
15537 Print the path to OS libraries for the selected
15538 multilib, relative to some @file{lib} subdirectory. If OS libraries are
15539 present in the @file{lib} subdirectory and no multilibs are used, this is
15540 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
15541 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
15542 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
15543 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
15545 @item -print-multiarch
15546 @opindex print-multiarch
15547 Print the path to OS libraries for the selected multiarch,
15548 relative to some @file{lib} subdirectory.
15550 @item -print-prog-name=@var{program}
15551 @opindex print-prog-name
15552 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
15554 @item -print-libgcc-file-name
15555 @opindex print-libgcc-file-name
15556 Same as @option{-print-file-name=libgcc.a}.
15558 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
15559 but you do want to link with @file{libgcc.a}. You can do:
15562 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
15565 @item -print-search-dirs
15566 @opindex print-search-dirs
15567 Print the name of the configured installation directory and a list of
15568 program and library directories @command{gcc} searches---and don't do anything else.
15570 This is useful when @command{gcc} prints the error message
15571 @samp{installation problem, cannot exec cpp0: No such file or directory}.
15572 To resolve this you either need to put @file{cpp0} and the other compiler
15573 components where @command{gcc} expects to find them, or you can set the environment
15574 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
15575 Don't forget the trailing @samp{/}.
15576 @xref{Environment Variables}.
15578 @item -print-sysroot
15579 @opindex print-sysroot
15580 Print the target sysroot directory that is used during
15581 compilation. This is the target sysroot specified either at configure
15582 time or using the @option{--sysroot} option, possibly with an extra
15583 suffix that depends on compilation options. If no target sysroot is
15584 specified, the option prints nothing.
15586 @item -print-sysroot-headers-suffix
15587 @opindex print-sysroot-headers-suffix
15588 Print the suffix added to the target sysroot when searching for
15589 headers, or give an error if the compiler is not configured with such
15590 a suffix---and don't do anything else.
15593 @opindex dumpmachine
15594 Print the compiler's target machine (for example,
15595 @samp{i686-pc-linux-gnu})---and don't do anything else.
15598 @opindex dumpversion
15599 Print the compiler version (for example, @code{3.0}, @code{6.3.0} or @code{7})---and don't do
15600 anything else. This is the compiler version used in filesystem paths and
15601 specs. Depending on how the compiler has been configured it can be just
15602 a single number (major version), two numbers separated by a dot (major and
15603 minor version) or three numbers separated by dots (major, minor and patchlevel
15606 @item -dumpfullversion
15607 @opindex dumpfullversion
15608 Print the full compiler version---and don't do anything else. The output is
15609 always three numbers separated by dots, major, minor and patchlevel version.
15613 Print the compiler's built-in specs---and don't do anything else. (This
15614 is used when GCC itself is being built.) @xref{Spec Files}.
15617 @node Submodel Options
15618 @section Machine-Dependent Options
15619 @cindex submodel options
15620 @cindex specifying hardware config
15621 @cindex hardware models and configurations, specifying
15622 @cindex target-dependent options
15623 @cindex machine-dependent options
15625 Each target machine supported by GCC can have its own options---for
15626 example, to allow you to compile for a particular processor variant or
15627 ABI, or to control optimizations specific to that machine. By
15628 convention, the names of machine-specific options start with
15631 Some configurations of the compiler also support additional target-specific
15632 options, usually for compatibility with other compilers on the same
15635 @c This list is ordered alphanumerically by subsection name.
15636 @c It should be the same order and spelling as these options are listed
15637 @c in Machine Dependent Options
15640 * AArch64 Options::
15641 * Adapteva Epiphany Options::
15642 * AMD GCN Options::
15646 * Blackfin Options::
15652 * DEC Alpha Options::
15657 * GNU/Linux Options::
15667 * MicroBlaze Options::
15670 * MN10300 Options::
15674 * Nios II Options::
15675 * Nvidia PTX Options::
15676 * OpenRISC Options::
15678 * picoChip Options::
15679 * PowerPC Options::
15683 * RS/6000 and PowerPC Options::
15685 * S/390 and zSeries Options::
15688 * Solaris 2 Options::
15690 * System V Options::
15691 * TILE-Gx Options::
15692 * TILEPro Options::
15697 * VxWorks Options::
15699 * x86 Windows Options::
15700 * Xstormy16 Options::
15702 * zSeries Options::
15705 @node AArch64 Options
15706 @subsection AArch64 Options
15707 @cindex AArch64 Options
15709 These options are defined for AArch64 implementations:
15713 @item -mabi=@var{name}
15715 Generate code for the specified data model. Permissible values
15716 are @samp{ilp32} for SysV-like data model where int, long int and pointers
15717 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
15718 but long int and pointers are 64 bits.
15720 The default depends on the specific target configuration. Note that
15721 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
15722 entire program with the same ABI, and link with a compatible set of libraries.
15725 @opindex mbig-endian
15726 Generate big-endian code. This is the default when GCC is configured for an
15727 @samp{aarch64_be-*-*} target.
15729 @item -mgeneral-regs-only
15730 @opindex mgeneral-regs-only
15731 Generate code which uses only the general-purpose registers. This will prevent
15732 the compiler from using floating-point and Advanced SIMD registers but will not
15733 impose any restrictions on the assembler.
15735 @item -mlittle-endian
15736 @opindex mlittle-endian
15737 Generate little-endian code. This is the default when GCC is configured for an
15738 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
15740 @item -mcmodel=tiny
15741 @opindex mcmodel=tiny
15742 Generate code for the tiny code model. The program and its statically defined
15743 symbols must be within 1MB of each other. Programs can be statically or
15744 dynamically linked.
15746 @item -mcmodel=small
15747 @opindex mcmodel=small
15748 Generate code for the small code model. The program and its statically defined
15749 symbols must be within 4GB of each other. Programs can be statically or
15750 dynamically linked. This is the default code model.
15752 @item -mcmodel=large
15753 @opindex mcmodel=large
15754 Generate code for the large code model. This makes no assumptions about
15755 addresses and sizes of sections. Programs can be statically linked only.
15757 @item -mstrict-align
15758 @itemx -mno-strict-align
15759 @opindex mstrict-align
15760 @opindex mno-strict-align
15761 Avoid or allow generating memory accesses that may not be aligned on a natural
15762 object boundary as described in the architecture specification.
15764 @item -momit-leaf-frame-pointer
15765 @itemx -mno-omit-leaf-frame-pointer
15766 @opindex momit-leaf-frame-pointer
15767 @opindex mno-omit-leaf-frame-pointer
15768 Omit or keep the frame pointer in leaf functions. The former behavior is the
15771 @item -mstack-protector-guard=@var{guard}
15772 @itemx -mstack-protector-guard-reg=@var{reg}
15773 @itemx -mstack-protector-guard-offset=@var{offset}
15774 @opindex mstack-protector-guard
15775 @opindex mstack-protector-guard-reg
15776 @opindex mstack-protector-guard-offset
15777 Generate stack protection code using canary at @var{guard}. Supported
15778 locations are @samp{global} for a global canary or @samp{sysreg} for a
15779 canary in an appropriate system register.
15781 With the latter choice the options
15782 @option{-mstack-protector-guard-reg=@var{reg}} and
15783 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15784 which system register to use as base register for reading the canary,
15785 and from what offset from that base register. There is no default
15786 register or offset as this is entirely for use within the Linux
15789 @item -mstack-protector-guard=@var{guard}
15790 @itemx -mstack-protector-guard-reg=@var{reg}
15791 @itemx -mstack-protector-guard-offset=@var{offset}
15792 @opindex mstack-protector-guard
15793 @opindex mstack-protector-guard-reg
15794 @opindex mstack-protector-guard-offset
15795 Generate stack protection code using canary at @var{guard}. Supported
15796 locations are @samp{global} for a global canary or @samp{sysreg} for a
15797 canary in an appropriate system register.
15799 With the latter choice the options
15800 @option{-mstack-protector-guard-reg=@var{reg}} and
15801 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
15802 which system register to use as base register for reading the canary,
15803 and from what offset from that base register. There is no default
15804 register or offset as this is entirely for use within the Linux
15807 @item -mtls-dialect=desc
15808 @opindex mtls-dialect=desc
15809 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
15810 of TLS variables. This is the default.
15812 @item -mtls-dialect=traditional
15813 @opindex mtls-dialect=traditional
15814 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
15817 @item -mtls-size=@var{size}
15819 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
15820 This option requires binutils 2.26 or newer.
15822 @item -mfix-cortex-a53-835769
15823 @itemx -mno-fix-cortex-a53-835769
15824 @opindex mfix-cortex-a53-835769
15825 @opindex mno-fix-cortex-a53-835769
15826 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
15827 This involves inserting a NOP instruction between memory instructions and
15828 64-bit integer multiply-accumulate instructions.
15830 @item -mfix-cortex-a53-843419
15831 @itemx -mno-fix-cortex-a53-843419
15832 @opindex mfix-cortex-a53-843419
15833 @opindex mno-fix-cortex-a53-843419
15834 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
15835 This erratum workaround is made at link time and this will only pass the
15836 corresponding flag to the linker.
15838 @item -mlow-precision-recip-sqrt
15839 @itemx -mno-low-precision-recip-sqrt
15840 @opindex mlow-precision-recip-sqrt
15841 @opindex mno-low-precision-recip-sqrt
15842 Enable or disable the reciprocal square root approximation.
15843 This option only has an effect if @option{-ffast-math} or
15844 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15845 precision of reciprocal square root results to about 16 bits for
15846 single precision and to 32 bits for double precision.
15848 @item -mlow-precision-sqrt
15849 @itemx -mno-low-precision-sqrt
15850 @opindex mlow-precision-sqrt
15851 @opindex mno-low-precision-sqrt
15852 Enable or disable the square root approximation.
15853 This option only has an effect if @option{-ffast-math} or
15854 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15855 precision of square root results to about 16 bits for
15856 single precision and to 32 bits for double precision.
15857 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
15859 @item -mlow-precision-div
15860 @itemx -mno-low-precision-div
15861 @opindex mlow-precision-div
15862 @opindex mno-low-precision-div
15863 Enable or disable the division approximation.
15864 This option only has an effect if @option{-ffast-math} or
15865 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
15866 precision of division results to about 16 bits for
15867 single precision and to 32 bits for double precision.
15869 @item -mtrack-speculation
15870 @itemx -mno-track-speculation
15871 Enable or disable generation of additional code to track speculative
15872 execution through conditional branches. The tracking state can then
15873 be used by the compiler when expanding calls to
15874 @code{__builtin_speculation_safe_copy} to permit a more efficient code
15875 sequence to be generated.
15877 @item -march=@var{name}
15879 Specify the name of the target architecture and, optionally, one or
15880 more feature modifiers. This option has the form
15881 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
15883 The permissible values for @var{arch} are @samp{armv8-a},
15884 @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a}, @samp{armv8.4-a},
15885 @samp{armv8.5-a} or @var{native}.
15887 The value @samp{armv8.5-a} implies @samp{armv8.4-a} and enables compiler
15888 support for the ARMv8.5-A architecture extensions.
15890 The value @samp{armv8.4-a} implies @samp{armv8.3-a} and enables compiler
15891 support for the ARMv8.4-A architecture extensions.
15893 The value @samp{armv8.3-a} implies @samp{armv8.2-a} and enables compiler
15894 support for the ARMv8.3-A architecture extensions.
15896 The value @samp{armv8.2-a} implies @samp{armv8.1-a} and enables compiler
15897 support for the ARMv8.2-A architecture extensions.
15899 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
15900 support for the ARMv8.1-A architecture extension. In particular, it
15901 enables the @samp{+crc}, @samp{+lse}, and @samp{+rdma} features.
15903 The value @samp{native} is available on native AArch64 GNU/Linux and
15904 causes the compiler to pick the architecture of the host system. This
15905 option has no effect if the compiler is unable to recognize the
15906 architecture of the host system,
15908 The permissible values for @var{feature} are listed in the sub-section
15909 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15910 Feature Modifiers}. Where conflicting feature modifiers are
15911 specified, the right-most feature is used.
15913 GCC uses @var{name} to determine what kind of instructions it can emit
15914 when generating assembly code. If @option{-march} is specified
15915 without either of @option{-mtune} or @option{-mcpu} also being
15916 specified, the code is tuned to perform well across a range of target
15917 processors implementing the target architecture.
15919 @item -mtune=@var{name}
15921 Specify the name of the target processor for which GCC should tune the
15922 performance of the code. Permissible values for this option are:
15923 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
15924 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
15925 @samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77},
15926 @samp{cortex-a65}, @samp{cortex-a65ae}, @samp{cortex-a34},
15927 @samp{ares}, @samp{exynos-m1}, @samp{emag}, @samp{falkor},
15928 @samp{neoverse-e1},@samp{neoverse-n1},@samp{qdf24xx}, @samp{saphira},
15929 @samp{phecda}, @samp{xgene1}, @samp{vulcan}, @samp{octeontx},
15930 @samp{octeontx81}, @samp{octeontx83}, @samp{thunderx}, @samp{thunderxt88},
15931 @samp{thunderxt88p1}, @samp{thunderxt81}, @samp{tsv110},
15932 @samp{thunderxt83}, @samp{thunderx2t99},
15933 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15934 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15935 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}
15938 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
15939 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53},
15940 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55} specify that GCC
15941 should tune for a big.LITTLE system.
15943 Additionally on native AArch64 GNU/Linux systems the value
15944 @samp{native} tunes performance to the host system. This option has no effect
15945 if the compiler is unable to recognize the processor of the host system.
15947 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
15948 are specified, the code is tuned to perform well across a range
15949 of target processors.
15951 This option cannot be suffixed by feature modifiers.
15953 @item -mcpu=@var{name}
15955 Specify the name of the target processor, optionally suffixed by one
15956 or more feature modifiers. This option has the form
15957 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
15958 the permissible values for @var{cpu} are the same as those available
15959 for @option{-mtune}. The permissible values for @var{feature} are
15960 documented in the sub-section on
15961 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
15962 Feature Modifiers}. Where conflicting feature modifiers are
15963 specified, the right-most feature is used.
15965 GCC uses @var{name} to determine what kind of instructions it can emit when
15966 generating assembly code (as if by @option{-march}) and to determine
15967 the target processor for which to tune for performance (as if
15968 by @option{-mtune}). Where this option is used in conjunction
15969 with @option{-march} or @option{-mtune}, those options take precedence
15970 over the appropriate part of this option.
15972 @item -moverride=@var{string}
15974 Override tuning decisions made by the back-end in response to a
15975 @option{-mtune=} switch. The syntax, semantics, and accepted values
15976 for @var{string} in this option are not guaranteed to be consistent
15979 This option is only intended to be useful when developing GCC.
15981 @item -mverbose-cost-dump
15982 @opindex mverbose-cost-dump
15983 Enable verbose cost model dumping in the debug dump files. This option is
15984 provided for use in debugging the compiler.
15986 @item -mpc-relative-literal-loads
15987 @itemx -mno-pc-relative-literal-loads
15988 @opindex mpc-relative-literal-loads
15989 @opindex mno-pc-relative-literal-loads
15990 Enable or disable PC-relative literal loads. With this option literal pools are
15991 accessed using a single instruction and emitted after each function. This
15992 limits the maximum size of functions to 1MB. This is enabled by default for
15993 @option{-mcmodel=tiny}.
15995 @item -msign-return-address=@var{scope}
15996 @opindex msign-return-address
15997 Select the function scope on which return address signing will be applied.
15998 Permissible values are @samp{none}, which disables return address signing,
15999 @samp{non-leaf}, which enables pointer signing for functions which are not leaf
16000 functions, and @samp{all}, which enables pointer signing for all functions. The
16001 default value is @samp{none}. This option has been deprecated by
16002 -mbranch-protection.
16004 @item -mbranch-protection=@var{none}|@var{standard}|@var{pac-ret}[+@var{leaf}+@var{b-key}]|@var{bti}
16005 @opindex mbranch-protection
16006 Select the branch protection features to use.
16007 @samp{none} is the default and turns off all types of branch protection.
16008 @samp{standard} turns on all types of branch protection features. If a feature
16009 has additional tuning options, then @samp{standard} sets it to its standard
16011 @samp{pac-ret[+@var{leaf}]} turns on return address signing to its standard
16012 level: signing functions that save the return address to memory (non-leaf
16013 functions will practically always do this) using the a-key. The optional
16014 argument @samp{leaf} can be used to extend the signing to include leaf
16015 functions. The optional argument @samp{b-key} can be used to sign the functions
16016 with the B-key instead of the A-key.
16017 @samp{bti} turns on branch target identification mechanism.
16019 @item -msve-vector-bits=@var{bits}
16020 @opindex msve-vector-bits
16021 Specify the number of bits in an SVE vector register. This option only has
16022 an effect when SVE is enabled.
16024 GCC supports two forms of SVE code generation: ``vector-length
16025 agnostic'' output that works with any size of vector register and
16026 ``vector-length specific'' output that allows GCC to make assumptions
16027 about the vector length when it is useful for optimization reasons.
16028 The possible values of @samp{bits} are: @samp{scalable}, @samp{128},
16029 @samp{256}, @samp{512}, @samp{1024} and @samp{2048}.
16030 Specifying @samp{scalable} selects vector-length agnostic
16031 output. At present @samp{-msve-vector-bits=128} also generates vector-length
16032 agnostic output. All other values generate vector-length specific code.
16033 The behavior of these values may change in future releases and no value except
16034 @samp{scalable} should be relied on for producing code that is portable across
16035 different hardware SVE vector lengths.
16037 The default is @samp{-msve-vector-bits=scalable}, which produces
16038 vector-length agnostic code.
16041 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
16042 @anchor{aarch64-feature-modifiers}
16043 @cindex @option{-march} feature modifiers
16044 @cindex @option{-mcpu} feature modifiers
16045 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
16046 the following and their inverses @option{no@var{feature}}:
16050 Enable CRC extension. This is on by default for
16051 @option{-march=armv8.1-a}.
16053 Enable Crypto extension. This also enables Advanced SIMD and floating-point
16056 Enable floating-point instructions. This is on by default for all possible
16057 values for options @option{-march} and @option{-mcpu}.
16059 Enable Advanced SIMD instructions. This also enables floating-point
16060 instructions. This is on by default for all possible values for options
16061 @option{-march} and @option{-mcpu}.
16063 Enable Scalable Vector Extension instructions. This also enables Advanced
16064 SIMD and floating-point instructions.
16066 Enable Large System Extension instructions. This is on by default for
16067 @option{-march=armv8.1-a}.
16069 Enable Round Double Multiply Accumulate instructions. This is on by default
16070 for @option{-march=armv8.1-a}.
16072 Enable FP16 extension. This also enables floating-point instructions.
16074 Enable FP16 fmla extension. This also enables FP16 extensions and
16075 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.
16078 Enable the RcPc extension. This does not change code generation from GCC,
16079 but is passed on to the assembler, enabling inline asm statements to use
16080 instructions from the RcPc extension.
16082 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
16084 Enable the Armv8-a aes and pmull crypto extension. This also enables Advanced
16087 Enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
16089 Enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD
16090 instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
16092 Enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions.
16093 Use of this option with architectures prior to Armv8.2-A is not supported.
16095 Enable the Statistical Profiling extension. This option is only to enable the
16096 extension at the assembler level and does not affect code generation.
16098 Enable the Armv8.5-a Random Number instructions. This option is only to
16099 enable the extension at the assembler level and does not affect code
16102 Enable the Armv8.5-a Memory Tagging Extensions. This option is only to
16103 enable the extension at the assembler level and does not affect code
16106 Enable the Armv8-a Speculation Barrier instruction. This option is only to
16107 enable the extension at the assembler level and does not affect code
16108 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16110 Enable the Armv8-a Speculative Store Bypass Safe instruction. This option
16111 is only to enable the extension at the assembler level and does not affect code
16112 generation. This option is enabled by default for @option{-march=armv8.5-a}.
16114 Enable the Armv8-a Execution and Data Prediction Restriction instructions.
16115 This option is only to enable the extension at the assembler level and does
16116 not affect code generation. This option is enabled by default for
16118 Enable the Armv8-a Scalable Vector Extension 2. This also enables SVE
16121 Enable SVE2 bitperm instructions. This also enables SVE2 instructions.
16123 Enable SVE2 sm4 instructions. This also enables SVE2 instructions.
16125 Enable SVE2 aes instructions. This also enables SVE2 instructions.
16127 Enable SVE2 sha3 instructions. This also enables SVE2 instructions.
16128 @option{-march=armv8.5-a}.
16130 Enable the Transactional Memory Extension.
16134 Feature @option{crypto} implies @option{aes}, @option{sha2}, and @option{simd},
16135 which implies @option{fp}.
16136 Conversely, @option{nofp} implies @option{nosimd}, which implies
16137 @option{nocrypto}, @option{noaes} and @option{nosha2}.
16139 @node Adapteva Epiphany Options
16140 @subsection Adapteva Epiphany Options
16142 These @samp{-m} options are defined for Adapteva Epiphany:
16145 @item -mhalf-reg-file
16146 @opindex mhalf-reg-file
16147 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
16148 That allows code to run on hardware variants that lack these registers.
16150 @item -mprefer-short-insn-regs
16151 @opindex mprefer-short-insn-regs
16152 Preferentially allocate registers that allow short instruction generation.
16153 This can result in increased instruction count, so this may either reduce or
16154 increase overall code size.
16156 @item -mbranch-cost=@var{num}
16157 @opindex mbranch-cost
16158 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16159 This cost is only a heuristic and is not guaranteed to produce
16160 consistent results across releases.
16164 Enable the generation of conditional moves.
16166 @item -mnops=@var{num}
16168 Emit @var{num} NOPs before every other generated instruction.
16170 @item -mno-soft-cmpsf
16171 @opindex mno-soft-cmpsf
16172 @opindex msoft-cmpsf
16173 For single-precision floating-point comparisons, emit an @code{fsub} instruction
16174 and test the flags. This is faster than a software comparison, but can
16175 get incorrect results in the presence of NaNs, or when two different small
16176 numbers are compared such that their difference is calculated as zero.
16177 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
16178 software comparisons.
16180 @item -mstack-offset=@var{num}
16181 @opindex mstack-offset
16182 Set the offset between the top of the stack and the stack pointer.
16183 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
16184 can be used by leaf functions without stack allocation.
16185 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
16186 Note also that this option changes the ABI; compiling a program with a
16187 different stack offset than the libraries have been compiled with
16188 generally does not work.
16189 This option can be useful if you want to evaluate if a different stack
16190 offset would give you better code, but to actually use a different stack
16191 offset to build working programs, it is recommended to configure the
16192 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
16194 @item -mno-round-nearest
16195 @opindex mno-round-nearest
16196 @opindex mround-nearest
16197 Make the scheduler assume that the rounding mode has been set to
16198 truncating. The default is @option{-mround-nearest}.
16201 @opindex mlong-calls
16202 If not otherwise specified by an attribute, assume all calls might be beyond
16203 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
16204 function address into a register before performing a (otherwise direct) call.
16205 This is the default.
16207 @item -mshort-calls
16208 @opindex short-calls
16209 If not otherwise specified by an attribute, assume all direct calls are
16210 in the range of the @code{b} / @code{bl} instructions, so use these instructions
16211 for direct calls. The default is @option{-mlong-calls}.
16215 Assume addresses can be loaded as 16-bit unsigned values. This does not
16216 apply to function addresses for which @option{-mlong-calls} semantics
16219 @item -mfp-mode=@var{mode}
16221 Set the prevailing mode of the floating-point unit.
16222 This determines the floating-point mode that is provided and expected
16223 at function call and return time. Making this mode match the mode you
16224 predominantly need at function start can make your programs smaller and
16225 faster by avoiding unnecessary mode switches.
16227 @var{mode} can be set to one the following values:
16231 Any mode at function entry is valid, and retained or restored when
16232 the function returns, and when it calls other functions.
16233 This mode is useful for compiling libraries or other compilation units
16234 you might want to incorporate into different programs with different
16235 prevailing FPU modes, and the convenience of being able to use a single
16236 object file outweighs the size and speed overhead for any extra
16237 mode switching that might be needed, compared with what would be needed
16238 with a more specific choice of prevailing FPU mode.
16241 This is the mode used for floating-point calculations with
16242 truncating (i.e.@: round towards zero) rounding mode. That includes
16243 conversion from floating point to integer.
16245 @item round-nearest
16246 This is the mode used for floating-point calculations with
16247 round-to-nearest-or-even rounding mode.
16250 This is the mode used to perform integer calculations in the FPU, e.g.@:
16251 integer multiply, or integer multiply-and-accumulate.
16254 The default is @option{-mfp-mode=caller}
16256 @item -mno-split-lohi
16257 @itemx -mno-postinc
16258 @itemx -mno-postmodify
16259 @opindex mno-split-lohi
16260 @opindex msplit-lohi
16261 @opindex mno-postinc
16263 @opindex mno-postmodify
16264 @opindex mpostmodify
16265 Code generation tweaks that disable, respectively, splitting of 32-bit
16266 loads, generation of post-increment addresses, and generation of
16267 post-modify addresses. The defaults are @option{msplit-lohi},
16268 @option{-mpost-inc}, and @option{-mpost-modify}.
16270 @item -mnovect-double
16271 @opindex mno-vect-double
16272 @opindex mvect-double
16273 Change the preferred SIMD mode to SImode. The default is
16274 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
16276 @item -max-vect-align=@var{num}
16277 @opindex max-vect-align
16278 The maximum alignment for SIMD vector mode types.
16279 @var{num} may be 4 or 8. The default is 8.
16280 Note that this is an ABI change, even though many library function
16281 interfaces are unaffected if they don't use SIMD vector modes
16282 in places that affect size and/or alignment of relevant types.
16284 @item -msplit-vecmove-early
16285 @opindex msplit-vecmove-early
16286 Split vector moves into single word moves before reload. In theory this
16287 can give better register allocation, but so far the reverse seems to be
16288 generally the case.
16290 @item -m1reg-@var{reg}
16292 Specify a register to hold the constant @minus{}1, which makes loading small negative
16293 constants and certain bitmasks faster.
16294 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
16295 which specify use of that register as a fixed register,
16296 and @samp{none}, which means that no register is used for this
16297 purpose. The default is @option{-m1reg-none}.
16301 @node AMD GCN Options
16302 @subsection AMD GCN Options
16303 @cindex AMD GCN Options
16305 These options are defined specifically for the AMD GCN port.
16309 @item -march=@var{gpu}
16311 @itemx -mtune=@var{gpu}
16313 Set architecture type or tuning for @var{gpu}. Supported values for @var{gpu}
16319 Compile for GCN3 Fiji devices (gfx803).
16322 Compile for GCN5 Vega 10 devices (gfx900).
16325 Compile for GCN5 Vega 20 devices (gfx906).
16329 @item -mstack-size=@var{bytes}
16330 @opindex mstack-size
16331 Specify how many @var{bytes} of stack space will be requested for each GPU
16332 thread (wave-front). Beware that there may be many threads and limited memory
16333 available. The size of the stack allocation may also have an impact on
16334 run-time performance. The default is 32KB when using OpenACC or OpenMP, and
16340 @subsection ARC Options
16341 @cindex ARC options
16343 The following options control the architecture variant for which code
16346 @c architecture variants
16349 @item -mbarrel-shifter
16350 @opindex mbarrel-shifter
16351 Generate instructions supported by barrel shifter. This is the default
16352 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
16355 @opindex mjli-alawys
16356 Force to call a function using jli_s instruction. This option is
16357 valid only for ARCv2 architecture.
16359 @item -mcpu=@var{cpu}
16361 Set architecture type, register usage, and instruction scheduling
16362 parameters for @var{cpu}. There are also shortcut alias options
16363 available for backward compatibility and convenience. Supported
16364 values for @var{cpu} are
16370 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
16374 Compile for ARC601. Alias: @option{-mARC601}.
16379 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
16380 This is the default when configured with @option{--with-cpu=arc700}@.
16383 Compile for ARC EM.
16386 Compile for ARC HS.
16389 Compile for ARC EM CPU with no hardware extensions.
16392 Compile for ARC EM4 CPU.
16395 Compile for ARC EM4 DMIPS CPU.
16398 Compile for ARC EM4 DMIPS CPU with the single-precision floating-point
16402 Compile for ARC EM4 DMIPS CPU with single-precision floating-point and
16403 double assist instructions.
16406 Compile for ARC HS CPU with no hardware extensions except the atomic
16410 Compile for ARC HS34 CPU.
16413 Compile for ARC HS38 CPU.
16416 Compile for ARC HS38 CPU with all hardware extensions on.
16419 Compile for ARC 600 CPU with @code{norm} instructions enabled.
16421 @item arc600_mul32x16
16422 Compile for ARC 600 CPU with @code{norm} and 32x16-bit multiply
16423 instructions enabled.
16426 Compile for ARC 600 CPU with @code{norm} and @code{mul64}-family
16427 instructions enabled.
16430 Compile for ARC 601 CPU with @code{norm} instructions enabled.
16432 @item arc601_mul32x16
16433 Compile for ARC 601 CPU with @code{norm} and 32x16-bit multiply
16434 instructions enabled.
16437 Compile for ARC 601 CPU with @code{norm} and @code{mul64}-family
16438 instructions enabled.
16441 Compile for ARC 700 on NPS400 chip.
16444 Compile for ARC EM minimalist configuration featuring reduced register
16451 @itemx -mdpfp-compact
16452 @opindex mdpfp-compact
16453 Generate double-precision FPX instructions, tuned for the compact
16457 @opindex mdpfp-fast
16458 Generate double-precision FPX instructions, tuned for the fast
16461 @item -mno-dpfp-lrsr
16462 @opindex mno-dpfp-lrsr
16463 Disable @code{lr} and @code{sr} instructions from using FPX extension
16468 Generate extended arithmetic instructions. Currently only
16469 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
16470 supported. This is always enabled for @option{-mcpu=ARC700}.
16475 Do not generate @code{mpy}-family instructions for ARC700. This option is
16480 Generate 32x16-bit multiply and multiply-accumulate instructions.
16484 Generate @code{mul64} and @code{mulu64} instructions.
16485 Only valid for @option{-mcpu=ARC600}.
16489 Generate @code{norm} instructions. This is the default if @option{-mcpu=ARC700}
16494 @itemx -mspfp-compact
16495 @opindex mspfp-compact
16496 Generate single-precision FPX instructions, tuned for the compact
16500 @opindex mspfp-fast
16501 Generate single-precision FPX instructions, tuned for the fast
16506 Enable generation of ARC SIMD instructions via target-specific
16507 builtins. Only valid for @option{-mcpu=ARC700}.
16510 @opindex msoft-float
16511 This option ignored; it is provided for compatibility purposes only.
16512 Software floating-point code is emitted by default, and this default
16513 can overridden by FPX options; @option{-mspfp}, @option{-mspfp-compact}, or
16514 @option{-mspfp-fast} for single precision, and @option{-mdpfp},
16515 @option{-mdpfp-compact}, or @option{-mdpfp-fast} for double precision.
16519 Generate @code{swap} instructions.
16523 This enables use of the locked load/store conditional extension to implement
16524 atomic memory built-in functions. Not available for ARC 6xx or ARC
16529 Enable @code{div} and @code{rem} instructions for ARCv2 cores.
16531 @item -mcode-density
16532 @opindex mcode-density
16533 Enable code density instructions for ARC EM.
16534 This option is on by default for ARC HS.
16538 Enable double load/store operations for ARC HS cores.
16540 @item -mtp-regno=@var{regno}
16542 Specify thread pointer register number.
16544 @item -mmpy-option=@var{multo}
16545 @opindex mmpy-option
16546 Compile ARCv2 code with a multiplier design option. You can specify
16547 the option using either a string or numeric value for @var{multo}.
16548 @samp{wlh1} is the default value. The recognized values are:
16553 No multiplier available.
16557 16x16 multiplier, fully pipelined.
16558 The following instructions are enabled: @code{mpyw} and @code{mpyuw}.
16562 32x32 multiplier, fully
16563 pipelined (1 stage). The following instructions are additionally
16564 enabled: @code{mpy}, @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16568 32x32 multiplier, fully pipelined
16569 (2 stages). The following instructions are additionally enabled: @code{mpy},
16570 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16574 Two 16x16 multipliers, blocking,
16575 sequential. The following instructions are additionally enabled: @code{mpy},
16576 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16580 One 16x16 multiplier, blocking,
16581 sequential. The following instructions are additionally enabled: @code{mpy},
16582 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16586 One 32x4 multiplier, blocking,
16587 sequential. The following instructions are additionally enabled: @code{mpy},
16588 @code{mpyu}, @code{mpym}, @code{mpymu}, and @code{mpy_s}.
16592 ARC HS SIMD support.
16596 ARC HS SIMD support.
16600 ARC HS SIMD support.
16604 This option is only available for ARCv2 cores@.
16606 @item -mfpu=@var{fpu}
16608 Enables support for specific floating-point hardware extensions for ARCv2
16609 cores. Supported values for @var{fpu} are:
16614 Enables support for single-precision floating-point hardware
16618 Enables support for double-precision floating-point hardware
16619 extensions. The single-precision floating-point extension is also
16620 enabled. Not available for ARC EM@.
16623 Enables support for double-precision floating-point hardware
16624 extensions using double-precision assist instructions. The single-precision
16625 floating-point extension is also enabled. This option is
16626 only available for ARC EM@.
16629 Enables support for double-precision floating-point hardware
16630 extensions using double-precision assist instructions.
16631 The single-precision floating-point, square-root, and divide
16632 extensions are also enabled. This option is
16633 only available for ARC EM@.
16636 Enables support for double-precision floating-point hardware
16637 extensions using double-precision assist instructions.
16638 The single-precision floating-point and fused multiply and add
16639 hardware extensions are also enabled. This option is
16640 only available for ARC EM@.
16643 Enables support for double-precision floating-point hardware
16644 extensions using double-precision assist instructions.
16645 All single-precision floating-point hardware extensions are also
16646 enabled. This option is only available for ARC EM@.
16649 Enables support for single-precision floating-point, square-root and divide
16650 hardware extensions@.
16653 Enables support for double-precision floating-point, square-root and divide
16654 hardware extensions. This option
16655 includes option @samp{fpus_div}. Not available for ARC EM@.
16658 Enables support for single-precision floating-point and
16659 fused multiply and add hardware extensions@.
16662 Enables support for double-precision floating-point and
16663 fused multiply and add hardware extensions. This option
16664 includes option @samp{fpus_fma}. Not available for ARC EM@.
16667 Enables support for all single-precision floating-point hardware
16671 Enables support for all single- and double-precision floating-point
16672 hardware extensions. Not available for ARC EM@.
16676 @item -mirq-ctrl-saved=@var{register-range}, @var{blink}, @var{lp_count}
16677 @opindex mirq-ctrl-saved
16678 Specifies general-purposes registers that the processor automatically
16679 saves/restores on interrupt entry and exit. @var{register-range} is
16680 specified as two registers separated by a dash. The register range
16681 always starts with @code{r0}, the upper limit is @code{fp} register.
16682 @var{blink} and @var{lp_count} are optional. This option is only
16683 valid for ARC EM and ARC HS cores.
16685 @item -mrgf-banked-regs=@var{number}
16686 @opindex mrgf-banked-regs
16687 Specifies the number of registers replicated in second register bank
16688 on entry to fast interrupt. Fast interrupts are interrupts with the
16689 highest priority level P0. These interrupts save only PC and STATUS32
16690 registers to avoid memory transactions during interrupt entry and exit
16691 sequences. Use this option when you are using fast interrupts in an
16692 ARC V2 family processor. Permitted values are 4, 8, 16, and 32.
16694 @item -mlpc-width=@var{width}
16695 @opindex mlpc-width
16696 Specify the width of the @code{lp_count} register. Valid values for
16697 @var{width} are 8, 16, 20, 24, 28 and 32 bits. The default width is
16698 fixed to 32 bits. If the width is less than 32, the compiler does not
16699 attempt to transform loops in your program to use the zero-delay loop
16700 mechanism unless it is known that the @code{lp_count} register can
16701 hold the required loop-counter value. Depending on the width
16702 specified, the compiler and run-time library might continue to use the
16703 loop mechanism for various needs. This option defines macro
16704 @code{__ARC_LPC_WIDTH__} with the value of @var{width}.
16708 This option instructs the compiler to generate code for a 16-entry
16709 register file. This option defines the @code{__ARC_RF16__}
16710 preprocessor macro.
16712 @item -mbranch-index
16713 @opindex mbranch-index
16714 Enable use of @code{bi} or @code{bih} instructions to implement jump
16719 The following options are passed through to the assembler, and also
16720 define preprocessor macro symbols.
16722 @c Flags used by the assembler, but for which we define preprocessor
16723 @c macro symbols as well.
16726 @opindex mdsp-packa
16727 Passed down to the assembler to enable the DSP Pack A extensions.
16728 Also sets the preprocessor symbol @code{__Xdsp_packa}. This option is
16733 Passed down to the assembler to enable the dual Viterbi butterfly
16734 extension. Also sets the preprocessor symbol @code{__Xdvbf}. This
16735 option is deprecated.
16737 @c ARC700 4.10 extension instruction
16740 Passed down to the assembler to enable the locked load/store
16741 conditional extension. Also sets the preprocessor symbol
16746 Passed down to the assembler. Also sets the preprocessor symbol
16747 @code{__Xxmac_d16}. This option is deprecated.
16751 Passed down to the assembler. Also sets the preprocessor symbol
16752 @code{__Xxmac_24}. This option is deprecated.
16754 @c ARC700 4.10 extension instruction
16757 Passed down to the assembler to enable the 64-bit time-stamp counter
16758 extension instruction. Also sets the preprocessor symbol
16759 @code{__Xrtsc}. This option is deprecated.
16761 @c ARC700 4.10 extension instruction
16764 Passed down to the assembler to enable the swap byte ordering
16765 extension instruction. Also sets the preprocessor symbol
16769 @opindex mtelephony
16770 Passed down to the assembler to enable dual- and single-operand
16771 instructions for telephony. Also sets the preprocessor symbol
16772 @code{__Xtelephony}. This option is deprecated.
16776 Passed down to the assembler to enable the XY memory extension. Also
16777 sets the preprocessor symbol @code{__Xxy}.
16781 The following options control how the assembly code is annotated:
16783 @c Assembly annotation options
16787 Annotate assembler instructions with estimated addresses.
16789 @item -mannotate-align
16790 @opindex mannotate-align
16791 Explain what alignment considerations lead to the decision to make an
16792 instruction short or long.
16796 The following options are passed through to the linker:
16798 @c options passed through to the linker
16802 Passed through to the linker, to specify use of the @code{arclinux} emulation.
16803 This option is enabled by default in tool chains built for
16804 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
16805 when profiling is not requested.
16807 @item -marclinux_prof
16808 @opindex marclinux_prof
16809 Passed through to the linker, to specify use of the
16810 @code{arclinux_prof} emulation. This option is enabled by default in
16811 tool chains built for @w{@code{arc-linux-uclibc}} and
16812 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
16816 The following options control the semantics of generated code:
16818 @c semantically relevant code generation options
16821 @opindex mlong-calls
16822 Generate calls as register indirect calls, thus providing access
16823 to the full 32-bit address range.
16825 @item -mmedium-calls
16826 @opindex mmedium-calls
16827 Don't use less than 25-bit addressing range for calls, which is the
16828 offset available for an unconditional branch-and-link
16829 instruction. Conditional execution of function calls is suppressed, to
16830 allow use of the 25-bit range, rather than the 21-bit range with
16831 conditional branch-and-link. This is the default for tool chains built
16832 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
16836 Put definitions of externally-visible data in a small data section if
16837 that data is no bigger than @var{num} bytes. The default value of
16838 @var{num} is 4 for any ARC configuration, or 8 when we have double
16839 load/store operations.
16844 Do not generate sdata references. This is the default for tool chains
16845 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
16848 @item -mvolatile-cache
16849 @opindex mvolatile-cache
16850 Use ordinarily cached memory accesses for volatile references. This is the
16853 @item -mno-volatile-cache
16854 @opindex mno-volatile-cache
16855 @opindex mvolatile-cache
16856 Enable cache bypass for volatile references.
16860 The following options fine tune code generation:
16861 @c code generation tuning options
16864 @opindex malign-call
16865 Do alignment optimizations for call instructions.
16867 @item -mauto-modify-reg
16868 @opindex mauto-modify-reg
16869 Enable the use of pre/post modify with register displacement.
16871 @item -mbbit-peephole
16872 @opindex mbbit-peephole
16873 Enable bbit peephole2.
16877 This option disables a target-specific pass in @file{arc_reorg} to
16878 generate compare-and-branch (@code{br@var{cc}}) instructions.
16879 It has no effect on
16880 generation of these instructions driven by the combiner pass.
16882 @item -mcase-vector-pcrel
16883 @opindex mcase-vector-pcrel
16884 Use PC-relative switch case tables to enable case table shortening.
16885 This is the default for @option{-Os}.
16887 @item -mcompact-casesi
16888 @opindex mcompact-casesi
16889 Enable compact @code{casesi} pattern. This is the default for @option{-Os},
16890 and only available for ARCv1 cores. This option is deprecated.
16892 @item -mno-cond-exec
16893 @opindex mno-cond-exec
16894 Disable the ARCompact-specific pass to generate conditional
16895 execution instructions.
16897 Due to delay slot scheduling and interactions between operand numbers,
16898 literal sizes, instruction lengths, and the support for conditional execution,
16899 the target-independent pass to generate conditional execution is often lacking,
16900 so the ARC port has kept a special pass around that tries to find more
16901 conditional execution generation opportunities after register allocation,
16902 branch shortening, and delay slot scheduling have been done. This pass
16903 generally, but not always, improves performance and code size, at the cost of
16904 extra compilation time, which is why there is an option to switch it off.
16905 If you have a problem with call instructions exceeding their allowable
16906 offset range because they are conditionalized, you should consider using
16907 @option{-mmedium-calls} instead.
16909 @item -mearly-cbranchsi
16910 @opindex mearly-cbranchsi
16911 Enable pre-reload use of the @code{cbranchsi} pattern.
16913 @item -mexpand-adddi
16914 @opindex mexpand-adddi
16915 Expand @code{adddi3} and @code{subdi3} at RTL generation time into
16916 @code{add.f}, @code{adc} etc. This option is deprecated.
16918 @item -mindexed-loads
16919 @opindex mindexed-loads
16920 Enable the use of indexed loads. This can be problematic because some
16921 optimizers then assume that indexed stores exist, which is not
16926 Enable Local Register Allocation. This is still experimental for ARC,
16927 so by default the compiler uses standard reload
16928 (i.e.@: @option{-mno-lra}).
16930 @item -mlra-priority-none
16931 @opindex mlra-priority-none
16932 Don't indicate any priority for target registers.
16934 @item -mlra-priority-compact
16935 @opindex mlra-priority-compact
16936 Indicate target register priority for r0..r3 / r12..r15.
16938 @item -mlra-priority-noncompact
16939 @opindex mlra-priority-noncompact
16940 Reduce target register priority for r0..r3 / r12..r15.
16943 @opindex mmillicode
16944 When optimizing for size (using @option{-Os}), prologues and epilogues
16945 that have to save or restore a large number of registers are often
16946 shortened by using call to a special function in libgcc; this is
16947 referred to as a @emph{millicode} call. As these calls can pose
16948 performance issues, and/or cause linking issues when linking in a
16949 nonstandard way, this option is provided to turn on or off millicode
16952 @item -mcode-density-frame
16953 @opindex mcode-density-frame
16954 This option enable the compiler to emit @code{enter} and @code{leave}
16955 instructions. These instructions are only valid for CPUs with
16956 code-density feature.
16959 @opindex mmixed-code
16960 Tweak register allocation to help 16-bit instruction generation.
16961 This generally has the effect of decreasing the average instruction size
16962 while increasing the instruction count.
16966 Enable @samp{q} instruction alternatives.
16967 This is the default for @option{-Os}.
16971 Enable @samp{Rcq} constraint handling.
16972 Most short code generation depends on this.
16973 This is the default.
16977 Enable @samp{Rcw} constraint handling.
16978 Most ccfsm condexec mostly depends on this.
16979 This is the default.
16981 @item -msize-level=@var{level}
16982 @opindex msize-level
16983 Fine-tune size optimization with regards to instruction lengths and alignment.
16984 The recognized values for @var{level} are:
16987 No size optimization. This level is deprecated and treated like @samp{1}.
16990 Short instructions are used opportunistically.
16993 In addition, alignment of loops and of code after barriers are dropped.
16996 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
17000 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
17001 the behavior when this is not set is equivalent to level @samp{1}.
17003 @item -mtune=@var{cpu}
17005 Set instruction scheduling parameters for @var{cpu}, overriding any implied
17006 by @option{-mcpu=}.
17008 Supported values for @var{cpu} are
17012 Tune for ARC600 CPU.
17015 Tune for ARC601 CPU.
17018 Tune for ARC700 CPU with standard multiplier block.
17021 Tune for ARC700 CPU with XMAC block.
17024 Tune for ARC725D CPU.
17027 Tune for ARC750D CPU.
17031 @item -mmultcost=@var{num}
17033 Cost to assume for a multiply instruction, with @samp{4} being equal to a
17034 normal instruction.
17036 @item -munalign-prob-threshold=@var{probability}
17037 @opindex munalign-prob-threshold
17038 Set probability threshold for unaligning branches.
17039 When tuning for @samp{ARC700} and optimizing for speed, branches without
17040 filled delay slot are preferably emitted unaligned and long, unless
17041 profiling indicates that the probability for the branch to be taken
17042 is below @var{probability}. @xref{Cross-profiling}.
17043 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
17047 The following options are maintained for backward compatibility, but
17048 are now deprecated and will be removed in a future release:
17050 @c Deprecated options
17058 @opindex mbig-endian
17061 Compile code for big-endian targets. Use of these options is now
17062 deprecated. Big-endian code is supported by configuring GCC to build
17063 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets,
17064 for which big endian is the default.
17066 @item -mlittle-endian
17067 @opindex mlittle-endian
17070 Compile code for little-endian targets. Use of these options is now
17071 deprecated. Little-endian code is supported by configuring GCC to build
17072 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets,
17073 for which little endian is the default.
17075 @item -mbarrel_shifter
17076 @opindex mbarrel_shifter
17077 Replaced by @option{-mbarrel-shifter}.
17079 @item -mdpfp_compact
17080 @opindex mdpfp_compact
17081 Replaced by @option{-mdpfp-compact}.
17084 @opindex mdpfp_fast
17085 Replaced by @option{-mdpfp-fast}.
17088 @opindex mdsp_packa
17089 Replaced by @option{-mdsp-packa}.
17093 Replaced by @option{-mea}.
17097 Replaced by @option{-mmac-24}.
17101 Replaced by @option{-mmac-d16}.
17103 @item -mspfp_compact
17104 @opindex mspfp_compact
17105 Replaced by @option{-mspfp-compact}.
17108 @opindex mspfp_fast
17109 Replaced by @option{-mspfp-fast}.
17111 @item -mtune=@var{cpu}
17113 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
17114 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
17115 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively.
17117 @item -multcost=@var{num}
17119 Replaced by @option{-mmultcost}.
17124 @subsection ARM Options
17125 @cindex ARM options
17127 These @samp{-m} options are defined for the ARM port:
17130 @item -mabi=@var{name}
17132 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
17133 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
17136 @opindex mapcs-frame
17137 Generate a stack frame that is compliant with the ARM Procedure Call
17138 Standard for all functions, even if this is not strictly necessary for
17139 correct execution of the code. Specifying @option{-fomit-frame-pointer}
17140 with this option causes the stack frames not to be generated for
17141 leaf functions. The default is @option{-mno-apcs-frame}.
17142 This option is deprecated.
17146 This is a synonym for @option{-mapcs-frame} and is deprecated.
17149 @c not currently implemented
17150 @item -mapcs-stack-check
17151 @opindex mapcs-stack-check
17152 Generate code to check the amount of stack space available upon entry to
17153 every function (that actually uses some stack space). If there is
17154 insufficient space available then either the function
17155 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
17156 called, depending upon the amount of stack space required. The runtime
17157 system is required to provide these functions. The default is
17158 @option{-mno-apcs-stack-check}, since this produces smaller code.
17160 @c not currently implemented
17161 @item -mapcs-reentrant
17162 @opindex mapcs-reentrant
17163 Generate reentrant, position-independent code. The default is
17164 @option{-mno-apcs-reentrant}.
17167 @item -mthumb-interwork
17168 @opindex mthumb-interwork
17169 Generate code that supports calling between the ARM and Thumb
17170 instruction sets. Without this option, on pre-v5 architectures, the
17171 two instruction sets cannot be reliably used inside one program. The
17172 default is @option{-mno-thumb-interwork}, since slightly larger code
17173 is generated when @option{-mthumb-interwork} is specified. In AAPCS
17174 configurations this option is meaningless.
17176 @item -mno-sched-prolog
17177 @opindex mno-sched-prolog
17178 @opindex msched-prolog
17179 Prevent the reordering of instructions in the function prologue, or the
17180 merging of those instruction with the instructions in the function's
17181 body. This means that all functions start with a recognizable set
17182 of instructions (or in fact one of a choice from a small set of
17183 different function prologues), and this information can be used to
17184 locate the start of functions inside an executable piece of code. The
17185 default is @option{-msched-prolog}.
17187 @item -mfloat-abi=@var{name}
17188 @opindex mfloat-abi
17189 Specifies which floating-point ABI to use. Permissible values
17190 are: @samp{soft}, @samp{softfp} and @samp{hard}.
17192 Specifying @samp{soft} causes GCC to generate output containing
17193 library calls for floating-point operations.
17194 @samp{softfp} allows the generation of code using hardware floating-point
17195 instructions, but still uses the soft-float calling conventions.
17196 @samp{hard} allows generation of floating-point instructions
17197 and uses FPU-specific calling conventions.
17199 The default depends on the specific target configuration. Note that
17200 the hard-float and soft-float ABIs are not link-compatible; you must
17201 compile your entire program with the same ABI, and link with a
17202 compatible set of libraries.
17204 @item -mgeneral-regs-only
17205 @opindex mgeneral-regs-only
17206 Generate code which uses only the general-purpose registers. This will prevent
17207 the compiler from using floating-point and Advanced SIMD registers but will not
17208 impose any restrictions on the assembler.
17210 @item -mlittle-endian
17211 @opindex mlittle-endian
17212 Generate code for a processor running in little-endian mode. This is
17213 the default for all standard configurations.
17216 @opindex mbig-endian
17217 Generate code for a processor running in big-endian mode; the default is
17218 to compile code for a little-endian processor.
17223 When linking a big-endian image select between BE8 and BE32 formats.
17224 The option has no effect for little-endian images and is ignored. The
17225 default is dependent on the selected target architecture. For ARMv6
17226 and later architectures the default is BE8, for older architectures
17227 the default is BE32. BE32 format has been deprecated by ARM.
17229 @item -march=@var{name}@r{[}+extension@dots{}@r{]}
17231 This specifies the name of the target ARM architecture. GCC uses this
17232 name to determine what kind of instructions it can emit when generating
17233 assembly code. This option can be used in conjunction with or instead
17234 of the @option{-mcpu=} option.
17236 Permissible names are:
17238 @samp{armv5t}, @samp{armv5te},
17239 @samp{armv6}, @samp{armv6j}, @samp{armv6k}, @samp{armv6kz}, @samp{armv6t2},
17240 @samp{armv6z}, @samp{armv6zk},
17241 @samp{armv7}, @samp{armv7-a}, @samp{armv7ve},
17242 @samp{armv8-a}, @samp{armv8.1-a}, @samp{armv8.2-a}, @samp{armv8.3-a},
17247 @samp{armv6-m}, @samp{armv6s-m},
17248 @samp{armv7-m}, @samp{armv7e-m},
17249 @samp{armv8-m.base}, @samp{armv8-m.main},
17250 @samp{iwmmxt} and @samp{iwmmxt2}.
17252 Additionally, the following architectures, which lack support for the
17253 Thumb execution state, are recognized but support is deprecated: @samp{armv4}.
17255 Many of the architectures support extensions. These can be added by
17256 appending @samp{+@var{extension}} to the architecture name. Extension
17257 options are processed in order and capabilities accumulate. An extension
17258 will also enable any necessary base extensions
17259 upon which it depends. For example, the @samp{+crypto} extension
17260 will always enable the @samp{+simd} extension. The exception to the
17261 additive construction is for extensions that are prefixed with
17262 @samp{+no@dots{}}: these extensions disable the specified option and
17263 any other extensions that may depend on the presence of that
17266 For example, @samp{-march=armv7-a+simd+nofp+vfpv4} is equivalent to
17267 writing @samp{-march=armv7-a+vfpv4} since the @samp{+simd} option is
17268 entirely disabled by the @samp{+nofp} option that follows it.
17270 Most extension names are generically named, but have an effect that is
17271 dependent upon the architecture to which it is applied. For example,
17272 the @samp{+simd} option can be applied to both @samp{armv7-a} and
17273 @samp{armv8-a} architectures, but will enable the original ARMv7-A
17274 Advanced SIMD (Neon) extensions for @samp{armv7-a} and the ARMv8-A
17275 variant for @samp{armv8-a}.
17277 The table below lists the supported extensions for each architecture.
17278 Architectures not mentioned do not support any extensions.
17291 The VFPv2 floating-point instructions. The extension @samp{+vfpv2} can be
17292 used as an alias for this extension.
17295 Disable the floating-point instructions.
17299 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
17302 The VFPv3 floating-point instructions, with 16 double-precision
17303 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17304 for this extension. Note that floating-point is not supported by the
17305 base ARMv7-M architecture, but is compatible with both the ARMv7-A and
17306 ARMv7-R architectures.
17309 Disable the floating-point instructions.
17315 The multiprocessing extension.
17318 The security extension.
17321 The VFPv3 floating-point instructions, with 16 double-precision
17322 registers. The extension @samp{+vfpv3-d16} can be used as an alias
17323 for this extension.
17326 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17327 The extensions @samp{+neon} and @samp{+neon-vfpv3} can be used as aliases
17328 for this extension.
17331 The VFPv3 floating-point instructions, with 32 double-precision
17334 @item +vfpv3-d16-fp16
17335 The VFPv3 floating-point instructions, with 16 double-precision
17336 registers and the half-precision floating-point conversion operations.
17339 The VFPv3 floating-point instructions, with 32 double-precision
17340 registers and the half-precision floating-point conversion operations.
17343 The VFPv4 floating-point instructions, with 16 double-precision
17347 The VFPv4 floating-point instructions, with 32 double-precision
17351 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17352 the half-precision floating-point conversion operations.
17355 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
17358 Disable the Advanced SIMD instructions (does not disable floating point).
17361 Disable the floating-point and Advanced SIMD instructions.
17365 The extended version of the ARMv7-A architecture with support for
17369 The VFPv4 floating-point instructions, with 16 double-precision registers.
17370 The extension @samp{+vfpv4-d16} can be used as an alias for this extension.
17373 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The
17374 extension @samp{+neon-vfpv4} can be used as an alias for this extension.
17377 The VFPv3 floating-point instructions, with 16 double-precision
17381 The VFPv3 floating-point instructions, with 32 double-precision
17384 @item +vfpv3-d16-fp16
17385 The VFPv3 floating-point instructions, with 16 double-precision
17386 registers and the half-precision floating-point conversion operations.
17389 The VFPv3 floating-point instructions, with 32 double-precision
17390 registers and the half-precision floating-point conversion operations.
17393 The VFPv4 floating-point instructions, with 16 double-precision
17397 The VFPv4 floating-point instructions, with 32 double-precision
17401 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions.
17402 The extension @samp{+neon-vfpv3} can be used as an alias for this extension.
17405 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with
17406 the half-precision floating-point conversion operations.
17409 Disable the Advanced SIMD instructions (does not disable floating point).
17412 Disable the floating-point and Advanced SIMD instructions.
17418 The Cyclic Redundancy Check (CRC) instructions.
17420 The ARMv8-A Advanced SIMD and floating-point instructions.
17422 The cryptographic instructions.
17424 Disable the cryptographic instructions.
17426 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17428 Speculation Barrier Instruction.
17430 Execution and Data Prediction Restriction Instructions.
17436 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17439 The cryptographic instructions. This also enables the Advanced SIMD and
17440 floating-point instructions.
17443 Disable the cryptographic instructions.
17446 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17449 Speculation Barrier Instruction.
17452 Execution and Data Prediction Restriction Instructions.
17459 The half-precision floating-point data processing instructions.
17460 This also enables the Advanced SIMD and floating-point instructions.
17463 The half-precision floating-point fmla extension. This also enables
17464 the half-precision floating-point extension and Advanced SIMD and
17465 floating-point instructions.
17468 The ARMv8.1-A Advanced SIMD and floating-point instructions.
17471 The cryptographic instructions. This also enables the Advanced SIMD and
17472 floating-point instructions.
17475 Enable the Dot Product extension. This also enables Advanced SIMD instructions.
17478 Disable the cryptographic extension.
17481 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17484 Speculation Barrier Instruction.
17487 Execution and Data Prediction Restriction Instructions.
17493 The half-precision floating-point data processing instructions.
17494 This also enables the Advanced SIMD and floating-point instructions as well
17495 as the Dot Product extension and the half-precision floating-point fmla
17499 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17500 Dot Product extension.
17503 The cryptographic instructions. This also enables the Advanced SIMD and
17504 floating-point instructions as well as the Dot Product extension.
17507 Disable the cryptographic extension.
17510 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17513 Speculation Barrier Instruction.
17516 Execution and Data Prediction Restriction Instructions.
17522 The half-precision floating-point data processing instructions.
17523 This also enables the Advanced SIMD and floating-point instructions as well
17524 as the Dot Product extension and the half-precision floating-point fmla
17528 The ARMv8.3-A Advanced SIMD and floating-point instructions as well as the
17529 Dot Product extension.
17532 The cryptographic instructions. This also enables the Advanced SIMD and
17533 floating-point instructions as well as the Dot Product extension.
17536 Disable the cryptographic extension.
17539 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17545 The single-precision VFPv3 floating-point instructions. The extension
17546 @samp{+vfpv3xd} can be used as an alias for this extension.
17549 The VFPv3 floating-point instructions with 16 double-precision registers.
17550 The extension +vfpv3-d16 can be used as an alias for this extension.
17552 @item +vfpv3xd-d16-fp16
17553 The single-precision VFPv3 floating-point instructions with 16 double-precision
17554 registers and the half-precision floating-point conversion operations.
17556 @item +vfpv3-d16-fp16
17557 The VFPv3 floating-point instructions with 16 double-precision
17558 registers and the half-precision floating-point conversion operations.
17561 Disable the floating-point extension.
17564 The ARM-state integer division instructions.
17567 Disable the ARM-state integer division extension.
17573 The single-precision VFPv4 floating-point instructions.
17576 The single-precision FPv5 floating-point instructions.
17579 The single- and double-precision FPv5 floating-point instructions.
17582 Disable the floating-point extensions.
17588 The DSP instructions.
17591 Disable the DSP extension.
17594 The single-precision floating-point instructions.
17597 The single- and double-precision floating-point instructions.
17600 Disable the floating-point extension.
17606 The Cyclic Redundancy Check (CRC) instructions.
17608 The single-precision FPv5 floating-point instructions.
17610 The ARMv8-A Advanced SIMD and floating-point instructions.
17612 The cryptographic instructions.
17614 Disable the cryptographic instructions.
17616 Disable the floating-point, Advanced SIMD and cryptographic instructions.
17621 @option{-march=native} causes the compiler to auto-detect the architecture
17622 of the build computer. At present, this feature is only supported on
17623 GNU/Linux, and not all architectures are recognized. If the auto-detect
17624 is unsuccessful the option has no effect.
17626 @item -mtune=@var{name}
17628 This option specifies the name of the target ARM processor for
17629 which GCC should tune the performance of the code.
17630 For some ARM implementations better performance can be obtained by using
17632 Permissible names are: @samp{arm7tdmi}, @samp{arm7tdmi-s}, @samp{arm710t},
17633 @samp{arm720t}, @samp{arm740t}, @samp{strongarm}, @samp{strongarm110},
17634 @samp{strongarm1100}, 0@samp{strongarm1110}, @samp{arm8}, @samp{arm810},
17635 @samp{arm9}, @samp{arm9e}, @samp{arm920}, @samp{arm920t}, @samp{arm922t},
17636 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm926ej-s},
17637 @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi}, @samp{arm1020t},
17638 @samp{arm1026ej-s}, @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
17639 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
17640 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
17641 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
17642 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
17643 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55},
17644 @samp{cortex-a57}, @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75},
17645 @samp{cortex-a76}, @samp{cortex-a76ae}, @samp{cortex-a77},
17646 @samp{ares}, @samp{cortex-r4}, @samp{cortex-r4f},
17647 @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52},
17648 @samp{cortex-m0}, @samp{cortex-m0plus}, @samp{cortex-m1}, @samp{cortex-m3},
17649 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m23}, @samp{cortex-m33},
17650 @samp{cortex-m35p},
17651 @samp{cortex-m1.small-multiply}, @samp{cortex-m0.small-multiply},
17652 @samp{cortex-m0plus.small-multiply}, @samp{exynos-m1}, @samp{marvell-pj4},
17653 @samp{neoverse-n1}, @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2},
17654 @samp{ep9312}, @samp{fa526}, @samp{fa626}, @samp{fa606te}, @samp{fa626te},
17655 @samp{fmp626}, @samp{fa726te}, @samp{xgene1}.
17657 Additionally, this option can specify that GCC should tune the performance
17658 of the code for a big.LITTLE system. Permissible names are:
17659 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
17660 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17661 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53},
17662 @samp{cortex-a75.cortex-a55}, @samp{cortex-a76.cortex-a55}.
17664 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
17665 performance for a blend of processors within architecture @var{arch}.
17666 The aim is to generate code that run well on the current most popular
17667 processors, balancing between optimizations that benefit some CPUs in the
17668 range, and avoiding performance pitfalls of other CPUs. The effects of
17669 this option may change in future GCC versions as CPU models come and go.
17671 @option{-mtune} permits the same extension options as @option{-mcpu}, but
17672 the extension options do not affect the tuning of the generated code.
17674 @option{-mtune=native} causes the compiler to auto-detect the CPU
17675 of the build computer. At present, this feature is only supported on
17676 GNU/Linux, and not all architectures are recognized. If the auto-detect is
17677 unsuccessful the option has no effect.
17679 @item -mcpu=@var{name}@r{[}+extension@dots{}@r{]}
17681 This specifies the name of the target ARM processor. GCC uses this name
17682 to derive the name of the target ARM architecture (as if specified
17683 by @option{-march}) and the ARM processor type for which to tune for
17684 performance (as if specified by @option{-mtune}). Where this option
17685 is used in conjunction with @option{-march} or @option{-mtune},
17686 those options take precedence over the appropriate part of this option.
17688 Many of the supported CPUs implement optional architectural
17689 extensions. Where this is so the architectural extensions are
17690 normally enabled by default. If implementations that lack the
17691 extension exist, then the extension syntax can be used to disable
17692 those extensions that have been omitted. For floating-point and
17693 Advanced SIMD (Neon) instructions, the settings of the options
17694 @option{-mfloat-abi} and @option{-mfpu} must also be considered:
17695 floating-point and Advanced SIMD instructions will only be used if
17696 @option{-mfloat-abi} is not set to @samp{soft}; and any setting of
17697 @option{-mfpu} other than @samp{auto} will override the available
17698 floating-point and SIMD extension instructions.
17700 For example, @samp{cortex-a9} can be found in three major
17701 configurations: integer only, with just a floating-point unit or with
17702 floating-point and Advanced SIMD. The default is to enable all the
17703 instructions, but the extensions @samp{+nosimd} and @samp{+nofp} can
17704 be used to disable just the SIMD or both the SIMD and floating-point
17705 instructions respectively.
17707 Permissible names for this option are the same as those for
17710 The following extension options are common to the listed CPUs:
17714 Disable the DSP instructions on @samp{cortex-m33}, @samp{cortex-m35p}.
17717 Disables the floating-point instructions on @samp{arm9e},
17718 @samp{arm946e-s}, @samp{arm966e-s}, @samp{arm968e-s}, @samp{arm10e},
17719 @samp{arm1020e}, @samp{arm1022e}, @samp{arm926ej-s},
17720 @samp{arm1026ej-s}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
17721 @samp{cortex-m4}, @samp{cortex-m7}, @samp{cortex-m33} and @samp{cortex-m35p}.
17722 Disables the floating-point and SIMD instructions on
17723 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7},
17724 @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a12},
17725 @samp{cortex-a15}, @samp{cortex-a17}, @samp{cortex-a15.cortex-a7},
17726 @samp{cortex-a17.cortex-a7}, @samp{cortex-a32}, @samp{cortex-a35},
17727 @samp{cortex-a53} and @samp{cortex-a55}.
17730 Disables the double-precision component of the floating-point instructions
17731 on @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8}, @samp{cortex-r52} and
17735 Disables the SIMD (but not floating-point) instructions on
17736 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}
17737 and @samp{cortex-a9}.
17740 Enables the cryptographic instructions on @samp{cortex-a32},
17741 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a55}, @samp{cortex-a57},
17742 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-a75}, @samp{exynos-m1},
17743 @samp{xgene1}, @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
17744 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53} and
17745 @samp{cortex-a75.cortex-a55}.
17748 Additionally the @samp{generic-armv7-a} pseudo target defaults to
17749 VFPv3 with 16 double-precision registers. It supports the following
17750 extension options: @samp{mp}, @samp{sec}, @samp{vfpv3-d16},
17751 @samp{vfpv3}, @samp{vfpv3-d16-fp16}, @samp{vfpv3-fp16},
17752 @samp{vfpv4-d16}, @samp{vfpv4}, @samp{neon}, @samp{neon-vfpv3},
17753 @samp{neon-fp16}, @samp{neon-vfpv4}. The meanings are the same as for
17754 the extensions to @option{-march=armv7-a}.
17756 @option{-mcpu=generic-@var{arch}} is also permissible, and is
17757 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
17758 See @option{-mtune} for more information.
17760 @option{-mcpu=native} causes the compiler to auto-detect the CPU
17761 of the build computer. At present, this feature is only supported on
17762 GNU/Linux, and not all architectures are recognized. If the auto-detect
17763 is unsuccessful the option has no effect.
17765 @item -mfpu=@var{name}
17767 This specifies what floating-point hardware (or hardware emulation) is
17768 available on the target. Permissible names are: @samp{auto}, @samp{vfpv2},
17770 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
17771 @samp{vfpv3xd-fp16}, @samp{neon-vfpv3}, @samp{neon-fp16}, @samp{vfpv4},
17772 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
17773 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
17774 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
17775 Note that @samp{neon} is an alias for @samp{neon-vfpv3} and @samp{vfp}
17776 is an alias for @samp{vfpv2}.
17778 The setting @samp{auto} is the default and is special. It causes the
17779 compiler to select the floating-point and Advanced SIMD instructions
17780 based on the settings of @option{-mcpu} and @option{-march}.
17782 If the selected floating-point hardware includes the NEON extension
17783 (e.g.@: @option{-mfpu=neon}), note that floating-point
17784 operations are not generated by GCC's auto-vectorization pass unless
17785 @option{-funsafe-math-optimizations} is also specified. This is
17786 because NEON hardware does not fully implement the IEEE 754 standard for
17787 floating-point arithmetic (in particular denormal values are treated as
17788 zero), so the use of NEON instructions may lead to a loss of precision.
17790 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}).
17792 @item -mfp16-format=@var{name}
17793 @opindex mfp16-format
17794 Specify the format of the @code{__fp16} half-precision floating-point type.
17795 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
17796 the default is @samp{none}, in which case the @code{__fp16} type is not
17797 defined. @xref{Half-Precision}, for more information.
17799 @item -mstructure-size-boundary=@var{n}
17800 @opindex mstructure-size-boundary
17801 The sizes of all structures and unions are rounded up to a multiple
17802 of the number of bits set by this option. Permissible values are 8, 32
17803 and 64. The default value varies for different toolchains. For the COFF
17804 targeted toolchain the default value is 8. A value of 64 is only allowed
17805 if the underlying ABI supports it.
17807 Specifying a larger number can produce faster, more efficient code, but
17808 can also increase the size of the program. Different values are potentially
17809 incompatible. Code compiled with one value cannot necessarily expect to
17810 work with code or libraries compiled with another value, if they exchange
17811 information using structures or unions.
17813 This option is deprecated.
17815 @item -mabort-on-noreturn
17816 @opindex mabort-on-noreturn
17817 Generate a call to the function @code{abort} at the end of a
17818 @code{noreturn} function. It is executed if the function tries to
17822 @itemx -mno-long-calls
17823 @opindex mlong-calls
17824 @opindex mno-long-calls
17825 Tells the compiler to perform function calls by first loading the
17826 address of the function into a register and then performing a subroutine
17827 call on this register. This switch is needed if the target function
17828 lies outside of the 64-megabyte addressing range of the offset-based
17829 version of subroutine call instruction.
17831 Even if this switch is enabled, not all function calls are turned
17832 into long calls. The heuristic is that static functions, functions
17833 that have the @code{short_call} attribute, functions that are inside
17834 the scope of a @code{#pragma no_long_calls} directive, and functions whose
17835 definitions have already been compiled within the current compilation
17836 unit are not turned into long calls. The exceptions to this rule are
17837 that weak function definitions, functions with the @code{long_call}
17838 attribute or the @code{section} attribute, and functions that are within
17839 the scope of a @code{#pragma long_calls} directive are always
17840 turned into long calls.
17842 This feature is not enabled by default. Specifying
17843 @option{-mno-long-calls} restores the default behavior, as does
17844 placing the function calls within the scope of a @code{#pragma
17845 long_calls_off} directive. Note these switches have no effect on how
17846 the compiler generates code to handle function calls via function
17849 @item -msingle-pic-base
17850 @opindex msingle-pic-base
17851 Treat the register used for PIC addressing as read-only, rather than
17852 loading it in the prologue for each function. The runtime system is
17853 responsible for initializing this register with an appropriate value
17854 before execution begins.
17856 @item -mpic-register=@var{reg}
17857 @opindex mpic-register
17858 Specify the register to be used for PIC addressing.
17859 For standard PIC base case, the default is any suitable register
17860 determined by compiler. For single PIC base case, the default is
17861 @samp{R9} if target is EABI based or stack-checking is enabled,
17862 otherwise the default is @samp{R10}.
17864 @item -mpic-data-is-text-relative
17865 @opindex mpic-data-is-text-relative
17866 Assume that the displacement between the text and data segments is fixed
17867 at static link time. This permits using PC-relative addressing
17868 operations to access data known to be in the data segment. For
17869 non-VxWorks RTP targets, this option is enabled by default. When
17870 disabled on such targets, it will enable @option{-msingle-pic-base} by
17873 @item -mpoke-function-name
17874 @opindex mpoke-function-name
17875 Write the name of each function into the text section, directly
17876 preceding the function prologue. The generated code is similar to this:
17880 .ascii "arm_poke_function_name", 0
17883 .word 0xff000000 + (t1 - t0)
17884 arm_poke_function_name
17886 stmfd sp!, @{fp, ip, lr, pc@}
17890 When performing a stack backtrace, code can inspect the value of
17891 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
17892 location @code{pc - 12} and the top 8 bits are set, then we know that
17893 there is a function name embedded immediately preceding this location
17894 and has length @code{((pc[-3]) & 0xff000000)}.
17901 Select between generating code that executes in ARM and Thumb
17902 states. The default for most configurations is to generate code
17903 that executes in ARM state, but the default can be changed by
17904 configuring GCC with the @option{--with-mode=}@var{state}
17907 You can also override the ARM and Thumb mode for each function
17908 by using the @code{target("thumb")} and @code{target("arm")} function attributes
17909 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
17912 @opindex mflip-thumb
17913 Switch ARM/Thumb modes on alternating functions.
17914 This option is provided for regression testing of mixed Thumb/ARM code
17915 generation, and is not intended for ordinary use in compiling code.
17918 @opindex mtpcs-frame
17919 Generate a stack frame that is compliant with the Thumb Procedure Call
17920 Standard for all non-leaf functions. (A leaf function is one that does
17921 not call any other functions.) The default is @option{-mno-tpcs-frame}.
17923 @item -mtpcs-leaf-frame
17924 @opindex mtpcs-leaf-frame
17925 Generate a stack frame that is compliant with the Thumb Procedure Call
17926 Standard for all leaf functions. (A leaf function is one that does
17927 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
17929 @item -mcallee-super-interworking
17930 @opindex mcallee-super-interworking
17931 Gives all externally visible functions in the file being compiled an ARM
17932 instruction set header which switches to Thumb mode before executing the
17933 rest of the function. This allows these functions to be called from
17934 non-interworking code. This option is not valid in AAPCS configurations
17935 because interworking is enabled by default.
17937 @item -mcaller-super-interworking
17938 @opindex mcaller-super-interworking
17939 Allows calls via function pointers (including virtual functions) to
17940 execute correctly regardless of whether the target code has been
17941 compiled for interworking or not. There is a small overhead in the cost
17942 of executing a function pointer if this option is enabled. This option
17943 is not valid in AAPCS configurations because interworking is enabled
17946 @item -mtp=@var{name}
17948 Specify the access model for the thread local storage pointer. The valid
17949 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
17950 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
17951 (supported in the arm6k architecture), and @samp{auto}, which uses the
17952 best available method for the selected processor. The default setting is
17955 @item -mtls-dialect=@var{dialect}
17956 @opindex mtls-dialect
17957 Specify the dialect to use for accessing thread local storage. Two
17958 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
17959 @samp{gnu} dialect selects the original GNU scheme for supporting
17960 local and global dynamic TLS models. The @samp{gnu2} dialect
17961 selects the GNU descriptor scheme, which provides better performance
17962 for shared libraries. The GNU descriptor scheme is compatible with
17963 the original scheme, but does require new assembler, linker and
17964 library support. Initial and local exec TLS models are unaffected by
17965 this option and always use the original scheme.
17967 @item -mword-relocations
17968 @opindex mword-relocations
17969 Only generate absolute relocations on word-sized values (i.e.@: R_ARM_ABS32).
17970 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
17971 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
17972 is specified. This option conflicts with @option{-mslow-flash-data}.
17974 @item -mfix-cortex-m3-ldrd
17975 @opindex mfix-cortex-m3-ldrd
17976 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
17977 with overlapping destination and base registers are used. This option avoids
17978 generating these instructions. This option is enabled by default when
17979 @option{-mcpu=cortex-m3} is specified.
17981 @item -munaligned-access
17982 @itemx -mno-unaligned-access
17983 @opindex munaligned-access
17984 @opindex mno-unaligned-access
17985 Enables (or disables) reading and writing of 16- and 32- bit values
17986 from addresses that are not 16- or 32- bit aligned. By default
17987 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for
17988 ARMv8-M Baseline architectures, and enabled for all other
17989 architectures. If unaligned access is not enabled then words in packed
17990 data structures are accessed a byte at a time.
17992 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
17993 generated object file to either true or false, depending upon the
17994 setting of this option. If unaligned access is enabled then the
17995 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
17998 @item -mneon-for-64bits
17999 @opindex mneon-for-64bits
18000 This option is deprecated and has no effect.
18002 @item -mslow-flash-data
18003 @opindex mslow-flash-data
18004 Assume loading data from flash is slower than fetching instruction.
18005 Therefore literal load is minimized for better performance.
18006 This option is only supported when compiling for ARMv7 M-profile and
18007 off by default. It conflicts with @option{-mword-relocations}.
18009 @item -masm-syntax-unified
18010 @opindex masm-syntax-unified
18011 Assume inline assembler is using unified asm syntax. The default is
18012 currently off which implies divided syntax. This option has no impact
18013 on Thumb2. However, this may change in future releases of GCC.
18014 Divided syntax should be considered deprecated.
18016 @item -mrestrict-it
18017 @opindex mrestrict-it
18018 Restricts generation of IT blocks to conform to the rules of ARMv8-A.
18019 IT blocks can only contain a single 16-bit instruction from a select
18020 set of instructions. This option is on by default for ARMv8-A Thumb mode.
18022 @item -mprint-tune-info
18023 @opindex mprint-tune-info
18024 Print CPU tuning information as comment in assembler file. This is
18025 an option used only for regression testing of the compiler and not
18026 intended for ordinary use in compiling code. This option is disabled
18029 @item -mverbose-cost-dump
18030 @opindex mverbose-cost-dump
18031 Enable verbose cost model dumping in the debug dump files. This option is
18032 provided for use in debugging the compiler.
18035 @opindex mpure-code
18036 Do not allow constant data to be placed in code sections.
18037 Additionally, when compiling for ELF object format give all text sections the
18038 ELF processor-specific section attribute @code{SHF_ARM_PURECODE}. This option
18039 is only available when generating non-pic code for M-profile targets with the
18044 Generate secure code as per the "ARMv8-M Security Extensions: Requirements on
18045 Development Tools Engineering Specification", which can be found on
18046 @url{http://infocenter.arm.com/help/topic/com.arm.doc.ecm0359818/ECM0359818_armv8m_security_extensions_reqs_on_dev_tools_1_0.pdf}.
18052 Select the FDPIC ABI, which uses 64-bit function descriptors to
18053 represent pointers to functions. When the compiler is configured for
18054 @code{arm-*-uclinuxfdpiceabi} targets, this option is on by default
18055 and implies @option{-fPIE} if none of the PIC/PIE-related options is
18056 provided. On other targets, it only enables the FDPIC-specific code
18057 generation features, and the user should explicitly provide the
18058 PIC/PIE-related options as needed.
18060 Note that static linking is not supported because it would still
18061 involve the dynamic linker when the program self-relocates. If such
18062 behavior is acceptable, use -static and -Wl,-dynamic-linker options.
18064 The opposite @option{-mno-fdpic} option is useful (and required) to
18065 build the Linux kernel using the same (@code{arm-*-uclinuxfdpiceabi})
18066 toolchain as the one used to build the userland programs.
18071 @subsection AVR Options
18072 @cindex AVR Options
18074 These options are defined for AVR implementations:
18077 @item -mmcu=@var{mcu}
18079 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
18081 The default for this option is@tie{}@samp{avr2}.
18083 GCC supports the following AVR devices and ISAs:
18085 @include avr-mmcu.texi
18090 Assume that all data in static storage can be accessed by LDS / STS
18091 instructions. This option has only an effect on reduced Tiny devices like
18092 ATtiny40. See also the @code{absdata}
18093 @ref{AVR Variable Attributes,variable attribute}.
18095 @item -maccumulate-args
18096 @opindex maccumulate-args
18097 Accumulate outgoing function arguments and acquire/release the needed
18098 stack space for outgoing function arguments once in function
18099 prologue/epilogue. Without this option, outgoing arguments are pushed
18100 before calling a function and popped afterwards.
18102 Popping the arguments after the function call can be expensive on
18103 AVR so that accumulating the stack space might lead to smaller
18104 executables because arguments need not be removed from the
18105 stack after such a function call.
18107 This option can lead to reduced code size for functions that perform
18108 several calls to functions that get their arguments on the stack like
18109 calls to printf-like functions.
18111 @item -mbranch-cost=@var{cost}
18112 @opindex mbranch-cost
18113 Set the branch costs for conditional branch instructions to
18114 @var{cost}. Reasonable values for @var{cost} are small, non-negative
18115 integers. The default branch cost is 0.
18117 @item -mcall-prologues
18118 @opindex mcall-prologues
18119 Functions prologues/epilogues are expanded as calls to appropriate
18120 subroutines. Code size is smaller.
18122 @item -mgas-isr-prologues
18123 @opindex mgas-isr-prologues
18124 Interrupt service routines (ISRs) may use the @code{__gcc_isr} pseudo
18125 instruction supported by GNU Binutils.
18126 If this option is on, the feature can still be disabled for individual
18127 ISRs by means of the @ref{AVR Function Attributes,,@code{no_gccisr}}
18128 function attribute. This feature is activated per default
18129 if optimization is on (but not with @option{-Og}, @pxref{Optimize Options}),
18130 and if GNU Binutils support @w{@uref{https://sourceware.org/PR21683,PR21683}}.
18134 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
18135 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
18136 and @code{long long} is 4 bytes. Please note that this option does not
18137 conform to the C standards, but it results in smaller code
18140 @item -mmain-is-OS_task
18141 @opindex mmain-is-OS_task
18142 Do not save registers in @code{main}. The effect is the same like
18143 attaching attribute @ref{AVR Function Attributes,,@code{OS_task}}
18144 to @code{main}. It is activated per default if optimization is on.
18146 @item -mn-flash=@var{num}
18148 Assume that the flash memory has a size of
18149 @var{num} times 64@tie{}KiB.
18151 @item -mno-interrupts
18152 @opindex mno-interrupts
18153 Generated code is not compatible with hardware interrupts.
18154 Code size is smaller.
18158 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
18159 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
18160 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
18161 the assembler's command line and the @option{--relax} option to the
18162 linker's command line.
18164 Jump relaxing is performed by the linker because jump offsets are not
18165 known before code is located. Therefore, the assembler code generated by the
18166 compiler is the same, but the instructions in the executable may
18167 differ from instructions in the assembler code.
18169 Relaxing must be turned on if linker stubs are needed, see the
18170 section on @code{EIND} and linker stubs below.
18174 Assume that the device supports the Read-Modify-Write
18175 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
18177 @item -mshort-calls
18178 @opindex mshort-calls
18180 Assume that @code{RJMP} and @code{RCALL} can target the whole
18183 This option is used internally for multilib selection. It is
18184 not an optimization option, and you don't need to set it by hand.
18188 Treat the stack pointer register as an 8-bit register,
18189 i.e.@: assume the high byte of the stack pointer is zero.
18190 In general, you don't need to set this option by hand.
18192 This option is used internally by the compiler to select and
18193 build multilibs for architectures @code{avr2} and @code{avr25}.
18194 These architectures mix devices with and without @code{SPH}.
18195 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
18196 the compiler driver adds or removes this option from the compiler
18197 proper's command line, because the compiler then knows if the device
18198 or architecture has an 8-bit stack pointer and thus no @code{SPH}
18203 Use address register @code{X} in a way proposed by the hardware. This means
18204 that @code{X} is only used in indirect, post-increment or
18205 pre-decrement addressing.
18207 Without this option, the @code{X} register may be used in the same way
18208 as @code{Y} or @code{Z} which then is emulated by additional
18210 For example, loading a value with @code{X+const} addressing with a
18211 small non-negative @code{const < 64} to a register @var{Rn} is
18215 adiw r26, const ; X += const
18216 ld @var{Rn}, X ; @var{Rn} = *X
18217 sbiw r26, const ; X -= const
18221 @opindex mtiny-stack
18222 Only change the lower 8@tie{}bits of the stack pointer.
18224 @item -mfract-convert-truncate
18225 @opindex mfract-convert-truncate
18226 Allow to use truncation instead of rounding towards zero for fractional fixed-point types.
18229 @opindex nodevicelib
18230 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
18232 @item -Waddr-space-convert
18233 @opindex Waddr-space-convert
18234 @opindex Wno-addr-space-convert
18235 Warn about conversions between address spaces in the case where the
18236 resulting address space is not contained in the incoming address space.
18238 @item -Wmisspelled-isr
18239 @opindex Wmisspelled-isr
18240 @opindex Wno-misspelled-isr
18241 Warn if the ISR is misspelled, i.e.@: without __vector prefix.
18242 Enabled by default.
18245 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
18246 @cindex @code{EIND}
18247 Pointers in the implementation are 16@tie{}bits wide.
18248 The address of a function or label is represented as word address so
18249 that indirect jumps and calls can target any code address in the
18250 range of 64@tie{}Ki words.
18252 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
18253 bytes of program memory space, there is a special function register called
18254 @code{EIND} that serves as most significant part of the target address
18255 when @code{EICALL} or @code{EIJMP} instructions are used.
18257 Indirect jumps and calls on these devices are handled as follows by
18258 the compiler and are subject to some limitations:
18263 The compiler never sets @code{EIND}.
18266 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
18267 instructions or might read @code{EIND} directly in order to emulate an
18268 indirect call/jump by means of a @code{RET} instruction.
18271 The compiler assumes that @code{EIND} never changes during the startup
18272 code or during the application. In particular, @code{EIND} is not
18273 saved/restored in function or interrupt service routine
18277 For indirect calls to functions and computed goto, the linker
18278 generates @emph{stubs}. Stubs are jump pads sometimes also called
18279 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
18280 The stub contains a direct jump to the desired address.
18283 Linker relaxation must be turned on so that the linker generates
18284 the stubs correctly in all situations. See the compiler option
18285 @option{-mrelax} and the linker option @option{--relax}.
18286 There are corner cases where the linker is supposed to generate stubs
18287 but aborts without relaxation and without a helpful error message.
18290 The default linker script is arranged for code with @code{EIND = 0}.
18291 If code is supposed to work for a setup with @code{EIND != 0}, a custom
18292 linker script has to be used in order to place the sections whose
18293 name start with @code{.trampolines} into the segment where @code{EIND}
18297 The startup code from libgcc never sets @code{EIND}.
18298 Notice that startup code is a blend of code from libgcc and AVR-LibC.
18299 For the impact of AVR-LibC on @code{EIND}, see the
18300 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
18303 It is legitimate for user-specific startup code to set up @code{EIND}
18304 early, for example by means of initialization code located in
18305 section @code{.init3}. Such code runs prior to general startup code
18306 that initializes RAM and calls constructors, but after the bit
18307 of startup code from AVR-LibC that sets @code{EIND} to the segment
18308 where the vector table is located.
18310 #include <avr/io.h>
18313 __attribute__((section(".init3"),naked,used,no_instrument_function))
18314 init3_set_eind (void)
18316 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
18317 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
18322 The @code{__trampolines_start} symbol is defined in the linker script.
18325 Stubs are generated automatically by the linker if
18326 the following two conditions are met:
18329 @item The address of a label is taken by means of the @code{gs} modifier
18330 (short for @emph{generate stubs}) like so:
18332 LDI r24, lo8(gs(@var{func}))
18333 LDI r25, hi8(gs(@var{func}))
18335 @item The final location of that label is in a code segment
18336 @emph{outside} the segment where the stubs are located.
18340 The compiler emits such @code{gs} modifiers for code labels in the
18341 following situations:
18343 @item Taking address of a function or code label.
18344 @item Computed goto.
18345 @item If prologue-save function is used, see @option{-mcall-prologues}
18346 command-line option.
18347 @item Switch/case dispatch tables. If you do not want such dispatch
18348 tables you can specify the @option{-fno-jump-tables} command-line option.
18349 @item C and C++ constructors/destructors called during startup/shutdown.
18350 @item If the tools hit a @code{gs()} modifier explained above.
18354 Jumping to non-symbolic addresses like so is @emph{not} supported:
18359 /* Call function at word address 0x2 */
18360 return ((int(*)(void)) 0x2)();
18364 Instead, a stub has to be set up, i.e.@: the function has to be called
18365 through a symbol (@code{func_4} in the example):
18370 extern int func_4 (void);
18372 /* Call function at byte address 0x4 */
18377 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
18378 Alternatively, @code{func_4} can be defined in the linker script.
18381 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
18382 @cindex @code{RAMPD}
18383 @cindex @code{RAMPX}
18384 @cindex @code{RAMPY}
18385 @cindex @code{RAMPZ}
18386 Some AVR devices support memories larger than the 64@tie{}KiB range
18387 that can be accessed with 16-bit pointers. To access memory locations
18388 outside this 64@tie{}KiB range, the content of a @code{RAMP}
18389 register is used as high part of the address:
18390 The @code{X}, @code{Y}, @code{Z} address register is concatenated
18391 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
18392 register, respectively, to get a wide address. Similarly,
18393 @code{RAMPD} is used together with direct addressing.
18397 The startup code initializes the @code{RAMP} special function
18398 registers with zero.
18401 If a @ref{AVR Named Address Spaces,named address space} other than
18402 generic or @code{__flash} is used, then @code{RAMPZ} is set
18403 as needed before the operation.
18406 If the device supports RAM larger than 64@tie{}KiB and the compiler
18407 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
18408 is reset to zero after the operation.
18411 If the device comes with a specific @code{RAMP} register, the ISR
18412 prologue/epilogue saves/restores that SFR and initializes it with
18413 zero in case the ISR code might (implicitly) use it.
18416 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
18417 If you use inline assembler to read from locations outside the
18418 16-bit address range and change one of the @code{RAMP} registers,
18419 you must reset it to zero after the access.
18423 @subsubsection AVR Built-in Macros
18425 GCC defines several built-in macros so that the user code can test
18426 for the presence or absence of features. Almost any of the following
18427 built-in macros are deduced from device capabilities and thus
18428 triggered by the @option{-mmcu=} command-line option.
18430 For even more AVR-specific built-in macros see
18431 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
18436 Build-in macro that resolves to a decimal number that identifies the
18437 architecture and depends on the @option{-mmcu=@var{mcu}} option.
18438 Possible values are:
18440 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
18441 @code{4}, @code{5}, @code{51}, @code{6}
18443 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
18444 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
18449 @code{102}, @code{103}, @code{104},
18450 @code{105}, @code{106}, @code{107}
18452 for @var{mcu}=@code{avrtiny},
18453 @code{avrxmega2}, @code{avrxmega3}, @code{avrxmega4},
18454 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
18455 If @var{mcu} specifies a device, this built-in macro is set
18456 accordingly. For example, with @option{-mmcu=atmega8} the macro is
18457 defined to @code{4}.
18459 @item __AVR_@var{Device}__
18460 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
18461 the device's name. For example, @option{-mmcu=atmega8} defines the
18462 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
18463 @code{__AVR_ATtiny261A__}, etc.
18465 The built-in macros' names follow
18466 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
18467 the device name as from the AVR user manual. The difference between
18468 @var{Device} in the built-in macro and @var{device} in
18469 @option{-mmcu=@var{device}} is that the latter is always lowercase.
18471 If @var{device} is not a device but only a core architecture like
18472 @samp{avr51}, this macro is not defined.
18474 @item __AVR_DEVICE_NAME__
18475 Setting @option{-mmcu=@var{device}} defines this built-in macro to
18476 the device's name. For example, with @option{-mmcu=atmega8} the macro
18477 is defined to @code{atmega8}.
18479 If @var{device} is not a device but only a core architecture like
18480 @samp{avr51}, this macro is not defined.
18482 @item __AVR_XMEGA__
18483 The device / architecture belongs to the XMEGA family of devices.
18485 @item __AVR_HAVE_ELPM__
18486 The device has the @code{ELPM} instruction.
18488 @item __AVR_HAVE_ELPMX__
18489 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
18490 R@var{n},Z+} instructions.
18492 @item __AVR_HAVE_MOVW__
18493 The device has the @code{MOVW} instruction to perform 16-bit
18494 register-register moves.
18496 @item __AVR_HAVE_LPMX__
18497 The device has the @code{LPM R@var{n},Z} and
18498 @code{LPM R@var{n},Z+} instructions.
18500 @item __AVR_HAVE_MUL__
18501 The device has a hardware multiplier.
18503 @item __AVR_HAVE_JMP_CALL__
18504 The device has the @code{JMP} and @code{CALL} instructions.
18505 This is the case for devices with more than 8@tie{}KiB of program
18508 @item __AVR_HAVE_EIJMP_EICALL__
18509 @itemx __AVR_3_BYTE_PC__
18510 The device has the @code{EIJMP} and @code{EICALL} instructions.
18511 This is the case for devices with more than 128@tie{}KiB of program memory.
18512 This also means that the program counter
18513 (PC) is 3@tie{}bytes wide.
18515 @item __AVR_2_BYTE_PC__
18516 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
18517 with up to 128@tie{}KiB of program memory.
18519 @item __AVR_HAVE_8BIT_SP__
18520 @itemx __AVR_HAVE_16BIT_SP__
18521 The stack pointer (SP) register is treated as 8-bit respectively
18522 16-bit register by the compiler.
18523 The definition of these macros is affected by @option{-mtiny-stack}.
18525 @item __AVR_HAVE_SPH__
18527 The device has the SPH (high part of stack pointer) special function
18528 register or has an 8-bit stack pointer, respectively.
18529 The definition of these macros is affected by @option{-mmcu=} and
18530 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
18533 @item __AVR_HAVE_RAMPD__
18534 @itemx __AVR_HAVE_RAMPX__
18535 @itemx __AVR_HAVE_RAMPY__
18536 @itemx __AVR_HAVE_RAMPZ__
18537 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
18538 @code{RAMPZ} special function register, respectively.
18540 @item __NO_INTERRUPTS__
18541 This macro reflects the @option{-mno-interrupts} command-line option.
18543 @item __AVR_ERRATA_SKIP__
18544 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
18545 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
18546 instructions because of a hardware erratum. Skip instructions are
18547 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
18548 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
18551 @item __AVR_ISA_RMW__
18552 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
18554 @item __AVR_SFR_OFFSET__=@var{offset}
18555 Instructions that can address I/O special function registers directly
18556 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
18557 address as if addressed by an instruction to access RAM like @code{LD}
18558 or @code{STS}. This offset depends on the device architecture and has
18559 to be subtracted from the RAM address in order to get the
18560 respective I/O@tie{}address.
18562 @item __AVR_SHORT_CALLS__
18563 The @option{-mshort-calls} command line option is set.
18565 @item __AVR_PM_BASE_ADDRESS__=@var{addr}
18566 Some devices support reading from flash memory by means of @code{LD*}
18567 instructions. The flash memory is seen in the data address space
18568 at an offset of @code{__AVR_PM_BASE_ADDRESS__}. If this macro
18569 is not defined, this feature is not available. If defined,
18570 the address space is linear and there is no need to put
18571 @code{.rodata} into RAM. This is handled by the default linker
18572 description file, and is currently available for
18573 @code{avrtiny} and @code{avrxmega3}. Even more convenient,
18574 there is no need to use address spaces like @code{__flash} or
18575 features like attribute @code{progmem} and @code{pgm_read_*}.
18577 @item __WITH_AVRLIBC__
18578 The compiler is configured to be used together with AVR-Libc.
18579 See the @option{--with-avrlibc} configure option.
18583 @node Blackfin Options
18584 @subsection Blackfin Options
18585 @cindex Blackfin Options
18588 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
18590 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
18591 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
18592 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
18593 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
18594 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
18595 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
18596 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
18597 @samp{bf561}, @samp{bf592}.
18599 The optional @var{sirevision} specifies the silicon revision of the target
18600 Blackfin processor. Any workarounds available for the targeted silicon revision
18601 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
18602 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
18603 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
18604 hexadecimal digits representing the major and minor numbers in the silicon
18605 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
18606 is not defined. If @var{sirevision} is @samp{any}, the
18607 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
18608 If this optional @var{sirevision} is not used, GCC assumes the latest known
18609 silicon revision of the targeted Blackfin processor.
18611 GCC defines a preprocessor macro for the specified @var{cpu}.
18612 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
18613 provided by libgloss to be linked in if @option{-msim} is not given.
18615 Without this option, @samp{bf532} is used as the processor by default.
18617 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
18618 only the preprocessor macro is defined.
18622 Specifies that the program will be run on the simulator. This causes
18623 the simulator BSP provided by libgloss to be linked in. This option
18624 has effect only for @samp{bfin-elf} toolchain.
18625 Certain other options, such as @option{-mid-shared-library} and
18626 @option{-mfdpic}, imply @option{-msim}.
18628 @item -momit-leaf-frame-pointer
18629 @opindex momit-leaf-frame-pointer
18630 Don't keep the frame pointer in a register for leaf functions. This
18631 avoids the instructions to save, set up and restore frame pointers and
18632 makes an extra register available in leaf functions.
18634 @item -mspecld-anomaly
18635 @opindex mspecld-anomaly
18636 When enabled, the compiler ensures that the generated code does not
18637 contain speculative loads after jump instructions. If this option is used,
18638 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
18640 @item -mno-specld-anomaly
18641 @opindex mno-specld-anomaly
18642 @opindex mspecld-anomaly
18643 Don't generate extra code to prevent speculative loads from occurring.
18645 @item -mcsync-anomaly
18646 @opindex mcsync-anomaly
18647 When enabled, the compiler ensures that the generated code does not
18648 contain CSYNC or SSYNC instructions too soon after conditional branches.
18649 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
18651 @item -mno-csync-anomaly
18652 @opindex mno-csync-anomaly
18653 @opindex mcsync-anomaly
18654 Don't generate extra code to prevent CSYNC or SSYNC instructions from
18655 occurring too soon after a conditional branch.
18659 When enabled, the compiler is free to take advantage of the knowledge that
18660 the entire program fits into the low 64k of memory.
18663 @opindex mno-low64k
18664 Assume that the program is arbitrarily large. This is the default.
18666 @item -mstack-check-l1
18667 @opindex mstack-check-l1
18668 Do stack checking using information placed into L1 scratchpad memory by the
18671 @item -mid-shared-library
18672 @opindex mid-shared-library
18673 Generate code that supports shared libraries via the library ID method.
18674 This allows for execute in place and shared libraries in an environment
18675 without virtual memory management. This option implies @option{-fPIC}.
18676 With a @samp{bfin-elf} target, this option implies @option{-msim}.
18678 @item -mno-id-shared-library
18679 @opindex mno-id-shared-library
18680 @opindex mid-shared-library
18681 Generate code that doesn't assume ID-based shared libraries are being used.
18682 This is the default.
18684 @item -mleaf-id-shared-library
18685 @opindex mleaf-id-shared-library
18686 Generate code that supports shared libraries via the library ID method,
18687 but assumes that this library or executable won't link against any other
18688 ID shared libraries. That allows the compiler to use faster code for jumps
18691 @item -mno-leaf-id-shared-library
18692 @opindex mno-leaf-id-shared-library
18693 @opindex mleaf-id-shared-library
18694 Do not assume that the code being compiled won't link against any ID shared
18695 libraries. Slower code is generated for jump and call insns.
18697 @item -mshared-library-id=n
18698 @opindex mshared-library-id
18699 Specifies the identification number of the ID-based shared library being
18700 compiled. Specifying a value of 0 generates more compact code; specifying
18701 other values forces the allocation of that number to the current
18702 library but is no more space- or time-efficient than omitting this option.
18706 Generate code that allows the data segment to be located in a different
18707 area of memory from the text segment. This allows for execute in place in
18708 an environment without virtual memory management by eliminating relocations
18709 against the text section.
18711 @item -mno-sep-data
18712 @opindex mno-sep-data
18714 Generate code that assumes that the data segment follows the text segment.
18715 This is the default.
18718 @itemx -mno-long-calls
18719 @opindex mlong-calls
18720 @opindex mno-long-calls
18721 Tells the compiler to perform function calls by first loading the
18722 address of the function into a register and then performing a subroutine
18723 call on this register. This switch is needed if the target function
18724 lies outside of the 24-bit addressing range of the offset-based
18725 version of subroutine call instruction.
18727 This feature is not enabled by default. Specifying
18728 @option{-mno-long-calls} restores the default behavior. Note these
18729 switches have no effect on how the compiler generates code to handle
18730 function calls via function pointers.
18734 Link with the fast floating-point library. This library relaxes some of
18735 the IEEE floating-point standard's rules for checking inputs against
18736 Not-a-Number (NAN), in the interest of performance.
18739 @opindex minline-plt
18740 Enable inlining of PLT entries in function calls to functions that are
18741 not known to bind locally. It has no effect without @option{-mfdpic}.
18744 @opindex mmulticore
18745 Build a standalone application for multicore Blackfin processors.
18746 This option causes proper start files and link scripts supporting
18747 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
18748 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
18750 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
18751 selects the one-application-per-core programming model. Without
18752 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
18753 programming model is used. In this model, the main function of Core B
18754 should be named as @code{coreb_main}.
18756 If this option is not used, the single-core application programming
18761 Build a standalone application for Core A of BF561 when using
18762 the one-application-per-core programming model. Proper start files
18763 and link scripts are used to support Core A, and the macro
18764 @code{__BFIN_COREA} is defined.
18765 This option can only be used in conjunction with @option{-mmulticore}.
18769 Build a standalone application for Core B of BF561 when using
18770 the one-application-per-core programming model. Proper start files
18771 and link scripts are used to support Core B, and the macro
18772 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
18773 should be used instead of @code{main}.
18774 This option can only be used in conjunction with @option{-mmulticore}.
18778 Build a standalone application for SDRAM. Proper start files and
18779 link scripts are used to put the application into SDRAM, and the macro
18780 @code{__BFIN_SDRAM} is defined.
18781 The loader should initialize SDRAM before loading the application.
18785 Assume that ICPLBs are enabled at run time. This has an effect on certain
18786 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
18787 are enabled; for standalone applications the default is off.
18791 @subsection C6X Options
18792 @cindex C6X Options
18795 @item -march=@var{name}
18797 This specifies the name of the target architecture. GCC uses this
18798 name to determine what kind of instructions it can emit when generating
18799 assembly code. Permissible names are: @samp{c62x},
18800 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
18803 @opindex mbig-endian
18804 Generate code for a big-endian target.
18806 @item -mlittle-endian
18807 @opindex mlittle-endian
18808 Generate code for a little-endian target. This is the default.
18812 Choose startup files and linker script suitable for the simulator.
18814 @item -msdata=default
18815 @opindex msdata=default
18816 Put small global and static data in the @code{.neardata} section,
18817 which is pointed to by register @code{B14}. Put small uninitialized
18818 global and static data in the @code{.bss} section, which is adjacent
18819 to the @code{.neardata} section. Put small read-only data into the
18820 @code{.rodata} section. The corresponding sections used for large
18821 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
18824 @opindex msdata=all
18825 Put all data, not just small objects, into the sections reserved for
18826 small data, and use addressing relative to the @code{B14} register to
18830 @opindex msdata=none
18831 Make no use of the sections reserved for small data, and use absolute
18832 addresses to access all data. Put all initialized global and static
18833 data in the @code{.fardata} section, and all uninitialized data in the
18834 @code{.far} section. Put all constant data into the @code{.const}
18839 @subsection CRIS Options
18840 @cindex CRIS Options
18842 These options are defined specifically for the CRIS ports.
18845 @item -march=@var{architecture-type}
18846 @itemx -mcpu=@var{architecture-type}
18849 Generate code for the specified architecture. The choices for
18850 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
18851 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
18852 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
18855 @item -mtune=@var{architecture-type}
18857 Tune to @var{architecture-type} everything applicable about the generated
18858 code, except for the ABI and the set of available instructions. The
18859 choices for @var{architecture-type} are the same as for
18860 @option{-march=@var{architecture-type}}.
18862 @item -mmax-stack-frame=@var{n}
18863 @opindex mmax-stack-frame
18864 Warn when the stack frame of a function exceeds @var{n} bytes.
18870 The options @option{-metrax4} and @option{-metrax100} are synonyms for
18871 @option{-march=v3} and @option{-march=v8} respectively.
18873 @item -mmul-bug-workaround
18874 @itemx -mno-mul-bug-workaround
18875 @opindex mmul-bug-workaround
18876 @opindex mno-mul-bug-workaround
18877 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
18878 models where it applies. This option is active by default.
18882 Enable CRIS-specific verbose debug-related information in the assembly
18883 code. This option also has the effect of turning off the @samp{#NO_APP}
18884 formatted-code indicator to the assembler at the beginning of the
18889 Do not use condition-code results from previous instruction; always emit
18890 compare and test instructions before use of condition codes.
18892 @item -mno-side-effects
18893 @opindex mno-side-effects
18894 @opindex mside-effects
18895 Do not emit instructions with side effects in addressing modes other than
18898 @item -mstack-align
18899 @itemx -mno-stack-align
18900 @itemx -mdata-align
18901 @itemx -mno-data-align
18902 @itemx -mconst-align
18903 @itemx -mno-const-align
18904 @opindex mstack-align
18905 @opindex mno-stack-align
18906 @opindex mdata-align
18907 @opindex mno-data-align
18908 @opindex mconst-align
18909 @opindex mno-const-align
18910 These options (@samp{no-} options) arrange (eliminate arrangements) for the
18911 stack frame, individual data and constants to be aligned for the maximum
18912 single data access size for the chosen CPU model. The default is to
18913 arrange for 32-bit alignment. ABI details such as structure layout are
18914 not affected by these options.
18922 Similar to the stack- data- and const-align options above, these options
18923 arrange for stack frame, writable data and constants to all be 32-bit,
18924 16-bit or 8-bit aligned. The default is 32-bit alignment.
18926 @item -mno-prologue-epilogue
18927 @itemx -mprologue-epilogue
18928 @opindex mno-prologue-epilogue
18929 @opindex mprologue-epilogue
18930 With @option{-mno-prologue-epilogue}, the normal function prologue and
18931 epilogue which set up the stack frame are omitted and no return
18932 instructions or return sequences are generated in the code. Use this
18933 option only together with visual inspection of the compiled code: no
18934 warnings or errors are generated when call-saved registers must be saved,
18935 or storage for local variables needs to be allocated.
18939 @opindex mno-gotplt
18941 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
18942 instruction sequences that load addresses for functions from the PLT part
18943 of the GOT rather than (traditional on other architectures) calls to the
18944 PLT@. The default is @option{-mgotplt}.
18948 Legacy no-op option only recognized with the cris-axis-elf and
18949 cris-axis-linux-gnu targets.
18953 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
18957 This option, recognized for the cris-axis-elf, arranges
18958 to link with input-output functions from a simulator library. Code,
18959 initialized data and zero-initialized data are allocated consecutively.
18963 Like @option{-sim}, but pass linker options to locate initialized data at
18964 0x40000000 and zero-initialized data at 0x80000000.
18968 @subsection CR16 Options
18969 @cindex CR16 Options
18971 These options are defined specifically for the CR16 ports.
18977 Enable the use of multiply-accumulate instructions. Disabled by default.
18981 @opindex mcr16cplus
18983 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
18988 Links the library libsim.a which is in compatible with simulator. Applicable
18989 to ELF compiler only.
18993 Choose integer type as 32-bit wide.
18997 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
18999 @item -mdata-model=@var{model}
19000 @opindex mdata-model
19001 Choose a data model. The choices for @var{model} are @samp{near},
19002 @samp{far} or @samp{medium}. @samp{medium} is default.
19003 However, @samp{far} is not valid with @option{-mcr16c}, as the
19004 CR16C architecture does not support the far data model.
19007 @node C-SKY Options
19008 @subsection C-SKY Options
19009 @cindex C-SKY Options
19011 GCC supports these options when compiling for C-SKY V2 processors.
19015 @item -march=@var{arch}
19017 Specify the C-SKY target architecture. Valid values for @var{arch} are:
19018 @samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
19019 The default is @samp{ck810}.
19021 @item -mcpu=@var{cpu}
19023 Specify the C-SKY target processor. Valid values for @var{cpu} are:
19024 @samp{ck801}, @samp{ck801t},
19025 @samp{ck802}, @samp{ck802t}, @samp{ck802j},
19026 @samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
19027 @samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
19028 @samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
19029 @samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
19030 @samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
19031 @samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
19032 @samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
19033 @samp{ck803eftr1}, @samp{ck803efhtr1},
19034 @samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
19035 @samp{ck803sef}, @samp{ck803seft},
19036 @samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
19037 @samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
19038 @samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
19039 @samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
19042 @opindex mbig-endian
19045 @itemx -mlittle-endian
19046 @opindex mlittle-endian
19050 Select big- or little-endian code. The default is little-endian.
19053 @opindex mhard-float
19054 @itemx -msoft-float
19055 @opindex msoft-float
19057 Select hardware or software floating-point implementations.
19058 The default is soft float.
19060 @item -mdouble-float
19061 @itemx -mno-double-float
19062 @opindex mdouble-float
19063 When @option{-mhard-float} is in effect, enable generation of
19064 double-precision float instructions. This is the default except
19065 when compiling for CK803.
19070 When @option{-mhard-float} is in effect, enable generation of
19071 @code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
19072 This is the default except when compiling for CK803.
19074 @item -mfpu=@var{fpu}
19076 Select the floating-point processor. This option can only be used with
19077 @option{-mhard-float}.
19078 Values for @var{fpu} are
19079 @samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
19080 @samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
19081 @samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
19086 Enable the extended @code{lrw} instruction. This option defaults to on
19087 for CK801 and off otherwise.
19092 Enable interrupt stack instructions; the default is off.
19094 The @option{-mistack} option is required to handle the
19095 @code{interrupt} and @code{isr} function attributes
19096 (@pxref{C-SKY Function Attributes}).
19100 Enable multiprocessor instructions; the default is off.
19104 Enable coprocessor instructions; the default is off.
19108 Enable coprocessor instructions; the default is off.
19112 Enable C-SKY security instructions; the default is off.
19116 Enable C-SKY trust instructions; the default is off.
19124 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
19125 All of these options default to off.
19130 Generate divide instructions. Default is off.
19135 Generate code for Smart Mode, using only registers numbered 0-7 to allow
19136 use of 16-bit instructions. This option is ignored for CK801 where this
19137 is the required behavior, and it defaults to on for CK802.
19138 For other targets, the default is off.
19140 @item -mhigh-registers
19141 @itemx -mno-high-registers
19142 @opindex mhigh-registers
19143 Generate code using the high registers numbered 16-31. This option
19144 is not supported on CK801, CK802, or CK803, and is enabled by default
19145 for other processors.
19150 Generate code using global anchor symbol addresses.
19153 @itemx -mno-pushpop
19155 Generate code using @code{push} and @code{pop} instructions. This option
19158 @item -mmultiple-stld
19160 @itemx -mno-multiple-stld
19162 @opindex mmultiple-stld
19163 Generate code using @code{stm} and @code{ldm} instructions. This option
19164 isn't supported on CK801 but is enabled by default on other processors.
19167 @itemx -mno-constpool
19168 @opindex mconstpool
19169 Create constant pools in the compiler instead of deferring it to the
19170 assembler. This option is the default and required for correct code
19171 generation on CK801 and CK802, and is optional on other processors.
19174 @item -mno-stack-size
19175 @opindex mstack-size
19176 Emit @code{.stack_size} directives for each function in the assembly
19177 output. This option defaults to off.
19182 Generate code for the C-SKY compiler runtime instead of libgcc. This
19183 option defaults to off.
19185 @item -mbranch-cost=@var{n}
19186 @opindex mbranch-cost=
19187 Set the branch costs to roughly @code{n} instructions. The default is 1.
19189 @item -msched-prolog
19190 @itemx -mno-sched-prolog
19191 @opindex msched-prolog
19192 Permit scheduling of function prologue and epilogue sequences. Using
19193 this option can result in code that is not compliant with the C-SKY V2 ABI
19194 prologue requirements and that cannot be debugged or backtraced.
19195 It is disabled by default.
19199 @node Darwin Options
19200 @subsection Darwin Options
19201 @cindex Darwin options
19203 These options are defined for all architectures running the Darwin operating
19206 FSF GCC on Darwin does not create ``fat'' object files; it creates
19207 an object file for the single architecture that GCC was built to
19208 target. Apple's GCC on Darwin does create ``fat'' files if multiple
19209 @option{-arch} options are used; it does so by running the compiler or
19210 linker multiple times and joining the results together with
19213 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
19214 @samp{i686}) is determined by the flags that specify the ISA
19215 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
19216 @option{-force_cpusubtype_ALL} option can be used to override this.
19218 The Darwin tools vary in their behavior when presented with an ISA
19219 mismatch. The assembler, @file{as}, only permits instructions to
19220 be used that are valid for the subtype of the file it is generating,
19221 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
19222 The linker for shared libraries, @file{/usr/bin/libtool}, fails
19223 and prints an error if asked to create a shared library with a less
19224 restrictive subtype than its input files (for instance, trying to put
19225 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
19226 for executables, @command{ld}, quietly gives the executable the most
19227 restrictive subtype of any of its input files.
19232 Add the framework directory @var{dir} to the head of the list of
19233 directories to be searched for header files. These directories are
19234 interleaved with those specified by @option{-I} options and are
19235 scanned in a left-to-right order.
19237 A framework directory is a directory with frameworks in it. A
19238 framework is a directory with a @file{Headers} and/or
19239 @file{PrivateHeaders} directory contained directly in it that ends
19240 in @file{.framework}. The name of a framework is the name of this
19241 directory excluding the @file{.framework}. Headers associated with
19242 the framework are found in one of those two directories, with
19243 @file{Headers} being searched first. A subframework is a framework
19244 directory that is in a framework's @file{Frameworks} directory.
19245 Includes of subframework headers can only appear in a header of a
19246 framework that contains the subframework, or in a sibling subframework
19247 header. Two subframeworks are siblings if they occur in the same
19248 framework. A subframework should not have the same name as a
19249 framework; a warning is issued if this is violated. Currently a
19250 subframework cannot have subframeworks; in the future, the mechanism
19251 may be extended to support this. The standard frameworks can be found
19252 in @file{/System/Library/Frameworks} and
19253 @file{/Library/Frameworks}. An example include looks like
19254 @code{#include <Framework/header.h>}, where @file{Framework} denotes
19255 the name of the framework and @file{header.h} is found in the
19256 @file{PrivateHeaders} or @file{Headers} directory.
19258 @item -iframework@var{dir}
19259 @opindex iframework
19260 Like @option{-F} except the directory is a treated as a system
19261 directory. The main difference between this @option{-iframework} and
19262 @option{-F} is that with @option{-iframework} the compiler does not
19263 warn about constructs contained within header files found via
19264 @var{dir}. This option is valid only for the C family of languages.
19268 Emit debugging information for symbols that are used. For stabs
19269 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
19270 This is by default ON@.
19274 Emit debugging information for all symbols and types.
19276 @item -mmacosx-version-min=@var{version}
19277 The earliest version of MacOS X that this executable will run on
19278 is @var{version}. Typical values of @var{version} include @code{10.1},
19279 @code{10.2}, and @code{10.3.9}.
19281 If the compiler was built to use the system's headers by default,
19282 then the default for this option is the system version on which the
19283 compiler is running, otherwise the default is to make choices that
19284 are compatible with as many systems and code bases as possible.
19288 Enable kernel development mode. The @option{-mkernel} option sets
19289 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
19290 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
19291 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
19292 applicable. This mode also sets @option{-mno-altivec},
19293 @option{-msoft-float}, @option{-fno-builtin} and
19294 @option{-mlong-branch} for PowerPC targets.
19296 @item -mone-byte-bool
19297 @opindex mone-byte-bool
19298 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
19299 By default @code{sizeof(bool)} is @code{4} when compiling for
19300 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
19301 option has no effect on x86.
19303 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
19304 to generate code that is not binary compatible with code generated
19305 without that switch. Using this switch may require recompiling all
19306 other modules in a program, including system libraries. Use this
19307 switch to conform to a non-default data model.
19309 @item -mfix-and-continue
19310 @itemx -ffix-and-continue
19311 @itemx -findirect-data
19312 @opindex mfix-and-continue
19313 @opindex ffix-and-continue
19314 @opindex findirect-data
19315 Generate code suitable for fast turnaround development, such as to
19316 allow GDB to dynamically load @file{.o} files into already-running
19317 programs. @option{-findirect-data} and @option{-ffix-and-continue}
19318 are provided for backwards compatibility.
19322 Loads all members of static archive libraries.
19323 See man ld(1) for more information.
19325 @item -arch_errors_fatal
19326 @opindex arch_errors_fatal
19327 Cause the errors having to do with files that have the wrong architecture
19330 @item -bind_at_load
19331 @opindex bind_at_load
19332 Causes the output file to be marked such that the dynamic linker will
19333 bind all undefined references when the file is loaded or launched.
19337 Produce a Mach-o bundle format file.
19338 See man ld(1) for more information.
19340 @item -bundle_loader @var{executable}
19341 @opindex bundle_loader
19342 This option specifies the @var{executable} that will load the build
19343 output file being linked. See man ld(1) for more information.
19346 @opindex dynamiclib
19347 When passed this option, GCC produces a dynamic library instead of
19348 an executable when linking, using the Darwin @file{libtool} command.
19350 @item -force_cpusubtype_ALL
19351 @opindex force_cpusubtype_ALL
19352 This causes GCC's output file to have the @samp{ALL} subtype, instead of
19353 one controlled by the @option{-mcpu} or @option{-march} option.
19355 @item -allowable_client @var{client_name}
19356 @itemx -client_name
19357 @itemx -compatibility_version
19358 @itemx -current_version
19360 @itemx -dependency-file
19362 @itemx -dylinker_install_name
19364 @itemx -exported_symbols_list
19367 @itemx -flat_namespace
19368 @itemx -force_flat_namespace
19369 @itemx -headerpad_max_install_names
19372 @itemx -install_name
19373 @itemx -keep_private_externs
19374 @itemx -multi_module
19375 @itemx -multiply_defined
19376 @itemx -multiply_defined_unused
19379 @itemx -no_dead_strip_inits_and_terms
19380 @itemx -nofixprebinding
19381 @itemx -nomultidefs
19383 @itemx -noseglinkedit
19384 @itemx -pagezero_size
19386 @itemx -prebind_all_twolevel_modules
19387 @itemx -private_bundle
19389 @itemx -read_only_relocs
19391 @itemx -sectobjectsymbols
19395 @itemx -sectobjectsymbols
19398 @itemx -segs_read_only_addr
19400 @itemx -segs_read_write_addr
19401 @itemx -seg_addr_table
19402 @itemx -seg_addr_table_filename
19403 @itemx -seglinkedit
19405 @itemx -segs_read_only_addr
19406 @itemx -segs_read_write_addr
19407 @itemx -single_module
19409 @itemx -sub_library
19411 @itemx -sub_umbrella
19412 @itemx -twolevel_namespace
19415 @itemx -unexported_symbols_list
19416 @itemx -weak_reference_mismatches
19417 @itemx -whatsloaded
19418 @opindex allowable_client
19419 @opindex client_name
19420 @opindex compatibility_version
19421 @opindex current_version
19422 @opindex dead_strip
19423 @opindex dependency-file
19424 @opindex dylib_file
19425 @opindex dylinker_install_name
19427 @opindex exported_symbols_list
19429 @opindex flat_namespace
19430 @opindex force_flat_namespace
19431 @opindex headerpad_max_install_names
19432 @opindex image_base
19434 @opindex install_name
19435 @opindex keep_private_externs
19436 @opindex multi_module
19437 @opindex multiply_defined
19438 @opindex multiply_defined_unused
19439 @opindex noall_load
19440 @opindex no_dead_strip_inits_and_terms
19441 @opindex nofixprebinding
19442 @opindex nomultidefs
19444 @opindex noseglinkedit
19445 @opindex pagezero_size
19447 @opindex prebind_all_twolevel_modules
19448 @opindex private_bundle
19449 @opindex read_only_relocs
19451 @opindex sectobjectsymbols
19454 @opindex sectcreate
19455 @opindex sectobjectsymbols
19458 @opindex segs_read_only_addr
19459 @opindex segs_read_write_addr
19460 @opindex seg_addr_table
19461 @opindex seg_addr_table_filename
19462 @opindex seglinkedit
19464 @opindex segs_read_only_addr
19465 @opindex segs_read_write_addr
19466 @opindex single_module
19468 @opindex sub_library
19469 @opindex sub_umbrella
19470 @opindex twolevel_namespace
19473 @opindex unexported_symbols_list
19474 @opindex weak_reference_mismatches
19475 @opindex whatsloaded
19476 These options are passed to the Darwin linker. The Darwin linker man page
19477 describes them in detail.
19480 @node DEC Alpha Options
19481 @subsection DEC Alpha Options
19483 These @samp{-m} options are defined for the DEC Alpha implementations:
19486 @item -mno-soft-float
19487 @itemx -msoft-float
19488 @opindex mno-soft-float
19489 @opindex msoft-float
19490 Use (do not use) the hardware floating-point instructions for
19491 floating-point operations. When @option{-msoft-float} is specified,
19492 functions in @file{libgcc.a} are used to perform floating-point
19493 operations. Unless they are replaced by routines that emulate the
19494 floating-point operations, or compiled in such a way as to call such
19495 emulations routines, these routines issue floating-point
19496 operations. If you are compiling for an Alpha without floating-point
19497 operations, you must ensure that the library is built so as not to call
19500 Note that Alpha implementations without floating-point operations are
19501 required to have floating-point registers.
19504 @itemx -mno-fp-regs
19506 @opindex mno-fp-regs
19507 Generate code that uses (does not use) the floating-point register set.
19508 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
19509 register set is not used, floating-point operands are passed in integer
19510 registers as if they were integers and floating-point results are passed
19511 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
19512 so any function with a floating-point argument or return value called by code
19513 compiled with @option{-mno-fp-regs} must also be compiled with that
19516 A typical use of this option is building a kernel that does not use,
19517 and hence need not save and restore, any floating-point registers.
19521 The Alpha architecture implements floating-point hardware optimized for
19522 maximum performance. It is mostly compliant with the IEEE floating-point
19523 standard. However, for full compliance, software assistance is
19524 required. This option generates code fully IEEE-compliant code
19525 @emph{except} that the @var{inexact-flag} is not maintained (see below).
19526 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
19527 defined during compilation. The resulting code is less efficient but is
19528 able to correctly support denormalized numbers and exceptional IEEE
19529 values such as not-a-number and plus/minus infinity. Other Alpha
19530 compilers call this option @option{-ieee_with_no_inexact}.
19532 @item -mieee-with-inexact
19533 @opindex mieee-with-inexact
19534 This is like @option{-mieee} except the generated code also maintains
19535 the IEEE @var{inexact-flag}. Turning on this option causes the
19536 generated code to implement fully-compliant IEEE math. In addition to
19537 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
19538 macro. On some Alpha implementations the resulting code may execute
19539 significantly slower than the code generated by default. Since there is
19540 very little code that depends on the @var{inexact-flag}, you should
19541 normally not specify this option. Other Alpha compilers call this
19542 option @option{-ieee_with_inexact}.
19544 @item -mfp-trap-mode=@var{trap-mode}
19545 @opindex mfp-trap-mode
19546 This option controls what floating-point related traps are enabled.
19547 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
19548 The trap mode can be set to one of four values:
19552 This is the default (normal) setting. The only traps that are enabled
19553 are the ones that cannot be disabled in software (e.g., division by zero
19557 In addition to the traps enabled by @samp{n}, underflow traps are enabled
19561 Like @samp{u}, but the instructions are marked to be safe for software
19562 completion (see Alpha architecture manual for details).
19565 Like @samp{su}, but inexact traps are enabled as well.
19568 @item -mfp-rounding-mode=@var{rounding-mode}
19569 @opindex mfp-rounding-mode
19570 Selects the IEEE rounding mode. Other Alpha compilers call this option
19571 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
19576 Normal IEEE rounding mode. Floating-point numbers are rounded towards
19577 the nearest machine number or towards the even machine number in case
19581 Round towards minus infinity.
19584 Chopped rounding mode. Floating-point numbers are rounded towards zero.
19587 Dynamic rounding mode. A field in the floating-point control register
19588 (@var{fpcr}, see Alpha architecture reference manual) controls the
19589 rounding mode in effect. The C library initializes this register for
19590 rounding towards plus infinity. Thus, unless your program modifies the
19591 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
19594 @item -mtrap-precision=@var{trap-precision}
19595 @opindex mtrap-precision
19596 In the Alpha architecture, floating-point traps are imprecise. This
19597 means without software assistance it is impossible to recover from a
19598 floating trap and program execution normally needs to be terminated.
19599 GCC can generate code that can assist operating system trap handlers
19600 in determining the exact location that caused a floating-point trap.
19601 Depending on the requirements of an application, different levels of
19602 precisions can be selected:
19606 Program precision. This option is the default and means a trap handler
19607 can only identify which program caused a floating-point exception.
19610 Function precision. The trap handler can determine the function that
19611 caused a floating-point exception.
19614 Instruction precision. The trap handler can determine the exact
19615 instruction that caused a floating-point exception.
19618 Other Alpha compilers provide the equivalent options called
19619 @option{-scope_safe} and @option{-resumption_safe}.
19621 @item -mieee-conformant
19622 @opindex mieee-conformant
19623 This option marks the generated code as IEEE conformant. You must not
19624 use this option unless you also specify @option{-mtrap-precision=i} and either
19625 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
19626 is to emit the line @samp{.eflag 48} in the function prologue of the
19627 generated assembly file.
19629 @item -mbuild-constants
19630 @opindex mbuild-constants
19631 Normally GCC examines a 32- or 64-bit integer constant to
19632 see if it can construct it from smaller constants in two or three
19633 instructions. If it cannot, it outputs the constant as a literal and
19634 generates code to load it from the data segment at run time.
19636 Use this option to require GCC to construct @emph{all} integer constants
19637 using code, even if it takes more instructions (the maximum is six).
19639 You typically use this option to build a shared library dynamic
19640 loader. Itself a shared library, it must relocate itself in memory
19641 before it can find the variables and constants in its own data segment.
19659 Indicate whether GCC should generate code to use the optional BWX,
19660 CIX, FIX and MAX instruction sets. The default is to use the instruction
19661 sets supported by the CPU type specified via @option{-mcpu=} option or that
19662 of the CPU on which GCC was built if none is specified.
19665 @itemx -mfloat-ieee
19666 @opindex mfloat-vax
19667 @opindex mfloat-ieee
19668 Generate code that uses (does not use) VAX F and G floating-point
19669 arithmetic instead of IEEE single and double precision.
19671 @item -mexplicit-relocs
19672 @itemx -mno-explicit-relocs
19673 @opindex mexplicit-relocs
19674 @opindex mno-explicit-relocs
19675 Older Alpha assemblers provided no way to generate symbol relocations
19676 except via assembler macros. Use of these macros does not allow
19677 optimal instruction scheduling. GNU binutils as of version 2.12
19678 supports a new syntax that allows the compiler to explicitly mark
19679 which relocations should apply to which instructions. This option
19680 is mostly useful for debugging, as GCC detects the capabilities of
19681 the assembler when it is built and sets the default accordingly.
19684 @itemx -mlarge-data
19685 @opindex msmall-data
19686 @opindex mlarge-data
19687 When @option{-mexplicit-relocs} is in effect, static data is
19688 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
19689 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
19690 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
19691 16-bit relocations off of the @code{$gp} register. This limits the
19692 size of the small data area to 64KB, but allows the variables to be
19693 directly accessed via a single instruction.
19695 The default is @option{-mlarge-data}. With this option the data area
19696 is limited to just below 2GB@. Programs that require more than 2GB of
19697 data must use @code{malloc} or @code{mmap} to allocate the data in the
19698 heap instead of in the program's data segment.
19700 When generating code for shared libraries, @option{-fpic} implies
19701 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
19704 @itemx -mlarge-text
19705 @opindex msmall-text
19706 @opindex mlarge-text
19707 When @option{-msmall-text} is used, the compiler assumes that the
19708 code of the entire program (or shared library) fits in 4MB, and is
19709 thus reachable with a branch instruction. When @option{-msmall-data}
19710 is used, the compiler can assume that all local symbols share the
19711 same @code{$gp} value, and thus reduce the number of instructions
19712 required for a function call from 4 to 1.
19714 The default is @option{-mlarge-text}.
19716 @item -mcpu=@var{cpu_type}
19718 Set the instruction set and instruction scheduling parameters for
19719 machine type @var{cpu_type}. You can specify either the @samp{EV}
19720 style name or the corresponding chip number. GCC supports scheduling
19721 parameters for the EV4, EV5 and EV6 family of processors and
19722 chooses the default values for the instruction set from the processor
19723 you specify. If you do not specify a processor type, GCC defaults
19724 to the processor on which the compiler was built.
19726 Supported values for @var{cpu_type} are
19732 Schedules as an EV4 and has no instruction set extensions.
19736 Schedules as an EV5 and has no instruction set extensions.
19740 Schedules as an EV5 and supports the BWX extension.
19745 Schedules as an EV5 and supports the BWX and MAX extensions.
19749 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
19753 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
19756 Native toolchains also support the value @samp{native},
19757 which selects the best architecture option for the host processor.
19758 @option{-mcpu=native} has no effect if GCC does not recognize
19761 @item -mtune=@var{cpu_type}
19763 Set only the instruction scheduling parameters for machine type
19764 @var{cpu_type}. The instruction set is not changed.
19766 Native toolchains also support the value @samp{native},
19767 which selects the best architecture option for the host processor.
19768 @option{-mtune=native} has no effect if GCC does not recognize
19771 @item -mmemory-latency=@var{time}
19772 @opindex mmemory-latency
19773 Sets the latency the scheduler should assume for typical memory
19774 references as seen by the application. This number is highly
19775 dependent on the memory access patterns used by the application
19776 and the size of the external cache on the machine.
19778 Valid options for @var{time} are
19782 A decimal number representing clock cycles.
19788 The compiler contains estimates of the number of clock cycles for
19789 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
19790 (also called Dcache, Scache, and Bcache), as well as to main memory.
19791 Note that L3 is only valid for EV5.
19797 @subsection eBPF Options
19798 @cindex eBPF Options
19801 @item -mframe-limit=@var{bytes}
19802 This specifies the hard limit for frame sizes, in bytes. Currently,
19803 the value that can be specified should be less than or equal to
19804 @samp{32767}. Defaults to whatever limit is imposed by the version of
19805 the Linux kernel targeted.
19807 @item -mkernel=@var{version}
19809 This specifies the minimum version of the kernel that will run the
19810 compiled program. GCC uses this version to determine which
19811 instructions to use, what kernel helpers to allow, etc. Currently,
19812 @var{version} can be one of @samp{4.0}, @samp{4.1}, @samp{4.2},
19813 @samp{4.3}, @samp{4.4}, @samp{4.5}, @samp{4.6}, @samp{4.7},
19814 @samp{4.8}, @samp{4.9}, @samp{4.10}, @samp{4.11}, @samp{4.12},
19815 @samp{4.13}, @samp{4.14}, @samp{4.15}, @samp{4.16}, @samp{4.17},
19816 @samp{4.18}, @samp{4.19}, @samp{4.20}, @samp{5.0}, @samp{5.1},
19817 @samp{5.2}, @samp{latest} and @samp{native}.
19820 @opindex mbig-endian
19821 Generate code for a big-endian target.
19823 @item -mlittle-endian
19824 @opindex mlittle-endian
19825 Generate code for a little-endian target. This is the default.
19829 @subsection FR30 Options
19830 @cindex FR30 Options
19832 These options are defined specifically for the FR30 port.
19836 @item -msmall-model
19837 @opindex msmall-model
19838 Use the small address space model. This can produce smaller code, but
19839 it does assume that all symbolic values and addresses fit into a
19844 Assume that runtime support has been provided and so there is no need
19845 to include the simulator library (@file{libsim.a}) on the linker
19851 @subsection FT32 Options
19852 @cindex FT32 Options
19854 These options are defined specifically for the FT32 port.
19860 Specifies that the program will be run on the simulator. This causes
19861 an alternate runtime startup and library to be linked.
19862 You must not use this option when generating programs that will run on
19863 real hardware; you must provide your own runtime library for whatever
19864 I/O functions are needed.
19868 Enable Local Register Allocation. This is still experimental for FT32,
19869 so by default the compiler uses standard reload.
19873 Do not use div and mod instructions.
19877 Enable use of the extended instructions of the FT32B processor.
19881 Compress all code using the Ft32B code compression scheme.
19885 Do not generate code that reads program memory.
19890 @subsection FRV Options
19891 @cindex FRV Options
19897 Only use the first 32 general-purpose registers.
19902 Use all 64 general-purpose registers.
19907 Use only the first 32 floating-point registers.
19912 Use all 64 floating-point registers.
19915 @opindex mhard-float
19917 Use hardware instructions for floating-point operations.
19920 @opindex msoft-float
19922 Use library routines for floating-point operations.
19927 Dynamically allocate condition code registers.
19932 Do not try to dynamically allocate condition code registers, only
19933 use @code{icc0} and @code{fcc0}.
19938 Change ABI to use double word insns.
19944 Do not use double word instructions.
19949 Use floating-point double instructions.
19952 @opindex mno-double
19954 Do not use floating-point double instructions.
19959 Use media instructions.
19964 Do not use media instructions.
19969 Use multiply and add/subtract instructions.
19972 @opindex mno-muladd
19974 Do not use multiply and add/subtract instructions.
19979 Select the FDPIC ABI, which uses function descriptors to represent
19980 pointers to functions. Without any PIC/PIE-related options, it
19981 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
19982 assumes GOT entries and small data are within a 12-bit range from the
19983 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
19984 are computed with 32 bits.
19985 With a @samp{bfin-elf} target, this option implies @option{-msim}.
19988 @opindex minline-plt
19990 Enable inlining of PLT entries in function calls to functions that are
19991 not known to bind locally. It has no effect without @option{-mfdpic}.
19992 It's enabled by default if optimizing for speed and compiling for
19993 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
19994 optimization option such as @option{-O3} or above is present in the
20000 Assume a large TLS segment when generating thread-local code.
20005 Do not assume a large TLS segment when generating thread-local code.
20010 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
20011 that is known to be in read-only sections. It's enabled by default,
20012 except for @option{-fpic} or @option{-fpie}: even though it may help
20013 make the global offset table smaller, it trades 1 instruction for 4.
20014 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
20015 one of which may be shared by multiple symbols, and it avoids the need
20016 for a GOT entry for the referenced symbol, so it's more likely to be a
20017 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
20019 @item -multilib-library-pic
20020 @opindex multilib-library-pic
20022 Link with the (library, not FD) pic libraries. It's implied by
20023 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
20024 @option{-fpic} without @option{-mfdpic}. You should never have to use
20028 @opindex mlinked-fp
20030 Follow the EABI requirement of always creating a frame pointer whenever
20031 a stack frame is allocated. This option is enabled by default and can
20032 be disabled with @option{-mno-linked-fp}.
20035 @opindex mlong-calls
20037 Use indirect addressing to call functions outside the current
20038 compilation unit. This allows the functions to be placed anywhere
20039 within the 32-bit address space.
20041 @item -malign-labels
20042 @opindex malign-labels
20044 Try to align labels to an 8-byte boundary by inserting NOPs into the
20045 previous packet. This option only has an effect when VLIW packing
20046 is enabled. It doesn't create new packets; it merely adds NOPs to
20049 @item -mlibrary-pic
20050 @opindex mlibrary-pic
20052 Generate position-independent EABI code.
20057 Use only the first four media accumulator registers.
20062 Use all eight media accumulator registers.
20067 Pack VLIW instructions.
20072 Do not pack VLIW instructions.
20075 @opindex mno-eflags
20077 Do not mark ABI switches in e_flags.
20080 @opindex mcond-move
20082 Enable the use of conditional-move instructions (default).
20084 This switch is mainly for debugging the compiler and will likely be removed
20085 in a future version.
20087 @item -mno-cond-move
20088 @opindex mno-cond-move
20090 Disable the use of conditional-move instructions.
20092 This switch is mainly for debugging the compiler and will likely be removed
20093 in a future version.
20098 Enable the use of conditional set instructions (default).
20100 This switch is mainly for debugging the compiler and will likely be removed
20101 in a future version.
20106 Disable the use of conditional set instructions.
20108 This switch is mainly for debugging the compiler and will likely be removed
20109 in a future version.
20112 @opindex mcond-exec
20114 Enable the use of conditional execution (default).
20116 This switch is mainly for debugging the compiler and will likely be removed
20117 in a future version.
20119 @item -mno-cond-exec
20120 @opindex mno-cond-exec
20122 Disable the use of conditional execution.
20124 This switch is mainly for debugging the compiler and will likely be removed
20125 in a future version.
20127 @item -mvliw-branch
20128 @opindex mvliw-branch
20130 Run a pass to pack branches into VLIW instructions (default).
20132 This switch is mainly for debugging the compiler and will likely be removed
20133 in a future version.
20135 @item -mno-vliw-branch
20136 @opindex mno-vliw-branch
20138 Do not run a pass to pack branches into VLIW instructions.
20140 This switch is mainly for debugging the compiler and will likely be removed
20141 in a future version.
20143 @item -mmulti-cond-exec
20144 @opindex mmulti-cond-exec
20146 Enable optimization of @code{&&} and @code{||} in conditional execution
20149 This switch is mainly for debugging the compiler and will likely be removed
20150 in a future version.
20152 @item -mno-multi-cond-exec
20153 @opindex mno-multi-cond-exec
20155 Disable optimization of @code{&&} and @code{||} in conditional execution.
20157 This switch is mainly for debugging the compiler and will likely be removed
20158 in a future version.
20160 @item -mnested-cond-exec
20161 @opindex mnested-cond-exec
20163 Enable nested conditional execution optimizations (default).
20165 This switch is mainly for debugging the compiler and will likely be removed
20166 in a future version.
20168 @item -mno-nested-cond-exec
20169 @opindex mno-nested-cond-exec
20171 Disable nested conditional execution optimizations.
20173 This switch is mainly for debugging the compiler and will likely be removed
20174 in a future version.
20176 @item -moptimize-membar
20177 @opindex moptimize-membar
20179 This switch removes redundant @code{membar} instructions from the
20180 compiler-generated code. It is enabled by default.
20182 @item -mno-optimize-membar
20183 @opindex mno-optimize-membar
20184 @opindex moptimize-membar
20186 This switch disables the automatic removal of redundant @code{membar}
20187 instructions from the generated code.
20189 @item -mtomcat-stats
20190 @opindex mtomcat-stats
20192 Cause gas to print out tomcat statistics.
20194 @item -mcpu=@var{cpu}
20197 Select the processor type for which to generate code. Possible values are
20198 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
20199 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
20203 @node GNU/Linux Options
20204 @subsection GNU/Linux Options
20206 These @samp{-m} options are defined for GNU/Linux targets:
20211 Use the GNU C library. This is the default except
20212 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
20213 @samp{*-*-linux-*android*} targets.
20217 Use uClibc C library. This is the default on
20218 @samp{*-*-linux-*uclibc*} targets.
20222 Use the musl C library. This is the default on
20223 @samp{*-*-linux-*musl*} targets.
20227 Use Bionic C library. This is the default on
20228 @samp{*-*-linux-*android*} targets.
20232 Compile code compatible with Android platform. This is the default on
20233 @samp{*-*-linux-*android*} targets.
20235 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
20236 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
20237 this option makes the GCC driver pass Android-specific options to the linker.
20238 Finally, this option causes the preprocessor macro @code{__ANDROID__}
20241 @item -tno-android-cc
20242 @opindex tno-android-cc
20243 Disable compilation effects of @option{-mandroid}, i.e., do not enable
20244 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
20245 @option{-fno-rtti} by default.
20247 @item -tno-android-ld
20248 @opindex tno-android-ld
20249 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
20250 linking options to the linker.
20254 @node H8/300 Options
20255 @subsection H8/300 Options
20257 These @samp{-m} options are defined for the H8/300 implementations:
20262 Shorten some address references at link time, when possible; uses the
20263 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
20264 ld, Using ld}, for a fuller description.
20268 Generate code for the H8/300H@.
20272 Generate code for the H8S@.
20276 Generate code for the H8S and H8/300H in the normal mode. This switch
20277 must be used either with @option{-mh} or @option{-ms}.
20281 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
20285 Extended registers are stored on stack before execution of function
20286 with monitor attribute. Default option is @option{-mexr}.
20287 This option is valid only for H8S targets.
20292 Extended registers are not stored on stack before execution of function
20293 with monitor attribute. Default option is @option{-mno-exr}.
20294 This option is valid only for H8S targets.
20298 Make @code{int} data 32 bits by default.
20301 @opindex malign-300
20302 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
20303 The default for the H8/300H and H8S is to align longs and floats on
20305 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
20306 This option has no effect on the H8/300.
20310 @subsection HPPA Options
20311 @cindex HPPA Options
20313 These @samp{-m} options are defined for the HPPA family of computers:
20316 @item -march=@var{architecture-type}
20318 Generate code for the specified architecture. The choices for
20319 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
20320 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
20321 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
20322 architecture option for your machine. Code compiled for lower numbered
20323 architectures runs on higher numbered architectures, but not the
20326 @item -mpa-risc-1-0
20327 @itemx -mpa-risc-1-1
20328 @itemx -mpa-risc-2-0
20329 @opindex mpa-risc-1-0
20330 @opindex mpa-risc-1-1
20331 @opindex mpa-risc-2-0
20332 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
20334 @item -mcaller-copies
20335 @opindex mcaller-copies
20336 The caller copies function arguments passed by hidden reference. This
20337 option should be used with care as it is not compatible with the default
20338 32-bit runtime. However, only aggregates larger than eight bytes are
20339 passed by hidden reference and the option provides better compatibility
20342 @item -mjump-in-delay
20343 @opindex mjump-in-delay
20344 This option is ignored and provided for compatibility purposes only.
20346 @item -mdisable-fpregs
20347 @opindex mdisable-fpregs
20348 Prevent floating-point registers from being used in any manner. This is
20349 necessary for compiling kernels that perform lazy context switching of
20350 floating-point registers. If you use this option and attempt to perform
20351 floating-point operations, the compiler aborts.
20353 @item -mdisable-indexing
20354 @opindex mdisable-indexing
20355 Prevent the compiler from using indexing address modes. This avoids some
20356 rather obscure problems when compiling MIG generated code under MACH@.
20358 @item -mno-space-regs
20359 @opindex mno-space-regs
20360 @opindex mspace-regs
20361 Generate code that assumes the target has no space registers. This allows
20362 GCC to generate faster indirect calls and use unscaled index address modes.
20364 Such code is suitable for level 0 PA systems and kernels.
20366 @item -mfast-indirect-calls
20367 @opindex mfast-indirect-calls
20368 Generate code that assumes calls never cross space boundaries. This
20369 allows GCC to emit code that performs faster indirect calls.
20371 This option does not work in the presence of shared libraries or nested
20374 @item -mfixed-range=@var{register-range}
20375 @opindex mfixed-range
20376 Generate code treating the given register range as fixed registers.
20377 A fixed register is one that the register allocator cannot use. This is
20378 useful when compiling kernel code. A register range is specified as
20379 two registers separated by a dash. Multiple register ranges can be
20380 specified separated by a comma.
20382 @item -mlong-load-store
20383 @opindex mlong-load-store
20384 Generate 3-instruction load and store sequences as sometimes required by
20385 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
20388 @item -mportable-runtime
20389 @opindex mportable-runtime
20390 Use the portable calling conventions proposed by HP for ELF systems.
20394 Enable the use of assembler directives only GAS understands.
20396 @item -mschedule=@var{cpu-type}
20398 Schedule code according to the constraints for the machine type
20399 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
20400 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
20401 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
20402 proper scheduling option for your machine. The default scheduling is
20406 @opindex mlinker-opt
20407 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
20408 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
20409 linkers in which they give bogus error messages when linking some programs.
20412 @opindex msoft-float
20413 Generate output containing library calls for floating point.
20414 @strong{Warning:} the requisite libraries are not available for all HPPA
20415 targets. Normally the facilities of the machine's usual C compiler are
20416 used, but this cannot be done directly in cross-compilation. You must make
20417 your own arrangements to provide suitable library functions for
20420 @option{-msoft-float} changes the calling convention in the output file;
20421 therefore, it is only useful if you compile @emph{all} of a program with
20422 this option. In particular, you need to compile @file{libgcc.a}, the
20423 library that comes with GCC, with @option{-msoft-float} in order for
20428 Generate the predefine, @code{_SIO}, for server IO@. The default is
20429 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
20430 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
20431 options are available under HP-UX and HI-UX@.
20435 Use options specific to GNU @command{ld}.
20436 This passes @option{-shared} to @command{ld} when
20437 building a shared library. It is the default when GCC is configured,
20438 explicitly or implicitly, with the GNU linker. This option does not
20439 affect which @command{ld} is called; it only changes what parameters
20440 are passed to that @command{ld}.
20441 The @command{ld} that is called is determined by the
20442 @option{--with-ld} configure option, GCC's program search path, and
20443 finally by the user's @env{PATH}. The linker used by GCC can be printed
20444 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
20445 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20449 Use options specific to HP @command{ld}.
20450 This passes @option{-b} to @command{ld} when building
20451 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
20452 links. It is the default when GCC is configured, explicitly or
20453 implicitly, with the HP linker. This option does not affect
20454 which @command{ld} is called; it only changes what parameters are passed to that
20456 The @command{ld} that is called is determined by the @option{--with-ld}
20457 configure option, GCC's program search path, and finally by the user's
20458 @env{PATH}. The linker used by GCC can be printed using @samp{which
20459 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
20460 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
20463 @opindex mno-long-calls
20464 @opindex mlong-calls
20465 Generate code that uses long call sequences. This ensures that a call
20466 is always able to reach linker generated stubs. The default is to generate
20467 long calls only when the distance from the call site to the beginning
20468 of the function or translation unit, as the case may be, exceeds a
20469 predefined limit set by the branch type being used. The limits for
20470 normal calls are 7,600,000 and 240,000 bytes, respectively for the
20471 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
20474 Distances are measured from the beginning of functions when using the
20475 @option{-ffunction-sections} option, or when using the @option{-mgas}
20476 and @option{-mno-portable-runtime} options together under HP-UX with
20479 It is normally not desirable to use this option as it degrades
20480 performance. However, it may be useful in large applications,
20481 particularly when partial linking is used to build the application.
20483 The types of long calls used depends on the capabilities of the
20484 assembler and linker, and the type of code being generated. The
20485 impact on systems that support long absolute calls, and long pic
20486 symbol-difference or pc-relative calls should be relatively small.
20487 However, an indirect call is used on 32-bit ELF systems in pic code
20488 and it is quite long.
20490 @item -munix=@var{unix-std}
20492 Generate compiler predefines and select a startfile for the specified
20493 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
20494 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
20495 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
20496 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
20497 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
20500 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
20501 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
20502 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
20503 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
20504 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
20505 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
20507 It is @emph{important} to note that this option changes the interfaces
20508 for various library routines. It also affects the operational behavior
20509 of the C library. Thus, @emph{extreme} care is needed in using this
20512 Library code that is intended to operate with more than one UNIX
20513 standard must test, set and restore the variable @code{__xpg4_extended_mask}
20514 as appropriate. Most GNU software doesn't provide this capability.
20518 Suppress the generation of link options to search libdld.sl when the
20519 @option{-static} option is specified on HP-UX 10 and later.
20523 The HP-UX implementation of setlocale in libc has a dependency on
20524 libdld.sl. There isn't an archive version of libdld.sl. Thus,
20525 when the @option{-static} option is specified, special link options
20526 are needed to resolve this dependency.
20528 On HP-UX 10 and later, the GCC driver adds the necessary options to
20529 link with libdld.sl when the @option{-static} option is specified.
20530 This causes the resulting binary to be dynamic. On the 64-bit port,
20531 the linkers generate dynamic binaries by default in any case. The
20532 @option{-nolibdld} option can be used to prevent the GCC driver from
20533 adding these link options.
20537 Add support for multithreading with the @dfn{dce thread} library
20538 under HP-UX@. This option sets flags for both the preprocessor and
20542 @node IA-64 Options
20543 @subsection IA-64 Options
20544 @cindex IA-64 Options
20546 These are the @samp{-m} options defined for the Intel IA-64 architecture.
20550 @opindex mbig-endian
20551 Generate code for a big-endian target. This is the default for HP-UX@.
20553 @item -mlittle-endian
20554 @opindex mlittle-endian
20555 Generate code for a little-endian target. This is the default for AIX5
20561 @opindex mno-gnu-as
20562 Generate (or don't) code for the GNU assembler. This is the default.
20563 @c Also, this is the default if the configure option @option{--with-gnu-as}
20569 @opindex mno-gnu-ld
20570 Generate (or don't) code for the GNU linker. This is the default.
20571 @c Also, this is the default if the configure option @option{--with-gnu-ld}
20576 Generate code that does not use a global pointer register. The result
20577 is not position independent code, and violates the IA-64 ABI@.
20579 @item -mvolatile-asm-stop
20580 @itemx -mno-volatile-asm-stop
20581 @opindex mvolatile-asm-stop
20582 @opindex mno-volatile-asm-stop
20583 Generate (or don't) a stop bit immediately before and after volatile asm
20586 @item -mregister-names
20587 @itemx -mno-register-names
20588 @opindex mregister-names
20589 @opindex mno-register-names
20590 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
20591 the stacked registers. This may make assembler output more readable.
20597 Disable (or enable) optimizations that use the small data section. This may
20598 be useful for working around optimizer bugs.
20600 @item -mconstant-gp
20601 @opindex mconstant-gp
20602 Generate code that uses a single constant global pointer value. This is
20603 useful when compiling kernel code.
20607 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
20608 This is useful when compiling firmware code.
20610 @item -minline-float-divide-min-latency
20611 @opindex minline-float-divide-min-latency
20612 Generate code for inline divides of floating-point values
20613 using the minimum latency algorithm.
20615 @item -minline-float-divide-max-throughput
20616 @opindex minline-float-divide-max-throughput
20617 Generate code for inline divides of floating-point values
20618 using the maximum throughput algorithm.
20620 @item -mno-inline-float-divide
20621 @opindex mno-inline-float-divide
20622 Do not generate inline code for divides of floating-point values.
20624 @item -minline-int-divide-min-latency
20625 @opindex minline-int-divide-min-latency
20626 Generate code for inline divides of integer values
20627 using the minimum latency algorithm.
20629 @item -minline-int-divide-max-throughput
20630 @opindex minline-int-divide-max-throughput
20631 Generate code for inline divides of integer values
20632 using the maximum throughput algorithm.
20634 @item -mno-inline-int-divide
20635 @opindex mno-inline-int-divide
20636 @opindex minline-int-divide
20637 Do not generate inline code for divides of integer values.
20639 @item -minline-sqrt-min-latency
20640 @opindex minline-sqrt-min-latency
20641 Generate code for inline square roots
20642 using the minimum latency algorithm.
20644 @item -minline-sqrt-max-throughput
20645 @opindex minline-sqrt-max-throughput
20646 Generate code for inline square roots
20647 using the maximum throughput algorithm.
20649 @item -mno-inline-sqrt
20650 @opindex mno-inline-sqrt
20651 Do not generate inline code for @code{sqrt}.
20654 @itemx -mno-fused-madd
20655 @opindex mfused-madd
20656 @opindex mno-fused-madd
20657 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
20658 instructions. The default is to use these instructions.
20660 @item -mno-dwarf2-asm
20661 @itemx -mdwarf2-asm
20662 @opindex mno-dwarf2-asm
20663 @opindex mdwarf2-asm
20664 Don't (or do) generate assembler code for the DWARF line number debugging
20665 info. This may be useful when not using the GNU assembler.
20667 @item -mearly-stop-bits
20668 @itemx -mno-early-stop-bits
20669 @opindex mearly-stop-bits
20670 @opindex mno-early-stop-bits
20671 Allow stop bits to be placed earlier than immediately preceding the
20672 instruction that triggered the stop bit. This can improve instruction
20673 scheduling, but does not always do so.
20675 @item -mfixed-range=@var{register-range}
20676 @opindex mfixed-range
20677 Generate code treating the given register range as fixed registers.
20678 A fixed register is one that the register allocator cannot use. This is
20679 useful when compiling kernel code. A register range is specified as
20680 two registers separated by a dash. Multiple register ranges can be
20681 specified separated by a comma.
20683 @item -mtls-size=@var{tls-size}
20685 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
20688 @item -mtune=@var{cpu-type}
20690 Tune the instruction scheduling for a particular CPU, Valid values are
20691 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
20692 and @samp{mckinley}.
20698 Generate code for a 32-bit or 64-bit environment.
20699 The 32-bit environment sets int, long and pointer to 32 bits.
20700 The 64-bit environment sets int to 32 bits and long and pointer
20701 to 64 bits. These are HP-UX specific flags.
20703 @item -mno-sched-br-data-spec
20704 @itemx -msched-br-data-spec
20705 @opindex mno-sched-br-data-spec
20706 @opindex msched-br-data-spec
20707 (Dis/En)able data speculative scheduling before reload.
20708 This results in generation of @code{ld.a} instructions and
20709 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20710 The default setting is disabled.
20712 @item -msched-ar-data-spec
20713 @itemx -mno-sched-ar-data-spec
20714 @opindex msched-ar-data-spec
20715 @opindex mno-sched-ar-data-spec
20716 (En/Dis)able data speculative scheduling after reload.
20717 This results in generation of @code{ld.a} instructions and
20718 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
20719 The default setting is enabled.
20721 @item -mno-sched-control-spec
20722 @itemx -msched-control-spec
20723 @opindex mno-sched-control-spec
20724 @opindex msched-control-spec
20725 (Dis/En)able control speculative scheduling. This feature is
20726 available only during region scheduling (i.e.@: before reload).
20727 This results in generation of the @code{ld.s} instructions and
20728 the corresponding check instructions @code{chk.s}.
20729 The default setting is disabled.
20731 @item -msched-br-in-data-spec
20732 @itemx -mno-sched-br-in-data-spec
20733 @opindex msched-br-in-data-spec
20734 @opindex mno-sched-br-in-data-spec
20735 (En/Dis)able speculative scheduling of the instructions that
20736 are dependent on the data speculative loads before reload.
20737 This is effective only with @option{-msched-br-data-spec} enabled.
20738 The default setting is enabled.
20740 @item -msched-ar-in-data-spec
20741 @itemx -mno-sched-ar-in-data-spec
20742 @opindex msched-ar-in-data-spec
20743 @opindex mno-sched-ar-in-data-spec
20744 (En/Dis)able speculative scheduling of the instructions that
20745 are dependent on the data speculative loads after reload.
20746 This is effective only with @option{-msched-ar-data-spec} enabled.
20747 The default setting is enabled.
20749 @item -msched-in-control-spec
20750 @itemx -mno-sched-in-control-spec
20751 @opindex msched-in-control-spec
20752 @opindex mno-sched-in-control-spec
20753 (En/Dis)able speculative scheduling of the instructions that
20754 are dependent on the control speculative loads.
20755 This is effective only with @option{-msched-control-spec} enabled.
20756 The default setting is enabled.
20758 @item -mno-sched-prefer-non-data-spec-insns
20759 @itemx -msched-prefer-non-data-spec-insns
20760 @opindex mno-sched-prefer-non-data-spec-insns
20761 @opindex msched-prefer-non-data-spec-insns
20762 If enabled, data-speculative instructions are chosen for schedule
20763 only if there are no other choices at the moment. This makes
20764 the use of the data speculation much more conservative.
20765 The default setting is disabled.
20767 @item -mno-sched-prefer-non-control-spec-insns
20768 @itemx -msched-prefer-non-control-spec-insns
20769 @opindex mno-sched-prefer-non-control-spec-insns
20770 @opindex msched-prefer-non-control-spec-insns
20771 If enabled, control-speculative instructions are chosen for schedule
20772 only if there are no other choices at the moment. This makes
20773 the use of the control speculation much more conservative.
20774 The default setting is disabled.
20776 @item -mno-sched-count-spec-in-critical-path
20777 @itemx -msched-count-spec-in-critical-path
20778 @opindex mno-sched-count-spec-in-critical-path
20779 @opindex msched-count-spec-in-critical-path
20780 If enabled, speculative dependencies are considered during
20781 computation of the instructions priorities. This makes the use of the
20782 speculation a bit more conservative.
20783 The default setting is disabled.
20785 @item -msched-spec-ldc
20786 @opindex msched-spec-ldc
20787 Use a simple data speculation check. This option is on by default.
20789 @item -msched-control-spec-ldc
20790 @opindex msched-spec-ldc
20791 Use a simple check for control speculation. This option is on by default.
20793 @item -msched-stop-bits-after-every-cycle
20794 @opindex msched-stop-bits-after-every-cycle
20795 Place a stop bit after every cycle when scheduling. This option is on
20798 @item -msched-fp-mem-deps-zero-cost
20799 @opindex msched-fp-mem-deps-zero-cost
20800 Assume that floating-point stores and loads are not likely to cause a conflict
20801 when placed into the same instruction group. This option is disabled by
20804 @item -msel-sched-dont-check-control-spec
20805 @opindex msel-sched-dont-check-control-spec
20806 Generate checks for control speculation in selective scheduling.
20807 This flag is disabled by default.
20809 @item -msched-max-memory-insns=@var{max-insns}
20810 @opindex msched-max-memory-insns
20811 Limit on the number of memory insns per instruction group, giving lower
20812 priority to subsequent memory insns attempting to schedule in the same
20813 instruction group. Frequently useful to prevent cache bank conflicts.
20814 The default value is 1.
20816 @item -msched-max-memory-insns-hard-limit
20817 @opindex msched-max-memory-insns-hard-limit
20818 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
20819 disallowing more than that number in an instruction group.
20820 Otherwise, the limit is ``soft'', meaning that non-memory operations
20821 are preferred when the limit is reached, but memory operations may still
20827 @subsection LM32 Options
20828 @cindex LM32 options
20830 These @option{-m} options are defined for the LatticeMico32 architecture:
20833 @item -mbarrel-shift-enabled
20834 @opindex mbarrel-shift-enabled
20835 Enable barrel-shift instructions.
20837 @item -mdivide-enabled
20838 @opindex mdivide-enabled
20839 Enable divide and modulus instructions.
20841 @item -mmultiply-enabled
20842 @opindex multiply-enabled
20843 Enable multiply instructions.
20845 @item -msign-extend-enabled
20846 @opindex msign-extend-enabled
20847 Enable sign extend instructions.
20849 @item -muser-enabled
20850 @opindex muser-enabled
20851 Enable user-defined instructions.
20856 @subsection M32C Options
20857 @cindex M32C options
20860 @item -mcpu=@var{name}
20862 Select the CPU for which code is generated. @var{name} may be one of
20863 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
20864 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
20865 the M32C/80 series.
20869 Specifies that the program will be run on the simulator. This causes
20870 an alternate runtime library to be linked in which supports, for
20871 example, file I/O@. You must not use this option when generating
20872 programs that will run on real hardware; you must provide your own
20873 runtime library for whatever I/O functions are needed.
20875 @item -memregs=@var{number}
20877 Specifies the number of memory-based pseudo-registers GCC uses
20878 during code generation. These pseudo-registers are used like real
20879 registers, so there is a tradeoff between GCC's ability to fit the
20880 code into available registers, and the performance penalty of using
20881 memory instead of registers. Note that all modules in a program must
20882 be compiled with the same value for this option. Because of that, you
20883 must not use this option with GCC's default runtime libraries.
20887 @node M32R/D Options
20888 @subsection M32R/D Options
20889 @cindex M32R/D options
20891 These @option{-m} options are defined for Renesas M32R/D architectures:
20896 Generate code for the M32R/2@.
20900 Generate code for the M32R/X@.
20904 Generate code for the M32R@. This is the default.
20906 @item -mmodel=small
20907 @opindex mmodel=small
20908 Assume all objects live in the lower 16MB of memory (so that their addresses
20909 can be loaded with the @code{ld24} instruction), and assume all subroutines
20910 are reachable with the @code{bl} instruction.
20911 This is the default.
20913 The addressability of a particular object can be set with the
20914 @code{model} attribute.
20916 @item -mmodel=medium
20917 @opindex mmodel=medium
20918 Assume objects may be anywhere in the 32-bit address space (the compiler
20919 generates @code{seth/add3} instructions to load their addresses), and
20920 assume all subroutines are reachable with the @code{bl} instruction.
20922 @item -mmodel=large
20923 @opindex mmodel=large
20924 Assume objects may be anywhere in the 32-bit address space (the compiler
20925 generates @code{seth/add3} instructions to load their addresses), and
20926 assume subroutines may not be reachable with the @code{bl} instruction
20927 (the compiler generates the much slower @code{seth/add3/jl}
20928 instruction sequence).
20931 @opindex msdata=none
20932 Disable use of the small data area. Variables are put into
20933 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
20934 @code{section} attribute has been specified).
20935 This is the default.
20937 The small data area consists of sections @code{.sdata} and @code{.sbss}.
20938 Objects may be explicitly put in the small data area with the
20939 @code{section} attribute using one of these sections.
20941 @item -msdata=sdata
20942 @opindex msdata=sdata
20943 Put small global and static data in the small data area, but do not
20944 generate special code to reference them.
20947 @opindex msdata=use
20948 Put small global and static data in the small data area, and generate
20949 special instructions to reference them.
20953 @cindex smaller data references
20954 Put global and static objects less than or equal to @var{num} bytes
20955 into the small data or BSS sections instead of the normal data or BSS
20956 sections. The default value of @var{num} is 8.
20957 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
20958 for this option to have any effect.
20960 All modules should be compiled with the same @option{-G @var{num}} value.
20961 Compiling with different values of @var{num} may or may not work; if it
20962 doesn't the linker gives an error message---incorrect code is not
20967 Makes the M32R-specific code in the compiler display some statistics
20968 that might help in debugging programs.
20970 @item -malign-loops
20971 @opindex malign-loops
20972 Align all loops to a 32-byte boundary.
20974 @item -mno-align-loops
20975 @opindex mno-align-loops
20976 Do not enforce a 32-byte alignment for loops. This is the default.
20978 @item -missue-rate=@var{number}
20979 @opindex missue-rate=@var{number}
20980 Issue @var{number} instructions per cycle. @var{number} can only be 1
20983 @item -mbranch-cost=@var{number}
20984 @opindex mbranch-cost=@var{number}
20985 @var{number} can only be 1 or 2. If it is 1 then branches are
20986 preferred over conditional code, if it is 2, then the opposite applies.
20988 @item -mflush-trap=@var{number}
20989 @opindex mflush-trap=@var{number}
20990 Specifies the trap number to use to flush the cache. The default is
20991 12. Valid numbers are between 0 and 15 inclusive.
20993 @item -mno-flush-trap
20994 @opindex mno-flush-trap
20995 Specifies that the cache cannot be flushed by using a trap.
20997 @item -mflush-func=@var{name}
20998 @opindex mflush-func=@var{name}
20999 Specifies the name of the operating system function to call to flush
21000 the cache. The default is @samp{_flush_cache}, but a function call
21001 is only used if a trap is not available.
21003 @item -mno-flush-func
21004 @opindex mno-flush-func
21005 Indicates that there is no OS function for flushing the cache.
21009 @node M680x0 Options
21010 @subsection M680x0 Options
21011 @cindex M680x0 options
21013 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
21014 The default settings depend on which architecture was selected when
21015 the compiler was configured; the defaults for the most common choices
21019 @item -march=@var{arch}
21021 Generate code for a specific M680x0 or ColdFire instruction set
21022 architecture. Permissible values of @var{arch} for M680x0
21023 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
21024 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
21025 architectures are selected according to Freescale's ISA classification
21026 and the permissible values are: @samp{isaa}, @samp{isaaplus},
21027 @samp{isab} and @samp{isac}.
21029 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
21030 code for a ColdFire target. The @var{arch} in this macro is one of the
21031 @option{-march} arguments given above.
21033 When used together, @option{-march} and @option{-mtune} select code
21034 that runs on a family of similar processors but that is optimized
21035 for a particular microarchitecture.
21037 @item -mcpu=@var{cpu}
21039 Generate code for a specific M680x0 or ColdFire processor.
21040 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
21041 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
21042 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
21043 below, which also classifies the CPUs into families:
21045 @multitable @columnfractions 0.20 0.80
21046 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
21047 @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}
21048 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
21049 @item @samp{5206e} @tab @samp{5206e}
21050 @item @samp{5208} @tab @samp{5207} @samp{5208}
21051 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
21052 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
21053 @item @samp{5216} @tab @samp{5214} @samp{5216}
21054 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
21055 @item @samp{5225} @tab @samp{5224} @samp{5225}
21056 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
21057 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
21058 @item @samp{5249} @tab @samp{5249}
21059 @item @samp{5250} @tab @samp{5250}
21060 @item @samp{5271} @tab @samp{5270} @samp{5271}
21061 @item @samp{5272} @tab @samp{5272}
21062 @item @samp{5275} @tab @samp{5274} @samp{5275}
21063 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
21064 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
21065 @item @samp{5307} @tab @samp{5307}
21066 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
21067 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
21068 @item @samp{5407} @tab @samp{5407}
21069 @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}
21072 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
21073 @var{arch} is compatible with @var{cpu}. Other combinations of
21074 @option{-mcpu} and @option{-march} are rejected.
21076 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
21077 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
21078 where the value of @var{family} is given by the table above.
21080 @item -mtune=@var{tune}
21082 Tune the code for a particular microarchitecture within the
21083 constraints set by @option{-march} and @option{-mcpu}.
21084 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
21085 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
21086 and @samp{cpu32}. The ColdFire microarchitectures
21087 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
21089 You can also use @option{-mtune=68020-40} for code that needs
21090 to run relatively well on 68020, 68030 and 68040 targets.
21091 @option{-mtune=68020-60} is similar but includes 68060 targets
21092 as well. These two options select the same tuning decisions as
21093 @option{-m68020-40} and @option{-m68020-60} respectively.
21095 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
21096 when tuning for 680x0 architecture @var{arch}. It also defines
21097 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
21098 option is used. If GCC is tuning for a range of architectures,
21099 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
21100 it defines the macros for every architecture in the range.
21102 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
21103 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
21104 of the arguments given above.
21110 Generate output for a 68000. This is the default
21111 when the compiler is configured for 68000-based systems.
21112 It is equivalent to @option{-march=68000}.
21114 Use this option for microcontrollers with a 68000 or EC000 core,
21115 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
21119 Generate output for a 68010. This is the default
21120 when the compiler is configured for 68010-based systems.
21121 It is equivalent to @option{-march=68010}.
21127 Generate output for a 68020. This is the default
21128 when the compiler is configured for 68020-based systems.
21129 It is equivalent to @option{-march=68020}.
21133 Generate output for a 68030. This is the default when the compiler is
21134 configured for 68030-based systems. It is equivalent to
21135 @option{-march=68030}.
21139 Generate output for a 68040. This is the default when the compiler is
21140 configured for 68040-based systems. It is equivalent to
21141 @option{-march=68040}.
21143 This option inhibits the use of 68881/68882 instructions that have to be
21144 emulated by software on the 68040. Use this option if your 68040 does not
21145 have code to emulate those instructions.
21149 Generate output for a 68060. This is the default when the compiler is
21150 configured for 68060-based systems. It is equivalent to
21151 @option{-march=68060}.
21153 This option inhibits the use of 68020 and 68881/68882 instructions that
21154 have to be emulated by software on the 68060. Use this option if your 68060
21155 does not have code to emulate those instructions.
21159 Generate output for a CPU32. This is the default
21160 when the compiler is configured for CPU32-based systems.
21161 It is equivalent to @option{-march=cpu32}.
21163 Use this option for microcontrollers with a
21164 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
21165 68336, 68340, 68341, 68349 and 68360.
21169 Generate output for a 520X ColdFire CPU@. This is the default
21170 when the compiler is configured for 520X-based systems.
21171 It is equivalent to @option{-mcpu=5206}, and is now deprecated
21172 in favor of that option.
21174 Use this option for microcontroller with a 5200 core, including
21175 the MCF5202, MCF5203, MCF5204 and MCF5206.
21179 Generate output for a 5206e ColdFire CPU@. The option is now
21180 deprecated in favor of the equivalent @option{-mcpu=5206e}.
21184 Generate output for a member of the ColdFire 528X family.
21185 The option is now deprecated in favor of the equivalent
21186 @option{-mcpu=528x}.
21190 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
21191 in favor of the equivalent @option{-mcpu=5307}.
21195 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
21196 in favor of the equivalent @option{-mcpu=5407}.
21200 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
21201 This includes use of hardware floating-point instructions.
21202 The option is equivalent to @option{-mcpu=547x}, and is now
21203 deprecated in favor of that option.
21207 Generate output for a 68040, without using any of the new instructions.
21208 This results in code that can run relatively efficiently on either a
21209 68020/68881 or a 68030 or a 68040. The generated code does use the
21210 68881 instructions that are emulated on the 68040.
21212 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
21216 Generate output for a 68060, without using any of the new instructions.
21217 This results in code that can run relatively efficiently on either a
21218 68020/68881 or a 68030 or a 68040. The generated code does use the
21219 68881 instructions that are emulated on the 68060.
21221 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
21225 @opindex mhard-float
21227 Generate floating-point instructions. This is the default for 68020
21228 and above, and for ColdFire devices that have an FPU@. It defines the
21229 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
21230 on ColdFire targets.
21233 @opindex msoft-float
21234 Do not generate floating-point instructions; use library calls instead.
21235 This is the default for 68000, 68010, and 68832 targets. It is also
21236 the default for ColdFire devices that have no FPU.
21242 Generate (do not generate) ColdFire hardware divide and remainder
21243 instructions. If @option{-march} is used without @option{-mcpu},
21244 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
21245 architectures. Otherwise, the default is taken from the target CPU
21246 (either the default CPU, or the one specified by @option{-mcpu}). For
21247 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
21248 @option{-mcpu=5206e}.
21250 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
21254 Consider type @code{int} to be 16 bits wide, like @code{short int}.
21255 Additionally, parameters passed on the stack are also aligned to a
21256 16-bit boundary even on targets whose API mandates promotion to 32-bit.
21260 Do not consider type @code{int} to be 16 bits wide. This is the default.
21263 @itemx -mno-bitfield
21264 @opindex mnobitfield
21265 @opindex mno-bitfield
21266 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
21267 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
21271 Do use the bit-field instructions. The @option{-m68020} option implies
21272 @option{-mbitfield}. This is the default if you use a configuration
21273 designed for a 68020.
21277 Use a different function-calling convention, in which functions
21278 that take a fixed number of arguments return with the @code{rtd}
21279 instruction, which pops their arguments while returning. This
21280 saves one instruction in the caller since there is no need to pop
21281 the arguments there.
21283 This calling convention is incompatible with the one normally
21284 used on Unix, so you cannot use it if you need to call libraries
21285 compiled with the Unix compiler.
21287 Also, you must provide function prototypes for all functions that
21288 take variable numbers of arguments (including @code{printf});
21289 otherwise incorrect code is generated for calls to those
21292 In addition, seriously incorrect code results if you call a
21293 function with too many arguments. (Normally, extra arguments are
21294 harmlessly ignored.)
21296 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
21297 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
21299 The default is @option{-mno-rtd}.
21302 @itemx -mno-align-int
21303 @opindex malign-int
21304 @opindex mno-align-int
21305 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
21306 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
21307 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
21308 Aligning variables on 32-bit boundaries produces code that runs somewhat
21309 faster on processors with 32-bit busses at the expense of more memory.
21311 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
21312 aligns structures containing the above types differently than
21313 most published application binary interface specifications for the m68k.
21317 Use the pc-relative addressing mode of the 68000 directly, instead of
21318 using a global offset table. At present, this option implies @option{-fpic},
21319 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
21320 not presently supported with @option{-mpcrel}, though this could be supported for
21321 68020 and higher processors.
21323 @item -mno-strict-align
21324 @itemx -mstrict-align
21325 @opindex mno-strict-align
21326 @opindex mstrict-align
21327 Do not (do) assume that unaligned memory references are handled by
21331 Generate code that allows the data segment to be located in a different
21332 area of memory from the text segment. This allows for execute-in-place in
21333 an environment without virtual memory management. This option implies
21336 @item -mno-sep-data
21337 Generate code that assumes that the data segment follows the text segment.
21338 This is the default.
21340 @item -mid-shared-library
21341 Generate code that supports shared libraries via the library ID method.
21342 This allows for execute-in-place and shared libraries in an environment
21343 without virtual memory management. This option implies @option{-fPIC}.
21345 @item -mno-id-shared-library
21346 Generate code that doesn't assume ID-based shared libraries are being used.
21347 This is the default.
21349 @item -mshared-library-id=n
21350 Specifies the identification number of the ID-based shared library being
21351 compiled. Specifying a value of 0 generates more compact code; specifying
21352 other values forces the allocation of that number to the current
21353 library, but is no more space- or time-efficient than omitting this option.
21359 When generating position-independent code for ColdFire, generate code
21360 that works if the GOT has more than 8192 entries. This code is
21361 larger and slower than code generated without this option. On M680x0
21362 processors, this option is not needed; @option{-fPIC} suffices.
21364 GCC normally uses a single instruction to load values from the GOT@.
21365 While this is relatively efficient, it only works if the GOT
21366 is smaller than about 64k. Anything larger causes the linker
21367 to report an error such as:
21369 @cindex relocation truncated to fit (ColdFire)
21371 relocation truncated to fit: R_68K_GOT16O foobar
21374 If this happens, you should recompile your code with @option{-mxgot}.
21375 It should then work with very large GOTs. However, code generated with
21376 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
21377 the value of a global symbol.
21379 Note that some linkers, including newer versions of the GNU linker,
21380 can create multiple GOTs and sort GOT entries. If you have such a linker,
21381 you should only need to use @option{-mxgot} when compiling a single
21382 object file that accesses more than 8192 GOT entries. Very few do.
21384 These options have no effect unless GCC is generating
21385 position-independent code.
21387 @item -mlong-jump-table-offsets
21388 @opindex mlong-jump-table-offsets
21389 Use 32-bit offsets in @code{switch} tables. The default is to use
21394 @node MCore Options
21395 @subsection MCore Options
21396 @cindex MCore options
21398 These are the @samp{-m} options defined for the Motorola M*Core
21404 @itemx -mno-hardlit
21406 @opindex mno-hardlit
21407 Inline constants into the code stream if it can be done in two
21408 instructions or less.
21414 Use the divide instruction. (Enabled by default).
21416 @item -mrelax-immediate
21417 @itemx -mno-relax-immediate
21418 @opindex mrelax-immediate
21419 @opindex mno-relax-immediate
21420 Allow arbitrary-sized immediates in bit operations.
21422 @item -mwide-bitfields
21423 @itemx -mno-wide-bitfields
21424 @opindex mwide-bitfields
21425 @opindex mno-wide-bitfields
21426 Always treat bit-fields as @code{int}-sized.
21428 @item -m4byte-functions
21429 @itemx -mno-4byte-functions
21430 @opindex m4byte-functions
21431 @opindex mno-4byte-functions
21432 Force all functions to be aligned to a 4-byte boundary.
21434 @item -mcallgraph-data
21435 @itemx -mno-callgraph-data
21436 @opindex mcallgraph-data
21437 @opindex mno-callgraph-data
21438 Emit callgraph information.
21441 @itemx -mno-slow-bytes
21442 @opindex mslow-bytes
21443 @opindex mno-slow-bytes
21444 Prefer word access when reading byte quantities.
21446 @item -mlittle-endian
21447 @itemx -mbig-endian
21448 @opindex mlittle-endian
21449 @opindex mbig-endian
21450 Generate code for a little-endian target.
21456 Generate code for the 210 processor.
21460 Assume that runtime support has been provided and so omit the
21461 simulator library (@file{libsim.a)} from the linker command line.
21463 @item -mstack-increment=@var{size}
21464 @opindex mstack-increment
21465 Set the maximum amount for a single stack increment operation. Large
21466 values can increase the speed of programs that contain functions
21467 that need a large amount of stack space, but they can also trigger a
21468 segmentation fault if the stack is extended too much. The default
21474 @subsection MeP Options
21475 @cindex MeP options
21481 Enables the @code{abs} instruction, which is the absolute difference
21482 between two registers.
21486 Enables all the optional instructions---average, multiply, divide, bit
21487 operations, leading zero, absolute difference, min/max, clip, and
21493 Enables the @code{ave} instruction, which computes the average of two
21496 @item -mbased=@var{n}
21498 Variables of size @var{n} bytes or smaller are placed in the
21499 @code{.based} section by default. Based variables use the @code{$tp}
21500 register as a base register, and there is a 128-byte limit to the
21501 @code{.based} section.
21505 Enables the bit operation instructions---bit test (@code{btstm}), set
21506 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
21507 test-and-set (@code{tas}).
21509 @item -mc=@var{name}
21511 Selects which section constant data is placed in. @var{name} may
21512 be @samp{tiny}, @samp{near}, or @samp{far}.
21516 Enables the @code{clip} instruction. Note that @option{-mclip} is not
21517 useful unless you also provide @option{-mminmax}.
21519 @item -mconfig=@var{name}
21521 Selects one of the built-in core configurations. Each MeP chip has
21522 one or more modules in it; each module has a core CPU and a variety of
21523 coprocessors, optional instructions, and peripherals. The
21524 @code{MeP-Integrator} tool, not part of GCC, provides these
21525 configurations through this option; using this option is the same as
21526 using all the corresponding command-line options. The default
21527 configuration is @samp{default}.
21531 Enables the coprocessor instructions. By default, this is a 32-bit
21532 coprocessor. Note that the coprocessor is normally enabled via the
21533 @option{-mconfig=} option.
21537 Enables the 32-bit coprocessor's instructions.
21541 Enables the 64-bit coprocessor's instructions.
21545 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
21549 Causes constant variables to be placed in the @code{.near} section.
21553 Enables the @code{div} and @code{divu} instructions.
21557 Generate big-endian code.
21561 Generate little-endian code.
21563 @item -mio-volatile
21564 @opindex mio-volatile
21565 Tells the compiler that any variable marked with the @code{io}
21566 attribute is to be considered volatile.
21570 Causes variables to be assigned to the @code{.far} section by default.
21574 Enables the @code{leadz} (leading zero) instruction.
21578 Causes variables to be assigned to the @code{.near} section by default.
21582 Enables the @code{min} and @code{max} instructions.
21586 Enables the multiplication and multiply-accumulate instructions.
21590 Disables all the optional instructions enabled by @option{-mall-opts}.
21594 Enables the @code{repeat} and @code{erepeat} instructions, used for
21595 low-overhead looping.
21599 Causes all variables to default to the @code{.tiny} section. Note
21600 that there is a 65536-byte limit to this section. Accesses to these
21601 variables use the @code{%gp} base register.
21605 Enables the saturation instructions. Note that the compiler does not
21606 currently generate these itself, but this option is included for
21607 compatibility with other tools, like @code{as}.
21611 Link the SDRAM-based runtime instead of the default ROM-based runtime.
21615 Link the simulator run-time libraries.
21619 Link the simulator runtime libraries, excluding built-in support
21620 for reset and exception vectors and tables.
21624 Causes all functions to default to the @code{.far} section. Without
21625 this option, functions default to the @code{.near} section.
21627 @item -mtiny=@var{n}
21629 Variables that are @var{n} bytes or smaller are allocated to the
21630 @code{.tiny} section. These variables use the @code{$gp} base
21631 register. The default for this option is 4, but note that there's a
21632 65536-byte limit to the @code{.tiny} section.
21636 @node MicroBlaze Options
21637 @subsection MicroBlaze Options
21638 @cindex MicroBlaze Options
21643 @opindex msoft-float
21644 Use software emulation for floating point (default).
21647 @opindex mhard-float
21648 Use hardware floating-point instructions.
21652 Do not optimize block moves, use @code{memcpy}.
21654 @item -mno-clearbss
21655 @opindex mno-clearbss
21656 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
21658 @item -mcpu=@var{cpu-type}
21660 Use features of, and schedule code for, the given CPU.
21661 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
21662 where @var{X} is a major version, @var{YY} is the minor version, and
21663 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
21664 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v6.00.a}.
21666 @item -mxl-soft-mul
21667 @opindex mxl-soft-mul
21668 Use software multiply emulation (default).
21670 @item -mxl-soft-div
21671 @opindex mxl-soft-div
21672 Use software emulation for divides (default).
21674 @item -mxl-barrel-shift
21675 @opindex mxl-barrel-shift
21676 Use the hardware barrel shifter.
21678 @item -mxl-pattern-compare
21679 @opindex mxl-pattern-compare
21680 Use pattern compare instructions.
21682 @item -msmall-divides
21683 @opindex msmall-divides
21684 Use table lookup optimization for small signed integer divisions.
21686 @item -mxl-stack-check
21687 @opindex mxl-stack-check
21688 This option is deprecated. Use @option{-fstack-check} instead.
21691 @opindex mxl-gp-opt
21692 Use GP-relative @code{.sdata}/@code{.sbss} sections.
21694 @item -mxl-multiply-high
21695 @opindex mxl-multiply-high
21696 Use multiply high instructions for high part of 32x32 multiply.
21698 @item -mxl-float-convert
21699 @opindex mxl-float-convert
21700 Use hardware floating-point conversion instructions.
21702 @item -mxl-float-sqrt
21703 @opindex mxl-float-sqrt
21704 Use hardware floating-point square root instruction.
21707 @opindex mbig-endian
21708 Generate code for a big-endian target.
21710 @item -mlittle-endian
21711 @opindex mlittle-endian
21712 Generate code for a little-endian target.
21715 @opindex mxl-reorder
21716 Use reorder instructions (swap and byte reversed load/store).
21718 @item -mxl-mode-@var{app-model}
21719 Select application model @var{app-model}. Valid models are
21722 normal executable (default), uses startup code @file{crt0.o}.
21724 @item -mpic-data-is-text-relative
21725 @opindex mpic-data-is-text-relative
21726 Assume that the displacement between the text and data segments is fixed
21727 at static link time. This allows data to be referenced by offset from start of
21728 text address instead of GOT since PC-relative addressing is not supported.
21731 for use with Xilinx Microprocessor Debugger (XMD) based
21732 software intrusive debug agent called xmdstub. This uses startup file
21733 @file{crt1.o} and sets the start address of the program to 0x800.
21736 for applications that are loaded using a bootloader.
21737 This model uses startup file @file{crt2.o} which does not contain a processor
21738 reset vector handler. This is suitable for transferring control on a
21739 processor reset to the bootloader rather than the application.
21742 for applications that do not require any of the
21743 MicroBlaze vectors. This option may be useful for applications running
21744 within a monitoring application. This model uses @file{crt3.o} as a startup file.
21747 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
21748 @option{-mxl-mode-@var{app-model}}.
21753 @subsection MIPS Options
21754 @cindex MIPS options
21760 Generate big-endian code.
21764 Generate little-endian code. This is the default for @samp{mips*el-*-*}
21767 @item -march=@var{arch}
21769 Generate code that runs on @var{arch}, which can be the name of a
21770 generic MIPS ISA, or the name of a particular processor.
21772 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
21773 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
21774 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
21775 @samp{mips64r5} and @samp{mips64r6}.
21776 The processor names are:
21777 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
21778 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
21779 @samp{5kc}, @samp{5kf},
21781 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
21782 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
21783 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
21784 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
21785 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
21786 @samp{i6400}, @samp{i6500},
21788 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a}, @samp{gs464},
21789 @samp{gs464e}, @samp{gs264e},
21791 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
21792 @samp{m5100}, @samp{m5101},
21793 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
21795 @samp{p5600}, @samp{p6600},
21796 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
21797 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r5900},
21798 @samp{r6000}, @samp{r8000},
21799 @samp{rm7000}, @samp{rm9000},
21800 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
21803 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
21804 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
21805 @samp{xlr} and @samp{xlp}.
21806 The special value @samp{from-abi} selects the
21807 most compatible architecture for the selected ABI (that is,
21808 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
21810 The native Linux/GNU toolchain also supports the value @samp{native},
21811 which selects the best architecture option for the host processor.
21812 @option{-march=native} has no effect if GCC does not recognize
21815 In processor names, a final @samp{000} can be abbreviated as @samp{k}
21816 (for example, @option{-march=r2k}). Prefixes are optional, and
21817 @samp{vr} may be written @samp{r}.
21819 Names of the form @samp{@var{n}f2_1} refer to processors with
21820 FPUs clocked at half the rate of the core, names of the form
21821 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
21822 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
21823 processors with FPUs clocked a ratio of 3:2 with respect to the core.
21824 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
21825 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
21826 accepted as synonyms for @samp{@var{n}f1_1}.
21828 GCC defines two macros based on the value of this option. The first
21829 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
21830 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
21831 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
21832 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
21833 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
21835 Note that the @code{_MIPS_ARCH} macro uses the processor names given
21836 above. In other words, it has the full prefix and does not
21837 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
21838 the macro names the resolved architecture (either @code{"mips1"} or
21839 @code{"mips3"}). It names the default architecture when no
21840 @option{-march} option is given.
21842 @item -mtune=@var{arch}
21844 Optimize for @var{arch}. Among other things, this option controls
21845 the way instructions are scheduled, and the perceived cost of arithmetic
21846 operations. The list of @var{arch} values is the same as for
21849 When this option is not used, GCC optimizes for the processor
21850 specified by @option{-march}. By using @option{-march} and
21851 @option{-mtune} together, it is possible to generate code that
21852 runs on a family of processors, but optimize the code for one
21853 particular member of that family.
21855 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
21856 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
21857 @option{-march} ones described above.
21861 Equivalent to @option{-march=mips1}.
21865 Equivalent to @option{-march=mips2}.
21869 Equivalent to @option{-march=mips3}.
21873 Equivalent to @option{-march=mips4}.
21877 Equivalent to @option{-march=mips32}.
21881 Equivalent to @option{-march=mips32r3}.
21885 Equivalent to @option{-march=mips32r5}.
21889 Equivalent to @option{-march=mips32r6}.
21893 Equivalent to @option{-march=mips64}.
21897 Equivalent to @option{-march=mips64r2}.
21901 Equivalent to @option{-march=mips64r3}.
21905 Equivalent to @option{-march=mips64r5}.
21909 Equivalent to @option{-march=mips64r6}.
21914 @opindex mno-mips16
21915 Generate (do not generate) MIPS16 code. If GCC is targeting a
21916 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
21918 MIPS16 code generation can also be controlled on a per-function basis
21919 by means of @code{mips16} and @code{nomips16} attributes.
21920 @xref{Function Attributes}, for more information.
21922 @item -mflip-mips16
21923 @opindex mflip-mips16
21924 Generate MIPS16 code on alternating functions. This option is provided
21925 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
21926 not intended for ordinary use in compiling user code.
21928 @item -minterlink-compressed
21929 @itemx -mno-interlink-compressed
21930 @opindex minterlink-compressed
21931 @opindex mno-interlink-compressed
21932 Require (do not require) that code using the standard (uncompressed) MIPS ISA
21933 be link-compatible with MIPS16 and microMIPS code, and vice versa.
21935 For example, code using the standard ISA encoding cannot jump directly
21936 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
21937 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
21938 knows that the target of the jump is not compressed.
21940 @item -minterlink-mips16
21941 @itemx -mno-interlink-mips16
21942 @opindex minterlink-mips16
21943 @opindex mno-interlink-mips16
21944 Aliases of @option{-minterlink-compressed} and
21945 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
21946 and are retained for backwards compatibility.
21958 Generate code for the given ABI@.
21960 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
21961 generates 64-bit code when you select a 64-bit architecture, but you
21962 can use @option{-mgp32} to get 32-bit code instead.
21964 For information about the O64 ABI, see
21965 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
21967 GCC supports a variant of the o32 ABI in which floating-point registers
21968 are 64 rather than 32 bits wide. You can select this combination with
21969 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
21970 and @code{mfhc1} instructions and is therefore only supported for
21971 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
21973 The register assignments for arguments and return values remain the
21974 same, but each scalar value is passed in a single 64-bit register
21975 rather than a pair of 32-bit registers. For example, scalar
21976 floating-point values are returned in @samp{$f0} only, not a
21977 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
21978 remains the same in that the even-numbered double-precision registers
21981 Two additional variants of the o32 ABI are supported to enable
21982 a transition from 32-bit to 64-bit registers. These are FPXX
21983 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
21984 The FPXX extension mandates that all code must execute correctly
21985 when run using 32-bit or 64-bit registers. The code can be interlinked
21986 with either FP32 or FP64, but not both.
21987 The FP64A extension is similar to the FP64 extension but forbids the
21988 use of odd-numbered single-precision registers. This can be used
21989 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
21990 processors and allows both FP32 and FP64A code to interlink and
21991 run in the same process without changing FPU modes.
21994 @itemx -mno-abicalls
21996 @opindex mno-abicalls
21997 Generate (do not generate) code that is suitable for SVR4-style
21998 dynamic objects. @option{-mabicalls} is the default for SVR4-based
22003 Generate (do not generate) code that is fully position-independent,
22004 and that can therefore be linked into shared libraries. This option
22005 only affects @option{-mabicalls}.
22007 All @option{-mabicalls} code has traditionally been position-independent,
22008 regardless of options like @option{-fPIC} and @option{-fpic}. However,
22009 as an extension, the GNU toolchain allows executables to use absolute
22010 accesses for locally-binding symbols. It can also use shorter GP
22011 initialization sequences and generate direct calls to locally-defined
22012 functions. This mode is selected by @option{-mno-shared}.
22014 @option{-mno-shared} depends on binutils 2.16 or higher and generates
22015 objects that can only be linked by the GNU linker. However, the option
22016 does not affect the ABI of the final executable; it only affects the ABI
22017 of relocatable objects. Using @option{-mno-shared} generally makes
22018 executables both smaller and quicker.
22020 @option{-mshared} is the default.
22026 Assume (do not assume) that the static and dynamic linkers
22027 support PLTs and copy relocations. This option only affects
22028 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
22029 has no effect without @option{-msym32}.
22031 You can make @option{-mplt} the default by configuring
22032 GCC with @option{--with-mips-plt}. The default is
22033 @option{-mno-plt} otherwise.
22039 Lift (do not lift) the usual restrictions on the size of the global
22042 GCC normally uses a single instruction to load values from the GOT@.
22043 While this is relatively efficient, it only works if the GOT
22044 is smaller than about 64k. Anything larger causes the linker
22045 to report an error such as:
22047 @cindex relocation truncated to fit (MIPS)
22049 relocation truncated to fit: R_MIPS_GOT16 foobar
22052 If this happens, you should recompile your code with @option{-mxgot}.
22053 This works with very large GOTs, although the code is also
22054 less efficient, since it takes three instructions to fetch the
22055 value of a global symbol.
22057 Note that some linkers can create multiple GOTs. If you have such a
22058 linker, you should only need to use @option{-mxgot} when a single object
22059 file accesses more than 64k's worth of GOT entries. Very few do.
22061 These options have no effect unless GCC is generating position
22066 Assume that general-purpose registers are 32 bits wide.
22070 Assume that general-purpose registers are 64 bits wide.
22074 Assume that floating-point registers are 32 bits wide.
22078 Assume that floating-point registers are 64 bits wide.
22082 Do not assume the width of floating-point registers.
22085 @opindex mhard-float
22086 Use floating-point coprocessor instructions.
22089 @opindex msoft-float
22090 Do not use floating-point coprocessor instructions. Implement
22091 floating-point calculations using library calls instead.
22095 Equivalent to @option{-msoft-float}, but additionally asserts that the
22096 program being compiled does not perform any floating-point operations.
22097 This option is presently supported only by some bare-metal MIPS
22098 configurations, where it may select a special set of libraries
22099 that lack all floating-point support (including, for example, the
22100 floating-point @code{printf} formats).
22101 If code compiled with @option{-mno-float} accidentally contains
22102 floating-point operations, it is likely to suffer a link-time
22103 or run-time failure.
22105 @item -msingle-float
22106 @opindex msingle-float
22107 Assume that the floating-point coprocessor only supports single-precision
22110 @item -mdouble-float
22111 @opindex mdouble-float
22112 Assume that the floating-point coprocessor supports double-precision
22113 operations. This is the default.
22116 @itemx -mno-odd-spreg
22117 @opindex modd-spreg
22118 @opindex mno-odd-spreg
22119 Enable the use of odd-numbered single-precision floating-point registers
22120 for the o32 ABI. This is the default for processors that are known to
22121 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
22125 @itemx -mabs=legacy
22127 @opindex mabs=legacy
22128 These options control the treatment of the special not-a-number (NaN)
22129 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
22130 @code{neg.@i{fmt}} machine instructions.
22132 By default or when @option{-mabs=legacy} is used the legacy
22133 treatment is selected. In this case these instructions are considered
22134 arithmetic and avoided where correct operation is required and the
22135 input operand might be a NaN. A longer sequence of instructions that
22136 manipulate the sign bit of floating-point datum manually is used
22137 instead unless the @option{-ffinite-math-only} option has also been
22140 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
22141 this case these instructions are considered non-arithmetic and therefore
22142 operating correctly in all cases, including in particular where the
22143 input operand is a NaN. These instructions are therefore always used
22144 for the respective operations.
22147 @itemx -mnan=legacy
22149 @opindex mnan=legacy
22150 These options control the encoding of the special not-a-number (NaN)
22151 IEEE 754 floating-point data.
22153 The @option{-mnan=legacy} option selects the legacy encoding. In this
22154 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
22155 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
22156 by the first bit of their trailing significand field being 1.
22158 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
22159 this case qNaNs are denoted by the first bit of their trailing
22160 significand field being 1, whereas sNaNs are denoted by the first bit of
22161 their trailing significand field being 0.
22163 The default is @option{-mnan=legacy} unless GCC has been configured with
22164 @option{--with-nan=2008}.
22170 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
22171 implement atomic memory built-in functions. When neither option is
22172 specified, GCC uses the instructions if the target architecture
22175 @option{-mllsc} is useful if the runtime environment can emulate the
22176 instructions and @option{-mno-llsc} can be useful when compiling for
22177 nonstandard ISAs. You can make either option the default by
22178 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
22179 respectively. @option{--with-llsc} is the default for some
22180 configurations; see the installation documentation for details.
22186 Use (do not use) revision 1 of the MIPS DSP ASE@.
22187 @xref{MIPS DSP Built-in Functions}. This option defines the
22188 preprocessor macro @code{__mips_dsp}. It also defines
22189 @code{__mips_dsp_rev} to 1.
22195 Use (do not use) revision 2 of the MIPS DSP ASE@.
22196 @xref{MIPS DSP Built-in Functions}. This option defines the
22197 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
22198 It also defines @code{__mips_dsp_rev} to 2.
22201 @itemx -mno-smartmips
22202 @opindex msmartmips
22203 @opindex mno-smartmips
22204 Use (do not use) the MIPS SmartMIPS ASE.
22206 @item -mpaired-single
22207 @itemx -mno-paired-single
22208 @opindex mpaired-single
22209 @opindex mno-paired-single
22210 Use (do not use) paired-single floating-point instructions.
22211 @xref{MIPS Paired-Single Support}. This option requires
22212 hardware floating-point support to be enabled.
22218 Use (do not use) MIPS Digital Media Extension instructions.
22219 This option can only be used when generating 64-bit code and requires
22220 hardware floating-point support to be enabled.
22225 @opindex mno-mips3d
22226 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
22227 The option @option{-mips3d} implies @option{-mpaired-single}.
22230 @itemx -mno-micromips
22231 @opindex mmicromips
22232 @opindex mno-mmicromips
22233 Generate (do not generate) microMIPS code.
22235 MicroMIPS code generation can also be controlled on a per-function basis
22236 by means of @code{micromips} and @code{nomicromips} attributes.
22237 @xref{Function Attributes}, for more information.
22243 Use (do not use) MT Multithreading instructions.
22249 Use (do not use) the MIPS MCU ASE instructions.
22255 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
22261 Use (do not use) the MIPS Virtualization (VZ) instructions.
22267 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
22273 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) instructions.
22279 Use (do not use) the MIPS Global INValidate (GINV) instructions.
22281 @item -mloongson-mmi
22282 @itemx -mno-loongson-mmi
22283 @opindex mloongson-mmi
22284 @opindex mno-loongson-mmi
22285 Use (do not use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
22287 @item -mloongson-ext
22288 @itemx -mno-loongson-ext
22289 @opindex mloongson-ext
22290 @opindex mno-loongson-ext
22291 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions.
22293 @item -mloongson-ext2
22294 @itemx -mno-loongson-ext2
22295 @opindex mloongson-ext2
22296 @opindex mno-loongson-ext2
22297 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
22301 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
22302 an explanation of the default and the way that the pointer size is
22307 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
22309 The default size of @code{int}s, @code{long}s and pointers depends on
22310 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
22311 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
22312 32-bit @code{long}s. Pointers are the same size as @code{long}s,
22313 or the same size as integer registers, whichever is smaller.
22319 Assume (do not assume) that all symbols have 32-bit values, regardless
22320 of the selected ABI@. This option is useful in combination with
22321 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
22322 to generate shorter and faster references to symbolic addresses.
22326 Put definitions of externally-visible data in a small data section
22327 if that data is no bigger than @var{num} bytes. GCC can then generate
22328 more efficient accesses to the data; see @option{-mgpopt} for details.
22330 The default @option{-G} option depends on the configuration.
22332 @item -mlocal-sdata
22333 @itemx -mno-local-sdata
22334 @opindex mlocal-sdata
22335 @opindex mno-local-sdata
22336 Extend (do not extend) the @option{-G} behavior to local data too,
22337 such as to static variables in C@. @option{-mlocal-sdata} is the
22338 default for all configurations.
22340 If the linker complains that an application is using too much small data,
22341 you might want to try rebuilding the less performance-critical parts with
22342 @option{-mno-local-sdata}. You might also want to build large
22343 libraries with @option{-mno-local-sdata}, so that the libraries leave
22344 more room for the main program.
22346 @item -mextern-sdata
22347 @itemx -mno-extern-sdata
22348 @opindex mextern-sdata
22349 @opindex mno-extern-sdata
22350 Assume (do not assume) that externally-defined data is in
22351 a small data section if the size of that data is within the @option{-G} limit.
22352 @option{-mextern-sdata} is the default for all configurations.
22354 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
22355 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
22356 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
22357 is placed in a small data section. If @var{Var} is defined by another
22358 module, you must either compile that module with a high-enough
22359 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
22360 definition. If @var{Var} is common, you must link the application
22361 with a high-enough @option{-G} setting.
22363 The easiest way of satisfying these restrictions is to compile
22364 and link every module with the same @option{-G} option. However,
22365 you may wish to build a library that supports several different
22366 small data limits. You can do this by compiling the library with
22367 the highest supported @option{-G} setting and additionally using
22368 @option{-mno-extern-sdata} to stop the library from making assumptions
22369 about externally-defined data.
22375 Use (do not use) GP-relative accesses for symbols that are known to be
22376 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
22377 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
22380 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
22381 might not hold the value of @code{_gp}. For example, if the code is
22382 part of a library that might be used in a boot monitor, programs that
22383 call boot monitor routines pass an unknown value in @code{$gp}.
22384 (In such situations, the boot monitor itself is usually compiled
22385 with @option{-G0}.)
22387 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
22388 @option{-mno-extern-sdata}.
22390 @item -membedded-data
22391 @itemx -mno-embedded-data
22392 @opindex membedded-data
22393 @opindex mno-embedded-data
22394 Allocate variables to the read-only data section first if possible, then
22395 next in the small data section if possible, otherwise in data. This gives
22396 slightly slower code than the default, but reduces the amount of RAM required
22397 when executing, and thus may be preferred for some embedded systems.
22399 @item -muninit-const-in-rodata
22400 @itemx -mno-uninit-const-in-rodata
22401 @opindex muninit-const-in-rodata
22402 @opindex mno-uninit-const-in-rodata
22403 Put uninitialized @code{const} variables in the read-only data section.
22404 This option is only meaningful in conjunction with @option{-membedded-data}.
22406 @item -mcode-readable=@var{setting}
22407 @opindex mcode-readable
22408 Specify whether GCC may generate code that reads from executable sections.
22409 There are three possible settings:
22412 @item -mcode-readable=yes
22413 Instructions may freely access executable sections. This is the
22416 @item -mcode-readable=pcrel
22417 MIPS16 PC-relative load instructions can access executable sections,
22418 but other instructions must not do so. This option is useful on 4KSc
22419 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
22420 It is also useful on processors that can be configured to have a dual
22421 instruction/data SRAM interface and that, like the M4K, automatically
22422 redirect PC-relative loads to the instruction RAM.
22424 @item -mcode-readable=no
22425 Instructions must not access executable sections. This option can be
22426 useful on targets that are configured to have a dual instruction/data
22427 SRAM interface but that (unlike the M4K) do not automatically redirect
22428 PC-relative loads to the instruction RAM.
22431 @item -msplit-addresses
22432 @itemx -mno-split-addresses
22433 @opindex msplit-addresses
22434 @opindex mno-split-addresses
22435 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
22436 relocation operators. This option has been superseded by
22437 @option{-mexplicit-relocs} but is retained for backwards compatibility.
22439 @item -mexplicit-relocs
22440 @itemx -mno-explicit-relocs
22441 @opindex mexplicit-relocs
22442 @opindex mno-explicit-relocs
22443 Use (do not use) assembler relocation operators when dealing with symbolic
22444 addresses. The alternative, selected by @option{-mno-explicit-relocs},
22445 is to use assembler macros instead.
22447 @option{-mexplicit-relocs} is the default if GCC was configured
22448 to use an assembler that supports relocation operators.
22450 @item -mcheck-zero-division
22451 @itemx -mno-check-zero-division
22452 @opindex mcheck-zero-division
22453 @opindex mno-check-zero-division
22454 Trap (do not trap) on integer division by zero.
22456 The default is @option{-mcheck-zero-division}.
22458 @item -mdivide-traps
22459 @itemx -mdivide-breaks
22460 @opindex mdivide-traps
22461 @opindex mdivide-breaks
22462 MIPS systems check for division by zero by generating either a
22463 conditional trap or a break instruction. Using traps results in
22464 smaller code, but is only supported on MIPS II and later. Also, some
22465 versions of the Linux kernel have a bug that prevents trap from
22466 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
22467 allow conditional traps on architectures that support them and
22468 @option{-mdivide-breaks} to force the use of breaks.
22470 The default is usually @option{-mdivide-traps}, but this can be
22471 overridden at configure time using @option{--with-divide=breaks}.
22472 Divide-by-zero checks can be completely disabled using
22473 @option{-mno-check-zero-division}.
22475 @item -mload-store-pairs
22476 @itemx -mno-load-store-pairs
22477 @opindex mload-store-pairs
22478 @opindex mno-load-store-pairs
22479 Enable (disable) an optimization that pairs consecutive load or store
22480 instructions to enable load/store bonding. This option is enabled by
22481 default but only takes effect when the selected architecture is known
22482 to support bonding.
22487 @opindex mno-memcpy
22488 Force (do not force) the use of @code{memcpy} for non-trivial block
22489 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
22490 most constant-sized copies.
22493 @itemx -mno-long-calls
22494 @opindex mlong-calls
22495 @opindex mno-long-calls
22496 Disable (do not disable) use of the @code{jal} instruction. Calling
22497 functions using @code{jal} is more efficient but requires the caller
22498 and callee to be in the same 256 megabyte segment.
22500 This option has no effect on abicalls code. The default is
22501 @option{-mno-long-calls}.
22507 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
22508 instructions, as provided by the R4650 ISA@.
22514 Enable (disable) use of the @code{madd} and @code{msub} integer
22515 instructions. The default is @option{-mimadd} on architectures
22516 that support @code{madd} and @code{msub} except for the 74k
22517 architecture where it was found to generate slower code.
22520 @itemx -mno-fused-madd
22521 @opindex mfused-madd
22522 @opindex mno-fused-madd
22523 Enable (disable) use of the floating-point multiply-accumulate
22524 instructions, when they are available. The default is
22525 @option{-mfused-madd}.
22527 On the R8000 CPU when multiply-accumulate instructions are used,
22528 the intermediate product is calculated to infinite precision
22529 and is not subject to the FCSR Flush to Zero bit. This may be
22530 undesirable in some circumstances. On other processors the result
22531 is numerically identical to the equivalent computation using
22532 separate multiply, add, subtract and negate instructions.
22536 Tell the MIPS assembler to not run its preprocessor over user
22537 assembler files (with a @samp{.s} suffix) when assembling them.
22540 @itemx -mno-fix-24k
22542 @opindex mno-fix-24k
22543 Work around the 24K E48 (lost data on stores during refill) errata.
22544 The workarounds are implemented by the assembler rather than by GCC@.
22547 @itemx -mno-fix-r4000
22548 @opindex mfix-r4000
22549 @opindex mno-fix-r4000
22550 Work around certain R4000 CPU errata:
22553 A double-word or a variable shift may give an incorrect result if executed
22554 immediately after starting an integer division.
22556 A double-word or a variable shift may give an incorrect result if executed
22557 while an integer multiplication is in progress.
22559 An integer division may give an incorrect result if started in a delay slot
22560 of a taken branch or a jump.
22564 @itemx -mno-fix-r4400
22565 @opindex mfix-r4400
22566 @opindex mno-fix-r4400
22567 Work around certain R4400 CPU errata:
22570 A double-word or a variable shift may give an incorrect result if executed
22571 immediately after starting an integer division.
22575 @itemx -mno-fix-r10000
22576 @opindex mfix-r10000
22577 @opindex mno-fix-r10000
22578 Work around certain R10000 errata:
22581 @code{ll}/@code{sc} sequences may not behave atomically on revisions
22582 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
22585 This option can only be used if the target architecture supports
22586 branch-likely instructions. @option{-mfix-r10000} is the default when
22587 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
22591 @itemx -mno-fix-r5900
22592 @opindex mfix-r5900
22593 Do not attempt to schedule the preceding instruction into the delay slot
22594 of a branch instruction placed at the end of a short loop of six
22595 instructions or fewer and always schedule a @code{nop} instruction there
22596 instead. The short loop bug under certain conditions causes loops to
22597 execute only once or twice, due to a hardware bug in the R5900 chip. The
22598 workaround is implemented by the assembler rather than by GCC@.
22601 @itemx -mno-fix-rm7000
22602 @opindex mfix-rm7000
22603 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
22604 workarounds are implemented by the assembler rather than by GCC@.
22607 @itemx -mno-fix-vr4120
22608 @opindex mfix-vr4120
22609 Work around certain VR4120 errata:
22612 @code{dmultu} does not always produce the correct result.
22614 @code{div} and @code{ddiv} do not always produce the correct result if one
22615 of the operands is negative.
22617 The workarounds for the division errata rely on special functions in
22618 @file{libgcc.a}. At present, these functions are only provided by
22619 the @code{mips64vr*-elf} configurations.
22621 Other VR4120 errata require a NOP to be inserted between certain pairs of
22622 instructions. These errata are handled by the assembler, not by GCC itself.
22625 @opindex mfix-vr4130
22626 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
22627 workarounds are implemented by the assembler rather than by GCC,
22628 although GCC avoids using @code{mflo} and @code{mfhi} if the
22629 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
22630 instructions are available instead.
22633 @itemx -mno-fix-sb1
22635 Work around certain SB-1 CPU core errata.
22636 (This flag currently works around the SB-1 revision 2
22637 ``F1'' and ``F2'' floating-point errata.)
22639 @item -mr10k-cache-barrier=@var{setting}
22640 @opindex mr10k-cache-barrier
22641 Specify whether GCC should insert cache barriers to avoid the
22642 side effects of speculation on R10K processors.
22644 In common with many processors, the R10K tries to predict the outcome
22645 of a conditional branch and speculatively executes instructions from
22646 the ``taken'' branch. It later aborts these instructions if the
22647 predicted outcome is wrong. However, on the R10K, even aborted
22648 instructions can have side effects.
22650 This problem only affects kernel stores and, depending on the system,
22651 kernel loads. As an example, a speculatively-executed store may load
22652 the target memory into cache and mark the cache line as dirty, even if
22653 the store itself is later aborted. If a DMA operation writes to the
22654 same area of memory before the ``dirty'' line is flushed, the cached
22655 data overwrites the DMA-ed data. See the R10K processor manual
22656 for a full description, including other potential problems.
22658 One workaround is to insert cache barrier instructions before every memory
22659 access that might be speculatively executed and that might have side
22660 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
22661 controls GCC's implementation of this workaround. It assumes that
22662 aborted accesses to any byte in the following regions does not have
22667 the memory occupied by the current function's stack frame;
22670 the memory occupied by an incoming stack argument;
22673 the memory occupied by an object with a link-time-constant address.
22676 It is the kernel's responsibility to ensure that speculative
22677 accesses to these regions are indeed safe.
22679 If the input program contains a function declaration such as:
22685 then the implementation of @code{foo} must allow @code{j foo} and
22686 @code{jal foo} to be executed speculatively. GCC honors this
22687 restriction for functions it compiles itself. It expects non-GCC
22688 functions (such as hand-written assembly code) to do the same.
22690 The option has three forms:
22693 @item -mr10k-cache-barrier=load-store
22694 Insert a cache barrier before a load or store that might be
22695 speculatively executed and that might have side effects even
22698 @item -mr10k-cache-barrier=store
22699 Insert a cache barrier before a store that might be speculatively
22700 executed and that might have side effects even if aborted.
22702 @item -mr10k-cache-barrier=none
22703 Disable the insertion of cache barriers. This is the default setting.
22706 @item -mflush-func=@var{func}
22707 @itemx -mno-flush-func
22708 @opindex mflush-func
22709 Specifies the function to call to flush the I and D caches, or to not
22710 call any such function. If called, the function must take the same
22711 arguments as the common @code{_flush_func}, that is, the address of the
22712 memory range for which the cache is being flushed, the size of the
22713 memory range, and the number 3 (to flush both caches). The default
22714 depends on the target GCC was configured for, but commonly is either
22715 @code{_flush_func} or @code{__cpu_flush}.
22717 @item mbranch-cost=@var{num}
22718 @opindex mbranch-cost
22719 Set the cost of branches to roughly @var{num} ``simple'' instructions.
22720 This cost is only a heuristic and is not guaranteed to produce
22721 consistent results across releases. A zero cost redundantly selects
22722 the default, which is based on the @option{-mtune} setting.
22724 @item -mbranch-likely
22725 @itemx -mno-branch-likely
22726 @opindex mbranch-likely
22727 @opindex mno-branch-likely
22728 Enable or disable use of Branch Likely instructions, regardless of the
22729 default for the selected architecture. By default, Branch Likely
22730 instructions may be generated if they are supported by the selected
22731 architecture. An exception is for the MIPS32 and MIPS64 architectures
22732 and processors that implement those architectures; for those, Branch
22733 Likely instructions are not be generated by default because the MIPS32
22734 and MIPS64 architectures specifically deprecate their use.
22736 @item -mcompact-branches=never
22737 @itemx -mcompact-branches=optimal
22738 @itemx -mcompact-branches=always
22739 @opindex mcompact-branches=never
22740 @opindex mcompact-branches=optimal
22741 @opindex mcompact-branches=always
22742 These options control which form of branches will be generated. The
22743 default is @option{-mcompact-branches=optimal}.
22745 The @option{-mcompact-branches=never} option ensures that compact branch
22746 instructions will never be generated.
22748 The @option{-mcompact-branches=always} option ensures that a compact
22749 branch instruction will be generated if available. If a compact branch
22750 instruction is not available, a delay slot form of the branch will be
22753 This option is supported from MIPS Release 6 onwards.
22755 The @option{-mcompact-branches=optimal} option will cause a delay slot
22756 branch to be used if one is available in the current ISA and the delay
22757 slot is successfully filled. If the delay slot is not filled, a compact
22758 branch will be chosen if one is available.
22760 @item -mfp-exceptions
22761 @itemx -mno-fp-exceptions
22762 @opindex mfp-exceptions
22763 Specifies whether FP exceptions are enabled. This affects how
22764 FP instructions are scheduled for some processors.
22765 The default is that FP exceptions are
22768 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
22769 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
22772 @item -mvr4130-align
22773 @itemx -mno-vr4130-align
22774 @opindex mvr4130-align
22775 The VR4130 pipeline is two-way superscalar, but can only issue two
22776 instructions together if the first one is 8-byte aligned. When this
22777 option is enabled, GCC aligns pairs of instructions that it
22778 thinks should execute in parallel.
22780 This option only has an effect when optimizing for the VR4130.
22781 It normally makes code faster, but at the expense of making it bigger.
22782 It is enabled by default at optimization level @option{-O3}.
22787 Enable (disable) generation of @code{synci} instructions on
22788 architectures that support it. The @code{synci} instructions (if
22789 enabled) are generated when @code{__builtin___clear_cache} is
22792 This option defaults to @option{-mno-synci}, but the default can be
22793 overridden by configuring GCC with @option{--with-synci}.
22795 When compiling code for single processor systems, it is generally safe
22796 to use @code{synci}. However, on many multi-core (SMP) systems, it
22797 does not invalidate the instruction caches on all cores and may lead
22798 to undefined behavior.
22800 @item -mrelax-pic-calls
22801 @itemx -mno-relax-pic-calls
22802 @opindex mrelax-pic-calls
22803 Try to turn PIC calls that are normally dispatched via register
22804 @code{$25} into direct calls. This is only possible if the linker can
22805 resolve the destination at link time and if the destination is within
22806 range for a direct call.
22808 @option{-mrelax-pic-calls} is the default if GCC was configured to use
22809 an assembler and a linker that support the @code{.reloc} assembly
22810 directive and @option{-mexplicit-relocs} is in effect. With
22811 @option{-mno-explicit-relocs}, this optimization can be performed by the
22812 assembler and the linker alone without help from the compiler.
22814 @item -mmcount-ra-address
22815 @itemx -mno-mcount-ra-address
22816 @opindex mmcount-ra-address
22817 @opindex mno-mcount-ra-address
22818 Emit (do not emit) code that allows @code{_mcount} to modify the
22819 calling function's return address. When enabled, this option extends
22820 the usual @code{_mcount} interface with a new @var{ra-address}
22821 parameter, which has type @code{intptr_t *} and is passed in register
22822 @code{$12}. @code{_mcount} can then modify the return address by
22823 doing both of the following:
22826 Returning the new address in register @code{$31}.
22828 Storing the new address in @code{*@var{ra-address}},
22829 if @var{ra-address} is nonnull.
22832 The default is @option{-mno-mcount-ra-address}.
22834 @item -mframe-header-opt
22835 @itemx -mno-frame-header-opt
22836 @opindex mframe-header-opt
22837 Enable (disable) frame header optimization in the o32 ABI. When using the
22838 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
22839 function to write out register arguments. When enabled, this optimization
22840 will suppress the allocation of the frame header if it can be determined that
22843 This optimization is off by default at all optimization levels.
22846 @itemx -mno-lxc1-sxc1
22847 @opindex mlxc1-sxc1
22848 When applicable, enable (disable) the generation of @code{lwxc1},
22849 @code{swxc1}, @code{ldxc1}, @code{sdxc1} instructions. Enabled by default.
22854 When applicable, enable (disable) the generation of 4-operand @code{madd.s},
22855 @code{madd.d} and related instructions. Enabled by default.
22860 @subsection MMIX Options
22861 @cindex MMIX Options
22863 These options are defined for the MMIX:
22867 @itemx -mno-libfuncs
22869 @opindex mno-libfuncs
22870 Specify that intrinsic library functions are being compiled, passing all
22871 values in registers, no matter the size.
22874 @itemx -mno-epsilon
22876 @opindex mno-epsilon
22877 Generate floating-point comparison instructions that compare with respect
22878 to the @code{rE} epsilon register.
22880 @item -mabi=mmixware
22882 @opindex mabi=mmixware
22884 Generate code that passes function parameters and return values that (in
22885 the called function) are seen as registers @code{$0} and up, as opposed to
22886 the GNU ABI which uses global registers @code{$231} and up.
22888 @item -mzero-extend
22889 @itemx -mno-zero-extend
22890 @opindex mzero-extend
22891 @opindex mno-zero-extend
22892 When reading data from memory in sizes shorter than 64 bits, use (do not
22893 use) zero-extending load instructions by default, rather than
22894 sign-extending ones.
22897 @itemx -mno-knuthdiv
22899 @opindex mno-knuthdiv
22900 Make the result of a division yielding a remainder have the same sign as
22901 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
22902 remainder follows the sign of the dividend. Both methods are
22903 arithmetically valid, the latter being almost exclusively used.
22905 @item -mtoplevel-symbols
22906 @itemx -mno-toplevel-symbols
22907 @opindex mtoplevel-symbols
22908 @opindex mno-toplevel-symbols
22909 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
22910 code can be used with the @code{PREFIX} assembly directive.
22914 Generate an executable in the ELF format, rather than the default
22915 @samp{mmo} format used by the @command{mmix} simulator.
22917 @item -mbranch-predict
22918 @itemx -mno-branch-predict
22919 @opindex mbranch-predict
22920 @opindex mno-branch-predict
22921 Use (do not use) the probable-branch instructions, when static branch
22922 prediction indicates a probable branch.
22924 @item -mbase-addresses
22925 @itemx -mno-base-addresses
22926 @opindex mbase-addresses
22927 @opindex mno-base-addresses
22928 Generate (do not generate) code that uses @emph{base addresses}. Using a
22929 base address automatically generates a request (handled by the assembler
22930 and the linker) for a constant to be set up in a global register. The
22931 register is used for one or more base address requests within the range 0
22932 to 255 from the value held in the register. The generally leads to short
22933 and fast code, but the number of different data items that can be
22934 addressed is limited. This means that a program that uses lots of static
22935 data may require @option{-mno-base-addresses}.
22937 @item -msingle-exit
22938 @itemx -mno-single-exit
22939 @opindex msingle-exit
22940 @opindex mno-single-exit
22941 Force (do not force) generated code to have a single exit point in each
22945 @node MN10300 Options
22946 @subsection MN10300 Options
22947 @cindex MN10300 options
22949 These @option{-m} options are defined for Matsushita MN10300 architectures:
22954 Generate code to avoid bugs in the multiply instructions for the MN10300
22955 processors. This is the default.
22957 @item -mno-mult-bug
22958 @opindex mno-mult-bug
22959 Do not generate code to avoid bugs in the multiply instructions for the
22960 MN10300 processors.
22964 Generate code using features specific to the AM33 processor.
22968 Do not generate code using features specific to the AM33 processor. This
22973 Generate code using features specific to the AM33/2.0 processor.
22977 Generate code using features specific to the AM34 processor.
22979 @item -mtune=@var{cpu-type}
22981 Use the timing characteristics of the indicated CPU type when
22982 scheduling instructions. This does not change the targeted processor
22983 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
22984 @samp{am33-2} or @samp{am34}.
22986 @item -mreturn-pointer-on-d0
22987 @opindex mreturn-pointer-on-d0
22988 When generating a function that returns a pointer, return the pointer
22989 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
22990 only in @code{a0}, and attempts to call such functions without a prototype
22991 result in errors. Note that this option is on by default; use
22992 @option{-mno-return-pointer-on-d0} to disable it.
22996 Do not link in the C run-time initialization object file.
23000 Indicate to the linker that it should perform a relaxation optimization pass
23001 to shorten branches, calls and absolute memory addresses. This option only
23002 has an effect when used on the command line for the final link step.
23004 This option makes symbolic debugging impossible.
23008 Allow the compiler to generate @emph{Long Instruction Word}
23009 instructions if the target is the @samp{AM33} or later. This is the
23010 default. This option defines the preprocessor macro @code{__LIW__}.
23014 Do not allow the compiler to generate @emph{Long Instruction Word}
23015 instructions. This option defines the preprocessor macro
23020 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
23021 instructions if the target is the @samp{AM33} or later. This is the
23022 default. This option defines the preprocessor macro @code{__SETLB__}.
23026 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
23027 instructions. This option defines the preprocessor macro
23028 @code{__NO_SETLB__}.
23032 @node Moxie Options
23033 @subsection Moxie Options
23034 @cindex Moxie Options
23040 Generate big-endian code. This is the default for @samp{moxie-*-*}
23045 Generate little-endian code.
23049 Generate mul.x and umul.x instructions. This is the default for
23050 @samp{moxiebox-*-*} configurations.
23054 Do not link in the C run-time initialization object file.
23058 @node MSP430 Options
23059 @subsection MSP430 Options
23060 @cindex MSP430 Options
23062 These options are defined for the MSP430:
23068 Force assembly output to always use hex constants. Normally such
23069 constants are signed decimals, but this option is available for
23070 testsuite and/or aesthetic purposes.
23074 Select the MCU to target. This is used to create a C preprocessor
23075 symbol based upon the MCU name, converted to upper case and pre- and
23076 post-fixed with @samp{__}. This in turn is used by the
23077 @file{msp430.h} header file to select an MCU-specific supplementary
23080 The option also sets the ISA to use. If the MCU name is one that is
23081 known to only support the 430 ISA then that is selected, otherwise the
23082 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
23083 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
23084 name selects the 430X ISA.
23086 In addition an MCU-specific linker script is added to the linker
23087 command line. The script's name is the name of the MCU with
23088 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
23089 command line defines the C preprocessor symbol @code{__XXX__} and
23090 cause the linker to search for a script called @file{xxx.ld}.
23092 The ISA and hardware multiply supported for the different MCUs is hard-coded
23093 into GCC. However, an external @samp{devices.csv} file can be used to
23094 extend device support beyond those that have been hard-coded.
23096 GCC searches for the @samp{devices.csv} file on the paths specified
23097 with the @code{-I} and @code{-L} options.
23100 @itemx -mno-warn-mcu
23102 @opindex mno-warn-mcu
23103 This option enables or disables warnings about conflicts between the
23104 MCU name specified by the @option{-mmcu} option and the ISA set by the
23105 @option{-mcpu} option and/or the hardware multiply support set by the
23106 @option{-mhwmult} option. It also toggles warnings about unrecognized
23107 MCU names. This option is on by default.
23111 Specifies the ISA to use. Accepted values are @samp{msp430},
23112 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
23113 @option{-mmcu=} option should be used to select the ISA.
23117 Link to the simulator runtime libraries and linker script. Overrides
23118 any scripts that would be selected by the @option{-mmcu=} option.
23122 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
23126 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
23130 This option is passed to the assembler and linker, and allows the
23131 linker to perform certain optimizations that cannot be done until
23136 Describes the type of hardware multiply supported by the target.
23137 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
23138 for the original 16-bit-only multiply supported by early MCUs.
23139 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
23140 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
23141 A value of @samp{auto} can also be given. This tells GCC to deduce
23142 the hardware multiply support based upon the MCU name provided by the
23143 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
23144 the MCU name is not recognized then no hardware multiply support is
23145 assumed. @code{auto} is the default setting.
23147 Hardware multiplies are normally performed by calling a library
23148 routine. This saves space in the generated code. When compiling at
23149 @option{-O3} or higher however the hardware multiplier is invoked
23150 inline. This makes for bigger, but faster code.
23152 The hardware multiply routines disable interrupts whilst running and
23153 restore the previous interrupt state when they finish. This makes
23154 them safe to use inside interrupt handlers as well as in normal code.
23158 Enable the use of a minimum runtime environment - no static
23159 initializers or constructors. This is intended for memory-constrained
23160 devices. The compiler includes special symbols in some objects
23161 that tell the linker and runtime which code fragments are required.
23163 @item -mcode-region=
23164 @itemx -mdata-region=
23165 @opindex mcode-region
23166 @opindex mdata-region
23167 These options tell the compiler where to place functions and data that
23168 do not have one of the @code{lower}, @code{upper}, @code{either} or
23169 @code{section} attributes. Possible values are @code{lower},
23170 @code{upper}, @code{either} or @code{any}. The first three behave
23171 like the corresponding attribute. The fourth possible value -
23172 @code{any} - is the default. It leaves placement entirely up to the
23173 linker script and how it assigns the standard sections
23174 (@code{.text}, @code{.data}, etc) to the memory regions.
23176 @item -msilicon-errata=
23177 @opindex msilicon-errata
23178 This option passes on a request to assembler to enable the fixes for
23179 the named silicon errata.
23181 @item -msilicon-errata-warn=
23182 @opindex msilicon-errata-warn
23183 This option passes on a request to the assembler to enable warning
23184 messages when a silicon errata might need to be applied.
23186 @item -mwarn-devices-csv
23187 @itemx -mno-warn-devices-csv
23188 @opindex mwarn-devices-csv
23189 @opindex mno-warn-devices-csv
23190 Warn if @samp{devices.csv} is not found or there are problem parsing it
23195 @node NDS32 Options
23196 @subsection NDS32 Options
23197 @cindex NDS32 Options
23199 These options are defined for NDS32 implementations:
23204 @opindex mbig-endian
23205 Generate code in big-endian mode.
23207 @item -mlittle-endian
23208 @opindex mlittle-endian
23209 Generate code in little-endian mode.
23211 @item -mreduced-regs
23212 @opindex mreduced-regs
23213 Use reduced-set registers for register allocation.
23216 @opindex mfull-regs
23217 Use full-set registers for register allocation.
23221 Generate conditional move instructions.
23225 Do not generate conditional move instructions.
23229 Generate performance extension instructions.
23231 @item -mno-ext-perf
23232 @opindex mno-ext-perf
23233 Do not generate performance extension instructions.
23236 @opindex mext-perf2
23237 Generate performance extension 2 instructions.
23239 @item -mno-ext-perf2
23240 @opindex mno-ext-perf2
23241 Do not generate performance extension 2 instructions.
23244 @opindex mext-string
23245 Generate string extension instructions.
23247 @item -mno-ext-string
23248 @opindex mno-ext-string
23249 Do not generate string extension instructions.
23253 Generate v3 push25/pop25 instructions.
23256 @opindex mno-v3push
23257 Do not generate v3 push25/pop25 instructions.
23261 Generate 16-bit instructions.
23264 @opindex mno-16-bit
23265 Do not generate 16-bit instructions.
23267 @item -misr-vector-size=@var{num}
23268 @opindex misr-vector-size
23269 Specify the size of each interrupt vector, which must be 4 or 16.
23271 @item -mcache-block-size=@var{num}
23272 @opindex mcache-block-size
23273 Specify the size of each cache block,
23274 which must be a power of 2 between 4 and 512.
23276 @item -march=@var{arch}
23278 Specify the name of the target architecture.
23280 @item -mcmodel=@var{code-model}
23282 Set the code model to one of
23285 All the data and read-only data segments must be within 512KB addressing space.
23286 The text segment must be within 16MB addressing space.
23287 @item @samp{medium}
23288 The data segment must be within 512KB while the read-only data segment can be
23289 within 4GB addressing space. The text segment should be still within 16MB
23292 All the text and data segments can be within 4GB addressing space.
23296 @opindex mctor-dtor
23297 Enable constructor/destructor feature.
23301 Guide linker to relax instructions.
23305 @node Nios II Options
23306 @subsection Nios II Options
23307 @cindex Nios II options
23308 @cindex Altera Nios II options
23310 These are the options defined for the Altera Nios II processor.
23316 @cindex smaller data references
23317 Put global and static objects less than or equal to @var{num} bytes
23318 into the small data or BSS sections instead of the normal data or BSS
23319 sections. The default value of @var{num} is 8.
23321 @item -mgpopt=@var{option}
23326 Generate (do not generate) GP-relative accesses. The following
23327 @var{option} names are recognized:
23332 Do not generate GP-relative accesses.
23335 Generate GP-relative accesses for small data objects that are not
23336 external, weak, or uninitialized common symbols.
23337 Also use GP-relative addressing for objects that
23338 have been explicitly placed in a small data section via a @code{section}
23342 As for @samp{local}, but also generate GP-relative accesses for
23343 small data objects that are external, weak, or common. If you use this option,
23344 you must ensure that all parts of your program (including libraries) are
23345 compiled with the same @option{-G} setting.
23348 Generate GP-relative accesses for all data objects in the program. If you
23349 use this option, the entire data and BSS segments
23350 of your program must fit in 64K of memory and you must use an appropriate
23351 linker script to allocate them within the addressable range of the
23355 Generate GP-relative addresses for function pointers as well as data
23356 pointers. If you use this option, the entire text, data, and BSS segments
23357 of your program must fit in 64K of memory and you must use an appropriate
23358 linker script to allocate them within the addressable range of the
23363 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
23364 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
23366 The default is @option{-mgpopt} except when @option{-fpic} or
23367 @option{-fPIC} is specified to generate position-independent code.
23368 Note that the Nios II ABI does not permit GP-relative accesses from
23371 You may need to specify @option{-mno-gpopt} explicitly when building
23372 programs that include large amounts of small data, including large
23373 GOT data sections. In this case, the 16-bit offset for GP-relative
23374 addressing may not be large enough to allow access to the entire
23375 small data section.
23377 @item -mgprel-sec=@var{regexp}
23378 @opindex mgprel-sec
23379 This option specifies additional section names that can be accessed via
23380 GP-relative addressing. It is most useful in conjunction with
23381 @code{section} attributes on variable declarations
23382 (@pxref{Common Variable Attributes}) and a custom linker script.
23383 The @var{regexp} is a POSIX Extended Regular Expression.
23385 This option does not affect the behavior of the @option{-G} option, and
23386 the specified sections are in addition to the standard @code{.sdata}
23387 and @code{.sbss} small-data sections that are recognized by @option{-mgpopt}.
23389 @item -mr0rel-sec=@var{regexp}
23390 @opindex mr0rel-sec
23391 This option specifies names of sections that can be accessed via a
23392 16-bit offset from @code{r0}; that is, in the low 32K or high 32K
23393 of the 32-bit address space. It is most useful in conjunction with
23394 @code{section} attributes on variable declarations
23395 (@pxref{Common Variable Attributes}) and a custom linker script.
23396 The @var{regexp} is a POSIX Extended Regular Expression.
23398 In contrast to the use of GP-relative addressing for small data,
23399 zero-based addressing is never generated by default and there are no
23400 conventional section names used in standard linker scripts for sections
23401 in the low or high areas of memory.
23407 Generate little-endian (default) or big-endian (experimental) code,
23410 @item -march=@var{arch}
23412 This specifies the name of the target Nios II architecture. GCC uses this
23413 name to determine what kind of instructions it can emit when generating
23414 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
23416 The preprocessor macro @code{__nios2_arch__} is available to programs,
23417 with value 1 or 2, indicating the targeted ISA level.
23419 @item -mbypass-cache
23420 @itemx -mno-bypass-cache
23421 @opindex mno-bypass-cache
23422 @opindex mbypass-cache
23423 Force all load and store instructions to always bypass cache by
23424 using I/O variants of the instructions. The default is not to
23427 @item -mno-cache-volatile
23428 @itemx -mcache-volatile
23429 @opindex mcache-volatile
23430 @opindex mno-cache-volatile
23431 Volatile memory access bypass the cache using the I/O variants of
23432 the load and store instructions. The default is not to bypass the cache.
23434 @item -mno-fast-sw-div
23435 @itemx -mfast-sw-div
23436 @opindex mno-fast-sw-div
23437 @opindex mfast-sw-div
23438 Do not use table-based fast divide for small numbers. The default
23439 is to use the fast divide at @option{-O3} and above.
23443 @itemx -mno-hw-mulx
23447 @opindex mno-hw-mul
23449 @opindex mno-hw-mulx
23451 @opindex mno-hw-div
23453 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
23454 instructions by the compiler. The default is to emit @code{mul}
23455 and not emit @code{div} and @code{mulx}.
23461 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
23462 CDX (code density) instructions. Enabling these instructions also
23463 requires @option{-march=r2}. Since these instructions are optional
23464 extensions to the R2 architecture, the default is not to emit them.
23466 @item -mcustom-@var{insn}=@var{N}
23467 @itemx -mno-custom-@var{insn}
23468 @opindex mcustom-@var{insn}
23469 @opindex mno-custom-@var{insn}
23470 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
23471 custom instruction with encoding @var{N} when generating code that uses
23472 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
23473 instruction 253 for single-precision floating-point add operations instead
23474 of the default behavior of using a library call.
23476 The following values of @var{insn} are supported. Except as otherwise
23477 noted, floating-point operations are expected to be implemented with
23478 normal IEEE 754 semantics and correspond directly to the C operators or the
23479 equivalent GCC built-in functions (@pxref{Other Builtins}).
23481 Single-precision floating point:
23484 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
23485 Binary arithmetic operations.
23491 Unary absolute value.
23493 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
23494 Comparison operations.
23496 @item @samp{fmins}, @samp{fmaxs}
23497 Floating-point minimum and maximum. These instructions are only
23498 generated if @option{-ffinite-math-only} is specified.
23500 @item @samp{fsqrts}
23501 Unary square root operation.
23503 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
23504 Floating-point trigonometric and exponential functions. These instructions
23505 are only generated if @option{-funsafe-math-optimizations} is also specified.
23509 Double-precision floating point:
23512 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
23513 Binary arithmetic operations.
23519 Unary absolute value.
23521 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
23522 Comparison operations.
23524 @item @samp{fmind}, @samp{fmaxd}
23525 Double-precision minimum and maximum. These instructions are only
23526 generated if @option{-ffinite-math-only} is specified.
23528 @item @samp{fsqrtd}
23529 Unary square root operation.
23531 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
23532 Double-precision trigonometric and exponential functions. These instructions
23533 are only generated if @option{-funsafe-math-optimizations} is also specified.
23539 @item @samp{fextsd}
23540 Conversion from single precision to double precision.
23542 @item @samp{ftruncds}
23543 Conversion from double precision to single precision.
23545 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
23546 Conversion from floating point to signed or unsigned integer types, with
23547 truncation towards zero.
23550 Conversion from single-precision floating point to signed integer,
23551 rounding to the nearest integer and ties away from zero.
23552 This corresponds to the @code{__builtin_lroundf} function when
23553 @option{-fno-math-errno} is used.
23555 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
23556 Conversion from signed or unsigned integer types to floating-point types.
23560 In addition, all of the following transfer instructions for internal
23561 registers X and Y must be provided to use any of the double-precision
23562 floating-point instructions. Custom instructions taking two
23563 double-precision source operands expect the first operand in the
23564 64-bit register X. The other operand (or only operand of a unary
23565 operation) is given to the custom arithmetic instruction with the
23566 least significant half in source register @var{src1} and the most
23567 significant half in @var{src2}. A custom instruction that returns a
23568 double-precision result returns the most significant 32 bits in the
23569 destination register and the other half in 32-bit register Y.
23570 GCC automatically generates the necessary code sequences to write
23571 register X and/or read register Y when double-precision floating-point
23572 instructions are used.
23577 Write @var{src1} into the least significant half of X and @var{src2} into
23578 the most significant half of X.
23581 Write @var{src1} into Y.
23583 @item @samp{frdxhi}, @samp{frdxlo}
23584 Read the most or least (respectively) significant half of X and store it in
23588 Read the value of Y and store it into @var{dest}.
23591 Note that you can gain more local control over generation of Nios II custom
23592 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
23593 and @code{target("no-custom-@var{insn}")} function attributes
23594 (@pxref{Function Attributes})
23595 or pragmas (@pxref{Function Specific Option Pragmas}).
23597 @item -mcustom-fpu-cfg=@var{name}
23598 @opindex mcustom-fpu-cfg
23600 This option enables a predefined, named set of custom instruction encodings
23601 (see @option{-mcustom-@var{insn}} above).
23602 Currently, the following sets are defined:
23604 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
23605 @gccoptlist{-mcustom-fmuls=252 @gol
23606 -mcustom-fadds=253 @gol
23607 -mcustom-fsubs=254 @gol
23608 -fsingle-precision-constant}
23610 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
23611 @gccoptlist{-mcustom-fmuls=252 @gol
23612 -mcustom-fadds=253 @gol
23613 -mcustom-fsubs=254 @gol
23614 -mcustom-fdivs=255 @gol
23615 -fsingle-precision-constant}
23617 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
23618 @gccoptlist{-mcustom-floatus=243 @gol
23619 -mcustom-fixsi=244 @gol
23620 -mcustom-floatis=245 @gol
23621 -mcustom-fcmpgts=246 @gol
23622 -mcustom-fcmples=249 @gol
23623 -mcustom-fcmpeqs=250 @gol
23624 -mcustom-fcmpnes=251 @gol
23625 -mcustom-fmuls=252 @gol
23626 -mcustom-fadds=253 @gol
23627 -mcustom-fsubs=254 @gol
23628 -mcustom-fdivs=255 @gol
23629 -fsingle-precision-constant}
23631 Custom instruction assignments given by individual
23632 @option{-mcustom-@var{insn}=} options override those given by
23633 @option{-mcustom-fpu-cfg=}, regardless of the
23634 order of the options on the command line.
23636 Note that you can gain more local control over selection of a FPU
23637 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
23638 function attribute (@pxref{Function Attributes})
23639 or pragma (@pxref{Function Specific Option Pragmas}).
23643 These additional @samp{-m} options are available for the Altera Nios II
23644 ELF (bare-metal) target:
23650 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
23651 startup and termination code, and is typically used in conjunction with
23652 @option{-msys-crt0=} to specify the location of the alternate startup code
23653 provided by the HAL BSP.
23657 Link with a limited version of the C library, @option{-lsmallc}, rather than
23660 @item -msys-crt0=@var{startfile}
23662 @var{startfile} is the file name of the startfile (crt0) to use
23663 when linking. This option is only useful in conjunction with @option{-mhal}.
23665 @item -msys-lib=@var{systemlib}
23667 @var{systemlib} is the library name of the library that provides
23668 low-level system calls required by the C library,
23669 e.g.@: @code{read} and @code{write}.
23670 This option is typically used to link with a library provided by a HAL BSP.
23674 @node Nvidia PTX Options
23675 @subsection Nvidia PTX Options
23676 @cindex Nvidia PTX options
23677 @cindex nvptx options
23679 These options are defined for Nvidia PTX:
23687 Generate code for 32-bit or 64-bit ABI.
23689 @item -misa=@var{ISA-string}
23691 Generate code for given the specified PTX ISA (e.g.@: @samp{sm_35}). ISA
23692 strings must be lower-case. Valid ISA strings include @samp{sm_30} and
23693 @samp{sm_35}. The default ISA is sm_30.
23696 @opindex mmainkernel
23697 Link in code for a __main kernel. This is for stand-alone instead of
23698 offloading execution.
23702 Apply partitioned execution optimizations. This is the default when any
23703 level of optimization is selected.
23706 @opindex msoft-stack
23707 Generate code that does not use @code{.local} memory
23708 directly for stack storage. Instead, a per-warp stack pointer is
23709 maintained explicitly. This enables variable-length stack allocation (with
23710 variable-length arrays or @code{alloca}), and when global memory is used for
23711 underlying storage, makes it possible to access automatic variables from other
23712 threads, or with atomic instructions. This code generation variant is used
23713 for OpenMP offloading, but the option is exposed on its own for the purpose
23714 of testing the compiler; to generate code suitable for linking into programs
23715 using OpenMP offloading, use option @option{-mgomp}.
23717 @item -muniform-simt
23718 @opindex muniform-simt
23719 Switch to code generation variant that allows to execute all threads in each
23720 warp, while maintaining memory state and side effects as if only one thread
23721 in each warp was active outside of OpenMP SIMD regions. All atomic operations
23722 and calls to runtime (malloc, free, vprintf) are conditionally executed (iff
23723 current lane index equals the master lane index), and the register being
23724 assigned is copied via a shuffle instruction from the master lane. Outside of
23725 SIMD regions lane 0 is the master; inside, each thread sees itself as the
23726 master. Shared memory array @code{int __nvptx_uni[]} stores all-zeros or
23727 all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD
23728 regions). Each thread can bitwise-and the bitmask at position @code{tid.y}
23729 with current lane index to compute the master lane index.
23733 Generate code for use in OpenMP offloading: enables @option{-msoft-stack} and
23734 @option{-muniform-simt} options, and selects corresponding multilib variant.
23738 @node OpenRISC Options
23739 @subsection OpenRISC Options
23740 @cindex OpenRISC Options
23742 These options are defined for OpenRISC:
23746 @item -mboard=@var{name}
23748 Configure a board specific runtime. This will be passed to the linker for
23749 newlib board library linking. The default is @code{or1ksim}.
23753 This option is ignored; it is for compatibility purposes only. This used to
23754 select linker and preprocessor options for use with newlib.
23760 Select software or hardware divide (@code{l.div}, @code{l.divu}) instructions.
23761 This default is hardware divide.
23767 Select software or hardware multiply (@code{l.mul}, @code{l.muli}) instructions.
23768 This default is hardware multiply.
23771 @itemx -mhard-float
23772 @opindex msoft-float
23773 @opindex mhard-float
23774 Select software or hardware for floating point operations.
23775 The default is software.
23777 @item -mdouble-float
23778 @opindex mdouble-float
23779 When @option{-mhard-float} is selected, enables generation of double-precision
23780 floating point instructions. By default functions from @file{libgcc} are used
23781 to perform double-precision floating point operations.
23783 @item -munordered-float
23784 @opindex munordered-float
23785 When @option{-mhard-float} is selected, enables generation of unordered
23786 floating point compare and set flag (@code{lf.sfun*}) instructions. By default
23787 functions from @file{libgcc} are used to perform unordered floating point
23788 compare and set flag operations.
23792 Enable generation of conditional move (@code{l.cmov}) instructions. By
23793 default the equivalent will be generated using using set and branch.
23797 Enable generation of rotate right (@code{l.ror}) instructions. By default
23798 functions from @file{libgcc} are used to perform rotate right operations.
23802 Enable generation of rotate right with immediate (@code{l.rori}) instructions.
23803 By default functions from @file{libgcc} are used to perform rotate right with
23804 immediate operations.
23808 Enable generation of sign extension (@code{l.ext*}) instructions. By default
23809 memory loads are used to perform sign extension.
23813 Enable generation of compare and set flag with immediate (@code{l.sf*i})
23814 instructions. By default extra instructions will be generated to store the
23815 immediate to a register first.
23819 Enable generation of shift with immediate (@code{l.srai}, @code{l.srli},
23820 @code{l.slli}) instructions. By default extra instructions will be generated
23821 to store the immediate to a register first.
23826 @node PDP-11 Options
23827 @subsection PDP-11 Options
23828 @cindex PDP-11 Options
23830 These options are defined for the PDP-11:
23835 Use hardware FPP floating point. This is the default. (FIS floating
23836 point on the PDP-11/40 is not supported.) Implies -m45.
23839 @opindex msoft-float
23840 Do not use hardware floating point.
23844 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
23848 Return floating-point results in memory. This is the default.
23852 Generate code for a PDP-11/40. Implies -msoft-float -mno-split.
23856 Generate code for a PDP-11/45. This is the default.
23860 Generate code for a PDP-11/10. Implies -msoft-float -mno-split.
23866 Use 16-bit @code{int}. This is the default.
23872 Use 32-bit @code{int}.
23876 Target has split instruction and data space. Implies -m45.
23880 Use Unix assembler syntax.
23884 Use DEC assembler syntax.
23888 Use GNU assembler syntax. This is the default.
23892 Use the new LRA register allocator. By default, the old ``reload''
23896 @node picoChip Options
23897 @subsection picoChip Options
23898 @cindex picoChip options
23900 These @samp{-m} options are defined for picoChip implementations:
23904 @item -mae=@var{ae_type}
23906 Set the instruction set, register set, and instruction scheduling
23907 parameters for array element type @var{ae_type}. Supported values
23908 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
23910 @option{-mae=ANY} selects a completely generic AE type. Code
23911 generated with this option runs on any of the other AE types. The
23912 code is not as efficient as it would be if compiled for a specific
23913 AE type, and some types of operation (e.g., multiplication) do not
23914 work properly on all types of AE.
23916 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
23917 for compiled code, and is the default.
23919 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
23920 option may suffer from poor performance of byte (char) manipulation,
23921 since the DSP AE does not provide hardware support for byte load/stores.
23923 @item -msymbol-as-address
23924 Enable the compiler to directly use a symbol name as an address in a
23925 load/store instruction, without first loading it into a
23926 register. Typically, the use of this option generates larger
23927 programs, which run faster than when the option isn't used. However, the
23928 results vary from program to program, so it is left as a user option,
23929 rather than being permanently enabled.
23931 @item -mno-inefficient-warnings
23932 Disables warnings about the generation of inefficient code. These
23933 warnings can be generated, for example, when compiling code that
23934 performs byte-level memory operations on the MAC AE type. The MAC AE has
23935 no hardware support for byte-level memory operations, so all byte
23936 load/stores must be synthesized from word load/store operations. This is
23937 inefficient and a warning is generated to indicate
23938 that you should rewrite the code to avoid byte operations, or to target
23939 an AE type that has the necessary hardware support. This option disables
23944 @node PowerPC Options
23945 @subsection PowerPC Options
23946 @cindex PowerPC options
23948 These are listed under @xref{RS/6000 and PowerPC Options}.
23951 @subsection PRU Options
23952 @cindex PRU Options
23954 These command-line options are defined for PRU target:
23959 Link with a minimum runtime environment, with no support for static
23960 initializers and constructors. Using this option can significantly reduce
23961 the size of the final ELF binary. Beware that the compiler could still
23962 generate code with static initializers and constructors. It is up to the
23963 programmer to ensure that the source program will not use those features.
23965 @item -mmcu=@var{mcu}
23967 Specify the PRU MCU variant to use. Check Newlib for the exact list of
23972 Make GCC pass the @option{--no-relax} command-line option to the linker
23973 instead of the @option{--relax} option.
23977 Allow (or do not allow) GCC to use the LOOP instruction.
23979 @item -mabi=@var{variant}
23981 Specify the ABI variant to output code for. @option{-mabi=ti} selects the
23982 unmodified TI ABI while @option{-mabi=gnu} selects a GNU variant that copes
23983 more naturally with certain GCC assumptions. These are the differences:
23986 @item Function Pointer Size
23987 TI ABI specifies that function (code) pointers are 16-bit, whereas GNU
23988 supports only 32-bit data and code pointers.
23990 @item Optional Return Value Pointer
23991 Function return values larger than 64 bits are passed by using a hidden
23992 pointer as the first argument of the function. TI ABI, though, mandates that
23993 the pointer can be NULL in case the caller is not using the returned value.
23994 GNU always passes and expects a valid return value pointer.
23998 The current @option{-mabi=ti} implementation simply raises a compile error
23999 when any of the above code constructs is detected. As a consequence
24000 the standard C library cannot be built and it is omitted when linking with
24003 Relaxation is a GNU feature and for safety reasons is disabled when using
24004 @option{-mabi=ti}. The TI toolchain does not emit relocations for QBBx
24005 instructions, so the GNU linker cannot adjust them when shortening adjacent
24006 LDI32 pseudo instructions.
24010 @node RISC-V Options
24011 @subsection RISC-V Options
24012 @cindex RISC-V Options
24014 These command-line options are defined for RISC-V targets:
24017 @item -mbranch-cost=@var{n}
24018 @opindex mbranch-cost
24019 Set the cost of branches to roughly @var{n} instructions.
24024 When generating PIC code, do or don't allow the use of PLTs. Ignored for
24025 non-PIC. The default is @option{-mplt}.
24027 @item -mabi=@var{ABI-string}
24029 Specify integer and floating-point calling convention. @var{ABI-string}
24030 contains two parts: the size of integer types and the registers used for
24031 floating-point types. For example @samp{-march=rv64ifd -mabi=lp64d} means that
24032 @samp{long} and pointers are 64-bit (implicitly defining @samp{int} to be
24033 32-bit), and that floating-point values up to 64 bits wide are passed in F
24034 registers. Contrast this with @samp{-march=rv64ifd -mabi=lp64f}, which still
24035 allows the compiler to generate code that uses the F and D extensions but only
24036 allows floating-point values up to 32 bits long to be passed in registers; or
24037 @samp{-march=rv64ifd -mabi=lp64}, in which no floating-point arguments will be
24038 passed in registers.
24040 The default for this argument is system dependent, users who want a specific
24041 calling convention should specify one explicitly. The valid calling
24042 conventions are: @samp{ilp32}, @samp{ilp32f}, @samp{ilp32d}, @samp{lp64},
24043 @samp{lp64f}, and @samp{lp64d}. Some calling conventions are impossible to
24044 implement on some ISAs: for example, @samp{-march=rv32if -mabi=ilp32d} is
24045 invalid because the ABI requires 64-bit values be passed in F registers, but F
24046 registers are only 32 bits wide. There is also the @samp{ilp32e} ABI that can
24047 only be used with the @samp{rv32e} architecture. This ABI is not well
24048 specified at present, and is subject to change.
24053 Do or don't use hardware floating-point divide and square root instructions.
24054 This requires the F or D extensions for floating-point registers. The default
24055 is to use them if the specified architecture has these instructions.
24060 Do or don't use hardware instructions for integer division. This requires the
24061 M extension. The default is to use them if the specified architecture has
24062 these instructions.
24064 @item -march=@var{ISA-string}
24066 Generate code for given RISC-V ISA (e.g.@: @samp{rv64im}). ISA strings must be
24067 lower-case. Examples include @samp{rv64i}, @samp{rv32g}, @samp{rv32e}, and
24070 @item -mtune=@var{processor-string}
24072 Optimize the output for the given processor, specified by microarchitecture
24073 name. Permissible values for this option are: @samp{rocket},
24074 @samp{sifive-3-series}, @samp{sifive-5-series}, @samp{sifive-7-series},
24077 When @option{-mtune=} is not specified, the default is @samp{rocket}.
24079 The @samp{size} choice is not intended for use by end-users. This is used
24080 when @option{-Os} is specified. It overrides the instruction cost info
24081 provided by @option{-mtune=}, but does not override the pipeline info. This
24082 helps reduce code size while still giving good performance.
24084 @item -mpreferred-stack-boundary=@var{num}
24085 @opindex mpreferred-stack-boundary
24086 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
24087 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
24088 the default is 4 (16 bytes or 128-bits).
24090 @strong{Warning:} If you use this switch, then you must build all modules with
24091 the same value, including any libraries. This includes the system libraries
24092 and startup modules.
24094 @item -msmall-data-limit=@var{n}
24095 @opindex msmall-data-limit
24096 Put global and static data smaller than @var{n} bytes into a special section
24099 @item -msave-restore
24100 @itemx -mno-save-restore
24101 @opindex msave-restore
24102 Do or don't use smaller but slower prologue and epilogue code that uses
24103 library function calls. The default is to use fast inline prologues and
24106 @item -mstrict-align
24107 @itemx -mno-strict-align
24108 @opindex mstrict-align
24109 Do not or do generate unaligned memory accesses. The default is set depending
24110 on whether the processor we are optimizing for supports fast unaligned access
24113 @item -mcmodel=medlow
24114 @opindex mcmodel=medlow
24115 Generate code for the medium-low code model. The program and its statically
24116 defined symbols must lie within a single 2 GiB address range and must lie
24117 between absolute addresses @minus{}2 GiB and +2 GiB. Programs can be
24118 statically or dynamically linked. This is the default code model.
24120 @item -mcmodel=medany
24121 @opindex mcmodel=medany
24122 Generate code for the medium-any code model. The program and its statically
24123 defined symbols must be within any single 2 GiB address range. Programs can be
24124 statically or dynamically linked.
24126 @item -mexplicit-relocs
24127 @itemx -mno-exlicit-relocs
24128 Use or do not use assembler relocation operators when dealing with symbolic
24129 addresses. The alternative is to use assembler macros instead, which may
24130 limit optimization.
24134 Take advantage of linker relaxations to reduce the number of instructions
24135 required to materialize symbol addresses. The default is to take advantage of
24136 linker relaxations.
24138 @item -memit-attribute
24139 @itemx -mno-emit-attribute
24140 Emit (do not emit) RISC-V attribute to record extra information into ELF
24141 objects. This feature requires at least binutils 2.32.
24143 @item -malign-data=@var{type}
24144 @opindex malign-data
24145 Control how GCC aligns variables and constants of array, structure, or union
24146 types. Supported values for @var{type} are @samp{xlen} which uses x register
24147 width as the alignment value, and @samp{natural} which uses natural alignment.
24148 @samp{xlen} is the default.
24152 @subsection RL78 Options
24153 @cindex RL78 Options
24159 Links in additional target libraries to support operation within a
24168 Specifies the type of hardware multiplication and division support to
24169 be used. The simplest is @code{none}, which uses software for both
24170 multiplication and division. This is the default. The @code{g13}
24171 value is for the hardware multiply/divide peripheral found on the
24172 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
24173 the multiplication and division instructions supported by the RL78/G14
24174 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
24175 the value @code{mg10} is an alias for @code{none}.
24177 In addition a C preprocessor macro is defined, based upon the setting
24178 of this option. Possible values are: @code{__RL78_MUL_NONE__},
24179 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
24186 Specifies the RL78 core to target. The default is the G14 core, also
24187 known as an S3 core or just RL78. The G13 or S2 core does not have
24188 multiply or divide instructions, instead it uses a hardware peripheral
24189 for these operations. The G10 or S1 core does not have register
24190 banks, so it uses a different calling convention.
24192 If this option is set it also selects the type of hardware multiply
24193 support to use, unless this is overridden by an explicit
24194 @option{-mmul=none} option on the command line. Thus specifying
24195 @option{-mcpu=g13} enables the use of the G13 hardware multiply
24196 peripheral and specifying @option{-mcpu=g10} disables the use of
24197 hardware multiplications altogether.
24199 Note, although the RL78/G14 core is the default target, specifying
24200 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
24201 change the behavior of the toolchain since it also enables G14
24202 hardware multiply support. If these options are not specified on the
24203 command line then software multiplication routines will be used even
24204 though the code targets the RL78 core. This is for backwards
24205 compatibility with older toolchains which did not have hardware
24206 multiply and divide support.
24208 In addition a C preprocessor macro is defined, based upon the setting
24209 of this option. Possible values are: @code{__RL78_G10__},
24210 @code{__RL78_G13__} or @code{__RL78_G14__}.
24220 These are aliases for the corresponding @option{-mcpu=} option. They
24221 are provided for backwards compatibility.
24225 Allow the compiler to use all of the available registers. By default
24226 registers @code{r24..r31} are reserved for use in interrupt handlers.
24227 With this option enabled these registers can be used in ordinary
24230 @item -m64bit-doubles
24231 @itemx -m32bit-doubles
24232 @opindex m64bit-doubles
24233 @opindex m32bit-doubles
24234 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
24235 or 32 bits (@option{-m32bit-doubles}) in size. The default is
24236 @option{-m32bit-doubles}.
24238 @item -msave-mduc-in-interrupts
24239 @itemx -mno-save-mduc-in-interrupts
24240 @opindex msave-mduc-in-interrupts
24241 @opindex mno-save-mduc-in-interrupts
24242 Specifies that interrupt handler functions should preserve the
24243 MDUC registers. This is only necessary if normal code might use
24244 the MDUC registers, for example because it performs multiplication
24245 and division operations. The default is to ignore the MDUC registers
24246 as this makes the interrupt handlers faster. The target option -mg13
24247 needs to be passed for this to work as this feature is only available
24248 on the G13 target (S2 core). The MDUC registers will only be saved
24249 if the interrupt handler performs a multiplication or division
24250 operation or it calls another function.
24254 @node RS/6000 and PowerPC Options
24255 @subsection IBM RS/6000 and PowerPC Options
24256 @cindex RS/6000 and PowerPC Options
24257 @cindex IBM RS/6000 and PowerPC Options
24259 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
24261 @item -mpowerpc-gpopt
24262 @itemx -mno-powerpc-gpopt
24263 @itemx -mpowerpc-gfxopt
24264 @itemx -mno-powerpc-gfxopt
24267 @itemx -mno-powerpc64
24271 @itemx -mno-popcntb
24273 @itemx -mno-popcntd
24280 @itemx -mno-hard-dfp
24281 @opindex mpowerpc-gpopt
24282 @opindex mno-powerpc-gpopt
24283 @opindex mpowerpc-gfxopt
24284 @opindex mno-powerpc-gfxopt
24285 @opindex mpowerpc64
24286 @opindex mno-powerpc64
24290 @opindex mno-popcntb
24292 @opindex mno-popcntd
24298 @opindex mno-hard-dfp
24299 You use these options to specify which instructions are available on the
24300 processor you are using. The default value of these options is
24301 determined when configuring GCC@. Specifying the
24302 @option{-mcpu=@var{cpu_type}} overrides the specification of these
24303 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
24304 rather than the options listed above.
24306 Specifying @option{-mpowerpc-gpopt} allows
24307 GCC to use the optional PowerPC architecture instructions in the
24308 General Purpose group, including floating-point square root. Specifying
24309 @option{-mpowerpc-gfxopt} allows GCC to
24310 use the optional PowerPC architecture instructions in the Graphics
24311 group, including floating-point select.
24313 The @option{-mmfcrf} option allows GCC to generate the move from
24314 condition register field instruction implemented on the POWER4
24315 processor and other processors that support the PowerPC V2.01
24317 The @option{-mpopcntb} option allows GCC to generate the popcount and
24318 double-precision FP reciprocal estimate instruction implemented on the
24319 POWER5 processor and other processors that support the PowerPC V2.02
24321 The @option{-mpopcntd} option allows GCC to generate the popcount
24322 instruction implemented on the POWER7 processor and other processors
24323 that support the PowerPC V2.06 architecture.
24324 The @option{-mfprnd} option allows GCC to generate the FP round to
24325 integer instructions implemented on the POWER5+ processor and other
24326 processors that support the PowerPC V2.03 architecture.
24327 The @option{-mcmpb} option allows GCC to generate the compare bytes
24328 instruction implemented on the POWER6 processor and other processors
24329 that support the PowerPC V2.05 architecture.
24330 The @option{-mhard-dfp} option allows GCC to generate the decimal
24331 floating-point instructions implemented on some POWER processors.
24333 The @option{-mpowerpc64} option allows GCC to generate the additional
24334 64-bit instructions that are found in the full PowerPC64 architecture
24335 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
24336 @option{-mno-powerpc64}.
24338 @item -mcpu=@var{cpu_type}
24340 Set architecture type, register usage, and
24341 instruction scheduling parameters for machine type @var{cpu_type}.
24342 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
24343 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
24344 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
24345 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
24346 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
24347 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
24348 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
24349 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
24350 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
24351 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
24352 @samp{power9}, @samp{future}, @samp{powerpc}, @samp{powerpc64},
24353 @samp{powerpc64le}, @samp{rs64}, and @samp{native}.
24355 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
24356 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
24357 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
24358 architecture machine types, with an appropriate, generic processor
24359 model assumed for scheduling purposes.
24361 Specifying @samp{native} as cpu type detects and selects the
24362 architecture option that corresponds to the host processor of the
24363 system performing the compilation.
24364 @option{-mcpu=native} has no effect if GCC does not recognize the
24367 The other options specify a specific processor. Code generated under
24368 those options runs best on that processor, and may not run at all on
24371 The @option{-mcpu} options automatically enable or disable the
24374 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
24375 -mpopcntb -mpopcntd -mpowerpc64 @gol
24376 -mpowerpc-gpopt -mpowerpc-gfxopt @gol
24377 -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
24378 -mcrypto -mhtm -mpower8-fusion -mpower8-vector @gol
24379 -mquad-memory -mquad-memory-atomic -mfloat128 -mfloat128-hardware}
24381 The particular options set for any particular CPU varies between
24382 compiler versions, depending on what setting seems to produce optimal
24383 code for that CPU; it doesn't necessarily reflect the actual hardware's
24384 capabilities. If you wish to set an individual option to a particular
24385 value, you may specify it after the @option{-mcpu} option, like
24386 @option{-mcpu=970 -mno-altivec}.
24388 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
24389 not enabled or disabled by the @option{-mcpu} option at present because
24390 AIX does not have full support for these options. You may still
24391 enable or disable them individually if you're sure it'll work in your
24394 @item -mtune=@var{cpu_type}
24396 Set the instruction scheduling parameters for machine type
24397 @var{cpu_type}, but do not set the architecture type or register usage,
24398 as @option{-mcpu=@var{cpu_type}} does. The same
24399 values for @var{cpu_type} are used for @option{-mtune} as for
24400 @option{-mcpu}. If both are specified, the code generated uses the
24401 architecture and registers set by @option{-mcpu}, but the
24402 scheduling parameters set by @option{-mtune}.
24404 @item -mcmodel=small
24405 @opindex mcmodel=small
24406 Generate PowerPC64 code for the small model: The TOC is limited to
24409 @item -mcmodel=medium
24410 @opindex mcmodel=medium
24411 Generate PowerPC64 code for the medium model: The TOC and other static
24412 data may be up to a total of 4G in size. This is the default for 64-bit
24415 @item -mcmodel=large
24416 @opindex mcmodel=large
24417 Generate PowerPC64 code for the large model: The TOC may be up to 4G
24418 in size. Other data and code is only limited by the 64-bit address
24422 @itemx -mno-altivec
24424 @opindex mno-altivec
24425 Generate code that uses (does not use) AltiVec instructions, and also
24426 enable the use of built-in functions that allow more direct access to
24427 the AltiVec instruction set. You may also need to set
24428 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
24431 When @option{-maltivec} is used, the element order for AltiVec intrinsics
24432 such as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
24433 match array element order corresponding to the endianness of the
24434 target. That is, element zero identifies the leftmost element in a
24435 vector register when targeting a big-endian platform, and identifies
24436 the rightmost element in a vector register when targeting a
24437 little-endian platform.
24442 @opindex mno-vrsave
24443 Generate VRSAVE instructions when generating AltiVec code.
24446 @opindex msecure-plt
24447 Generate code that allows @command{ld} and @command{ld.so}
24448 to build executables and shared
24449 libraries with non-executable @code{.plt} and @code{.got} sections.
24451 32-bit SYSV ABI option.
24455 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
24457 requires @code{.plt} and @code{.got}
24458 sections that are both writable and executable.
24459 This is a PowerPC 32-bit SYSV ABI option.
24465 This switch enables or disables the generation of ISEL instructions.
24471 Generate code that uses (does not use) vector/scalar (VSX)
24472 instructions, and also enable the use of built-in functions that allow
24473 more direct access to the VSX instruction set.
24478 @opindex mno-crypto
24479 Enable the use (disable) of the built-in functions that allow direct
24480 access to the cryptographic instructions that were added in version
24481 2.07 of the PowerPC ISA.
24487 Enable (disable) the use of the built-in functions that allow direct
24488 access to the Hardware Transactional Memory (HTM) instructions that
24489 were added in version 2.07 of the PowerPC ISA.
24491 @item -mpower8-fusion
24492 @itemx -mno-power8-fusion
24493 @opindex mpower8-fusion
24494 @opindex mno-power8-fusion
24495 Generate code that keeps (does not keeps) some integer operations
24496 adjacent so that the instructions can be fused together on power8 and
24499 @item -mpower8-vector
24500 @itemx -mno-power8-vector
24501 @opindex mpower8-vector
24502 @opindex mno-power8-vector
24503 Generate code that uses (does not use) the vector and scalar
24504 instructions that were added in version 2.07 of the PowerPC ISA. Also
24505 enable the use of built-in functions that allow more direct access to
24506 the vector instructions.
24508 @item -mquad-memory
24509 @itemx -mno-quad-memory
24510 @opindex mquad-memory
24511 @opindex mno-quad-memory
24512 Generate code that uses (does not use) the non-atomic quad word memory
24513 instructions. The @option{-mquad-memory} option requires use of
24516 @item -mquad-memory-atomic
24517 @itemx -mno-quad-memory-atomic
24518 @opindex mquad-memory-atomic
24519 @opindex mno-quad-memory-atomic
24520 Generate code that uses (does not use) the atomic quad word memory
24521 instructions. The @option{-mquad-memory-atomic} option requires use of
24525 @itemx -mno-float128
24527 @opindex mno-float128
24528 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
24529 and use either software emulation for IEEE 128-bit floating point or
24530 hardware instructions.
24532 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7},
24533 @option{-mcpu=power8}), or @option{-mcpu=power9} must be enabled to
24534 use the IEEE 128-bit floating point support. The IEEE 128-bit
24535 floating point support only works on PowerPC Linux systems.
24537 The default for @option{-mfloat128} is enabled on PowerPC Linux
24538 systems using the VSX instruction set, and disabled on other systems.
24540 If you use the ISA 3.0 instruction set (@option{-mpower9-vector} or
24541 @option{-mcpu=power9}) on a 64-bit system, the IEEE 128-bit floating
24542 point support will also enable the generation of ISA 3.0 IEEE 128-bit
24543 floating point instructions. Otherwise, if you do not specify to
24544 generate ISA 3.0 instructions or you are targeting a 32-bit big endian
24545 system, IEEE 128-bit floating point will be done with software
24548 @item -mfloat128-hardware
24549 @itemx -mno-float128-hardware
24550 @opindex mfloat128-hardware
24551 @opindex mno-float128-hardware
24552 Enable/disable using ISA 3.0 hardware instructions to support the
24553 @var{__float128} data type.
24555 The default for @option{-mfloat128-hardware} is enabled on PowerPC
24556 Linux systems using the ISA 3.0 instruction set, and disabled on other
24563 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
24564 targets (including GNU/Linux). The 32-bit environment sets int, long
24565 and pointer to 32 bits and generates code that runs on any PowerPC
24566 variant. The 64-bit environment sets int to 32 bits and long and
24567 pointer to 64 bits, and generates code for PowerPC64, as for
24568 @option{-mpowerpc64}.
24571 @itemx -mno-fp-in-toc
24572 @itemx -mno-sum-in-toc
24573 @itemx -mminimal-toc
24575 @opindex mno-fp-in-toc
24576 @opindex mno-sum-in-toc
24577 @opindex mminimal-toc
24578 Modify generation of the TOC (Table Of Contents), which is created for
24579 every executable file. The @option{-mfull-toc} option is selected by
24580 default. In that case, GCC allocates at least one TOC entry for
24581 each unique non-automatic variable reference in your program. GCC
24582 also places floating-point constants in the TOC@. However, only
24583 16,384 entries are available in the TOC@.
24585 If you receive a linker error message that saying you have overflowed
24586 the available TOC space, you can reduce the amount of TOC space used
24587 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
24588 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
24589 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
24590 generate code to calculate the sum of an address and a constant at
24591 run time instead of putting that sum into the TOC@. You may specify one
24592 or both of these options. Each causes GCC to produce very slightly
24593 slower and larger code at the expense of conserving TOC space.
24595 If you still run out of space in the TOC even when you specify both of
24596 these options, specify @option{-mminimal-toc} instead. This option causes
24597 GCC to make only one TOC entry for every file. When you specify this
24598 option, GCC produces code that is slower and larger but which
24599 uses extremely little TOC space. You may wish to use this option
24600 only on files that contain less frequently-executed code.
24606 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
24607 @code{long} type, and the infrastructure needed to support them.
24608 Specifying @option{-maix64} implies @option{-mpowerpc64},
24609 while @option{-maix32} disables the 64-bit ABI and
24610 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
24613 @itemx -mno-xl-compat
24614 @opindex mxl-compat
24615 @opindex mno-xl-compat
24616 Produce code that conforms more closely to IBM XL compiler semantics
24617 when using AIX-compatible ABI@. Pass floating-point arguments to
24618 prototyped functions beyond the register save area (RSA) on the stack
24619 in addition to argument FPRs. Do not assume that most significant
24620 double in 128-bit long double value is properly rounded when comparing
24621 values and converting to double. Use XL symbol names for long double
24624 The AIX calling convention was extended but not initially documented to
24625 handle an obscure K&R C case of calling a function that takes the
24626 address of its arguments with fewer arguments than declared. IBM XL
24627 compilers access floating-point arguments that do not fit in the
24628 RSA from the stack when a subroutine is compiled without
24629 optimization. Because always storing floating-point arguments on the
24630 stack is inefficient and rarely needed, this option is not enabled by
24631 default and only is necessary when calling subroutines compiled by IBM
24632 XL compilers without optimization.
24636 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
24637 application written to use message passing with special startup code to
24638 enable the application to run. The system must have PE installed in the
24639 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
24640 must be overridden with the @option{-specs=} option to specify the
24641 appropriate directory location. The Parallel Environment does not
24642 support threads, so the @option{-mpe} option and the @option{-pthread}
24643 option are incompatible.
24645 @item -malign-natural
24646 @itemx -malign-power
24647 @opindex malign-natural
24648 @opindex malign-power
24649 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
24650 @option{-malign-natural} overrides the ABI-defined alignment of larger
24651 types, such as floating-point doubles, on their natural size-based boundary.
24652 The option @option{-malign-power} instructs GCC to follow the ABI-specified
24653 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
24655 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
24659 @itemx -mhard-float
24660 @opindex msoft-float
24661 @opindex mhard-float
24662 Generate code that does not use (uses) the floating-point register set.
24663 Software floating-point emulation is provided if you use the
24664 @option{-msoft-float} option, and pass the option to GCC when linking.
24667 @itemx -mno-multiple
24669 @opindex mno-multiple
24670 Generate code that uses (does not use) the load multiple word
24671 instructions and the store multiple word instructions. These
24672 instructions are generated by default on POWER systems, and not
24673 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
24674 PowerPC systems, since those instructions do not work when the
24675 processor is in little-endian mode. The exceptions are PPC740 and
24676 PPC750 which permit these instructions in little-endian mode.
24681 @opindex mno-update
24682 Generate code that uses (does not use) the load or store instructions
24683 that update the base register to the address of the calculated memory
24684 location. These instructions are generated by default. If you use
24685 @option{-mno-update}, there is a small window between the time that the
24686 stack pointer is updated and the address of the previous frame is
24687 stored, which means code that walks the stack frame across interrupts or
24688 signals may get corrupted data.
24690 @item -mavoid-indexed-addresses
24691 @itemx -mno-avoid-indexed-addresses
24692 @opindex mavoid-indexed-addresses
24693 @opindex mno-avoid-indexed-addresses
24694 Generate code that tries to avoid (not avoid) the use of indexed load
24695 or store instructions. These instructions can incur a performance
24696 penalty on Power6 processors in certain situations, such as when
24697 stepping through large arrays that cross a 16M boundary. This option
24698 is enabled by default when targeting Power6 and disabled otherwise.
24701 @itemx -mno-fused-madd
24702 @opindex mfused-madd
24703 @opindex mno-fused-madd
24704 Generate code that uses (does not use) the floating-point multiply and
24705 accumulate instructions. These instructions are generated by default
24706 if hardware floating point is used. The machine-dependent
24707 @option{-mfused-madd} option is now mapped to the machine-independent
24708 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
24709 mapped to @option{-ffp-contract=off}.
24715 Generate code that uses (does not use) the half-word multiply and
24716 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
24717 These instructions are generated by default when targeting those
24724 Generate code that uses (does not use) the string-search @samp{dlmzb}
24725 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
24726 generated by default when targeting those processors.
24728 @item -mno-bit-align
24730 @opindex mno-bit-align
24731 @opindex mbit-align
24732 On System V.4 and embedded PowerPC systems do not (do) force structures
24733 and unions that contain bit-fields to be aligned to the base type of the
24736 For example, by default a structure containing nothing but 8
24737 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
24738 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
24739 the structure is aligned to a 1-byte boundary and is 1 byte in
24742 @item -mno-strict-align
24743 @itemx -mstrict-align
24744 @opindex mno-strict-align
24745 @opindex mstrict-align
24746 On System V.4 and embedded PowerPC systems do not (do) assume that
24747 unaligned memory references are handled by the system.
24749 @item -mrelocatable
24750 @itemx -mno-relocatable
24751 @opindex mrelocatable
24752 @opindex mno-relocatable
24753 Generate code that allows (does not allow) a static executable to be
24754 relocated to a different address at run time. A simple embedded
24755 PowerPC system loader should relocate the entire contents of
24756 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
24757 a table of 32-bit addresses generated by this option. For this to
24758 work, all objects linked together must be compiled with
24759 @option{-mrelocatable} or @option{-mrelocatable-lib}.
24760 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
24762 @item -mrelocatable-lib
24763 @itemx -mno-relocatable-lib
24764 @opindex mrelocatable-lib
24765 @opindex mno-relocatable-lib
24766 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
24767 @code{.fixup} section to allow static executables to be relocated at
24768 run time, but @option{-mrelocatable-lib} does not use the smaller stack
24769 alignment of @option{-mrelocatable}. Objects compiled with
24770 @option{-mrelocatable-lib} may be linked with objects compiled with
24771 any combination of the @option{-mrelocatable} options.
24777 On System V.4 and embedded PowerPC systems do not (do) assume that
24778 register 2 contains a pointer to a global area pointing to the addresses
24779 used in the program.
24782 @itemx -mlittle-endian
24784 @opindex mlittle-endian
24785 On System V.4 and embedded PowerPC systems compile code for the
24786 processor in little-endian mode. The @option{-mlittle-endian} option is
24787 the same as @option{-mlittle}.
24790 @itemx -mbig-endian
24792 @opindex mbig-endian
24793 On System V.4 and embedded PowerPC systems compile code for the
24794 processor in big-endian mode. The @option{-mbig-endian} option is
24795 the same as @option{-mbig}.
24797 @item -mdynamic-no-pic
24798 @opindex mdynamic-no-pic
24799 On Darwin and Mac OS X systems, compile code so that it is not
24800 relocatable, but that its external references are relocatable. The
24801 resulting code is suitable for applications, but not shared
24804 @item -msingle-pic-base
24805 @opindex msingle-pic-base
24806 Treat the register used for PIC addressing as read-only, rather than
24807 loading it in the prologue for each function. The runtime system is
24808 responsible for initializing this register with an appropriate value
24809 before execution begins.
24811 @item -mprioritize-restricted-insns=@var{priority}
24812 @opindex mprioritize-restricted-insns
24813 This option controls the priority that is assigned to
24814 dispatch-slot restricted instructions during the second scheduling
24815 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
24816 or @samp{2} to assign no, highest, or second-highest (respectively)
24817 priority to dispatch-slot restricted
24820 @item -msched-costly-dep=@var{dependence_type}
24821 @opindex msched-costly-dep
24822 This option controls which dependences are considered costly
24823 by the target during instruction scheduling. The argument
24824 @var{dependence_type} takes one of the following values:
24828 No dependence is costly.
24831 All dependences are costly.
24833 @item @samp{true_store_to_load}
24834 A true dependence from store to load is costly.
24836 @item @samp{store_to_load}
24837 Any dependence from store to load is costly.
24840 Any dependence for which the latency is greater than or equal to
24841 @var{number} is costly.
24844 @item -minsert-sched-nops=@var{scheme}
24845 @opindex minsert-sched-nops
24846 This option controls which NOP insertion scheme is used during
24847 the second scheduling pass. The argument @var{scheme} takes one of the
24855 Pad with NOPs any dispatch group that has vacant issue slots,
24856 according to the scheduler's grouping.
24858 @item @samp{regroup_exact}
24859 Insert NOPs to force costly dependent insns into
24860 separate groups. Insert exactly as many NOPs as needed to force an insn
24861 to a new group, according to the estimated processor grouping.
24864 Insert NOPs to force costly dependent insns into
24865 separate groups. Insert @var{number} NOPs to force an insn to a new group.
24869 @opindex mcall-sysv
24870 On System V.4 and embedded PowerPC systems compile code using calling
24871 conventions that adhere to the March 1995 draft of the System V
24872 Application Binary Interface, PowerPC processor supplement. This is the
24873 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
24875 @item -mcall-sysv-eabi
24877 @opindex mcall-sysv-eabi
24878 @opindex mcall-eabi
24879 Specify both @option{-mcall-sysv} and @option{-meabi} options.
24881 @item -mcall-sysv-noeabi
24882 @opindex mcall-sysv-noeabi
24883 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
24885 @item -mcall-aixdesc
24887 On System V.4 and embedded PowerPC systems compile code for the AIX
24891 @opindex mcall-linux
24892 On System V.4 and embedded PowerPC systems compile code for the
24893 Linux-based GNU system.
24895 @item -mcall-freebsd
24896 @opindex mcall-freebsd
24897 On System V.4 and embedded PowerPC systems compile code for the
24898 FreeBSD operating system.
24900 @item -mcall-netbsd
24901 @opindex mcall-netbsd
24902 On System V.4 and embedded PowerPC systems compile code for the
24903 NetBSD operating system.
24905 @item -mcall-openbsd
24906 @opindex mcall-netbsd
24907 On System V.4 and embedded PowerPC systems compile code for the
24908 OpenBSD operating system.
24910 @item -mtraceback=@var{traceback_type}
24911 @opindex mtraceback
24912 Select the type of traceback table. Valid values for @var{traceback_type}
24913 are @samp{full}, @samp{part}, and @samp{no}.
24915 @item -maix-struct-return
24916 @opindex maix-struct-return
24917 Return all structures in memory (as specified by the AIX ABI)@.
24919 @item -msvr4-struct-return
24920 @opindex msvr4-struct-return
24921 Return structures smaller than 8 bytes in registers (as specified by the
24924 @item -mabi=@var{abi-type}
24926 Extend the current ABI with a particular extension, or remove such extension.
24927 Valid values are @samp{altivec}, @samp{no-altivec},
24928 @samp{ibmlongdouble}, @samp{ieeelongdouble},
24929 @samp{elfv1}, @samp{elfv2}@.
24931 @item -mabi=ibmlongdouble
24932 @opindex mabi=ibmlongdouble
24933 Change the current ABI to use IBM extended-precision long double.
24934 This is not likely to work if your system defaults to using IEEE
24935 extended-precision long double. If you change the long double type
24936 from IEEE extended-precision, the compiler will issue a warning unless
24937 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24940 @item -mabi=ieeelongdouble
24941 @opindex mabi=ieeelongdouble
24942 Change the current ABI to use IEEE extended-precision long double.
24943 This is not likely to work if your system defaults to using IBM
24944 extended-precision long double. If you change the long double type
24945 from IBM extended-precision, the compiler will issue a warning unless
24946 you use the @option{-Wno-psabi} option. Requires @option{-mlong-double-128}
24950 @opindex mabi=elfv1
24951 Change the current ABI to use the ELFv1 ABI.
24952 This is the default ABI for big-endian PowerPC 64-bit Linux.
24953 Overriding the default ABI requires special system support and is
24954 likely to fail in spectacular ways.
24957 @opindex mabi=elfv2
24958 Change the current ABI to use the ELFv2 ABI.
24959 This is the default ABI for little-endian PowerPC 64-bit Linux.
24960 Overriding the default ABI requires special system support and is
24961 likely to fail in spectacular ways.
24963 @item -mgnu-attribute
24964 @itemx -mno-gnu-attribute
24965 @opindex mgnu-attribute
24966 @opindex mno-gnu-attribute
24967 Emit .gnu_attribute assembly directives to set tag/value pairs in a
24968 .gnu.attributes section that specify ABI variations in function
24969 parameters or return values.
24972 @itemx -mno-prototype
24973 @opindex mprototype
24974 @opindex mno-prototype
24975 On System V.4 and embedded PowerPC systems assume that all calls to
24976 variable argument functions are properly prototyped. Otherwise, the
24977 compiler must insert an instruction before every non-prototyped call to
24978 set or clear bit 6 of the condition code register (@code{CR}) to
24979 indicate whether floating-point values are passed in the floating-point
24980 registers in case the function takes variable arguments. With
24981 @option{-mprototype}, only calls to prototyped variable argument functions
24982 set or clear the bit.
24986 On embedded PowerPC systems, assume that the startup module is called
24987 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
24988 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
24993 On embedded PowerPC systems, assume that the startup module is called
24994 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
24999 On embedded PowerPC systems, assume that the startup module is called
25000 @file{crt0.o} and the standard C libraries are @file{libads.a} and
25003 @item -myellowknife
25004 @opindex myellowknife
25005 On embedded PowerPC systems, assume that the startup module is called
25006 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
25011 On System V.4 and embedded PowerPC systems, specify that you are
25012 compiling for a VxWorks system.
25016 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
25017 header to indicate that @samp{eabi} extended relocations are used.
25023 On System V.4 and embedded PowerPC systems do (do not) adhere to the
25024 Embedded Applications Binary Interface (EABI), which is a set of
25025 modifications to the System V.4 specifications. Selecting @option{-meabi}
25026 means that the stack is aligned to an 8-byte boundary, a function
25027 @code{__eabi} is called from @code{main} to set up the EABI
25028 environment, and the @option{-msdata} option can use both @code{r2} and
25029 @code{r13} to point to two separate small data areas. Selecting
25030 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
25031 no EABI initialization function is called from @code{main}, and the
25032 @option{-msdata} option only uses @code{r13} to point to a single
25033 small data area. The @option{-meabi} option is on by default if you
25034 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
25037 @opindex msdata=eabi
25038 On System V.4 and embedded PowerPC systems, put small initialized
25039 @code{const} global and static data in the @code{.sdata2} section, which
25040 is pointed to by register @code{r2}. Put small initialized
25041 non-@code{const} global and static data in the @code{.sdata} section,
25042 which is pointed to by register @code{r13}. Put small uninitialized
25043 global and static data in the @code{.sbss} section, which is adjacent to
25044 the @code{.sdata} section. The @option{-msdata=eabi} option is
25045 incompatible with the @option{-mrelocatable} option. The
25046 @option{-msdata=eabi} option also sets the @option{-memb} option.
25049 @opindex msdata=sysv
25050 On System V.4 and embedded PowerPC systems, put small global and static
25051 data in the @code{.sdata} section, which is pointed to by register
25052 @code{r13}. Put small uninitialized global and static data in the
25053 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
25054 The @option{-msdata=sysv} option is incompatible with the
25055 @option{-mrelocatable} option.
25057 @item -msdata=default
25059 @opindex msdata=default
25061 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
25062 compile code the same as @option{-msdata=eabi}, otherwise compile code the
25063 same as @option{-msdata=sysv}.
25066 @opindex msdata=data
25067 On System V.4 and embedded PowerPC systems, put small global
25068 data in the @code{.sdata} section. Put small uninitialized global
25069 data in the @code{.sbss} section. Do not use register @code{r13}
25070 to address small data however. This is the default behavior unless
25071 other @option{-msdata} options are used.
25075 @opindex msdata=none
25077 On embedded PowerPC systems, put all initialized global and static data
25078 in the @code{.data} section, and all uninitialized data in the
25079 @code{.bss} section.
25081 @item -mreadonly-in-sdata
25082 @opindex mreadonly-in-sdata
25083 @opindex mno-readonly-in-sdata
25084 Put read-only objects in the @code{.sdata} section as well. This is the
25087 @item -mblock-move-inline-limit=@var{num}
25088 @opindex mblock-move-inline-limit
25089 Inline all block moves (such as calls to @code{memcpy} or structure
25090 copies) less than or equal to @var{num} bytes. The minimum value for
25091 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
25092 targets. The default value is target-specific.
25094 @item -mblock-compare-inline-limit=@var{num}
25095 @opindex mblock-compare-inline-limit
25096 Generate non-looping inline code for all block compares (such as calls
25097 to @code{memcmp} or structure compares) less than or equal to @var{num}
25098 bytes. If @var{num} is 0, all inline expansion (non-loop and loop) of
25099 block compare is disabled. The default value is target-specific.
25101 @item -mblock-compare-inline-loop-limit=@var{num}
25102 @opindex mblock-compare-inline-loop-limit
25103 Generate an inline expansion using loop code for all block compares that
25104 are less than or equal to @var{num} bytes, but greater than the limit
25105 for non-loop inline block compare expansion. If the block length is not
25106 constant, at most @var{num} bytes will be compared before @code{memcmp}
25107 is called to compare the remainder of the block. The default value is
25110 @item -mstring-compare-inline-limit=@var{num}
25111 @opindex mstring-compare-inline-limit
25112 Compare at most @var{num} string bytes with inline code.
25113 If the difference or end of string is not found at the
25114 end of the inline compare a call to @code{strcmp} or @code{strncmp} will
25115 take care of the rest of the comparison. The default is 64 bytes.
25119 @cindex smaller data references (PowerPC)
25120 @cindex .sdata/.sdata2 references (PowerPC)
25121 On embedded PowerPC systems, put global and static items less than or
25122 equal to @var{num} bytes into the small data or BSS sections instead of
25123 the normal data or BSS section. By default, @var{num} is 8. The
25124 @option{-G @var{num}} switch is also passed to the linker.
25125 All modules should be compiled with the same @option{-G @var{num}} value.
25128 @itemx -mno-regnames
25130 @opindex mno-regnames
25131 On System V.4 and embedded PowerPC systems do (do not) emit register
25132 names in the assembly language output using symbolic forms.
25135 @itemx -mno-longcall
25137 @opindex mno-longcall
25138 By default assume that all calls are far away so that a longer and more
25139 expensive calling sequence is required. This is required for calls
25140 farther than 32 megabytes (33,554,432 bytes) from the current location.
25141 A short call is generated if the compiler knows
25142 the call cannot be that far away. This setting can be overridden by
25143 the @code{shortcall} function attribute, or by @code{#pragma
25146 Some linkers are capable of detecting out-of-range calls and generating
25147 glue code on the fly. On these systems, long calls are unnecessary and
25148 generate slower code. As of this writing, the AIX linker can do this,
25149 as can the GNU linker for PowerPC/64. It is planned to add this feature
25150 to the GNU linker for 32-bit PowerPC systems as well.
25152 On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers,
25153 GCC can generate long calls using an inline PLT call sequence (see
25154 @option{-mpltseq}). PowerPC with @option{-mbss-plt} and PowerPC64
25155 ELFv1 (big-endian) do not support inline PLT calls.
25157 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
25158 callee, L42}, plus a @dfn{branch island} (glue code). The two target
25159 addresses represent the callee and the branch island. The
25160 Darwin/PPC linker prefers the first address and generates a @code{bl
25161 callee} if the PPC @code{bl} instruction reaches the callee directly;
25162 otherwise, the linker generates @code{bl L42} to call the branch
25163 island. The branch island is appended to the body of the
25164 calling function; it computes the full 32-bit address of the callee
25167 On Mach-O (Darwin) systems, this option directs the compiler emit to
25168 the glue for every direct call, and the Darwin linker decides whether
25169 to use or discard it.
25171 In the future, GCC may ignore all longcall specifications
25172 when the linker is known to generate glue.
25177 @opindex mno-pltseq
25178 Implement (do not implement) -fno-plt and long calls using an inline
25179 PLT call sequence that supports lazy linking and long calls to
25180 functions in dlopen'd shared libraries. Inline PLT calls are only
25181 supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU
25182 linkers, and are enabled by default if the support is detected when
25183 configuring GCC, and, in the case of 32-bit PowerPC, if GCC is
25184 configured with @option{--enable-secureplt}. @option{-mpltseq} code
25185 and @option{-mbss-plt} 32-bit PowerPC relocatable objects may not be
25188 @item -mtls-markers
25189 @itemx -mno-tls-markers
25190 @opindex mtls-markers
25191 @opindex mno-tls-markers
25192 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
25193 specifying the function argument. The relocation allows the linker to
25194 reliably associate function call with argument setup instructions for
25195 TLS optimization, which in turn allows GCC to better schedule the
25201 This option enables use of the reciprocal estimate and
25202 reciprocal square root estimate instructions with additional
25203 Newton-Raphson steps to increase precision instead of doing a divide or
25204 square root and divide for floating-point arguments. You should use
25205 the @option{-ffast-math} option when using @option{-mrecip} (or at
25206 least @option{-funsafe-math-optimizations},
25207 @option{-ffinite-math-only}, @option{-freciprocal-math} and
25208 @option{-fno-trapping-math}). Note that while the throughput of the
25209 sequence is generally higher than the throughput of the non-reciprocal
25210 instruction, the precision of the sequence can be decreased by up to 2
25211 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
25214 @item -mrecip=@var{opt}
25215 @opindex mrecip=opt
25216 This option controls which reciprocal estimate instructions
25217 may be used. @var{opt} is a comma-separated list of options, which may
25218 be preceded by a @code{!} to invert the option:
25223 Enable all estimate instructions.
25226 Enable the default instructions, equivalent to @option{-mrecip}.
25229 Disable all estimate instructions, equivalent to @option{-mno-recip}.
25232 Enable the reciprocal approximation instructions for both
25233 single and double precision.
25236 Enable the single-precision reciprocal approximation instructions.
25239 Enable the double-precision reciprocal approximation instructions.
25242 Enable the reciprocal square root approximation instructions for both
25243 single and double precision.
25246 Enable the single-precision reciprocal square root approximation instructions.
25249 Enable the double-precision reciprocal square root approximation instructions.
25253 So, for example, @option{-mrecip=all,!rsqrtd} enables
25254 all of the reciprocal estimate instructions, except for the
25255 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
25256 which handle the double-precision reciprocal square root calculations.
25258 @item -mrecip-precision
25259 @itemx -mno-recip-precision
25260 @opindex mrecip-precision
25261 Assume (do not assume) that the reciprocal estimate instructions
25262 provide higher-precision estimates than is mandated by the PowerPC
25263 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
25264 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
25265 The double-precision square root estimate instructions are not generated by
25266 default on low-precision machines, since they do not provide an
25267 estimate that converges after three steps.
25269 @item -mveclibabi=@var{type}
25270 @opindex mveclibabi
25271 Specifies the ABI type to use for vectorizing intrinsics using an
25272 external library. The only type supported at present is @samp{mass},
25273 which specifies to use IBM's Mathematical Acceleration Subsystem
25274 (MASS) libraries for vectorizing intrinsics using external libraries.
25275 GCC currently emits calls to @code{acosd2}, @code{acosf4},
25276 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
25277 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
25278 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
25279 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
25280 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
25281 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
25282 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
25283 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
25284 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
25285 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
25286 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
25287 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
25288 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
25289 for power7. Both @option{-ftree-vectorize} and
25290 @option{-funsafe-math-optimizations} must also be enabled. The MASS
25291 libraries must be specified at link time.
25296 Generate (do not generate) the @code{friz} instruction when the
25297 @option{-funsafe-math-optimizations} option is used to optimize
25298 rounding of floating-point values to 64-bit integer and back to floating
25299 point. The @code{friz} instruction does not return the same value if
25300 the floating-point number is too large to fit in an integer.
25302 @item -mpointers-to-nested-functions
25303 @itemx -mno-pointers-to-nested-functions
25304 @opindex mpointers-to-nested-functions
25305 Generate (do not generate) code to load up the static chain register
25306 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
25307 systems where a function pointer points to a 3-word descriptor giving
25308 the function address, TOC value to be loaded in register @code{r2}, and
25309 static chain value to be loaded in register @code{r11}. The
25310 @option{-mpointers-to-nested-functions} is on by default. You cannot
25311 call through pointers to nested functions or pointers
25312 to functions compiled in other languages that use the static chain if
25313 you use @option{-mno-pointers-to-nested-functions}.
25315 @item -msave-toc-indirect
25316 @itemx -mno-save-toc-indirect
25317 @opindex msave-toc-indirect
25318 Generate (do not generate) code to save the TOC value in the reserved
25319 stack location in the function prologue if the function calls through
25320 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
25321 saved in the prologue, it is saved just before the call through the
25322 pointer. The @option{-mno-save-toc-indirect} option is the default.
25324 @item -mcompat-align-parm
25325 @itemx -mno-compat-align-parm
25326 @opindex mcompat-align-parm
25327 Generate (do not generate) code to pass structure parameters with a
25328 maximum alignment of 64 bits, for compatibility with older versions
25331 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
25332 structure parameter on a 128-bit boundary when that structure contained
25333 a member requiring 128-bit alignment. This is corrected in more
25334 recent versions of GCC. This option may be used to generate code
25335 that is compatible with functions compiled with older versions of
25338 The @option{-mno-compat-align-parm} option is the default.
25340 @item -mstack-protector-guard=@var{guard}
25341 @itemx -mstack-protector-guard-reg=@var{reg}
25342 @itemx -mstack-protector-guard-offset=@var{offset}
25343 @itemx -mstack-protector-guard-symbol=@var{symbol}
25344 @opindex mstack-protector-guard
25345 @opindex mstack-protector-guard-reg
25346 @opindex mstack-protector-guard-offset
25347 @opindex mstack-protector-guard-symbol
25348 Generate stack protection code using canary at @var{guard}. Supported
25349 locations are @samp{global} for global canary or @samp{tls} for per-thread
25350 canary in the TLS block (the default with GNU libc version 2.4 or later).
25352 With the latter choice the options
25353 @option{-mstack-protector-guard-reg=@var{reg}} and
25354 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
25355 which register to use as base register for reading the canary, and from what
25356 offset from that base register. The default for those is as specified in the
25357 relevant ABI. @option{-mstack-protector-guard-symbol=@var{symbol}} overrides
25358 the offset with a symbol reference to a canary in the TLS block.
25364 Generate (do not generate) pc-relative addressing when the option
25365 @option{-mcpu=future} is used.
25369 @subsection RX Options
25372 These command-line options are defined for RX targets:
25375 @item -m64bit-doubles
25376 @itemx -m32bit-doubles
25377 @opindex m64bit-doubles
25378 @opindex m32bit-doubles
25379 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
25380 or 32 bits (@option{-m32bit-doubles}) in size. The default is
25381 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
25382 works on 32-bit values, which is why the default is
25383 @option{-m32bit-doubles}.
25389 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
25390 floating-point hardware. The default is enabled for the RX600
25391 series and disabled for the RX200 series.
25393 Floating-point instructions are only generated for 32-bit floating-point
25394 values, however, so the FPU hardware is not used for doubles if the
25395 @option{-m64bit-doubles} option is used.
25397 @emph{Note} If the @option{-fpu} option is enabled then
25398 @option{-funsafe-math-optimizations} is also enabled automatically.
25399 This is because the RX FPU instructions are themselves unsafe.
25401 @item -mcpu=@var{name}
25403 Selects the type of RX CPU to be targeted. Currently three types are
25404 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
25405 the specific @samp{RX610} CPU. The default is @samp{RX600}.
25407 The only difference between @samp{RX600} and @samp{RX610} is that the
25408 @samp{RX610} does not support the @code{MVTIPL} instruction.
25410 The @samp{RX200} series does not have a hardware floating-point unit
25411 and so @option{-nofpu} is enabled by default when this type is
25414 @item -mbig-endian-data
25415 @itemx -mlittle-endian-data
25416 @opindex mbig-endian-data
25417 @opindex mlittle-endian-data
25418 Store data (but not code) in the big-endian format. The default is
25419 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
25422 @item -msmall-data-limit=@var{N}
25423 @opindex msmall-data-limit
25424 Specifies the maximum size in bytes of global and static variables
25425 which can be placed into the small data area. Using the small data
25426 area can lead to smaller and faster code, but the size of area is
25427 limited and it is up to the programmer to ensure that the area does
25428 not overflow. Also when the small data area is used one of the RX's
25429 registers (usually @code{r13}) is reserved for use pointing to this
25430 area, so it is no longer available for use by the compiler. This
25431 could result in slower and/or larger code if variables are pushed onto
25432 the stack instead of being held in this register.
25434 Note, common variables (variables that have not been initialized) and
25435 constants are not placed into the small data area as they are assigned
25436 to other sections in the output executable.
25438 The default value is zero, which disables this feature. Note, this
25439 feature is not enabled by default with higher optimization levels
25440 (@option{-O2} etc) because of the potentially detrimental effects of
25441 reserving a register. It is up to the programmer to experiment and
25442 discover whether this feature is of benefit to their program. See the
25443 description of the @option{-mpid} option for a description of how the
25444 actual register to hold the small data area pointer is chosen.
25450 Use the simulator runtime. The default is to use the libgloss
25451 board-specific runtime.
25453 @item -mas100-syntax
25454 @itemx -mno-as100-syntax
25455 @opindex mas100-syntax
25456 @opindex mno-as100-syntax
25457 When generating assembler output use a syntax that is compatible with
25458 Renesas's AS100 assembler. This syntax can also be handled by the GAS
25459 assembler, but it has some restrictions so it is not generated by default.
25461 @item -mmax-constant-size=@var{N}
25462 @opindex mmax-constant-size
25463 Specifies the maximum size, in bytes, of a constant that can be used as
25464 an operand in a RX instruction. Although the RX instruction set does
25465 allow constants of up to 4 bytes in length to be used in instructions,
25466 a longer value equates to a longer instruction. Thus in some
25467 circumstances it can be beneficial to restrict the size of constants
25468 that are used in instructions. Constants that are too big are instead
25469 placed into a constant pool and referenced via register indirection.
25471 The value @var{N} can be between 0 and 4. A value of 0 (the default)
25472 or 4 means that constants of any size are allowed.
25476 Enable linker relaxation. Linker relaxation is a process whereby the
25477 linker attempts to reduce the size of a program by finding shorter
25478 versions of various instructions. Disabled by default.
25480 @item -mint-register=@var{N}
25481 @opindex mint-register
25482 Specify the number of registers to reserve for fast interrupt handler
25483 functions. The value @var{N} can be between 0 and 4. A value of 1
25484 means that register @code{r13} is reserved for the exclusive use
25485 of fast interrupt handlers. A value of 2 reserves @code{r13} and
25486 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
25487 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
25488 A value of 0, the default, does not reserve any registers.
25490 @item -msave-acc-in-interrupts
25491 @opindex msave-acc-in-interrupts
25492 Specifies that interrupt handler functions should preserve the
25493 accumulator register. This is only necessary if normal code might use
25494 the accumulator register, for example because it performs 64-bit
25495 multiplications. The default is to ignore the accumulator as this
25496 makes the interrupt handlers faster.
25502 Enables the generation of position independent data. When enabled any
25503 access to constant data is done via an offset from a base address
25504 held in a register. This allows the location of constant data to be
25505 determined at run time without requiring the executable to be
25506 relocated, which is a benefit to embedded applications with tight
25507 memory constraints. Data that can be modified is not affected by this
25510 Note, using this feature reserves a register, usually @code{r13}, for
25511 the constant data base address. This can result in slower and/or
25512 larger code, especially in complicated functions.
25514 The actual register chosen to hold the constant data base address
25515 depends upon whether the @option{-msmall-data-limit} and/or the
25516 @option{-mint-register} command-line options are enabled. Starting
25517 with register @code{r13} and proceeding downwards, registers are
25518 allocated first to satisfy the requirements of @option{-mint-register},
25519 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
25520 is possible for the small data area register to be @code{r8} if both
25521 @option{-mint-register=4} and @option{-mpid} are specified on the
25524 By default this feature is not enabled. The default can be restored
25525 via the @option{-mno-pid} command-line option.
25527 @item -mno-warn-multiple-fast-interrupts
25528 @itemx -mwarn-multiple-fast-interrupts
25529 @opindex mno-warn-multiple-fast-interrupts
25530 @opindex mwarn-multiple-fast-interrupts
25531 Prevents GCC from issuing a warning message if it finds more than one
25532 fast interrupt handler when it is compiling a file. The default is to
25533 issue a warning for each extra fast interrupt handler found, as the RX
25534 only supports one such interrupt.
25536 @item -mallow-string-insns
25537 @itemx -mno-allow-string-insns
25538 @opindex mallow-string-insns
25539 @opindex mno-allow-string-insns
25540 Enables or disables the use of the string manipulation instructions
25541 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
25542 @code{SWHILE} and also the @code{RMPA} instruction. These
25543 instructions may prefetch data, which is not safe to do if accessing
25544 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
25545 for more information).
25547 The default is to allow these instructions, but it is not possible for
25548 GCC to reliably detect all circumstances where a string instruction
25549 might be used to access an I/O register, so their use cannot be
25550 disabled automatically. Instead it is reliant upon the programmer to
25551 use the @option{-mno-allow-string-insns} option if their program
25552 accesses I/O space.
25554 When the instructions are enabled GCC defines the C preprocessor
25555 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
25556 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
25562 Use only (or not only) @code{JSR} instructions to access functions.
25563 This option can be used when code size exceeds the range of @code{BSR}
25564 instructions. Note that @option{-mno-jsr} does not mean to not use
25565 @code{JSR} but instead means that any type of branch may be used.
25568 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
25569 has special significance to the RX port when used with the
25570 @code{interrupt} function attribute. This attribute indicates a
25571 function intended to process fast interrupts. GCC ensures
25572 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
25573 and/or @code{r13} and only provided that the normal use of the
25574 corresponding registers have been restricted via the
25575 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
25578 @node S/390 and zSeries Options
25579 @subsection S/390 and zSeries Options
25580 @cindex S/390 and zSeries Options
25582 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
25586 @itemx -msoft-float
25587 @opindex mhard-float
25588 @opindex msoft-float
25589 Use (do not use) the hardware floating-point instructions and registers
25590 for floating-point operations. When @option{-msoft-float} is specified,
25591 functions in @file{libgcc.a} are used to perform floating-point
25592 operations. When @option{-mhard-float} is specified, the compiler
25593 generates IEEE floating-point instructions. This is the default.
25596 @itemx -mno-hard-dfp
25598 @opindex mno-hard-dfp
25599 Use (do not use) the hardware decimal-floating-point instructions for
25600 decimal-floating-point operations. When @option{-mno-hard-dfp} is
25601 specified, functions in @file{libgcc.a} are used to perform
25602 decimal-floating-point operations. When @option{-mhard-dfp} is
25603 specified, the compiler generates decimal-floating-point hardware
25604 instructions. This is the default for @option{-march=z9-ec} or higher.
25606 @item -mlong-double-64
25607 @itemx -mlong-double-128
25608 @opindex mlong-double-64
25609 @opindex mlong-double-128
25610 These switches control the size of @code{long double} type. A size
25611 of 64 bits makes the @code{long double} type equivalent to the @code{double}
25612 type. This is the default.
25615 @itemx -mno-backchain
25616 @opindex mbackchain
25617 @opindex mno-backchain
25618 Store (do not store) the address of the caller's frame as backchain pointer
25619 into the callee's stack frame.
25620 A backchain may be needed to allow debugging using tools that do not understand
25621 DWARF call frame information.
25622 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
25623 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
25624 the backchain is placed into the topmost word of the 96/160 byte register
25627 In general, code compiled with @option{-mbackchain} is call-compatible with
25628 code compiled with @option{-mmo-backchain}; however, use of the backchain
25629 for debugging purposes usually requires that the whole binary is built with
25630 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
25631 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25632 to build a linux kernel use @option{-msoft-float}.
25634 The default is to not maintain the backchain.
25636 @item -mpacked-stack
25637 @itemx -mno-packed-stack
25638 @opindex mpacked-stack
25639 @opindex mno-packed-stack
25640 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
25641 specified, the compiler uses the all fields of the 96/160 byte register save
25642 area only for their default purpose; unused fields still take up stack space.
25643 When @option{-mpacked-stack} is specified, register save slots are densely
25644 packed at the top of the register save area; unused space is reused for other
25645 purposes, allowing for more efficient use of the available stack space.
25646 However, when @option{-mbackchain} is also in effect, the topmost word of
25647 the save area is always used to store the backchain, and the return address
25648 register is always saved two words below the backchain.
25650 As long as the stack frame backchain is not used, code generated with
25651 @option{-mpacked-stack} is call-compatible with code generated with
25652 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
25653 S/390 or zSeries generated code that uses the stack frame backchain at run
25654 time, not just for debugging purposes. Such code is not call-compatible
25655 with code compiled with @option{-mpacked-stack}. Also, note that the
25656 combination of @option{-mbackchain},
25657 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
25658 to build a linux kernel use @option{-msoft-float}.
25660 The default is to not use the packed stack layout.
25663 @itemx -mno-small-exec
25664 @opindex msmall-exec
25665 @opindex mno-small-exec
25666 Generate (or do not generate) code using the @code{bras} instruction
25667 to do subroutine calls.
25668 This only works reliably if the total executable size does not
25669 exceed 64k. The default is to use the @code{basr} instruction instead,
25670 which does not have this limitation.
25676 When @option{-m31} is specified, generate code compliant to the
25677 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
25678 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
25679 particular to generate 64-bit instructions. For the @samp{s390}
25680 targets, the default is @option{-m31}, while the @samp{s390x}
25681 targets default to @option{-m64}.
25687 When @option{-mzarch} is specified, generate code using the
25688 instructions available on z/Architecture.
25689 When @option{-mesa} is specified, generate code using the
25690 instructions available on ESA/390. Note that @option{-mesa} is
25691 not possible with @option{-m64}.
25692 When generating code compliant to the GNU/Linux for S/390 ABI,
25693 the default is @option{-mesa}. When generating code compliant
25694 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
25700 The @option{-mhtm} option enables a set of builtins making use of
25701 instructions available with the transactional execution facility
25702 introduced with the IBM zEnterprise EC12 machine generation
25703 @ref{S/390 System z Built-in Functions}.
25704 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
25710 When @option{-mvx} is specified, generate code using the instructions
25711 available with the vector extension facility introduced with the IBM
25712 z13 machine generation.
25713 This option changes the ABI for some vector type values with regard to
25714 alignment and calling conventions. In case vector type values are
25715 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
25716 command will be added to mark the resulting binary with the ABI used.
25717 @option{-mvx} is enabled by default when using @option{-march=z13}.
25720 @itemx -mno-zvector
25722 @opindex mno-zvector
25723 The @option{-mzvector} option enables vector language extensions and
25724 builtins using instructions available with the vector extension
25725 facility introduced with the IBM z13 machine generation.
25726 This option adds support for @samp{vector} to be used as a keyword to
25727 define vector type variables and arguments. @samp{vector} is only
25728 available when GNU extensions are enabled. It will not be expanded
25729 when requesting strict standard compliance e.g.@: with @option{-std=c99}.
25730 In addition to the GCC low-level builtins @option{-mzvector} enables
25731 a set of builtins added for compatibility with AltiVec-style
25732 implementations like Power and Cell. In order to make use of these
25733 builtins the header file @file{vecintrin.h} needs to be included.
25734 @option{-mzvector} is disabled by default.
25740 Generate (or do not generate) code using the @code{mvcle} instruction
25741 to perform block moves. When @option{-mno-mvcle} is specified,
25742 use a @code{mvc} loop instead. This is the default unless optimizing for
25749 Print (or do not print) additional debug information when compiling.
25750 The default is to not print debug information.
25752 @item -march=@var{cpu-type}
25754 Generate code that runs on @var{cpu-type}, which is the name of a
25755 system representing a certain processor type. Possible values for
25756 @var{cpu-type} are @samp{z900}/@samp{arch5}, @samp{z990}/@samp{arch6},
25757 @samp{z9-109}, @samp{z9-ec}/@samp{arch7}, @samp{z10}/@samp{arch8},
25758 @samp{z196}/@samp{arch9}, @samp{zEC12}, @samp{z13}/@samp{arch11},
25759 @samp{z14}/@samp{arch12}, and @samp{native}.
25761 The default is @option{-march=z900}.
25763 Specifying @samp{native} as cpu type can be used to select the best
25764 architecture option for the host processor.
25765 @option{-march=native} has no effect if GCC does not recognize the
25768 @item -mtune=@var{cpu-type}
25770 Tune to @var{cpu-type} everything applicable about the generated code,
25771 except for the ABI and the set of available instructions.
25772 The list of @var{cpu-type} values is the same as for @option{-march}.
25773 The default is the value used for @option{-march}.
25776 @itemx -mno-tpf-trace
25777 @opindex mtpf-trace
25778 @opindex mno-tpf-trace
25779 Generate code that adds (does not add) in TPF OS specific branches to trace
25780 routines in the operating system. This option is off by default, even
25781 when compiling for the TPF OS@.
25784 @itemx -mno-fused-madd
25785 @opindex mfused-madd
25786 @opindex mno-fused-madd
25787 Generate code that uses (does not use) the floating-point multiply and
25788 accumulate instructions. These instructions are generated by default if
25789 hardware floating point is used.
25791 @item -mwarn-framesize=@var{framesize}
25792 @opindex mwarn-framesize
25793 Emit a warning if the current function exceeds the given frame size. Because
25794 this is a compile-time check it doesn't need to be a real problem when the program
25795 runs. It is intended to identify functions that most probably cause
25796 a stack overflow. It is useful to be used in an environment with limited stack
25797 size e.g.@: the linux kernel.
25799 @item -mwarn-dynamicstack
25800 @opindex mwarn-dynamicstack
25801 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
25802 arrays. This is generally a bad idea with a limited stack size.
25804 @item -mstack-guard=@var{stack-guard}
25805 @itemx -mstack-size=@var{stack-size}
25806 @opindex mstack-guard
25807 @opindex mstack-size
25808 If these options are provided the S/390 back end emits additional instructions in
25809 the function prologue that trigger a trap if the stack size is @var{stack-guard}
25810 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
25811 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
25812 the frame size of the compiled function is chosen.
25813 These options are intended to be used to help debugging stack overflow problems.
25814 The additionally emitted code causes only little overhead and hence can also be
25815 used in production-like systems without greater performance degradation. The given
25816 values have to be exact powers of 2 and @var{stack-size} has to be greater than
25817 @var{stack-guard} without exceeding 64k.
25818 In order to be efficient the extra code makes the assumption that the stack starts
25819 at an address aligned to the value given by @var{stack-size}.
25820 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
25822 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
25824 If the hotpatch option is enabled, a ``hot-patching'' function
25825 prologue is generated for all functions in the compilation unit.
25826 The funtion label is prepended with the given number of two-byte
25827 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
25828 the label, 2 * @var{post-halfwords} bytes are appended, using the
25829 largest NOP like instructions the architecture allows (maximum
25832 If both arguments are zero, hotpatching is disabled.
25834 This option can be overridden for individual functions with the
25835 @code{hotpatch} attribute.
25838 @node Score Options
25839 @subsection Score Options
25840 @cindex Score Options
25842 These options are defined for Score implementations:
25847 Compile code for big-endian mode. This is the default.
25851 Compile code for little-endian mode.
25855 Disable generation of @code{bcnz} instructions.
25859 Enable generation of unaligned load and store instructions.
25863 Enable the use of multiply-accumulate instructions. Disabled by default.
25867 Specify the SCORE5 as the target architecture.
25871 Specify the SCORE5U of the target architecture.
25875 Specify the SCORE7 as the target architecture. This is the default.
25879 Specify the SCORE7D as the target architecture.
25883 @subsection SH Options
25885 These @samp{-m} options are defined for the SH implementations:
25890 Generate code for the SH1.
25894 Generate code for the SH2.
25897 Generate code for the SH2e.
25901 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
25902 that the floating-point unit is not used.
25904 @item -m2a-single-only
25905 @opindex m2a-single-only
25906 Generate code for the SH2a-FPU, in such a way that no double-precision
25907 floating-point operations are used.
25910 @opindex m2a-single
25911 Generate code for the SH2a-FPU assuming the floating-point unit is in
25912 single-precision mode by default.
25916 Generate code for the SH2a-FPU assuming the floating-point unit is in
25917 double-precision mode by default.
25921 Generate code for the SH3.
25925 Generate code for the SH3e.
25929 Generate code for the SH4 without a floating-point unit.
25931 @item -m4-single-only
25932 @opindex m4-single-only
25933 Generate code for the SH4 with a floating-point unit that only
25934 supports single-precision arithmetic.
25938 Generate code for the SH4 assuming the floating-point unit is in
25939 single-precision mode by default.
25943 Generate code for the SH4.
25947 Generate code for SH4-100.
25949 @item -m4-100-nofpu
25950 @opindex m4-100-nofpu
25951 Generate code for SH4-100 in such a way that the
25952 floating-point unit is not used.
25954 @item -m4-100-single
25955 @opindex m4-100-single
25956 Generate code for SH4-100 assuming the floating-point unit is in
25957 single-precision mode by default.
25959 @item -m4-100-single-only
25960 @opindex m4-100-single-only
25961 Generate code for SH4-100 in such a way that no double-precision
25962 floating-point operations are used.
25966 Generate code for SH4-200.
25968 @item -m4-200-nofpu
25969 @opindex m4-200-nofpu
25970 Generate code for SH4-200 without in such a way that the
25971 floating-point unit is not used.
25973 @item -m4-200-single
25974 @opindex m4-200-single
25975 Generate code for SH4-200 assuming the floating-point unit is in
25976 single-precision mode by default.
25978 @item -m4-200-single-only
25979 @opindex m4-200-single-only
25980 Generate code for SH4-200 in such a way that no double-precision
25981 floating-point operations are used.
25985 Generate code for SH4-300.
25987 @item -m4-300-nofpu
25988 @opindex m4-300-nofpu
25989 Generate code for SH4-300 without in such a way that the
25990 floating-point unit is not used.
25992 @item -m4-300-single
25993 @opindex m4-300-single
25994 Generate code for SH4-300 in such a way that no double-precision
25995 floating-point operations are used.
25997 @item -m4-300-single-only
25998 @opindex m4-300-single-only
25999 Generate code for SH4-300 in such a way that no double-precision
26000 floating-point operations are used.
26004 Generate code for SH4-340 (no MMU, no FPU).
26008 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
26013 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
26014 floating-point unit is not used.
26016 @item -m4a-single-only
26017 @opindex m4a-single-only
26018 Generate code for the SH4a, in such a way that no double-precision
26019 floating-point operations are used.
26022 @opindex m4a-single
26023 Generate code for the SH4a assuming the floating-point unit is in
26024 single-precision mode by default.
26028 Generate code for the SH4a.
26032 Same as @option{-m4a-nofpu}, except that it implicitly passes
26033 @option{-dsp} to the assembler. GCC doesn't generate any DSP
26034 instructions at the moment.
26038 Compile code for the processor in big-endian mode.
26042 Compile code for the processor in little-endian mode.
26046 Align doubles at 64-bit boundaries. Note that this changes the calling
26047 conventions, and thus some functions from the standard C library do
26048 not work unless you recompile it first with @option{-mdalign}.
26052 Shorten some address references at link time, when possible; uses the
26053 linker option @option{-relax}.
26057 Use 32-bit offsets in @code{switch} tables. The default is to use
26062 Enable the use of bit manipulation instructions on SH2A.
26066 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
26067 alignment constraints.
26071 Comply with the calling conventions defined by Renesas.
26074 @opindex mno-renesas
26075 Comply with the calling conventions defined for GCC before the Renesas
26076 conventions were available. This option is the default for all
26077 targets of the SH toolchain.
26080 @opindex mnomacsave
26081 Mark the @code{MAC} register as call-clobbered, even if
26082 @option{-mrenesas} is given.
26088 Control the IEEE compliance of floating-point comparisons, which affects the
26089 handling of cases where the result of a comparison is unordered. By default
26090 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
26091 enabled @option{-mno-ieee} is implicitly set, which results in faster
26092 floating-point greater-equal and less-equal comparisons. The implicit settings
26093 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
26095 @item -minline-ic_invalidate
26096 @opindex minline-ic_invalidate
26097 Inline code to invalidate instruction cache entries after setting up
26098 nested function trampolines.
26099 This option has no effect if @option{-musermode} is in effect and the selected
26100 code generation option (e.g.@: @option{-m4}) does not allow the use of the @code{icbi}
26102 If the selected code generation option does not allow the use of the @code{icbi}
26103 instruction, and @option{-musermode} is not in effect, the inlined code
26104 manipulates the instruction cache address array directly with an associative
26105 write. This not only requires privileged mode at run time, but it also
26106 fails if the cache line had been mapped via the TLB and has become unmapped.
26110 Dump instruction size and location in the assembly code.
26113 @opindex mpadstruct
26114 This option is deprecated. It pads structures to multiple of 4 bytes,
26115 which is incompatible with the SH ABI@.
26117 @item -matomic-model=@var{model}
26118 @opindex matomic-model=@var{model}
26119 Sets the model of atomic operations and additional parameters as a comma
26120 separated list. For details on the atomic built-in functions see
26121 @ref{__atomic Builtins}. The following models and parameters are supported:
26126 Disable compiler generated atomic sequences and emit library calls for atomic
26127 operations. This is the default if the target is not @code{sh*-*-linux*}.
26130 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
26131 built-in functions. The generated atomic sequences require additional support
26132 from the interrupt/exception handling code of the system and are only suitable
26133 for SH3* and SH4* single-core systems. This option is enabled by default when
26134 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
26135 this option also partially utilizes the hardware atomic instructions
26136 @code{movli.l} and @code{movco.l} to create more efficient code, unless
26137 @samp{strict} is specified.
26140 Generate software atomic sequences that use a variable in the thread control
26141 block. This is a variation of the gUSA sequences which can also be used on
26142 SH1* and SH2* targets. The generated atomic sequences require additional
26143 support from the interrupt/exception handling code of the system and are only
26144 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
26145 parameter has to be specified as well.
26148 Generate software atomic sequences that temporarily disable interrupts by
26149 setting @code{SR.IMASK = 1111}. This model works only when the program runs
26150 in privileged mode and is only suitable for single-core systems. Additional
26151 support from the interrupt/exception handling code of the system is not
26152 required. This model is enabled by default when the target is
26153 @code{sh*-*-linux*} and SH1* or SH2*.
26156 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
26157 instructions only. This is only available on SH4A and is suitable for
26158 multi-core systems. Since the hardware instructions support only 32 bit atomic
26159 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
26160 Code compiled with this option is also compatible with other software
26161 atomic model interrupt/exception handling systems if executed on an SH4A
26162 system. Additional support from the interrupt/exception handling code of the
26163 system is not required for this model.
26166 This parameter specifies the offset in bytes of the variable in the thread
26167 control block structure that should be used by the generated atomic sequences
26168 when the @samp{soft-tcb} model has been selected. For other models this
26169 parameter is ignored. The specified value must be an integer multiple of four
26170 and in the range 0-1020.
26173 This parameter prevents mixed usage of multiple atomic models, even if they
26174 are compatible, and makes the compiler generate atomic sequences of the
26175 specified model only.
26181 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
26182 Notice that depending on the particular hardware and software configuration
26183 this can degrade overall performance due to the operand cache line flushes
26184 that are implied by the @code{tas.b} instruction. On multi-core SH4A
26185 processors the @code{tas.b} instruction must be used with caution since it
26186 can result in data corruption for certain cache configurations.
26189 @opindex mprefergot
26190 When generating position-independent code, emit function calls using
26191 the Global Offset Table instead of the Procedure Linkage Table.
26194 @itemx -mno-usermode
26196 @opindex mno-usermode
26197 Don't allow (allow) the compiler generating privileged mode code. Specifying
26198 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
26199 inlined code would not work in user mode. @option{-musermode} is the default
26200 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
26201 @option{-musermode} has no effect, since there is no user mode.
26203 @item -multcost=@var{number}
26204 @opindex multcost=@var{number}
26205 Set the cost to assume for a multiply insn.
26207 @item -mdiv=@var{strategy}
26208 @opindex mdiv=@var{strategy}
26209 Set the division strategy to be used for integer division operations.
26210 @var{strategy} can be one of:
26215 Calls a library function that uses the single-step division instruction
26216 @code{div1} to perform the operation. Division by zero calculates an
26217 unspecified result and does not trap. This is the default except for SH4,
26218 SH2A and SHcompact.
26221 Calls a library function that performs the operation in double precision
26222 floating point. Division by zero causes a floating-point exception. This is
26223 the default for SHcompact with FPU. Specifying this for targets that do not
26224 have a double precision FPU defaults to @code{call-div1}.
26227 Calls a library function that uses a lookup table for small divisors and
26228 the @code{div1} instruction with case distinction for larger divisors. Division
26229 by zero calculates an unspecified result and does not trap. This is the default
26230 for SH4. Specifying this for targets that do not have dynamic shift
26231 instructions defaults to @code{call-div1}.
26235 When a division strategy has not been specified the default strategy is
26236 selected based on the current target. For SH2A the default strategy is to
26237 use the @code{divs} and @code{divu} instructions instead of library function
26240 @item -maccumulate-outgoing-args
26241 @opindex maccumulate-outgoing-args
26242 Reserve space once for outgoing arguments in the function prologue rather
26243 than around each call. Generally beneficial for performance and size. Also
26244 needed for unwinding to avoid changing the stack frame around conditional code.
26246 @item -mdivsi3_libfunc=@var{name}
26247 @opindex mdivsi3_libfunc=@var{name}
26248 Set the name of the library function used for 32-bit signed division to
26250 This only affects the name used in the @samp{call} division strategies, and
26251 the compiler still expects the same sets of input/output/clobbered registers as
26252 if this option were not present.
26254 @item -mfixed-range=@var{register-range}
26255 @opindex mfixed-range
26256 Generate code treating the given register range as fixed registers.
26257 A fixed register is one that the register allocator cannot use. This is
26258 useful when compiling kernel code. A register range is specified as
26259 two registers separated by a dash. Multiple register ranges can be
26260 specified separated by a comma.
26262 @item -mbranch-cost=@var{num}
26263 @opindex mbranch-cost=@var{num}
26264 Assume @var{num} to be the cost for a branch instruction. Higher numbers
26265 make the compiler try to generate more branch-free code if possible.
26266 If not specified the value is selected depending on the processor type that
26267 is being compiled for.
26270 @itemx -mno-zdcbranch
26271 @opindex mzdcbranch
26272 @opindex mno-zdcbranch
26273 Assume (do not assume) that zero displacement conditional branch instructions
26274 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
26275 compiler prefers zero displacement branch code sequences. This is
26276 enabled by default when generating code for SH4 and SH4A. It can be explicitly
26277 disabled by specifying @option{-mno-zdcbranch}.
26279 @item -mcbranch-force-delay-slot
26280 @opindex mcbranch-force-delay-slot
26281 Force the usage of delay slots for conditional branches, which stuffs the delay
26282 slot with a @code{nop} if a suitable instruction cannot be found. By default
26283 this option is disabled. It can be enabled to work around hardware bugs as
26284 found in the original SH7055.
26287 @itemx -mno-fused-madd
26288 @opindex mfused-madd
26289 @opindex mno-fused-madd
26290 Generate code that uses (does not use) the floating-point multiply and
26291 accumulate instructions. These instructions are generated by default
26292 if hardware floating point is used. The machine-dependent
26293 @option{-mfused-madd} option is now mapped to the machine-independent
26294 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
26295 mapped to @option{-ffp-contract=off}.
26301 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
26302 and cosine approximations. The option @option{-mfsca} must be used in
26303 combination with @option{-funsafe-math-optimizations}. It is enabled by default
26304 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
26305 approximations even if @option{-funsafe-math-optimizations} is in effect.
26311 Allow or disallow the compiler to emit the @code{fsrra} instruction for
26312 reciprocal square root approximations. The option @option{-mfsrra} must be used
26313 in combination with @option{-funsafe-math-optimizations} and
26314 @option{-ffinite-math-only}. It is enabled by default when generating code for
26315 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
26316 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
26319 @item -mpretend-cmove
26320 @opindex mpretend-cmove
26321 Prefer zero-displacement conditional branches for conditional move instruction
26322 patterns. This can result in faster code on the SH4 processor.
26326 Generate code using the FDPIC ABI.
26330 @node Solaris 2 Options
26331 @subsection Solaris 2 Options
26332 @cindex Solaris 2 options
26334 These @samp{-m} options are supported on Solaris 2:
26337 @item -mclear-hwcap
26338 @opindex mclear-hwcap
26339 @option{-mclear-hwcap} tells the compiler to remove the hardware
26340 capabilities generated by the Solaris assembler. This is only necessary
26341 when object files use ISA extensions not supported by the current
26342 machine, but check at runtime whether or not to use them.
26344 @item -mimpure-text
26345 @opindex mimpure-text
26346 @option{-mimpure-text}, used in addition to @option{-shared}, tells
26347 the compiler to not pass @option{-z text} to the linker when linking a
26348 shared object. Using this option, you can link position-dependent
26349 code into a shared object.
26351 @option{-mimpure-text} suppresses the ``relocations remain against
26352 allocatable but non-writable sections'' linker error message.
26353 However, the necessary relocations trigger copy-on-write, and the
26354 shared object is not actually shared across processes. Instead of
26355 using @option{-mimpure-text}, you should compile all source code with
26356 @option{-fpic} or @option{-fPIC}.
26360 These switches are supported in addition to the above on Solaris 2:
26365 This is a synonym for @option{-pthread}.
26368 @node SPARC Options
26369 @subsection SPARC Options
26370 @cindex SPARC options
26372 These @samp{-m} options are supported on the SPARC:
26375 @item -mno-app-regs
26377 @opindex mno-app-regs
26379 Specify @option{-mapp-regs} to generate output using the global registers
26380 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
26381 global register 1, each global register 2 through 4 is then treated as an
26382 allocable register that is clobbered by function calls. This is the default.
26384 To be fully SVR4 ABI-compliant at the cost of some performance loss,
26385 specify @option{-mno-app-regs}. You should compile libraries and system
26386 software with this option.
26392 With @option{-mflat}, the compiler does not generate save/restore instructions
26393 and uses a ``flat'' or single register window model. This model is compatible
26394 with the regular register window model. The local registers and the input
26395 registers (0--5) are still treated as ``call-saved'' registers and are
26396 saved on the stack as needed.
26398 With @option{-mno-flat} (the default), the compiler generates save/restore
26399 instructions (except for leaf functions). This is the normal operating mode.
26402 @itemx -mhard-float
26404 @opindex mhard-float
26405 Generate output containing floating-point instructions. This is the
26409 @itemx -msoft-float
26411 @opindex msoft-float
26412 Generate output containing library calls for floating point.
26413 @strong{Warning:} the requisite libraries are not available for all SPARC
26414 targets. Normally the facilities of the machine's usual C compiler are
26415 used, but this cannot be done directly in cross-compilation. You must make
26416 your own arrangements to provide suitable library functions for
26417 cross-compilation. The embedded targets @samp{sparc-*-aout} and
26418 @samp{sparclite-*-*} do provide software floating-point support.
26420 @option{-msoft-float} changes the calling convention in the output file;
26421 therefore, it is only useful if you compile @emph{all} of a program with
26422 this option. In particular, you need to compile @file{libgcc.a}, the
26423 library that comes with GCC, with @option{-msoft-float} in order for
26426 @item -mhard-quad-float
26427 @opindex mhard-quad-float
26428 Generate output containing quad-word (long double) floating-point
26431 @item -msoft-quad-float
26432 @opindex msoft-quad-float
26433 Generate output containing library calls for quad-word (long double)
26434 floating-point instructions. The functions called are those specified
26435 in the SPARC ABI@. This is the default.
26437 As of this writing, there are no SPARC implementations that have hardware
26438 support for the quad-word floating-point instructions. They all invoke
26439 a trap handler for one of these instructions, and then the trap handler
26440 emulates the effect of the instruction. Because of the trap handler overhead,
26441 this is much slower than calling the ABI library routines. Thus the
26442 @option{-msoft-quad-float} option is the default.
26444 @item -mno-unaligned-doubles
26445 @itemx -munaligned-doubles
26446 @opindex mno-unaligned-doubles
26447 @opindex munaligned-doubles
26448 Assume that doubles have 8-byte alignment. This is the default.
26450 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
26451 alignment only if they are contained in another type, or if they have an
26452 absolute address. Otherwise, it assumes they have 4-byte alignment.
26453 Specifying this option avoids some rare compatibility problems with code
26454 generated by other compilers. It is not the default because it results
26455 in a performance loss, especially for floating-point code.
26458 @itemx -mno-user-mode
26459 @opindex muser-mode
26460 @opindex mno-user-mode
26461 Do not generate code that can only run in supervisor mode. This is relevant
26462 only for the @code{casa} instruction emitted for the LEON3 processor. This
26465 @item -mfaster-structs
26466 @itemx -mno-faster-structs
26467 @opindex mfaster-structs
26468 @opindex mno-faster-structs
26469 With @option{-mfaster-structs}, the compiler assumes that structures
26470 should have 8-byte alignment. This enables the use of pairs of
26471 @code{ldd} and @code{std} instructions for copies in structure
26472 assignment, in place of twice as many @code{ld} and @code{st} pairs.
26473 However, the use of this changed alignment directly violates the SPARC
26474 ABI@. Thus, it's intended only for use on targets where the developer
26475 acknowledges that their resulting code is not directly in line with
26476 the rules of the ABI@.
26478 @item -mstd-struct-return
26479 @itemx -mno-std-struct-return
26480 @opindex mstd-struct-return
26481 @opindex mno-std-struct-return
26482 With @option{-mstd-struct-return}, the compiler generates checking code
26483 in functions returning structures or unions to detect size mismatches
26484 between the two sides of function calls, as per the 32-bit ABI@.
26486 The default is @option{-mno-std-struct-return}. This option has no effect
26493 Enable Local Register Allocation. This is the default for SPARC since GCC 7
26494 so @option{-mno-lra} needs to be passed to get old Reload.
26496 @item -mcpu=@var{cpu_type}
26498 Set the instruction set, register set, and instruction scheduling parameters
26499 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
26500 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
26501 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
26502 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
26503 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
26504 @samp{niagara3}, @samp{niagara4}, @samp{niagara7} and @samp{m8}.
26506 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
26507 which selects the best architecture option for the host processor.
26508 @option{-mcpu=native} has no effect if GCC does not recognize
26511 Default instruction scheduling parameters are used for values that select
26512 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
26513 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
26515 Here is a list of each supported architecture and their supported
26523 supersparc, hypersparc, leon, leon3
26526 f930, f934, sparclite86x
26532 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4,
26536 By default (unless configured otherwise), GCC generates code for the V7
26537 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
26538 additionally optimizes it for the Cypress CY7C602 chip, as used in the
26539 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
26540 SPARCStation 1, 2, IPX etc.
26542 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
26543 architecture. The only difference from V7 code is that the compiler emits
26544 the integer multiply and integer divide instructions which exist in SPARC-V8
26545 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
26546 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
26549 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
26550 the SPARC architecture. This adds the integer multiply, integer divide step
26551 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
26552 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
26553 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
26554 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
26555 MB86934 chip, which is the more recent SPARClite with FPU@.
26557 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
26558 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
26559 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
26560 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
26561 optimizes it for the TEMIC SPARClet chip.
26563 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
26564 architecture. This adds 64-bit integer and floating-point move instructions,
26565 3 additional floating-point condition code registers and conditional move
26566 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
26567 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
26568 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
26569 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
26570 @option{-mcpu=niagara}, the compiler additionally optimizes it for
26571 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
26572 additionally optimizes it for Sun UltraSPARC T2 chips. With
26573 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
26574 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
26575 additionally optimizes it for Sun UltraSPARC T4 chips. With
26576 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
26577 Oracle SPARC M7 chips. With @option{-mcpu=m8}, the compiler
26578 additionally optimizes it for Oracle M8 chips.
26580 @item -mtune=@var{cpu_type}
26582 Set the instruction scheduling parameters for machine type
26583 @var{cpu_type}, but do not set the instruction set or register set that the
26584 option @option{-mcpu=@var{cpu_type}} does.
26586 The same values for @option{-mcpu=@var{cpu_type}} can be used for
26587 @option{-mtune=@var{cpu_type}}, but the only useful values are those
26588 that select a particular CPU implementation. Those are
26589 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
26590 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
26591 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
26592 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
26593 @samp{niagara4}, @samp{niagara7} and @samp{m8}. With native Solaris
26594 and GNU/Linux toolchains, @samp{native} can also be used.
26599 @opindex mno-v8plus
26600 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
26601 difference from the V8 ABI is that the global and out registers are
26602 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
26603 mode for all SPARC-V9 processors.
26609 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
26610 Visual Instruction Set extensions. The default is @option{-mno-vis}.
26616 With @option{-mvis2}, GCC generates code that takes advantage of
26617 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
26618 default is @option{-mvis2} when targeting a cpu that supports such
26619 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
26620 also sets @option{-mvis}.
26626 With @option{-mvis3}, GCC generates code that takes advantage of
26627 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
26628 default is @option{-mvis3} when targeting a cpu that supports such
26629 instructions, such as niagara-3 and later. Setting @option{-mvis3}
26630 also sets @option{-mvis2} and @option{-mvis}.
26636 With @option{-mvis4}, GCC generates code that takes advantage of
26637 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
26638 default is @option{-mvis4} when targeting a cpu that supports such
26639 instructions, such as niagara-7 and later. Setting @option{-mvis4}
26640 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
26646 With @option{-mvis4b}, GCC generates code that takes advantage of
26647 version 4.0 of the UltraSPARC Visual Instruction Set extensions, plus
26648 the additional VIS instructions introduced in the Oracle SPARC
26649 Architecture 2017. The default is @option{-mvis4b} when targeting a
26650 cpu that supports such instructions, such as m8 and later. Setting
26651 @option{-mvis4b} also sets @option{-mvis4}, @option{-mvis3},
26652 @option{-mvis2} and @option{-mvis}.
26657 @opindex mno-cbcond
26658 With @option{-mcbcond}, GCC generates code that takes advantage of the UltraSPARC
26659 Compare-and-Branch-on-Condition instructions. The default is @option{-mcbcond}
26660 when targeting a CPU that supports such instructions, such as Niagara-4 and
26667 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
26668 Fused Multiply-Add Floating-point instructions. The default is @option{-mfmaf}
26669 when targeting a CPU that supports such instructions, such as Niagara-3 and
26675 @opindex mno-fsmuld
26676 With @option{-mfsmuld}, GCC generates code that takes advantage of the
26677 Floating-point Multiply Single to Double (FsMULd) instruction. The default is
26678 @option{-mfsmuld} when targeting a CPU supporting the architecture versions V8
26679 or V9 with FPU except @option{-mcpu=leon}.
26685 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
26686 Population Count instruction. The default is @option{-mpopc}
26687 when targeting a CPU that supports such an instruction, such as Niagara-2 and
26694 With @option{-msubxc}, GCC generates code that takes advantage of the UltraSPARC
26695 Subtract-Extended-with-Carry instruction. The default is @option{-msubxc}
26696 when targeting a CPU that supports such an instruction, such as Niagara-7 and
26700 @opindex mfix-at697f
26701 Enable the documented workaround for the single erratum of the Atmel AT697F
26702 processor (which corresponds to erratum #13 of the AT697E processor).
26705 @opindex mfix-ut699
26706 Enable the documented workarounds for the floating-point errata and the data
26707 cache nullify errata of the UT699 processor.
26710 @opindex mfix-ut700
26711 Enable the documented workaround for the back-to-back store errata of
26712 the UT699E/UT700 processor.
26714 @item -mfix-gr712rc
26715 @opindex mfix-gr712rc
26716 Enable the documented workaround for the back-to-back store errata of
26717 the GR712RC processor.
26720 These @samp{-m} options are supported in addition to the above
26721 on SPARC-V9 processors in 64-bit environments:
26728 Generate code for a 32-bit or 64-bit environment.
26729 The 32-bit environment sets int, long and pointer to 32 bits.
26730 The 64-bit environment sets int to 32 bits and long and pointer
26733 @item -mcmodel=@var{which}
26735 Set the code model to one of
26739 The Medium/Low code model: 64-bit addresses, programs
26740 must be linked in the low 32 bits of memory. Programs can be statically
26741 or dynamically linked.
26744 The Medium/Middle code model: 64-bit addresses, programs
26745 must be linked in the low 44 bits of memory, the text and data segments must
26746 be less than 2GB in size and the data segment must be located within 2GB of
26750 The Medium/Anywhere code model: 64-bit addresses, programs
26751 may be linked anywhere in memory, the text and data segments must be less
26752 than 2GB in size and the data segment must be located within 2GB of the
26756 The Medium/Anywhere code model for embedded systems:
26757 64-bit addresses, the text and data segments must be less than 2GB in
26758 size, both starting anywhere in memory (determined at link time). The
26759 global register %g4 points to the base of the data segment. Programs
26760 are statically linked and PIC is not supported.
26763 @item -mmemory-model=@var{mem-model}
26764 @opindex mmemory-model
26765 Set the memory model in force on the processor to one of
26769 The default memory model for the processor and operating system.
26772 Relaxed Memory Order
26775 Partial Store Order
26781 Sequential Consistency
26784 These memory models are formally defined in Appendix D of the SPARC-V9
26785 architecture manual, as set in the processor's @code{PSTATE.MM} field.
26788 @itemx -mno-stack-bias
26789 @opindex mstack-bias
26790 @opindex mno-stack-bias
26791 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
26792 frame pointer if present, are offset by @minus{}2047 which must be added back
26793 when making stack frame references. This is the default in 64-bit mode.
26794 Otherwise, assume no such offset is present.
26797 @node System V Options
26798 @subsection Options for System V
26800 These additional options are available on System V Release 4 for
26801 compatibility with other compilers on those systems:
26806 Create a shared object.
26807 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
26811 Identify the versions of each tool used by the compiler, in a
26812 @code{.ident} assembler directive in the output.
26816 Refrain from adding @code{.ident} directives to the output file (this is
26819 @item -YP,@var{dirs}
26821 Search the directories @var{dirs}, and no others, for libraries
26822 specified with @option{-l}.
26824 @item -Ym,@var{dir}
26826 Look in the directory @var{dir} to find the M4 preprocessor.
26827 The assembler uses this option.
26828 @c This is supposed to go with a -Yd for predefined M4 macro files, but
26829 @c the generic assembler that comes with Solaris takes just -Ym.
26832 @node TILE-Gx Options
26833 @subsection TILE-Gx Options
26834 @cindex TILE-Gx options
26836 These @samp{-m} options are supported on the TILE-Gx:
26839 @item -mcmodel=small
26840 @opindex mcmodel=small
26841 Generate code for the small model. The distance for direct calls is
26842 limited to 500M in either direction. PC-relative addresses are 32
26843 bits. Absolute addresses support the full address range.
26845 @item -mcmodel=large
26846 @opindex mcmodel=large
26847 Generate code for the large model. There is no limitation on call
26848 distance, pc-relative addresses, or absolute addresses.
26850 @item -mcpu=@var{name}
26852 Selects the type of CPU to be targeted. Currently the only supported
26853 type is @samp{tilegx}.
26859 Generate code for a 32-bit or 64-bit environment. The 32-bit
26860 environment sets int, long, and pointer to 32 bits. The 64-bit
26861 environment sets int to 32 bits and long and pointer to 64 bits.
26864 @itemx -mlittle-endian
26865 @opindex mbig-endian
26866 @opindex mlittle-endian
26867 Generate code in big/little endian mode, respectively.
26870 @node TILEPro Options
26871 @subsection TILEPro Options
26872 @cindex TILEPro options
26874 These @samp{-m} options are supported on the TILEPro:
26877 @item -mcpu=@var{name}
26879 Selects the type of CPU to be targeted. Currently the only supported
26880 type is @samp{tilepro}.
26884 Generate code for a 32-bit environment, which sets int, long, and
26885 pointer to 32 bits. This is the only supported behavior so the flag
26886 is essentially ignored.
26890 @subsection V850 Options
26891 @cindex V850 Options
26893 These @samp{-m} options are defined for V850 implementations:
26897 @itemx -mno-long-calls
26898 @opindex mlong-calls
26899 @opindex mno-long-calls
26900 Treat all calls as being far away (near). If calls are assumed to be
26901 far away, the compiler always loads the function's address into a
26902 register, and calls indirect through the pointer.
26908 Do not optimize (do optimize) basic blocks that use the same index
26909 pointer 4 or more times to copy pointer into the @code{ep} register, and
26910 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
26911 option is on by default if you optimize.
26913 @item -mno-prolog-function
26914 @itemx -mprolog-function
26915 @opindex mno-prolog-function
26916 @opindex mprolog-function
26917 Do not use (do use) external functions to save and restore registers
26918 at the prologue and epilogue of a function. The external functions
26919 are slower, but use less code space if more than one function saves
26920 the same number of registers. The @option{-mprolog-function} option
26921 is on by default if you optimize.
26925 Try to make the code as small as possible. At present, this just turns
26926 on the @option{-mep} and @option{-mprolog-function} options.
26928 @item -mtda=@var{n}
26930 Put static or global variables whose size is @var{n} bytes or less into
26931 the tiny data area that register @code{ep} points to. The tiny data
26932 area can hold up to 256 bytes in total (128 bytes for byte references).
26934 @item -msda=@var{n}
26936 Put static or global variables whose size is @var{n} bytes or less into
26937 the small data area that register @code{gp} points to. The small data
26938 area can hold up to 64 kilobytes.
26940 @item -mzda=@var{n}
26942 Put static or global variables whose size is @var{n} bytes or less into
26943 the first 32 kilobytes of memory.
26947 Specify that the target processor is the V850.
26951 Specify that the target processor is the V850E3V5. The preprocessor
26952 constant @code{__v850e3v5__} is defined if this option is used.
26956 Specify that the target processor is the V850E3V5. This is an alias for
26957 the @option{-mv850e3v5} option.
26961 Specify that the target processor is the V850E2V3. The preprocessor
26962 constant @code{__v850e2v3__} is defined if this option is used.
26966 Specify that the target processor is the V850E2. The preprocessor
26967 constant @code{__v850e2__} is defined if this option is used.
26971 Specify that the target processor is the V850E1. The preprocessor
26972 constants @code{__v850e1__} and @code{__v850e__} are defined if
26973 this option is used.
26977 Specify that the target processor is the V850ES. This is an alias for
26978 the @option{-mv850e1} option.
26982 Specify that the target processor is the V850E@. The preprocessor
26983 constant @code{__v850e__} is defined if this option is used.
26985 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
26986 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
26987 are defined then a default target processor is chosen and the
26988 relevant @samp{__v850*__} preprocessor constant is defined.
26990 The preprocessor constants @code{__v850} and @code{__v851__} are always
26991 defined, regardless of which processor variant is the target.
26993 @item -mdisable-callt
26994 @itemx -mno-disable-callt
26995 @opindex mdisable-callt
26996 @opindex mno-disable-callt
26997 This option suppresses generation of the @code{CALLT} instruction for the
26998 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
27001 This option is enabled by default when the RH850 ABI is
27002 in use (see @option{-mrh850-abi}), and disabled by default when the
27003 GCC ABI is in use. If @code{CALLT} instructions are being generated
27004 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
27010 Pass on (or do not pass on) the @option{-mrelax} command-line option
27014 @itemx -mno-long-jumps
27015 @opindex mlong-jumps
27016 @opindex mno-long-jumps
27017 Disable (or re-enable) the generation of PC-relative jump instructions.
27020 @itemx -mhard-float
27021 @opindex msoft-float
27022 @opindex mhard-float
27023 Disable (or re-enable) the generation of hardware floating point
27024 instructions. This option is only significant when the target
27025 architecture is @samp{V850E2V3} or higher. If hardware floating point
27026 instructions are being generated then the C preprocessor symbol
27027 @code{__FPU_OK__} is defined, otherwise the symbol
27028 @code{__NO_FPU__} is defined.
27032 Enables the use of the e3v5 LOOP instruction. The use of this
27033 instruction is not enabled by default when the e3v5 architecture is
27034 selected because its use is still experimental.
27038 @opindex mrh850-abi
27040 Enables support for the RH850 version of the V850 ABI. This is the
27041 default. With this version of the ABI the following rules apply:
27045 Integer sized structures and unions are returned via a memory pointer
27046 rather than a register.
27049 Large structures and unions (more than 8 bytes in size) are passed by
27053 Functions are aligned to 16-bit boundaries.
27056 The @option{-m8byte-align} command-line option is supported.
27059 The @option{-mdisable-callt} command-line option is enabled by
27060 default. The @option{-mno-disable-callt} command-line option is not
27064 When this version of the ABI is enabled the C preprocessor symbol
27065 @code{__V850_RH850_ABI__} is defined.
27069 Enables support for the old GCC version of the V850 ABI. With this
27070 version of the ABI the following rules apply:
27074 Integer sized structures and unions are returned in register @code{r10}.
27077 Large structures and unions (more than 8 bytes in size) are passed by
27081 Functions are aligned to 32-bit boundaries, unless optimizing for
27085 The @option{-m8byte-align} command-line option is not supported.
27088 The @option{-mdisable-callt} command-line option is supported but not
27089 enabled by default.
27092 When this version of the ABI is enabled the C preprocessor symbol
27093 @code{__V850_GCC_ABI__} is defined.
27095 @item -m8byte-align
27096 @itemx -mno-8byte-align
27097 @opindex m8byte-align
27098 @opindex mno-8byte-align
27099 Enables support for @code{double} and @code{long long} types to be
27100 aligned on 8-byte boundaries. The default is to restrict the
27101 alignment of all objects to at most 4-bytes. When
27102 @option{-m8byte-align} is in effect the C preprocessor symbol
27103 @code{__V850_8BYTE_ALIGN__} is defined.
27106 @opindex mbig-switch
27107 Generate code suitable for big switch tables. Use this option only if
27108 the assembler/linker complain about out of range branches within a switch
27113 This option causes r2 and r5 to be used in the code generated by
27114 the compiler. This setting is the default.
27116 @item -mno-app-regs
27117 @opindex mno-app-regs
27118 This option causes r2 and r5 to be treated as fixed registers.
27123 @subsection VAX Options
27124 @cindex VAX options
27126 These @samp{-m} options are defined for the VAX:
27131 Do not output certain jump instructions (@code{aobleq} and so on)
27132 that the Unix assembler for the VAX cannot handle across long
27137 Do output those jump instructions, on the assumption that the
27138 GNU assembler is being used.
27142 Output code for G-format floating-point numbers instead of D-format.
27145 @node Visium Options
27146 @subsection Visium Options
27147 @cindex Visium options
27153 A program which performs file I/O and is destined to run on an MCM target
27154 should be linked with this option. It causes the libraries libc.a and
27155 libdebug.a to be linked. The program should be run on the target under
27156 the control of the GDB remote debugging stub.
27160 A program which performs file I/O and is destined to run on the simulator
27161 should be linked with option. This causes libraries libc.a and libsim.a to
27165 @itemx -mhard-float
27167 @opindex mhard-float
27168 Generate code containing floating-point instructions. This is the
27172 @itemx -msoft-float
27174 @opindex msoft-float
27175 Generate code containing library calls for floating-point.
27177 @option{-msoft-float} changes the calling convention in the output file;
27178 therefore, it is only useful if you compile @emph{all} of a program with
27179 this option. In particular, you need to compile @file{libgcc.a}, the
27180 library that comes with GCC, with @option{-msoft-float} in order for
27183 @item -mcpu=@var{cpu_type}
27185 Set the instruction set, register set, and instruction scheduling parameters
27186 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
27187 @samp{mcm}, @samp{gr5} and @samp{gr6}.
27189 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
27191 By default (unless configured otherwise), GCC generates code for the GR5
27192 variant of the Visium architecture.
27194 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
27195 architecture. The only difference from GR5 code is that the compiler will
27196 generate block move instructions.
27198 @item -mtune=@var{cpu_type}
27200 Set the instruction scheduling parameters for machine type @var{cpu_type},
27201 but do not set the instruction set or register set that the option
27202 @option{-mcpu=@var{cpu_type}} would.
27206 Generate code for the supervisor mode, where there are no restrictions on
27207 the access to general registers. This is the default.
27210 @opindex muser-mode
27211 Generate code for the user mode, where the access to some general registers
27212 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
27213 mode; on the GR6, only registers r29 to r31 are affected.
27217 @subsection VMS Options
27219 These @samp{-m} options are defined for the VMS implementations:
27222 @item -mvms-return-codes
27223 @opindex mvms-return-codes
27224 Return VMS condition codes from @code{main}. The default is to return POSIX-style
27225 condition (e.g.@: error) codes.
27227 @item -mdebug-main=@var{prefix}
27228 @opindex mdebug-main=@var{prefix}
27229 Flag the first routine whose name starts with @var{prefix} as the main
27230 routine for the debugger.
27234 Default to 64-bit memory allocation routines.
27236 @item -mpointer-size=@var{size}
27237 @opindex mpointer-size=@var{size}
27238 Set the default size of pointers. Possible options for @var{size} are
27239 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
27240 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
27241 The later option disables @code{pragma pointer_size}.
27244 @node VxWorks Options
27245 @subsection VxWorks Options
27246 @cindex VxWorks Options
27248 The options in this section are defined for all VxWorks targets.
27249 Options specific to the target hardware are listed with the other
27250 options for that target.
27255 GCC can generate code for both VxWorks kernels and real time processes
27256 (RTPs). This option switches from the former to the latter. It also
27257 defines the preprocessor macro @code{__RTP__}.
27260 @opindex non-static
27261 Link an RTP executable against shared libraries rather than static
27262 libraries. The options @option{-static} and @option{-shared} can
27263 also be used for RTPs (@pxref{Link Options}); @option{-static}
27270 These options are passed down to the linker. They are defined for
27271 compatibility with Diab.
27274 @opindex Xbind-lazy
27275 Enable lazy binding of function calls. This option is equivalent to
27276 @option{-Wl,-z,now} and is defined for compatibility with Diab.
27280 Disable lazy binding of function calls. This option is the default and
27281 is defined for compatibility with Diab.
27285 @subsection x86 Options
27286 @cindex x86 Options
27288 These @samp{-m} options are defined for the x86 family of computers.
27292 @item -march=@var{cpu-type}
27294 Generate instructions for the machine type @var{cpu-type}. In contrast to
27295 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
27296 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
27297 to generate code that may not run at all on processors other than the one
27298 indicated. Specifying @option{-march=@var{cpu-type}} implies
27299 @option{-mtune=@var{cpu-type}}.
27301 The choices for @var{cpu-type} are:
27305 This selects the CPU to generate code for at compilation time by determining
27306 the processor type of the compiling machine. Using @option{-march=native}
27307 enables all instruction subsets supported by the local machine (hence
27308 the result might not run on different machines). Using @option{-mtune=native}
27309 produces code optimized for the local machine under the constraints
27310 of the selected instruction set.
27313 A generic CPU with 64-bit extensions.
27316 Original Intel i386 CPU@.
27319 Intel i486 CPU@. (No scheduling is implemented for this chip.)
27323 Intel Pentium CPU with no MMX support.
27326 Intel Lakemont MCU, based on Intel Pentium CPU.
27329 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
27332 Intel Pentium Pro CPU@.
27335 When used with @option{-march}, the Pentium Pro
27336 instruction set is used, so the code runs on all i686 family chips.
27337 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
27340 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
27345 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
27349 Intel Pentium M; low-power version of Intel Pentium III CPU
27350 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
27354 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
27357 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
27361 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
27362 SSE2 and SSE3 instruction set support.
27365 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
27366 instruction set support.
27369 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27370 SSE4.1, SSE4.2 and POPCNT instruction set support.
27373 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27374 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
27377 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27378 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
27381 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
27382 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
27383 instruction set support.
27386 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27387 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27388 BMI, BMI2 and F16C instruction set support.
27391 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27392 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27393 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
27396 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27397 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27398 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
27399 XSAVES instruction set support.
27402 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
27403 instruction set support.
27406 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27407 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
27410 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27411 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT and FSGSBASE
27412 instruction set support.
27414 @item goldmont-plus
27415 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27416 SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE,
27417 PTWRITE, RDPID, SGX and UMIP instruction set support.
27420 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27421 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, PTWRITE,
27422 RDPID, SGX, UMIP, GFNI-SSE, CLWB and ENCLV instruction set support.
27425 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27426 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27427 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
27428 AVX512CD instruction set support.
27431 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27432 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27433 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER, AVX512CD,
27434 AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ instruction set support.
27436 @item skylake-avx512
27437 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
27438 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
27439 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27440 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
27443 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27444 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27445 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27446 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27447 AVX512IFMA, SHA and UMIP instruction set support.
27449 @item icelake-client
27450 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27451 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27452 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27453 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27454 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27455 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES instruction set support.
27457 @item icelake-server
27458 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2,
27459 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE,
27460 RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC,
27461 XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI,
27462 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ,
27463 AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, PCONFIG and WBNOINVD instruction
27467 Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27468 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27469 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27470 AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support.
27473 Intel cooperlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27474 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27475 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB,
27476 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI and AVX512BF16 instruction
27480 Intel Tigerlake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
27481 SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI,
27482 BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
27483 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, AVX512IFMA, SHA, CLWB, UMIP,
27484 RDPID, GFNI, AVX512VBMI2, AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ,
27485 VAES, PCONFIG, WBNOINVD, MOVDIRI, MOVDIR64B and AVX512VP2INTERSECT instruction
27489 AMD K6 CPU with MMX instruction set support.
27493 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
27496 @itemx athlon-tbird
27497 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
27503 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
27504 instruction set support.
27510 Processors based on the AMD K8 core with x86-64 instruction set support,
27511 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
27512 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
27513 instruction set extensions.)
27516 @itemx opteron-sse3
27517 @itemx athlon64-sse3
27518 Improved versions of AMD K8 cores with SSE3 instruction set support.
27522 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
27523 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
27524 instruction set extensions.)
27527 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
27528 supersets FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
27529 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
27531 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27532 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX,
27533 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
27536 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27537 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
27538 PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
27539 64-bit instruction set extensions.
27541 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
27542 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
27543 AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
27544 SSE4.2, ABM and 64-bit instruction set extensions.
27547 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27548 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
27549 SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
27550 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27551 instruction set extensions.
27553 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
27554 supersets BMI, BMI2, ,CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED,
27555 MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A,
27556 SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
27557 instruction set extensions.)
27561 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
27562 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
27563 instruction set extensions.)
27566 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
27567 includes MOVBE, F16C, BMI, AVX, PCLMUL, AES, SSE4.2, SSE4.1, CX16, ABM,
27568 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
27571 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
27575 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
27576 instruction set support.
27579 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
27580 (No scheduling is implemented for this chip.)
27583 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
27584 (No scheduling is implemented for this chip.)
27587 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27588 (No scheduling is implemented for this chip.)
27591 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
27592 (No scheduling is implemented for this chip.)
27595 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
27596 (No scheduling is implemented for this chip.)
27599 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
27600 (No scheduling is implemented for this chip.)
27603 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
27604 (No scheduling is implemented for this chip.)
27607 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
27608 AVX and AVX2 instruction set support.
27609 (No scheduling is implemented for this chip.)
27612 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27613 instruction set support.
27614 (No scheduling is implemented for this chip.)
27617 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27618 instruction set support.
27619 (No scheduling is implemented for this chip.)
27622 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
27623 instruction set support.
27624 (No scheduling is implemented for this chip.)
27627 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27628 instruction set support.
27629 (No scheduling is implemented for this chip.)
27632 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27633 instruction set support.
27634 (No scheduling is implemented for this chip.)
27637 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
27638 instruction set support.
27639 (No scheduling is implemented for this chip.)
27642 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
27645 @item -mtune=@var{cpu-type}
27647 Tune to @var{cpu-type} everything applicable about the generated code, except
27648 for the ABI and the set of available instructions.
27649 While picking a specific @var{cpu-type} schedules things appropriately
27650 for that particular chip, the compiler does not generate any code that
27651 cannot run on the default machine type unless you use a
27652 @option{-march=@var{cpu-type}} option.
27653 For example, if GCC is configured for i686-pc-linux-gnu
27654 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
27655 but still runs on i686 machines.
27657 The choices for @var{cpu-type} are the same as for @option{-march}.
27658 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
27662 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
27663 If you know the CPU on which your code will run, then you should use
27664 the corresponding @option{-mtune} or @option{-march} option instead of
27665 @option{-mtune=generic}. But, if you do not know exactly what CPU users
27666 of your application will have, then you should use this option.
27668 As new processors are deployed in the marketplace, the behavior of this
27669 option will change. Therefore, if you upgrade to a newer version of
27670 GCC, code generation controlled by this option will change to reflect
27672 that are most common at the time that version of GCC is released.
27674 There is no @option{-march=generic} option because @option{-march}
27675 indicates the instruction set the compiler can use, and there is no
27676 generic instruction set applicable to all processors. In contrast,
27677 @option{-mtune} indicates the processor (or, in this case, collection of
27678 processors) for which the code is optimized.
27681 Produce code optimized for the most current Intel processors, which are
27682 Haswell and Silvermont for this version of GCC. If you know the CPU
27683 on which your code will run, then you should use the corresponding
27684 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
27685 But, if you want your application performs better on both Haswell and
27686 Silvermont, then you should use this option.
27688 As new Intel processors are deployed in the marketplace, the behavior of
27689 this option will change. Therefore, if you upgrade to a newer version of
27690 GCC, code generation controlled by this option will change to reflect
27691 the most current Intel processors at the time that version of GCC is
27694 There is no @option{-march=intel} option because @option{-march} indicates
27695 the instruction set the compiler can use, and there is no common
27696 instruction set applicable to all processors. In contrast,
27697 @option{-mtune} indicates the processor (or, in this case, collection of
27698 processors) for which the code is optimized.
27701 @item -mcpu=@var{cpu-type}
27703 A deprecated synonym for @option{-mtune}.
27705 @item -mfpmath=@var{unit}
27707 Generate floating-point arithmetic for selected unit @var{unit}. The choices
27708 for @var{unit} are:
27712 Use the standard 387 floating-point coprocessor present on the majority of chips and
27713 emulated otherwise. Code compiled with this option runs almost everywhere.
27714 The temporary results are computed in 80-bit precision instead of the precision
27715 specified by the type, resulting in slightly different results compared to most
27716 of other chips. See @option{-ffloat-store} for more detailed description.
27718 This is the default choice for non-Darwin x86-32 targets.
27721 Use scalar floating-point instructions present in the SSE instruction set.
27722 This instruction set is supported by Pentium III and newer chips,
27723 and in the AMD line
27724 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
27725 instruction set supports only single-precision arithmetic, thus the double and
27726 extended-precision arithmetic are still done using 387. A later version, present
27727 only in Pentium 4 and AMD x86-64 chips, supports double-precision
27730 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
27731 or @option{-msse2} switches to enable SSE extensions and make this option
27732 effective. For the x86-64 compiler, these extensions are enabled by default.
27734 The resulting code should be considerably faster in the majority of cases and avoid
27735 the numerical instability problems of 387 code, but may break some existing
27736 code that expects temporaries to be 80 bits.
27738 This is the default choice for the x86-64 compiler, Darwin x86-32 targets,
27739 and the default choice for x86-32 targets with the SSE2 instruction set
27740 when @option{-ffast-math} is enabled.
27745 Attempt to utilize both instruction sets at once. This effectively doubles the
27746 amount of available registers, and on chips with separate execution units for
27747 387 and SSE the execution resources too. Use this option with care, as it is
27748 still experimental, because the GCC register allocator does not model separate
27749 functional units well, resulting in unstable performance.
27752 @item -masm=@var{dialect}
27753 @opindex masm=@var{dialect}
27754 Output assembly instructions using selected @var{dialect}. Also affects
27755 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
27756 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
27757 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
27758 not support @samp{intel}.
27761 @itemx -mno-ieee-fp
27763 @opindex mno-ieee-fp
27764 Control whether or not the compiler uses IEEE floating-point
27765 comparisons. These correctly handle the case where the result of a
27766 comparison is unordered.
27769 @itemx -mhard-float
27771 @opindex mhard-float
27772 Generate output containing 80387 instructions for floating point.
27775 @itemx -msoft-float
27777 @opindex msoft-float
27778 Generate output containing library calls for floating point.
27780 @strong{Warning:} the requisite libraries are not part of GCC@.
27781 Normally the facilities of the machine's usual C compiler are used, but
27782 this cannot be done directly in cross-compilation. You must make your
27783 own arrangements to provide suitable library functions for
27786 On machines where a function returns floating-point results in the 80387
27787 register stack, some floating-point opcodes may be emitted even if
27788 @option{-msoft-float} is used.
27790 @item -mno-fp-ret-in-387
27791 @opindex mno-fp-ret-in-387
27792 @opindex mfp-ret-in-387
27793 Do not use the FPU registers for return values of functions.
27795 The usual calling convention has functions return values of types
27796 @code{float} and @code{double} in an FPU register, even if there
27797 is no FPU@. The idea is that the operating system should emulate
27800 The option @option{-mno-fp-ret-in-387} causes such values to be returned
27801 in ordinary CPU registers instead.
27803 @item -mno-fancy-math-387
27804 @opindex mno-fancy-math-387
27805 @opindex mfancy-math-387
27806 Some 387 emulators do not support the @code{sin}, @code{cos} and
27807 @code{sqrt} instructions for the 387. Specify this option to avoid
27808 generating those instructions.
27809 This option is overridden when @option{-march}
27810 indicates that the target CPU always has an FPU and so the
27811 instruction does not need emulation. These
27812 instructions are not generated unless you also use the
27813 @option{-funsafe-math-optimizations} switch.
27815 @item -malign-double
27816 @itemx -mno-align-double
27817 @opindex malign-double
27818 @opindex mno-align-double
27819 Control whether GCC aligns @code{double}, @code{long double}, and
27820 @code{long long} variables on a two-word boundary or a one-word
27821 boundary. Aligning @code{double} variables on a two-word boundary
27822 produces code that runs somewhat faster on a Pentium at the
27823 expense of more memory.
27825 On x86-64, @option{-malign-double} is enabled by default.
27827 @strong{Warning:} if you use the @option{-malign-double} switch,
27828 structures containing the above types are aligned differently than
27829 the published application binary interface specifications for the x86-32
27830 and are not binary compatible with structures in code compiled
27831 without that switch.
27833 @item -m96bit-long-double
27834 @itemx -m128bit-long-double
27835 @opindex m96bit-long-double
27836 @opindex m128bit-long-double
27837 These switches control the size of @code{long double} type. The x86-32
27838 application binary interface specifies the size to be 96 bits,
27839 so @option{-m96bit-long-double} is the default in 32-bit mode.
27841 Modern architectures (Pentium and newer) prefer @code{long double}
27842 to be aligned to an 8- or 16-byte boundary. In arrays or structures
27843 conforming to the ABI, this is not possible. So specifying
27844 @option{-m128bit-long-double} aligns @code{long double}
27845 to a 16-byte boundary by padding the @code{long double} with an additional
27848 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
27849 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
27851 Notice that neither of these options enable any extra precision over the x87
27852 standard of 80 bits for a @code{long double}.
27854 @strong{Warning:} if you override the default value for your target ABI, this
27855 changes the size of
27856 structures and arrays containing @code{long double} variables,
27857 as well as modifying the function calling convention for functions taking
27858 @code{long double}. Hence they are not binary-compatible
27859 with code compiled without that switch.
27861 @item -mlong-double-64
27862 @itemx -mlong-double-80
27863 @itemx -mlong-double-128
27864 @opindex mlong-double-64
27865 @opindex mlong-double-80
27866 @opindex mlong-double-128
27867 These switches control the size of @code{long double} type. A size
27868 of 64 bits makes the @code{long double} type equivalent to the @code{double}
27869 type. This is the default for 32-bit Bionic C library. A size
27870 of 128 bits makes the @code{long double} type equivalent to the
27871 @code{__float128} type. This is the default for 64-bit Bionic C library.
27873 @strong{Warning:} if you override the default value for your target ABI, this
27874 changes the size of
27875 structures and arrays containing @code{long double} variables,
27876 as well as modifying the function calling convention for functions taking
27877 @code{long double}. Hence they are not binary-compatible
27878 with code compiled without that switch.
27880 @item -malign-data=@var{type}
27881 @opindex malign-data
27882 Control how GCC aligns variables. Supported values for @var{type} are
27883 @samp{compat} uses increased alignment value compatible uses GCC 4.8
27884 and earlier, @samp{abi} uses alignment value as specified by the
27885 psABI, and @samp{cacheline} uses increased alignment value to match
27886 the cache line size. @samp{compat} is the default.
27888 @item -mlarge-data-threshold=@var{threshold}
27889 @opindex mlarge-data-threshold
27890 When @option{-mcmodel=medium} is specified, data objects larger than
27891 @var{threshold} are placed in the large data section. This value must be the
27892 same across all objects linked into the binary, and defaults to 65535.
27896 Use a different function-calling convention, in which functions that
27897 take a fixed number of arguments return with the @code{ret @var{num}}
27898 instruction, which pops their arguments while returning. This saves one
27899 instruction in the caller since there is no need to pop the arguments
27902 You can specify that an individual function is called with this calling
27903 sequence with the function attribute @code{stdcall}. You can also
27904 override the @option{-mrtd} option by using the function attribute
27905 @code{cdecl}. @xref{Function Attributes}.
27907 @strong{Warning:} this calling convention is incompatible with the one
27908 normally used on Unix, so you cannot use it if you need to call
27909 libraries compiled with the Unix compiler.
27911 Also, you must provide function prototypes for all functions that
27912 take variable numbers of arguments (including @code{printf});
27913 otherwise incorrect code is generated for calls to those
27916 In addition, seriously incorrect code results if you call a
27917 function with too many arguments. (Normally, extra arguments are
27918 harmlessly ignored.)
27920 @item -mregparm=@var{num}
27922 Control how many registers are used to pass integer arguments. By
27923 default, no registers are used to pass arguments, and at most 3
27924 registers can be used. You can control this behavior for a specific
27925 function by using the function attribute @code{regparm}.
27926 @xref{Function Attributes}.
27928 @strong{Warning:} if you use this switch, and
27929 @var{num} is nonzero, then you must build all modules with the same
27930 value, including any libraries. This includes the system libraries and
27934 @opindex msseregparm
27935 Use SSE register passing conventions for float and double arguments
27936 and return values. You can control this behavior for a specific
27937 function by using the function attribute @code{sseregparm}.
27938 @xref{Function Attributes}.
27940 @strong{Warning:} if you use this switch then you must build all
27941 modules with the same value, including any libraries. This includes
27942 the system libraries and startup modules.
27944 @item -mvect8-ret-in-mem
27945 @opindex mvect8-ret-in-mem
27946 Return 8-byte vectors in memory instead of MMX registers. This is the
27947 default on VxWorks to match the ABI of the Sun Studio compilers until
27948 version 12. @emph{Only} use this option if you need to remain
27949 compatible with existing code produced by those previous compiler
27950 versions or older versions of GCC@.
27959 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
27960 is specified, the significands of results of floating-point operations are
27961 rounded to 24 bits (single precision); @option{-mpc64} rounds the
27962 significands of results of floating-point operations to 53 bits (double
27963 precision) and @option{-mpc80} rounds the significands of results of
27964 floating-point operations to 64 bits (extended double precision), which is
27965 the default. When this option is used, floating-point operations in higher
27966 precisions are not available to the programmer without setting the FPU
27967 control word explicitly.
27969 Setting the rounding of floating-point operations to less than the default
27970 80 bits can speed some programs by 2% or more. Note that some mathematical
27971 libraries assume that extended-precision (80-bit) floating-point operations
27972 are enabled by default; routines in such libraries could suffer significant
27973 loss of accuracy, typically through so-called ``catastrophic cancellation'',
27974 when this option is used to set the precision to less than extended precision.
27976 @item -mstackrealign
27977 @opindex mstackrealign
27978 Realign the stack at entry. On the x86, the @option{-mstackrealign}
27979 option generates an alternate prologue and epilogue that realigns the
27980 run-time stack if necessary. This supports mixing legacy codes that keep
27981 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
27982 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
27983 applicable to individual functions.
27985 @item -mpreferred-stack-boundary=@var{num}
27986 @opindex mpreferred-stack-boundary
27987 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
27988 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
27989 the default is 4 (16 bytes or 128 bits).
27991 @strong{Warning:} When generating code for the x86-64 architecture with
27992 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
27993 used to keep the stack boundary aligned to 8 byte boundary. Since
27994 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
27995 intended to be used in controlled environment where stack space is
27996 important limitation. This option leads to wrong code when functions
27997 compiled with 16 byte stack alignment (such as functions from a standard
27998 library) are called with misaligned stack. In this case, SSE
27999 instructions may lead to misaligned memory access traps. In addition,
28000 variable arguments are handled incorrectly for 16 byte aligned
28001 objects (including x87 long double and __int128), leading to wrong
28002 results. You must build all modules with
28003 @option{-mpreferred-stack-boundary=3}, including any libraries. This
28004 includes the system libraries and startup modules.
28006 @item -mincoming-stack-boundary=@var{num}
28007 @opindex mincoming-stack-boundary
28008 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
28009 boundary. If @option{-mincoming-stack-boundary} is not specified,
28010 the one specified by @option{-mpreferred-stack-boundary} is used.
28012 On Pentium and Pentium Pro, @code{double} and @code{long double} values
28013 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
28014 suffer significant run time performance penalties. On Pentium III, the
28015 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
28016 properly if it is not 16-byte aligned.
28018 To ensure proper alignment of this values on the stack, the stack boundary
28019 must be as aligned as that required by any value stored on the stack.
28020 Further, every function must be generated such that it keeps the stack
28021 aligned. Thus calling a function compiled with a higher preferred
28022 stack boundary from a function compiled with a lower preferred stack
28023 boundary most likely misaligns the stack. It is recommended that
28024 libraries that use callbacks always use the default setting.
28026 This extra alignment does consume extra stack space, and generally
28027 increases code size. Code that is sensitive to stack space usage, such
28028 as embedded systems and operating system kernels, may want to reduce the
28029 preferred alignment to @option{-mpreferred-stack-boundary=2}.
28086 @itemx -mavx512ifma
28087 @opindex mavx512ifma
28089 @itemx -mavx512vbmi
28090 @opindex mavx512vbmi
28101 @itemx -mclflushopt
28102 @opindex mclflushopt
28137 @itemx -mprefetchwt1
28138 @opindex mprefetchwt1
28209 @itemx -mavx512vbmi2
28210 @opindex mavx512vbmi2
28212 @itemx -mavx512bf16
28213 @opindex mavx512bf16
28224 @itemx -mvpclmulqdq
28225 @opindex mvpclmulqdq
28227 @itemx -mavx512bitalg
28228 @opindex mavx512bitalg
28234 @opindex mmovdir64b
28239 @itemx -mavx512vpopcntdq
28240 @opindex mavx512vpopcntdq
28242 @itemx -mavx512vp2intersect
28243 @opindex mavx512vp2intersect
28245 @itemx -mavx5124fmaps
28246 @opindex mavx5124fmaps
28248 @itemx -mavx512vnni
28249 @opindex mavx512vnni
28251 @itemx -mavx5124vnniw
28252 @opindex mavx5124vnniw
28256 These switches enable the use of instructions in the MMX, SSE,
28257 SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, AVX512PF,
28258 AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA,
28259 AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG,
28260 WBNOINVD, FMA4, PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP,
28261 3DNow!@:, enhanced 3DNow!@:, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE,
28262 XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2,
28263 GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16,
28264 ENQCMD, AVX512VPOPCNTDQ, AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, or CLDEMOTE
28265 extended instruction sets. Each has a corresponding @option{-mno-} option to
28266 disable use of these instructions.
28268 These extensions are also available as built-in functions: see
28269 @ref{x86 Built-in Functions}, for details of the functions enabled and
28270 disabled by these switches.
28272 To generate SSE/SSE2 instructions automatically from floating-point
28273 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
28275 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
28276 generates new AVX instructions or AVX equivalence for all SSEx instructions
28279 These options enable GCC to use these extended instructions in
28280 generated code, even without @option{-mfpmath=sse}. Applications that
28281 perform run-time CPU detection must compile separate files for each
28282 supported architecture, using the appropriate flags. In particular,
28283 the file containing the CPU detection code should be compiled without
28286 @item -mdump-tune-features
28287 @opindex mdump-tune-features
28288 This option instructs GCC to dump the names of the x86 performance
28289 tuning features and default settings. The names can be used in
28290 @option{-mtune-ctrl=@var{feature-list}}.
28292 @item -mtune-ctrl=@var{feature-list}
28293 @opindex mtune-ctrl=@var{feature-list}
28294 This option is used to do fine grain control of x86 code generation features.
28295 @var{feature-list} is a comma separated list of @var{feature} names. See also
28296 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
28297 on if it is not preceded with @samp{^}, otherwise, it is turned off.
28298 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
28299 developers. Using it may lead to code paths not covered by testing and can
28300 potentially result in compiler ICEs or runtime errors.
28303 @opindex mno-default
28304 This option instructs GCC to turn off all tunable features. See also
28305 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
28309 This option instructs GCC to emit a @code{cld} instruction in the prologue
28310 of functions that use string instructions. String instructions depend on
28311 the DF flag to select between autoincrement or autodecrement mode. While the
28312 ABI specifies the DF flag to be cleared on function entry, some operating
28313 systems violate this specification by not clearing the DF flag in their
28314 exception dispatchers. The exception handler can be invoked with the DF flag
28315 set, which leads to wrong direction mode when string instructions are used.
28316 This option can be enabled by default on 32-bit x86 targets by configuring
28317 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
28318 instructions can be suppressed with the @option{-mno-cld} compiler option
28322 @opindex mvzeroupper
28323 This option instructs GCC to emit a @code{vzeroupper} instruction
28324 before a transfer of control flow out of the function to minimize
28325 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
28328 @item -mprefer-avx128
28329 @opindex mprefer-avx128
28330 This option instructs GCC to use 128-bit AVX instructions instead of
28331 256-bit AVX instructions in the auto-vectorizer.
28333 @item -mprefer-vector-width=@var{opt}
28334 @opindex mprefer-vector-width
28335 This option instructs GCC to use @var{opt}-bit vector width in instructions
28336 instead of default on the selected platform.
28340 No extra limitations applied to GCC other than defined by the selected platform.
28343 Prefer 128-bit vector width for instructions.
28346 Prefer 256-bit vector width for instructions.
28349 Prefer 512-bit vector width for instructions.
28354 This option enables GCC to generate @code{CMPXCHG16B} instructions in 64-bit
28355 code to implement compare-and-exchange operations on 16-byte aligned 128-bit
28356 objects. This is useful for atomic updates of data structures exceeding one
28357 machine word in size. The compiler uses this instruction to implement
28358 @ref{__sync Builtins}. However, for @ref{__atomic Builtins} operating on
28359 128-bit integers, a library call is always used.
28363 This option enables generation of @code{SAHF} instructions in 64-bit code.
28364 Early Intel Pentium 4 CPUs with Intel 64 support,
28365 prior to the introduction of Pentium 4 G1 step in December 2005,
28366 lacked the @code{LAHF} and @code{SAHF} instructions
28367 which are supported by AMD64.
28368 These are load and store instructions, respectively, for certain status flags.
28369 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
28370 @code{drem}, and @code{remainder} built-in functions;
28371 see @ref{Other Builtins} for details.
28375 This option enables use of the @code{movbe} instruction to implement
28376 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
28380 The @option{-mshstk} option enables shadow stack built-in functions
28381 from x86 Control-flow Enforcement Technology (CET).
28385 This option enables built-in functions @code{__builtin_ia32_crc32qi},
28386 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
28387 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
28391 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
28392 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
28393 with an additional Newton-Raphson step
28394 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
28395 (and their vectorized
28396 variants) for single-precision floating-point arguments. These instructions
28397 are generated only when @option{-funsafe-math-optimizations} is enabled
28398 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
28399 Note that while the throughput of the sequence is higher than the throughput
28400 of the non-reciprocal instruction, the precision of the sequence can be
28401 decreased by up to 2 ulp (i.e.@: the inverse of 1.0 equals 0.99999994).
28403 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
28404 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
28405 combination), and doesn't need @option{-mrecip}.
28407 Also note that GCC emits the above sequence with additional Newton-Raphson step
28408 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
28409 already with @option{-ffast-math} (or the above option combination), and
28410 doesn't need @option{-mrecip}.
28412 @item -mrecip=@var{opt}
28413 @opindex mrecip=opt
28414 This option controls which reciprocal estimate instructions
28415 may be used. @var{opt} is a comma-separated list of options, which may
28416 be preceded by a @samp{!} to invert the option:
28420 Enable all estimate instructions.
28423 Enable the default instructions, equivalent to @option{-mrecip}.
28426 Disable all estimate instructions, equivalent to @option{-mno-recip}.
28429 Enable the approximation for scalar division.
28432 Enable the approximation for vectorized division.
28435 Enable the approximation for scalar square root.
28438 Enable the approximation for vectorized square root.
28441 So, for example, @option{-mrecip=all,!sqrt} enables
28442 all of the reciprocal approximations, except for square root.
28444 @item -mveclibabi=@var{type}
28445 @opindex mveclibabi
28446 Specifies the ABI type to use for vectorizing intrinsics using an
28447 external library. Supported values for @var{type} are @samp{svml}
28448 for the Intel short
28449 vector math library and @samp{acml} for the AMD math core library.
28450 To use this option, both @option{-ftree-vectorize} and
28451 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
28452 ABI-compatible library must be specified at link time.
28454 GCC currently emits calls to @code{vmldExp2},
28455 @code{vmldLn2}, @code{vmldLog102}, @code{vmldPow2},
28456 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
28457 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
28458 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
28459 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4},
28460 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
28461 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
28462 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
28463 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
28464 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
28465 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
28466 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
28467 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
28468 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
28469 when @option{-mveclibabi=acml} is used.
28471 @item -mabi=@var{name}
28473 Generate code for the specified calling convention. Permissible values
28474 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
28475 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
28476 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
28477 You can control this behavior for specific functions by
28478 using the function attributes @code{ms_abi} and @code{sysv_abi}.
28479 @xref{Function Attributes}.
28481 @item -mforce-indirect-call
28482 @opindex mforce-indirect-call
28483 Force all calls to functions to be indirect. This is useful
28484 when using Intel Processor Trace where it generates more precise timing
28485 information for function calls.
28487 @item -mmanual-endbr
28488 @opindex mmanual-endbr
28489 Insert ENDBR instruction at function entry only via the @code{cf_check}
28490 function attribute. This is useful when used with the option
28491 @option{-fcf-protection=branch} to control ENDBR insertion at the
28494 @item -mcall-ms2sysv-xlogues
28495 @opindex mcall-ms2sysv-xlogues
28496 @opindex mno-call-ms2sysv-xlogues
28497 Due to differences in 64-bit ABIs, any Microsoft ABI function that calls a
28498 System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. By
28499 default, the code for saving and restoring these registers is emitted inline,
28500 resulting in fairly lengthy prologues and epilogues. Using
28501 @option{-mcall-ms2sysv-xlogues} emits prologues and epilogues that
28502 use stubs in the static portion of libgcc to perform these saves and restores,
28503 thus reducing function size at the cost of a few extra instructions.
28505 @item -mtls-dialect=@var{type}
28506 @opindex mtls-dialect
28507 Generate code to access thread-local storage using the @samp{gnu} or
28508 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
28509 @samp{gnu2} is more efficient, but it may add compile- and run-time
28510 requirements that cannot be satisfied on all systems.
28513 @itemx -mno-push-args
28514 @opindex mpush-args
28515 @opindex mno-push-args
28516 Use PUSH operations to store outgoing parameters. This method is shorter
28517 and usually equally fast as method using SUB/MOV operations and is enabled
28518 by default. In some cases disabling it may improve performance because of
28519 improved scheduling and reduced dependencies.
28521 @item -maccumulate-outgoing-args
28522 @opindex maccumulate-outgoing-args
28523 If enabled, the maximum amount of space required for outgoing arguments is
28524 computed in the function prologue. This is faster on most modern CPUs
28525 because of reduced dependencies, improved scheduling and reduced stack usage
28526 when the preferred stack boundary is not equal to 2. The drawback is a notable
28527 increase in code size. This switch implies @option{-mno-push-args}.
28531 Support thread-safe exception handling on MinGW. Programs that rely
28532 on thread-safe exception handling must compile and link all code with the
28533 @option{-mthreads} option. When compiling, @option{-mthreads} defines
28534 @option{-D_MT}; when linking, it links in a special thread helper library
28535 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
28537 @item -mms-bitfields
28538 @itemx -mno-ms-bitfields
28539 @opindex mms-bitfields
28540 @opindex mno-ms-bitfields
28542 Enable/disable bit-field layout compatible with the native Microsoft
28545 If @code{packed} is used on a structure, or if bit-fields are used,
28546 it may be that the Microsoft ABI lays out the structure differently
28547 than the way GCC normally does. Particularly when moving packed
28548 data between functions compiled with GCC and the native Microsoft compiler
28549 (either via function call or as data in a file), it may be necessary to access
28552 This option is enabled by default for Microsoft Windows
28553 targets. This behavior can also be controlled locally by use of variable
28554 or type attributes. For more information, see @ref{x86 Variable Attributes}
28555 and @ref{x86 Type Attributes}.
28557 The Microsoft structure layout algorithm is fairly simple with the exception
28558 of the bit-field packing.
28559 The padding and alignment of members of structures and whether a bit-field
28560 can straddle a storage-unit boundary are determine by these rules:
28563 @item Structure members are stored sequentially in the order in which they are
28564 declared: the first member has the lowest memory address and the last member
28567 @item Every data object has an alignment requirement. The alignment requirement
28568 for all data except structures, unions, and arrays is either the size of the
28569 object or the current packing size (specified with either the
28570 @code{aligned} attribute or the @code{pack} pragma),
28571 whichever is less. For structures, unions, and arrays,
28572 the alignment requirement is the largest alignment requirement of its members.
28573 Every object is allocated an offset so that:
28576 offset % alignment_requirement == 0
28579 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
28580 unit if the integral types are the same size and if the next bit-field fits
28581 into the current allocation unit without crossing the boundary imposed by the
28582 common alignment requirements of the bit-fields.
28585 MSVC interprets zero-length bit-fields in the following ways:
28588 @item If a zero-length bit-field is inserted between two bit-fields that
28589 are normally coalesced, the bit-fields are not coalesced.
28596 unsigned long bf_1 : 12;
28598 unsigned long bf_2 : 12;
28603 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
28604 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
28606 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
28607 alignment of the zero-length bit-field is greater than the member that follows it,
28608 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
28629 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
28630 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
28631 bit-field does not affect the alignment of @code{bar} or, as a result, the size
28634 Taking this into account, it is important to note the following:
28637 @item If a zero-length bit-field follows a normal bit-field, the type of the
28638 zero-length bit-field may affect the alignment of the structure as whole. For
28639 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
28640 normal bit-field, and is of type short.
28642 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
28643 still affect the alignment of the structure:
28654 Here, @code{t4} takes up 4 bytes.
28657 @item Zero-length bit-fields following non-bit-field members are ignored:
28669 Here, @code{t5} takes up 2 bytes.
28673 @item -mno-align-stringops
28674 @opindex mno-align-stringops
28675 @opindex malign-stringops
28676 Do not align the destination of inlined string operations. This switch reduces
28677 code size and improves performance in case the destination is already aligned,
28678 but GCC doesn't know about it.
28680 @item -minline-all-stringops
28681 @opindex minline-all-stringops
28682 By default GCC inlines string operations only when the destination is
28683 known to be aligned to least a 4-byte boundary.
28684 This enables more inlining and increases code
28685 size, but may improve performance of code that depends on fast
28686 @code{memcpy} and @code{memset} for short lengths.
28687 The option enables inline expansion of @code{strlen} for all
28688 pointer alignments.
28690 @item -minline-stringops-dynamically
28691 @opindex minline-stringops-dynamically
28692 For string operations of unknown size, use run-time checks with
28693 inline code for small blocks and a library call for large blocks.
28695 @item -mstringop-strategy=@var{alg}
28696 @opindex mstringop-strategy=@var{alg}
28697 Override the internal decision heuristic for the particular algorithm to use
28698 for inlining string operations. The allowed values for @var{alg} are:
28704 Expand using i386 @code{rep} prefix of the specified size.
28708 @itemx unrolled_loop
28709 Expand into an inline loop.
28712 Always use a library call.
28715 @item -mmemcpy-strategy=@var{strategy}
28716 @opindex mmemcpy-strategy=@var{strategy}
28717 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
28718 should be inlined and what inline algorithm to use when the expected size
28719 of the copy operation is known. @var{strategy}
28720 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
28721 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
28722 the max byte size with which inline algorithm @var{alg} is allowed. For the last
28723 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
28724 in the list must be specified in increasing order. The minimal byte size for
28725 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
28728 @item -mmemset-strategy=@var{strategy}
28729 @opindex mmemset-strategy=@var{strategy}
28730 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
28731 @code{__builtin_memset} expansion.
28733 @item -momit-leaf-frame-pointer
28734 @opindex momit-leaf-frame-pointer
28735 Don't keep the frame pointer in a register for leaf functions. This
28736 avoids the instructions to save, set up, and restore frame pointers and
28737 makes an extra register available in leaf functions. The option
28738 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
28739 which might make debugging harder.
28741 @item -mtls-direct-seg-refs
28742 @itemx -mno-tls-direct-seg-refs
28743 @opindex mtls-direct-seg-refs
28744 Controls whether TLS variables may be accessed with offsets from the
28745 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
28746 or whether the thread base pointer must be added. Whether or not this
28747 is valid depends on the operating system, and whether it maps the
28748 segment to cover the entire TLS area.
28750 For systems that use the GNU C Library, the default is on.
28753 @itemx -mno-sse2avx
28755 Specify that the assembler should encode SSE instructions with VEX
28756 prefix. The option @option{-mavx} turns this on by default.
28761 If profiling is active (@option{-pg}), put the profiling
28762 counter call before the prologue.
28763 Note: On x86 architectures the attribute @code{ms_hook_prologue}
28764 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
28766 @item -mrecord-mcount
28767 @itemx -mno-record-mcount
28768 @opindex mrecord-mcount
28769 If profiling is active (@option{-pg}), generate a __mcount_loc section
28770 that contains pointers to each profiling call. This is useful for
28771 automatically patching and out calls.
28774 @itemx -mno-nop-mcount
28775 @opindex mnop-mcount
28776 If profiling is active (@option{-pg}), generate the calls to
28777 the profiling functions as NOPs. This is useful when they
28778 should be patched in later dynamically. This is likely only
28779 useful together with @option{-mrecord-mcount}.
28781 @item -minstrument-return=@var{type}
28782 @opindex minstrument-return
28783 Instrument function exit in -pg -mfentry instrumented functions with
28784 call to specified function. This only instruments true returns ending
28785 with ret, but not sibling calls ending with jump. Valid types
28786 are @var{none} to not instrument, @var{call} to generate a call to __return__,
28787 or @var{nop5} to generate a 5 byte nop.
28789 @item -mrecord-return
28790 @itemx -mno-record-return
28791 @opindex mrecord-return
28792 Generate a __return_loc section pointing to all return instrumentation code.
28794 @item -mfentry-name=@var{name}
28795 @opindex mfentry-name
28796 Set name of __fentry__ symbol called at function entry for -pg -mfentry functions.
28798 @item -mfentry-section=@var{name}
28799 @opindex mfentry-section
28800 Set name of section to record -mrecord-mcount calls (default __mcount_loc).
28802 @item -mskip-rax-setup
28803 @itemx -mno-skip-rax-setup
28804 @opindex mskip-rax-setup
28805 When generating code for the x86-64 architecture with SSE extensions
28806 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
28807 register when there are no variable arguments passed in vector registers.
28809 @strong{Warning:} Since RAX register is used to avoid unnecessarily
28810 saving vector registers on stack when passing variable arguments, the
28811 impacts of this option are callees may waste some stack space,
28812 misbehave or jump to a random location. GCC 4.4 or newer don't have
28813 those issues, regardless the RAX register value.
28816 @itemx -mno-8bit-idiv
28817 @opindex m8bit-idiv
28818 On some processors, like Intel Atom, 8-bit unsigned integer divide is
28819 much faster than 32-bit/64-bit integer divide. This option generates a
28820 run-time check. If both dividend and divisor are within range of 0
28821 to 255, 8-bit unsigned integer divide is used instead of
28822 32-bit/64-bit integer divide.
28824 @item -mavx256-split-unaligned-load
28825 @itemx -mavx256-split-unaligned-store
28826 @opindex mavx256-split-unaligned-load
28827 @opindex mavx256-split-unaligned-store
28828 Split 32-byte AVX unaligned load and store.
28830 @item -mstack-protector-guard=@var{guard}
28831 @itemx -mstack-protector-guard-reg=@var{reg}
28832 @itemx -mstack-protector-guard-offset=@var{offset}
28833 @opindex mstack-protector-guard
28834 @opindex mstack-protector-guard-reg
28835 @opindex mstack-protector-guard-offset
28836 Generate stack protection code using canary at @var{guard}. Supported
28837 locations are @samp{global} for global canary or @samp{tls} for per-thread
28838 canary in the TLS block (the default). This option has effect only when
28839 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
28841 With the latter choice the options
28842 @option{-mstack-protector-guard-reg=@var{reg}} and
28843 @option{-mstack-protector-guard-offset=@var{offset}} furthermore specify
28844 which segment register (@code{%fs} or @code{%gs}) to use as base register
28845 for reading the canary, and from what offset from that base register.
28846 The default for those is as specified in the relevant ABI.
28848 @item -mgeneral-regs-only
28849 @opindex mgeneral-regs-only
28850 Generate code that uses only the general-purpose registers. This
28851 prevents the compiler from using floating-point, vector, mask and bound
28854 @item -mindirect-branch=@var{choice}
28855 @opindex mindirect-branch
28856 Convert indirect call and jump with @var{choice}. The default is
28857 @samp{keep}, which keeps indirect call and jump unmodified.
28858 @samp{thunk} converts indirect call and jump to call and return thunk.
28859 @samp{thunk-inline} converts indirect call and jump to inlined call
28860 and return thunk. @samp{thunk-extern} converts indirect call and jump
28861 to external call and return thunk provided in a separate object file.
28862 You can control this behavior for a specific function by using the
28863 function attribute @code{indirect_branch}. @xref{Function Attributes}.
28865 Note that @option{-mcmodel=large} is incompatible with
28866 @option{-mindirect-branch=thunk} and
28867 @option{-mindirect-branch=thunk-extern} since the thunk function may
28868 not be reachable in the large code model.
28870 Note that @option{-mindirect-branch=thunk-extern} is incompatible with
28871 @option{-fcf-protection=branch} since the external thunk cannot be modified
28872 to disable control-flow check.
28874 @item -mfunction-return=@var{choice}
28875 @opindex mfunction-return
28876 Convert function return with @var{choice}. The default is @samp{keep},
28877 which keeps function return unmodified. @samp{thunk} converts function
28878 return to call and return thunk. @samp{thunk-inline} converts function
28879 return to inlined call and return thunk. @samp{thunk-extern} converts
28880 function return to external call and return thunk provided in a separate
28881 object file. You can control this behavior for a specific function by
28882 using the function attribute @code{function_return}.
28883 @xref{Function Attributes}.
28885 Note that @option{-mcmodel=large} is incompatible with
28886 @option{-mfunction-return=thunk} and
28887 @option{-mfunction-return=thunk-extern} since the thunk function may
28888 not be reachable in the large code model.
28891 @item -mindirect-branch-register
28892 @opindex mindirect-branch-register
28893 Force indirect call and jump via register.
28897 These @samp{-m} switches are supported in addition to the above
28898 on x86-64 processors in 64-bit environments.
28911 Generate code for a 16-bit, 32-bit or 64-bit environment.
28912 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
28914 generates code that runs on any i386 system.
28916 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
28917 types to 64 bits, and generates code for the x86-64 architecture.
28918 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
28919 and @option{-mdynamic-no-pic} options.
28921 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
28923 generates code for the x86-64 architecture.
28925 The @option{-m16} option is the same as @option{-m32}, except for that
28926 it outputs the @code{.code16gcc} assembly directive at the beginning of
28927 the assembly output so that the binary can run in 16-bit mode.
28929 The @option{-miamcu} option generates code which conforms to Intel MCU
28930 psABI. It requires the @option{-m32} option to be turned on.
28932 @item -mno-red-zone
28933 @opindex mno-red-zone
28935 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
28936 by the x86-64 ABI; it is a 128-byte area beyond the location of the
28937 stack pointer that is not modified by signal or interrupt handlers
28938 and therefore can be used for temporary data without adjusting the stack
28939 pointer. The flag @option{-mno-red-zone} disables this red zone.
28941 @item -mcmodel=small
28942 @opindex mcmodel=small
28943 Generate code for the small code model: the program and its symbols must
28944 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
28945 Programs can be statically or dynamically linked. This is the default
28948 @item -mcmodel=kernel
28949 @opindex mcmodel=kernel
28950 Generate code for the kernel code model. The kernel runs in the
28951 negative 2 GB of the address space.
28952 This model has to be used for Linux kernel code.
28954 @item -mcmodel=medium
28955 @opindex mcmodel=medium
28956 Generate code for the medium model: the program is linked in the lower 2
28957 GB of the address space. Small symbols are also placed there. Symbols
28958 with sizes larger than @option{-mlarge-data-threshold} are put into
28959 large data or BSS sections and can be located above 2GB. Programs can
28960 be statically or dynamically linked.
28962 @item -mcmodel=large
28963 @opindex mcmodel=large
28964 Generate code for the large model. This model makes no assumptions
28965 about addresses and sizes of sections.
28967 @item -maddress-mode=long
28968 @opindex maddress-mode=long
28969 Generate code for long address mode. This is only supported for 64-bit
28970 and x32 environments. It is the default address mode for 64-bit
28973 @item -maddress-mode=short
28974 @opindex maddress-mode=short
28975 Generate code for short address mode. This is only supported for 32-bit
28976 and x32 environments. It is the default address mode for 32-bit and
28980 @node x86 Windows Options
28981 @subsection x86 Windows Options
28982 @cindex x86 Windows Options
28983 @cindex Windows Options for x86
28985 These additional options are available for Microsoft Windows targets:
28991 specifies that a console application is to be generated, by
28992 instructing the linker to set the PE header subsystem type
28993 required for console applications.
28994 This option is available for Cygwin and MinGW targets and is
28995 enabled by default on those targets.
28999 This option is available for Cygwin and MinGW targets. It
29000 specifies that a DLL---a dynamic link library---is to be
29001 generated, enabling the selection of the required runtime
29002 startup object and entry point.
29004 @item -mnop-fun-dllimport
29005 @opindex mnop-fun-dllimport
29006 This option is available for Cygwin and MinGW targets. It
29007 specifies that the @code{dllimport} attribute should be ignored.
29011 This option is available for MinGW targets. It specifies
29012 that MinGW-specific thread support is to be used.
29016 This option is available for MinGW-w64 targets. It causes
29017 the @code{UNICODE} preprocessor macro to be predefined, and
29018 chooses Unicode-capable runtime startup code.
29022 This option is available for Cygwin and MinGW targets. It
29023 specifies that the typical Microsoft Windows predefined macros are to
29024 be set in the pre-processor, but does not influence the choice
29025 of runtime library/startup code.
29029 This option is available for Cygwin and MinGW targets. It
29030 specifies that a GUI application is to be generated by
29031 instructing the linker to set the PE header subsystem type
29034 @item -fno-set-stack-executable
29035 @opindex fno-set-stack-executable
29036 @opindex fset-stack-executable
29037 This option is available for MinGW targets. It specifies that
29038 the executable flag for the stack used by nested functions isn't
29039 set. This is necessary for binaries running in kernel mode of
29040 Microsoft Windows, as there the User32 API, which is used to set executable
29041 privileges, isn't available.
29043 @item -fwritable-relocated-rdata
29044 @opindex fno-writable-relocated-rdata
29045 @opindex fwritable-relocated-rdata
29046 This option is available for MinGW and Cygwin targets. It specifies
29047 that relocated-data in read-only section is put into the @code{.data}
29048 section. This is a necessary for older runtimes not supporting
29049 modification of @code{.rdata} sections for pseudo-relocation.
29051 @item -mpe-aligned-commons
29052 @opindex mpe-aligned-commons
29053 This option is available for Cygwin and MinGW targets. It
29054 specifies that the GNU extension to the PE file format that
29055 permits the correct alignment of COMMON variables should be
29056 used when generating code. It is enabled by default if
29057 GCC detects that the target assembler found during configuration
29058 supports the feature.
29061 See also under @ref{x86 Options} for standard options.
29063 @node Xstormy16 Options
29064 @subsection Xstormy16 Options
29065 @cindex Xstormy16 Options
29067 These options are defined for Xstormy16:
29072 Choose startup files and linker script suitable for the simulator.
29075 @node Xtensa Options
29076 @subsection Xtensa Options
29077 @cindex Xtensa Options
29079 These options are supported for Xtensa targets:
29083 @itemx -mno-const16
29085 @opindex mno-const16
29086 Enable or disable use of @code{CONST16} instructions for loading
29087 constant values. The @code{CONST16} instruction is currently not a
29088 standard option from Tensilica. When enabled, @code{CONST16}
29089 instructions are always used in place of the standard @code{L32R}
29090 instructions. The use of @code{CONST16} is enabled by default only if
29091 the @code{L32R} instruction is not available.
29094 @itemx -mno-fused-madd
29095 @opindex mfused-madd
29096 @opindex mno-fused-madd
29097 Enable or disable use of fused multiply/add and multiply/subtract
29098 instructions in the floating-point option. This has no effect if the
29099 floating-point option is not also enabled. Disabling fused multiply/add
29100 and multiply/subtract instructions forces the compiler to use separate
29101 instructions for the multiply and add/subtract operations. This may be
29102 desirable in some cases where strict IEEE 754-compliant results are
29103 required: the fused multiply add/subtract instructions do not round the
29104 intermediate result, thereby producing results with @emph{more} bits of
29105 precision than specified by the IEEE standard. Disabling fused multiply
29106 add/subtract instructions also ensures that the program output is not
29107 sensitive to the compiler's ability to combine multiply and add/subtract
29110 @item -mserialize-volatile
29111 @itemx -mno-serialize-volatile
29112 @opindex mserialize-volatile
29113 @opindex mno-serialize-volatile
29114 When this option is enabled, GCC inserts @code{MEMW} instructions before
29115 @code{volatile} memory references to guarantee sequential consistency.
29116 The default is @option{-mserialize-volatile}. Use
29117 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
29119 @item -mforce-no-pic
29120 @opindex mforce-no-pic
29121 For targets, like GNU/Linux, where all user-mode Xtensa code must be
29122 position-independent code (PIC), this option disables PIC for compiling
29125 @item -mtext-section-literals
29126 @itemx -mno-text-section-literals
29127 @opindex mtext-section-literals
29128 @opindex mno-text-section-literals
29129 These options control the treatment of literal pools. The default is
29130 @option{-mno-text-section-literals}, which places literals in a separate
29131 section in the output file. This allows the literal pool to be placed
29132 in a data RAM/ROM, and it also allows the linker to combine literal
29133 pools from separate object files to remove redundant literals and
29134 improve code size. With @option{-mtext-section-literals}, the literals
29135 are interspersed in the text section in order to keep them as close as
29136 possible to their references. This may be necessary for large assembly
29137 files. Literals for each function are placed right before that function.
29139 @item -mauto-litpools
29140 @itemx -mno-auto-litpools
29141 @opindex mauto-litpools
29142 @opindex mno-auto-litpools
29143 These options control the treatment of literal pools. The default is
29144 @option{-mno-auto-litpools}, which places literals in a separate
29145 section in the output file unless @option{-mtext-section-literals} is
29146 used. With @option{-mauto-litpools} the literals are interspersed in
29147 the text section by the assembler. Compiler does not produce explicit
29148 @code{.literal} directives and loads literals into registers with
29149 @code{MOVI} instructions instead of @code{L32R} to let the assembler
29150 do relaxation and place literals as necessary. This option allows
29151 assembler to create several literal pools per function and assemble
29152 very big functions, which may not be possible with
29153 @option{-mtext-section-literals}.
29155 @item -mtarget-align
29156 @itemx -mno-target-align
29157 @opindex mtarget-align
29158 @opindex mno-target-align
29159 When this option is enabled, GCC instructs the assembler to
29160 automatically align instructions to reduce branch penalties at the
29161 expense of some code density. The assembler attempts to widen density
29162 instructions to align branch targets and the instructions following call
29163 instructions. If there are not enough preceding safe density
29164 instructions to align a target, no widening is performed. The
29165 default is @option{-mtarget-align}. These options do not affect the
29166 treatment of auto-aligned instructions like @code{LOOP}, which the
29167 assembler always aligns, either by widening density instructions or
29168 by inserting NOP instructions.
29171 @itemx -mno-longcalls
29172 @opindex mlongcalls
29173 @opindex mno-longcalls
29174 When this option is enabled, GCC instructs the assembler to translate
29175 direct calls to indirect calls unless it can determine that the target
29176 of a direct call is in the range allowed by the call instruction. This
29177 translation typically occurs for calls to functions in other source
29178 files. Specifically, the assembler translates a direct @code{CALL}
29179 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
29180 The default is @option{-mno-longcalls}. This option should be used in
29181 programs where the call target can potentially be out of range. This
29182 option is implemented in the assembler, not the compiler, so the
29183 assembly code generated by GCC still shows direct call
29184 instructions---look at the disassembled object code to see the actual
29185 instructions. Note that the assembler uses an indirect call for
29186 every cross-file call, not just those that really are out of range.
29189 @node zSeries Options
29190 @subsection zSeries Options
29191 @cindex zSeries options
29193 These are listed under @xref{S/390 and zSeries Options}.
29199 @section Specifying Subprocesses and the Switches to Pass to Them
29202 @command{gcc} is a driver program. It performs its job by invoking a
29203 sequence of other programs to do the work of compiling, assembling and
29204 linking. GCC interprets its command-line parameters and uses these to
29205 deduce which programs it should invoke, and which command-line options
29206 it ought to place on their command lines. This behavior is controlled
29207 by @dfn{spec strings}. In most cases there is one spec string for each
29208 program that GCC can invoke, but a few programs have multiple spec
29209 strings to control their behavior. The spec strings built into GCC can
29210 be overridden by using the @option{-specs=} command-line switch to specify
29213 @dfn{Spec files} are plain-text files that are used to construct spec
29214 strings. They consist of a sequence of directives separated by blank
29215 lines. The type of directive is determined by the first non-whitespace
29216 character on the line, which can be one of the following:
29219 @item %@var{command}
29220 Issues a @var{command} to the spec file processor. The commands that can
29224 @item %include <@var{file}>
29225 @cindex @code{%include}
29226 Search for @var{file} and insert its text at the current point in the
29229 @item %include_noerr <@var{file}>
29230 @cindex @code{%include_noerr}
29231 Just like @samp{%include}, but do not generate an error message if the include
29232 file cannot be found.
29234 @item %rename @var{old_name} @var{new_name}
29235 @cindex @code{%rename}
29236 Rename the spec string @var{old_name} to @var{new_name}.
29240 @item *[@var{spec_name}]:
29241 This tells the compiler to create, override or delete the named spec
29242 string. All lines after this directive up to the next directive or
29243 blank line are considered to be the text for the spec string. If this
29244 results in an empty string then the spec is deleted. (Or, if the
29245 spec did not exist, then nothing happens.) Otherwise, if the spec
29246 does not currently exist a new spec is created. If the spec does
29247 exist then its contents are overridden by the text of this
29248 directive, unless the first character of that text is the @samp{+}
29249 character, in which case the text is appended to the spec.
29251 @item [@var{suffix}]:
29252 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
29253 and up to the next directive or blank line are considered to make up the
29254 spec string for the indicated suffix. When the compiler encounters an
29255 input file with the named suffix, it processes the spec string in
29256 order to work out how to compile that file. For example:
29260 z-compile -input %i
29263 This says that any input file whose name ends in @samp{.ZZ} should be
29264 passed to the program @samp{z-compile}, which should be invoked with the
29265 command-line switch @option{-input} and with the result of performing the
29266 @samp{%i} substitution. (See below.)
29268 As an alternative to providing a spec string, the text following a
29269 suffix directive can be one of the following:
29272 @item @@@var{language}
29273 This says that the suffix is an alias for a known @var{language}. This is
29274 similar to using the @option{-x} command-line switch to GCC to specify a
29275 language explicitly. For example:
29282 Says that .ZZ files are, in fact, C++ source files.
29285 This causes an error messages saying:
29288 @var{name} compiler not installed on this system.
29292 GCC already has an extensive list of suffixes built into it.
29293 This directive adds an entry to the end of the list of suffixes, but
29294 since the list is searched from the end backwards, it is effectively
29295 possible to override earlier entries using this technique.
29299 GCC has the following spec strings built into it. Spec files can
29300 override these strings or create their own. Note that individual
29301 targets can also add their own spec strings to this list.
29304 asm Options to pass to the assembler
29305 asm_final Options to pass to the assembler post-processor
29306 cpp Options to pass to the C preprocessor
29307 cc1 Options to pass to the C compiler
29308 cc1plus Options to pass to the C++ compiler
29309 endfile Object files to include at the end of the link
29310 link Options to pass to the linker
29311 lib Libraries to include on the command line to the linker
29312 libgcc Decides which GCC support library to pass to the linker
29313 linker Sets the name of the linker
29314 predefines Defines to be passed to the C preprocessor
29315 signed_char Defines to pass to CPP to say whether @code{char} is signed
29317 startfile Object files to include at the start of the link
29320 Here is a small example of a spec file:
29323 %rename lib old_lib
29326 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
29329 This example renames the spec called @samp{lib} to @samp{old_lib} and
29330 then overrides the previous definition of @samp{lib} with a new one.
29331 The new definition adds in some extra command-line options before
29332 including the text of the old definition.
29334 @dfn{Spec strings} are a list of command-line options to be passed to their
29335 corresponding program. In addition, the spec strings can contain
29336 @samp{%}-prefixed sequences to substitute variable text or to
29337 conditionally insert text into the command line. Using these constructs
29338 it is possible to generate quite complex command lines.
29340 Here is a table of all defined @samp{%}-sequences for spec
29341 strings. Note that spaces are not generated automatically around the
29342 results of expanding these sequences. Therefore you can concatenate them
29343 together or combine them with constant text in a single argument.
29347 Substitute one @samp{%} into the program name or argument.
29350 Substitute the name of the input file being processed.
29353 Substitute the basename of the input file being processed.
29354 This is the substring up to (and not including) the last period
29355 and not including the directory.
29358 This is the same as @samp{%b}, but include the file suffix (text after
29362 Marks the argument containing or following the @samp{%d} as a
29363 temporary file name, so that that file is deleted if GCC exits
29364 successfully. Unlike @samp{%g}, this contributes no text to the
29367 @item %g@var{suffix}
29368 Substitute a file name that has suffix @var{suffix} and is chosen
29369 once per compilation, and mark the argument in the same way as
29370 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
29371 name is now chosen in a way that is hard to predict even when previously
29372 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
29373 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
29374 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
29375 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
29376 was simply substituted with a file name chosen once per compilation,
29377 without regard to any appended suffix (which was therefore treated
29378 just like ordinary text), making such attacks more likely to succeed.
29380 @item %u@var{suffix}
29381 Like @samp{%g}, but generates a new temporary file name
29382 each time it appears instead of once per compilation.
29384 @item %U@var{suffix}
29385 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
29386 new one if there is no such last file name. In the absence of any
29387 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
29388 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
29389 involves the generation of two distinct file names, one
29390 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
29391 simply substituted with a file name chosen for the previous @samp{%u},
29392 without regard to any appended suffix.
29394 @item %j@var{suffix}
29395 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
29396 writable, and if @option{-save-temps} is not used;
29397 otherwise, substitute the name
29398 of a temporary file, just like @samp{%u}. This temporary file is not
29399 meant for communication between processes, but rather as a junk
29400 disposal mechanism.
29402 @item %|@var{suffix}
29403 @itemx %m@var{suffix}
29404 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
29405 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
29406 all. These are the two most common ways to instruct a program that it
29407 should read from standard input or write to standard output. If you
29408 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
29409 construct: see for example @file{gcc/fortran/lang-specs.h}.
29411 @item %.@var{SUFFIX}
29412 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
29413 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
29414 terminated by the next space or %.
29417 Marks the argument containing or following the @samp{%w} as the
29418 designated output file of this compilation. This puts the argument
29419 into the sequence of arguments that @samp{%o} substitutes.
29422 Substitutes the names of all the output files, with spaces
29423 automatically placed around them. You should write spaces
29424 around the @samp{%o} as well or the results are undefined.
29425 @samp{%o} is for use in the specs for running the linker.
29426 Input files whose names have no recognized suffix are not compiled
29427 at all, but they are included among the output files, so they are
29431 Substitutes the suffix for object files. Note that this is
29432 handled specially when it immediately follows @samp{%g, %u, or %U},
29433 because of the need for those to form complete file names. The
29434 handling is such that @samp{%O} is treated exactly as if it had already
29435 been substituted, except that @samp{%g, %u, and %U} do not currently
29436 support additional @var{suffix} characters following @samp{%O} as they do
29437 following, for example, @samp{.o}.
29440 Substitutes the standard macro predefinitions for the
29441 current target machine. Use this when running @command{cpp}.
29444 Like @samp{%p}, but puts @samp{__} before and after the name of each
29445 predefined macro, except for macros that start with @samp{__} or with
29446 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
29450 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
29451 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
29452 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
29453 and @option{-imultilib} as necessary.
29456 Current argument is the name of a library or startup file of some sort.
29457 Search for that file in a standard list of directories and substitute
29458 the full name found. The current working directory is included in the
29459 list of directories scanned.
29462 Current argument is the name of a linker script. Search for that file
29463 in the current list of directories to scan for libraries. If the file
29464 is located insert a @option{--script} option into the command line
29465 followed by the full path name found. If the file is not found then
29466 generate an error message. Note: the current working directory is not
29470 Print @var{str} as an error message. @var{str} is terminated by a newline.
29471 Use this when inconsistent options are detected.
29473 @item %(@var{name})
29474 Substitute the contents of spec string @var{name} at this point.
29476 @item %x@{@var{option}@}
29477 Accumulate an option for @samp{%X}.
29480 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
29484 Output the accumulated assembler options specified by @option{-Wa}.
29487 Output the accumulated preprocessor options specified by @option{-Wp}.
29490 Process the @code{asm} spec. This is used to compute the
29491 switches to be passed to the assembler.
29494 Process the @code{asm_final} spec. This is a spec string for
29495 passing switches to an assembler post-processor, if such a program is
29499 Process the @code{link} spec. This is the spec for computing the
29500 command line passed to the linker. Typically it makes use of the
29501 @samp{%L %G %S %D and %E} sequences.
29504 Dump out a @option{-L} option for each directory that GCC believes might
29505 contain startup files. If the target supports multilibs then the
29506 current multilib directory is prepended to each of these paths.
29509 Process the @code{lib} spec. This is a spec string for deciding which
29510 libraries are included on the command line to the linker.
29513 Process the @code{libgcc} spec. This is a spec string for deciding
29514 which GCC support library is included on the command line to the linker.
29517 Process the @code{startfile} spec. This is a spec for deciding which
29518 object files are the first ones passed to the linker. Typically
29519 this might be a file named @file{crt0.o}.
29522 Process the @code{endfile} spec. This is a spec string that specifies
29523 the last object files that are passed to the linker.
29526 Process the @code{cpp} spec. This is used to construct the arguments
29527 to be passed to the C preprocessor.
29530 Process the @code{cc1} spec. This is used to construct the options to be
29531 passed to the actual C compiler (@command{cc1}).
29534 Process the @code{cc1plus} spec. This is used to construct the options to be
29535 passed to the actual C++ compiler (@command{cc1plus}).
29538 Substitute the variable part of a matched option. See below.
29539 Note that each comma in the substituted string is replaced by
29543 Remove all occurrences of @code{-S} from the command line. Note---this
29544 command is position dependent. @samp{%} commands in the spec string
29545 before this one see @code{-S}, @samp{%} commands in the spec string
29546 after this one do not.
29548 @item %:@var{function}(@var{args})
29549 Call the named function @var{function}, passing it @var{args}.
29550 @var{args} is first processed as a nested spec string, then split
29551 into an argument vector in the usual fashion. The function returns
29552 a string which is processed as if it had appeared literally as part
29553 of the current spec.
29555 The following built-in spec functions are provided:
29558 @item @code{getenv}
29559 The @code{getenv} spec function takes two arguments: an environment
29560 variable name and a string. If the environment variable is not
29561 defined, a fatal error is issued. Otherwise, the return value is the
29562 value of the environment variable concatenated with the string. For
29563 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
29566 %:getenv(TOPDIR /include)
29569 expands to @file{/path/to/top/include}.
29571 @item @code{if-exists}
29572 The @code{if-exists} spec function takes one argument, an absolute
29573 pathname to a file. If the file exists, @code{if-exists} returns the
29574 pathname. Here is a small example of its usage:
29578 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
29581 @item @code{if-exists-else}
29582 The @code{if-exists-else} spec function is similar to the @code{if-exists}
29583 spec function, except that it takes two arguments. The first argument is
29584 an absolute pathname to a file. If the file exists, @code{if-exists-else}
29585 returns the pathname. If it does not exist, it returns the second argument.
29586 This way, @code{if-exists-else} can be used to select one file or another,
29587 based on the existence of the first. Here is a small example of its usage:
29591 crt0%O%s %:if-exists(crti%O%s) \
29592 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
29595 @item @code{replace-outfile}
29596 The @code{replace-outfile} spec function takes two arguments. It looks for the
29597 first argument in the outfiles array and replaces it with the second argument. Here
29598 is a small example of its usage:
29601 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
29604 @item @code{remove-outfile}
29605 The @code{remove-outfile} spec function takes one argument. It looks for the
29606 first argument in the outfiles array and removes it. Here is a small example
29610 %:remove-outfile(-lm)
29613 @item @code{pass-through-libs}
29614 The @code{pass-through-libs} spec function takes any number of arguments. It
29615 finds any @option{-l} options and any non-options ending in @file{.a} (which it
29616 assumes are the names of linker input library archive files) and returns a
29617 result containing all the found arguments each prepended by
29618 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
29619 intended to be passed to the LTO linker plugin.
29622 %:pass-through-libs(%G %L %G)
29625 @item @code{print-asm-header}
29626 The @code{print-asm-header} function takes no arguments and simply
29627 prints a banner like:
29633 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
29636 It is used to separate compiler options from assembler options
29637 in the @option{--target-help} output.
29641 Substitutes the @code{-S} switch, if that switch is given to GCC@.
29642 If that switch is not specified, this substitutes nothing. Note that
29643 the leading dash is omitted when specifying this option, and it is
29644 automatically inserted if the substitution is performed. Thus the spec
29645 string @samp{%@{foo@}} matches the command-line option @option{-foo}
29646 and outputs the command-line option @option{-foo}.
29649 Like %@{@code{S}@} but mark last argument supplied within as a file to be
29650 deleted on failure.
29653 Substitutes all the switches specified to GCC whose names start
29654 with @code{-S}, but which also take an argument. This is used for
29655 switches like @option{-o}, @option{-D}, @option{-I}, etc.
29656 GCC considers @option{-o foo} as being
29657 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
29658 text, including the space. Thus two arguments are generated.
29661 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
29662 (the order of @code{S} and @code{T} in the spec is not significant).
29663 There can be any number of ampersand-separated variables; for each the
29664 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
29667 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
29670 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
29673 Substitutes @code{X} if one or more switches whose names start with
29674 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
29675 once, no matter how many such switches appeared. However, if @code{%*}
29676 appears somewhere in @code{X}, then @code{X} is substituted once
29677 for each matching switch, with the @code{%*} replaced by the part of
29678 that switch matching the @code{*}.
29680 If @code{%*} appears as the last part of a spec sequence then a space
29681 is added after the end of the last substitution. If there is more
29682 text in the sequence, however, then a space is not generated. This
29683 allows the @code{%*} substitution to be used as part of a larger
29684 string. For example, a spec string like this:
29687 %@{mcu=*:--script=%*/memory.ld@}
29691 when matching an option like @option{-mcu=newchip} produces:
29694 --script=newchip/memory.ld
29698 Substitutes @code{X}, if processing a file with suffix @code{S}.
29701 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
29704 Substitutes @code{X}, if processing a file for language @code{S}.
29707 Substitutes @code{X}, if not processing a file for language @code{S}.
29710 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
29711 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
29712 @code{*} sequences as well, although they have a stronger binding than
29713 the @samp{|}. If @code{%*} appears in @code{X}, all of the
29714 alternatives must be starred, and only the first matching alternative
29717 For example, a spec string like this:
29720 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
29724 outputs the following command-line options from the following input
29725 command-line options:
29730 -d fred.c -foo -baz -boggle
29731 -d jim.d -bar -baz -boggle
29734 @item %@{S:X; T:Y; :D@}
29736 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
29737 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
29738 be as many clauses as you need. This may be combined with @code{.},
29739 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
29744 The switch matching text @code{S} in a @samp{%@{S@}}, @samp{%@{S:X@}}
29745 or similar construct can use a backslash to ignore the special meaning
29746 of the character following it, thus allowing literal matching of a
29747 character that is otherwise specially treated. For example,
29748 @samp{%@{std=iso9899\:1999:X@}} substitutes @code{X} if the
29749 @option{-std=iso9899:1999} option is given.
29751 The conditional text @code{X} in a @samp{%@{S:X@}} or similar
29752 construct may contain other nested @samp{%} constructs or spaces, or
29753 even newlines. They are processed as usual, as described above.
29754 Trailing white space in @code{X} is ignored. White space may also
29755 appear anywhere on the left side of the colon in these constructs,
29756 except between @code{.} or @code{*} and the corresponding word.
29758 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
29759 handled specifically in these constructs. If another value of
29760 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
29761 @option{-W} switch is found later in the command line, the earlier
29762 switch value is ignored, except with @{@code{S}*@} where @code{S} is
29763 just one letter, which passes all matching options.
29765 The character @samp{|} at the beginning of the predicate text is used to
29766 indicate that a command should be piped to the following command, but
29767 only if @option{-pipe} is specified.
29769 It is built into GCC which switches take arguments and which do not.
29770 (You might think it would be useful to generalize this to allow each
29771 compiler's spec to say which switches take arguments. But this cannot
29772 be done in a consistent fashion. GCC cannot even decide which input
29773 files have been specified without knowing which switches take arguments,
29774 and it must know which input files to compile in order to tell which
29777 GCC also knows implicitly that arguments starting in @option{-l} are to be
29778 treated as compiler output files, and passed to the linker in their
29779 proper position among the other output files.
29781 @node Environment Variables
29782 @section Environment Variables Affecting GCC
29783 @cindex environment variables
29785 @c man begin ENVIRONMENT
29786 This section describes several environment variables that affect how GCC
29787 operates. Some of them work by specifying directories or prefixes to use
29788 when searching for various kinds of files. Some are used to specify other
29789 aspects of the compilation environment.
29791 Note that you can also specify places to search using options such as
29792 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
29793 take precedence over places specified using environment variables, which
29794 in turn take precedence over those specified by the configuration of GCC@.
29795 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
29796 GNU Compiler Collection (GCC) Internals}.
29801 @c @itemx LC_COLLATE
29803 @c @itemx LC_MONETARY
29804 @c @itemx LC_NUMERIC
29809 @c @findex LC_COLLATE
29810 @findex LC_MESSAGES
29811 @c @findex LC_MONETARY
29812 @c @findex LC_NUMERIC
29816 These environment variables control the way that GCC uses
29817 localization information which allows GCC to work with different
29818 national conventions. GCC inspects the locale categories
29819 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
29820 so. These locale categories can be set to any value supported by your
29821 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
29822 Kingdom encoded in UTF-8.
29824 The @env{LC_CTYPE} environment variable specifies character
29825 classification. GCC uses it to determine the character boundaries in
29826 a string; this is needed for some multibyte encodings that contain quote
29827 and escape characters that are otherwise interpreted as a string
29830 The @env{LC_MESSAGES} environment variable specifies the language to
29831 use in diagnostic messages.
29833 If the @env{LC_ALL} environment variable is set, it overrides the value
29834 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
29835 and @env{LC_MESSAGES} default to the value of the @env{LANG}
29836 environment variable. If none of these variables are set, GCC
29837 defaults to traditional C English behavior.
29841 If @env{TMPDIR} is set, it specifies the directory to use for temporary
29842 files. GCC uses temporary files to hold the output of one stage of
29843 compilation which is to be used as input to the next stage: for example,
29844 the output of the preprocessor, which is the input to the compiler
29847 @item GCC_COMPARE_DEBUG
29848 @findex GCC_COMPARE_DEBUG
29849 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
29850 @option{-fcompare-debug} to the compiler driver. See the documentation
29851 of this option for more details.
29853 @item GCC_EXEC_PREFIX
29854 @findex GCC_EXEC_PREFIX
29855 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
29856 names of the subprograms executed by the compiler. No slash is added
29857 when this prefix is combined with the name of a subprogram, but you can
29858 specify a prefix that ends with a slash if you wish.
29860 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
29861 an appropriate prefix to use based on the pathname it is invoked with.
29863 If GCC cannot find the subprogram using the specified prefix, it
29864 tries looking in the usual places for the subprogram.
29866 The default value of @env{GCC_EXEC_PREFIX} is
29867 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
29868 the installed compiler. In many cases @var{prefix} is the value
29869 of @code{prefix} when you ran the @file{configure} script.
29871 Other prefixes specified with @option{-B} take precedence over this prefix.
29873 This prefix is also used for finding files such as @file{crt0.o} that are
29876 In addition, the prefix is used in an unusual way in finding the
29877 directories to search for header files. For each of the standard
29878 directories whose name normally begins with @samp{/usr/local/lib/gcc}
29879 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
29880 replacing that beginning with the specified prefix to produce an
29881 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
29882 @file{foo/bar} just before it searches the standard directory
29883 @file{/usr/local/lib/bar}.
29884 If a standard directory begins with the configured
29885 @var{prefix} then the value of @var{prefix} is replaced by
29886 @env{GCC_EXEC_PREFIX} when looking for header files.
29888 @item COMPILER_PATH
29889 @findex COMPILER_PATH
29890 The value of @env{COMPILER_PATH} is a colon-separated list of
29891 directories, much like @env{PATH}. GCC tries the directories thus
29892 specified when searching for subprograms, if it cannot find the
29893 subprograms using @env{GCC_EXEC_PREFIX}.
29896 @findex LIBRARY_PATH
29897 The value of @env{LIBRARY_PATH} is a colon-separated list of
29898 directories, much like @env{PATH}. When configured as a native compiler,
29899 GCC tries the directories thus specified when searching for special
29900 linker files, if it cannot find them using @env{GCC_EXEC_PREFIX}. Linking
29901 using GCC also uses these directories when searching for ordinary
29902 libraries for the @option{-l} option (but directories specified with
29903 @option{-L} come first).
29907 @cindex locale definition
29908 This variable is used to pass locale information to the compiler. One way in
29909 which this information is used is to determine the character set to be used
29910 when character literals, string literals and comments are parsed in C and C++.
29911 When the compiler is configured to allow multibyte characters,
29912 the following values for @env{LANG} are recognized:
29916 Recognize JIS characters.
29918 Recognize SJIS characters.
29920 Recognize EUCJP characters.
29923 If @env{LANG} is not defined, or if it has some other value, then the
29924 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
29925 recognize and translate multibyte characters.
29929 Some additional environment variables affect the behavior of the
29932 @include cppenv.texi
29936 @node Precompiled Headers
29937 @section Using Precompiled Headers
29938 @cindex precompiled headers
29939 @cindex speed of compilation
29941 Often large projects have many header files that are included in every
29942 source file. The time the compiler takes to process these header files
29943 over and over again can account for nearly all of the time required to
29944 build the project. To make builds faster, GCC allows you to
29945 @dfn{precompile} a header file.
29947 To create a precompiled header file, simply compile it as you would any
29948 other file, if necessary using the @option{-x} option to make the driver
29949 treat it as a C or C++ header file. You may want to use a
29950 tool like @command{make} to keep the precompiled header up-to-date when
29951 the headers it contains change.
29953 A precompiled header file is searched for when @code{#include} is
29954 seen in the compilation. As it searches for the included file
29955 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
29956 compiler looks for a precompiled header in each directory just before it
29957 looks for the include file in that directory. The name searched for is
29958 the name specified in the @code{#include} with @samp{.gch} appended. If
29959 the precompiled header file cannot be used, it is ignored.
29961 For instance, if you have @code{#include "all.h"}, and you have
29962 @file{all.h.gch} in the same directory as @file{all.h}, then the
29963 precompiled header file is used if possible, and the original
29964 header is used otherwise.
29966 Alternatively, you might decide to put the precompiled header file in a
29967 directory and use @option{-I} to ensure that directory is searched
29968 before (or instead of) the directory containing the original header.
29969 Then, if you want to check that the precompiled header file is always
29970 used, you can put a file of the same name as the original header in this
29971 directory containing an @code{#error} command.
29973 This also works with @option{-include}. So yet another way to use
29974 precompiled headers, good for projects not designed with precompiled
29975 header files in mind, is to simply take most of the header files used by
29976 a project, include them from another header file, precompile that header
29977 file, and @option{-include} the precompiled header. If the header files
29978 have guards against multiple inclusion, they are skipped because
29979 they've already been included (in the precompiled header).
29981 If you need to precompile the same header file for different
29982 languages, targets, or compiler options, you can instead make a
29983 @emph{directory} named like @file{all.h.gch}, and put each precompiled
29984 header in the directory, perhaps using @option{-o}. It doesn't matter
29985 what you call the files in the directory; every precompiled header in
29986 the directory is considered. The first precompiled header
29987 encountered in the directory that is valid for this compilation is
29988 used; they're searched in no particular order.
29990 There are many other possibilities, limited only by your imagination,
29991 good sense, and the constraints of your build system.
29993 A precompiled header file can be used only when these conditions apply:
29997 Only one precompiled header can be used in a particular compilation.
30000 A precompiled header cannot be used once the first C token is seen. You
30001 can have preprocessor directives before a precompiled header; you cannot
30002 include a precompiled header from inside another header.
30005 The precompiled header file must be produced for the same language as
30006 the current compilation. You cannot use a C precompiled header for a C++
30010 The precompiled header file must have been produced by the same compiler
30011 binary as the current compilation is using.
30014 Any macros defined before the precompiled header is included must
30015 either be defined in the same way as when the precompiled header was
30016 generated, or must not affect the precompiled header, which usually
30017 means that they don't appear in the precompiled header at all.
30019 The @option{-D} option is one way to define a macro before a
30020 precompiled header is included; using a @code{#define} can also do it.
30021 There are also some options that define macros implicitly, like
30022 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
30025 @item If debugging information is output when using the precompiled
30026 header, using @option{-g} or similar, the same kind of debugging information
30027 must have been output when building the precompiled header. However,
30028 a precompiled header built using @option{-g} can be used in a compilation
30029 when no debugging information is being output.
30031 @item The same @option{-m} options must generally be used when building
30032 and using the precompiled header. @xref{Submodel Options},
30033 for any cases where this rule is relaxed.
30035 @item Each of the following options must be the same when building and using
30036 the precompiled header:
30038 @gccoptlist{-fexceptions}
30041 Some other command-line options starting with @option{-f},
30042 @option{-p}, or @option{-O} must be defined in the same way as when
30043 the precompiled header was generated. At present, it's not clear
30044 which options are safe to change and which are not; the safest choice
30045 is to use exactly the same options when generating and using the
30046 precompiled header. The following are known to be safe:
30048 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
30049 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
30050 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
30055 For all of these except the last, the compiler automatically
30056 ignores the precompiled header if the conditions aren't met. If you
30057 find an option combination that doesn't work and doesn't cause the
30058 precompiled header to be ignored, please consider filing a bug report,
30061 If you do use differing options when generating and using the
30062 precompiled header, the actual behavior is a mixture of the
30063 behavior for the options. For instance, if you use @option{-g} to
30064 generate the precompiled header but not when using it, you may or may
30065 not get debugging information for routines in the precompiled header.